Add files using upload-large-folder tool

ckpts/universal/global_step40/zero/16.mlp.dense_4h_to_h.weight/exp_avg_sq.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e89bf2e55e35ed0f05fbec26ef376f030a1b946484f40f2c9c0c827f184f1889
+size 33555627

ckpts/universal/global_step40/zero/17.mlp.dense_h_to_4h.weight/exp_avg_sq.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbc680c6b7aa23968e96f914f4e462357b4fb538d4a269593c227c67d4f4bd1c
+size 33555627

ckpts/universal/global_step40/zero/17.mlp.dense_h_to_4h.weight/fp32.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:808d89a48271846ad48d88b3d8c563a03d2650c4ba42a020a2b2e9e35f2294d1
+size 33555533

ckpts/universal/global_step40/zero/6.attention.dense.weight/exp_avg.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be0cdd8a27794a9b0901b27c872baccf9e5ff24b3ac7505f9b3235390c31e9ea
+size 16778396

ckpts/universal/global_step40/zero/6.attention.dense.weight/exp_avg_sq.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f05953525977728553a39a506e566c6611353f64cb94b6053cd02e152cef9964
+size 16778411

ckpts/universal/global_step40/zero/6.attention.dense.weight/fp32.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6c7ae9a86ef8973269e9560eb05c8eba5ea41f355259ba690fc2ac7c0efb0c2
+size 16778317

venv/lib/python3.10/site-packages/transformers/models/autoformer/__init__.py

@@ -0,0 +1,63 @@
+# Copyright 2023 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from typing import TYPE_CHECKING
+
+# rely on isort to merge the imports
+from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
+
+
+_import_structure = {
+    "configuration_autoformer": [
+        "AUTOFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP",
+        "AutoformerConfig",
+    ],
+}
+
+try:
+    if not is_torch_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["modeling_autoformer"] = [
+        "AUTOFORMER_PRETRAINED_MODEL_ARCHIVE_LIST",
+        "AutoformerForPrediction",
+        "AutoformerModel",
+        "AutoformerPreTrainedModel",
+    ]
+
+
+if TYPE_CHECKING:
+    from .configuration_autoformer import (
+        AUTOFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        AutoformerConfig,
+    )
+
+    try:
+        if not is_torch_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .modeling_autoformer import (
+            AUTOFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
+            AutoformerForPrediction,
+            AutoformerModel,
+            AutoformerPreTrainedModel,
+        )
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)

venv/lib/python3.10/site-packages/transformers/models/autoformer/configuration_autoformer.py

@@ -0,0 +1,245 @@
+# coding=utf-8
+# Copyright 2023 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" Autoformer model configuration"""
+
+from typing import List, Optional
+
+from ...configuration_utils import PretrainedConfig
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+
+from ..deprecated._archive_maps import AUTOFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP  # noqa: F401, E402
+
+
+class AutoformerConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of an [`AutoformerModel`]. It is used to instantiate an
+    Autoformer model according to the specified arguments, defining the model architecture. Instantiating a
+    configuration with the defaults will yield a similar configuration to that of the Autoformer
+    [huggingface/autoformer-tourism-monthly](https://huggingface.co/huggingface/autoformer-tourism-monthly)
+    architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        prediction_length (`int`):
+            The prediction length for the decoder. In other words, the prediction horizon of the model.
+        context_length (`int`, *optional*, defaults to `prediction_length`):
+            The context length for the encoder. If unset, the context length will be the same as the
+            `prediction_length`.
+        distribution_output (`string`, *optional*, defaults to `"student_t"`):
+            The distribution emission head for the model. Could be either "student_t", "normal" or "negative_binomial".
+        loss (`string`, *optional*, defaults to `"nll"`):
+            The loss function for the model corresponding to the `distribution_output` head. For parametric
+            distributions it is the negative log likelihood (nll) - which currently is the only supported one.
+        input_size (`int`, *optional*, defaults to 1):
+            The size of the target variable which by default is 1 for univariate targets. Would be > 1 in case of
+            multivariate targets.
+        lags_sequence (`list[int]`, *optional*, defaults to `[1, 2, 3, 4, 5, 6, 7]`):
+            The lags of the input time series as covariates often dictated by the frequency. Default is `[1, 2, 3, 4,
+            5, 6, 7]`.
+        scaling (`bool`, *optional* defaults to `True`):
+            Whether to scale the input targets.
+        num_time_features (`int`, *optional*, defaults to 0):
+            The number of time features in the input time series.
+        num_dynamic_real_features (`int`, *optional*, defaults to 0):
+            The number of dynamic real valued features.
+        num_static_categorical_features (`int`, *optional*, defaults to 0):
+            The number of static categorical features.
+        num_static_real_features (`int`, *optional*, defaults to 0):
+            The number of static real valued features.
+        cardinality (`list[int]`, *optional*):
+            The cardinality (number of different values) for each of the static categorical features. Should be a list
+            of integers, having the same length as `num_static_categorical_features`. Cannot be `None` if
+            `num_static_categorical_features` is > 0.
+        embedding_dimension (`list[int]`, *optional*):
+            The dimension of the embedding for each of the static categorical features. Should be a list of integers,
+            having the same length as `num_static_categorical_features`. Cannot be `None` if
+            `num_static_categorical_features` is > 0.
+        d_model (`int`, *optional*, defaults to 64):
+            Dimensionality of the transformer layers.
+        encoder_layers (`int`, *optional*, defaults to 2):
+            Number of encoder layers.
+        decoder_layers (`int`, *optional*, defaults to 2):
+            Number of decoder layers.
+        encoder_attention_heads (`int`, *optional*, defaults to 2):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        decoder_attention_heads (`int`, *optional*, defaults to 2):
+            Number of attention heads for each attention layer in the Transformer decoder.
+        encoder_ffn_dim (`int`, *optional*, defaults to 32):
+            Dimension of the "intermediate" (often named feed-forward) layer in encoder.
+        decoder_ffn_dim (`int`, *optional*, defaults to 32):
+            Dimension of the "intermediate" (often named feed-forward) layer in decoder.
+        activation_function (`str` or `function`, *optional*, defaults to `"gelu"`):
+            The non-linear activation function (function or string) in the encoder and decoder. If string, `"gelu"` and
+            `"relu"` are supported.
+        dropout (`float`, *optional*, defaults to 0.1):
+            The dropout probability for all fully connected layers in the encoder, and decoder.
+        encoder_layerdrop (`float`, *optional*, defaults to 0.1):
+            The dropout probability for the attention and fully connected layers for each encoder layer.
+        decoder_layerdrop (`float`, *optional*, defaults to 0.1):
+            The dropout probability for the attention and fully connected layers for each decoder layer.
+        attention_dropout (`float`, *optional*, defaults to 0.1):
+            The dropout probability for the attention probabilities.
+        activation_dropout (`float`, *optional*, defaults to 0.1):
+            The dropout probability used between the two layers of the feed-forward networks.
+        num_parallel_samples (`int`, *optional*, defaults to 100):
+            The number of samples to generate in parallel for each time step of inference.
+        init_std (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated normal weight initialization distribution.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether to use the past key/values attentions (if applicable to the model) to speed up decoding.
+        label_length (`int`, *optional*, defaults to 10):
+            Start token length of the Autoformer decoder, which is used for direct multi-step prediction (i.e.
+            non-autoregressive generation).
+        moving_average (`int`, defaults to 25):
+            The window size of the moving average. In practice, it's the kernel size in AvgPool1d of the Decomposition
+            Layer.
+        autocorrelation_factor (`int`, defaults to 3):
+            "Attention" (i.e. AutoCorrelation mechanism) factor which is used to find top k autocorrelations delays.
+            It's recommended in the paper to set it to a number between 1 and 5.
+
+
+        Example:
+
+    ```python
+    >>> from transformers import AutoformerConfig, AutoformerModel
+
+    >>> # Initializing a default Autoformer configuration
+    >>> configuration = AutoformerConfig()
+
+    >>> # Randomly initializing a model (with random weights) from the configuration
+    >>> model = AutoformerModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "autoformer"
+    attribute_map = {
+        "hidden_size": "d_model",
+        "num_attention_heads": "encoder_attention_heads",
+        "num_hidden_layers": "encoder_layers",
+    }
+
+    def __init__(
+        self,
+        prediction_length: Optional[int] = None,
+        context_length: Optional[int] = None,
+        distribution_output: str = "student_t",
+        loss: str = "nll",
+        input_size: int = 1,
+        lags_sequence: List[int] = [1, 2, 3, 4, 5, 6, 7],
+        scaling: bool = True,
+        num_time_features: int = 0,
+        num_dynamic_real_features: int = 0,
+        num_static_categorical_features: int = 0,
+        num_static_real_features: int = 0,
+        cardinality: Optional[List[int]] = None,
+        embedding_dimension: Optional[List[int]] = None,
+        d_model: int = 64,
+        encoder_attention_heads: int = 2,
+        decoder_attention_heads: int = 2,
+        encoder_layers: int = 2,
+        decoder_layers: int = 2,
+        encoder_ffn_dim: int = 32,
+        decoder_ffn_dim: int = 32,
+        activation_function: str = "gelu",
+        dropout: float = 0.1,
+        encoder_layerdrop: float = 0.1,
+        decoder_layerdrop: float = 0.1,
+        attention_dropout: float = 0.1,
+        activation_dropout: float = 0.1,
+        num_parallel_samples: int = 100,
+        init_std: float = 0.02,
+        use_cache: bool = True,
+        is_encoder_decoder=True,
+        # Autoformer arguments
+        label_length: int = 10,
+        moving_average: int = 25,
+        autocorrelation_factor: int = 3,
+        **kwargs,
+    ):
+        # time series specific configuration
+        self.prediction_length = prediction_length
+        self.context_length = context_length if context_length is not None else prediction_length
+        self.distribution_output = distribution_output
+        self.loss = loss
+        self.input_size = input_size
+        self.num_time_features = num_time_features
+        self.lags_sequence = lags_sequence
+        self.scaling = scaling
+        self.num_dynamic_real_features = num_dynamic_real_features
+        self.num_static_real_features = num_static_real_features
+        self.num_static_categorical_features = num_static_categorical_features
+        if cardinality is not None and num_static_categorical_features > 0:
+            if len(cardinality) != num_static_categorical_features:
+                raise ValueError(
+                    "The cardinality should be a list of the same length as `num_static_categorical_features`"
+                )
+            self.cardinality = cardinality
+        else:
+            self.cardinality = [0]
+        if embedding_dimension is not None and num_static_categorical_features > 0:
+            if len(embedding_dimension) != num_static_categorical_features:
+                raise ValueError(
+                    "The embedding dimension should be a list of the same length as `num_static_categorical_features`"
+                )
+            self.embedding_dimension = embedding_dimension
+        else:
+            self.embedding_dimension = [min(50, (cat + 1) // 2) for cat in self.cardinality]
+        self.num_parallel_samples = num_parallel_samples
+
+        # Transformer architecture configuration
+        self.feature_size = input_size * len(self.lags_sequence) + self._number_of_features
+        self.d_model = d_model
+        self.encoder_attention_heads = encoder_attention_heads
+        self.decoder_attention_heads = decoder_attention_heads
+        self.encoder_ffn_dim = encoder_ffn_dim
+        self.decoder_ffn_dim = decoder_ffn_dim
+        self.encoder_layers = encoder_layers
+        self.decoder_layers = decoder_layers
+
+        self.dropout = dropout
+        self.attention_dropout = attention_dropout
+        self.activation_dropout = activation_dropout
+        self.encoder_layerdrop = encoder_layerdrop
+        self.decoder_layerdrop = decoder_layerdrop
+
+        self.activation_function = activation_function
+        self.init_std = init_std
+
+        self.use_cache = use_cache
+
+        # Autoformer
+        self.label_length = label_length
+        self.moving_average = moving_average
+        self.autocorrelation_factor = autocorrelation_factor
+
+        super().__init__(is_encoder_decoder=is_encoder_decoder, **kwargs)
+
+    @property
+    def _number_of_features(self) -> int:
+        return (
+            sum(self.embedding_dimension)
+            + self.num_dynamic_real_features
+            + self.num_time_features
+            + self.num_static_real_features
+            + self.input_size * 2  # the log1p(abs(loc)) and log(scale) features
+        )

venv/lib/python3.10/site-packages/transformers/models/bartpho/__init__.py

@@ -0,0 +1,42 @@
+# Copyright 2021 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import TYPE_CHECKING
+
+from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_sentencepiece_available
+
+
+_import_structure = {}
+
+try:
+    if not is_sentencepiece_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["tokenization_bartpho"] = ["BartphoTokenizer"]
+
+if TYPE_CHECKING:
+    try:
+        if not is_sentencepiece_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .tokenization_bartpho import BartphoTokenizer
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)

venv/lib/python3.10/site-packages/transformers/models/bartpho/tokenization_bartpho.py

@@ -0,0 +1,314 @@
+# coding=utf-8
+# Copyright 2021 VinAI Research and the HuggingFace Inc. team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License
+""" Tokenization classes for BARTpho-syllable model."""
+
+
+import os
+from shutil import copyfile
+from typing import Any, Dict, List, Optional, Tuple
+
+import sentencepiece as spm
+
+from ...tokenization_utils import AddedToken, PreTrainedTokenizer
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+SPIECE_UNDERLINE = "▁"
+
+VOCAB_FILES_NAMES = {"vocab_file": "sentencepiece.bpe.model", "monolingual_vocab_file": "dict.txt"}
+
+
+class BartphoTokenizer(PreTrainedTokenizer):
+    """
+    Adapted from [`XLMRobertaTokenizer`]. Based on [SentencePiece](https://github.com/google/sentencepiece).
+
+    This tokenizer inherits from [`PreTrainedTokenizer`] which contains most of the main methods. Users should refer to
+    this superclass for more information regarding those methods.
+
+    Args:
+        vocab_file (`str`):
+            Path to the vocabulary file. This vocabulary is the pre-trained SentencePiece model available from the
+            multilingual XLM-RoBERTa, also used in mBART, consisting of 250K types.
+        monolingual_vocab_file (`str`):
+            Path to the monolingual vocabulary file. This monolingual vocabulary consists of Vietnamese-specialized
+            types extracted from the multilingual vocabulary vocab_file of 250K types.
+        bos_token (`str`, *optional*, defaults to `"<s>"`):
+            The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token.
+
+            <Tip>
+
+            When building a sequence using special tokens, this is not the token that is used for the beginning of
+            sequence. The token used is the `cls_token`.
+
+            </Tip>
+
+        eos_token (`str`, *optional*, defaults to `"</s>"`):
+            The end of sequence token.
+
+            <Tip>
+
+            When building a sequence using special tokens, this is not the token that is used for the end of sequence.
+            The token used is the `sep_token`.
+
+            </Tip>
+
+        sep_token (`str`, *optional*, defaults to `"</s>"`):
+            The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for
+            sequence classification or for a text and a question for question answering. It is also used as the last
+            token of a sequence built with special tokens.
+        cls_token (`str`, *optional*, defaults to `"<s>"`):
+            The classifier token which is used when doing sequence classification (classification of the whole sequence
+            instead of per-token classification). It is the first token of the sequence when built with special tokens.
+        unk_token (`str`, *optional*, defaults to `"<unk>"`):
+            The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
+            token instead.
+        pad_token (`str`, *optional*, defaults to `"<pad>"`):
+            The token used for padding, for example when batching sequences of different lengths.
+        mask_token (`str`, *optional*, defaults to `"<mask>"`):
+            The token used for masking values. This is the token used when training this model with masked language
+            modeling. This is the token which the model will try to predict.
+        sp_model_kwargs (`dict`, *optional*):
+            Will be passed to the `SentencePieceProcessor.__init__()` method. The [Python wrapper for
+            SentencePiece](https://github.com/google/sentencepiece/tree/master/python) can be used, among other things,
+            to set:
+
+            - `enable_sampling`: Enable subword regularization.
+            - `nbest_size`: Sampling parameters for unigram. Invalid for BPE-Dropout.
+
+              - `nbest_size = {0,1}`: No sampling is performed.
+              - `nbest_size > 1`: samples from the nbest_size results.
+              - `nbest_size < 0`: assuming that nbest_size is infinite and samples from the all hypothesis (lattice)
+                using forward-filtering-and-backward-sampling algorithm.
+
+            - `alpha`: Smoothing parameter for unigram sampling, and dropout probability of merge operations for
+              BPE-dropout.
+
+    Attributes:
+        sp_model (`SentencePieceProcessor`):
+            The *SentencePiece* processor that is used for every conversion (string, tokens and IDs).
+    """
+
+    vocab_files_names = VOCAB_FILES_NAMES
+    model_input_names = ["input_ids", "attention_mask"]
+
+    def __init__(
+        self,
+        vocab_file,
+        monolingual_vocab_file,
+        bos_token="<s>",
+        eos_token="</s>",
+        sep_token="</s>",
+        cls_token="<s>",
+        unk_token="<unk>",
+        pad_token="<pad>",
+        mask_token="<mask>",
+        sp_model_kwargs: Optional[Dict[str, Any]] = None,
+        **kwargs,
+    ) -> None:
+        # Mask token behave like a normal word, i.e. include the space before it
+        mask_token = AddedToken(mask_token, lstrip=True, rstrip=False) if isinstance(mask_token, str) else mask_token
+
+        self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
+
+        self.vocab_file = vocab_file
+        self.monolingual_vocab_file = monolingual_vocab_file
+        self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
+        self.sp_model.Load(str(vocab_file))
+
+        # Load the reduced vocab
+
+        # Keep order of special tokens for backward compatibility
+        self.fairseq_tokens_to_ids = {}
+        cnt = 0
+        for token in [bos_token, pad_token, eos_token, unk_token, sep_token, cls_token]:
+            if str(token) not in self.fairseq_tokens_to_ids:
+                self.fairseq_tokens_to_ids[str(token)] = cnt
+                cnt += 1
+        with open(monolingual_vocab_file, "r", encoding="utf-8") as f:
+            for line in f.readlines():
+                token = line.strip().split()[0]
+                self.fairseq_tokens_to_ids[token] = len(self.fairseq_tokens_to_ids)
+        if str(mask_token) not in self.fairseq_tokens_to_ids:
+            self.fairseq_tokens_to_ids[str(mask_token)] = len(self.fairseq_tokens_to_ids)
+
+        self.fairseq_ids_to_tokens = {v: k for k, v in self.fairseq_tokens_to_ids.items()}
+
+        super().__init__(
+            bos_token=bos_token,
+            eos_token=eos_token,
+            unk_token=unk_token,
+            sep_token=sep_token,
+            cls_token=cls_token,
+            pad_token=pad_token,
+            mask_token=mask_token,
+            sp_model_kwargs=self.sp_model_kwargs,
+            **kwargs,
+        )
+
+    def __getstate__(self):
+        state = self.__dict__.copy()
+        state["sp_model"] = None
+        state["sp_model_proto"] = self.sp_model.serialized_model_proto()
+        return state
+
+    def __setstate__(self, d):
+        self.__dict__ = d
+
+        # for backward compatibility
+        if not hasattr(self, "sp_model_kwargs"):
+            self.sp_model_kwargs = {}
+
+        self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
+        self.sp_model.LoadFromSerializedProto(self.sp_model_proto)
+
+    def build_inputs_with_special_tokens(
+        self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
+    ) -> List[int]:
+        """
+        Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and
+        adding special tokens. An BARTPho sequence has the following format:
+
+        - single sequence: `<s> X </s>`
+        - pair of sequences: `<s> A </s></s> B </s>`
+
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs to which the special tokens will be added.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+
+        Returns:
+            `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens.
+        """
+
+        if token_ids_1 is None:
+            return [self.cls_token_id] + token_ids_0 + [self.sep_token_id]
+        cls = [self.cls_token_id]
+        sep = [self.sep_token_id]
+        return cls + token_ids_0 + sep + sep + token_ids_1 + sep
+
+    def get_special_tokens_mask(
+        self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None, already_has_special_tokens: bool = False
+    ) -> List[int]:
+        """
+        Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
+        special tokens using the tokenizer `prepare_for_model` method.
+
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+            already_has_special_tokens (`bool`, *optional*, defaults to `False`):
+                Whether or not the token list is already formatted with special tokens for the model.
+
+        Returns:
+            `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
+        """
+
+        if already_has_special_tokens:
+            return super().get_special_tokens_mask(
+                token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
+            )
+
+        if token_ids_1 is None:
+            return [1] + ([0] * len(token_ids_0)) + [1]
+        return [1] + ([0] * len(token_ids_0)) + [1, 1] + ([0] * len(token_ids_1)) + [1]
+
+    def create_token_type_ids_from_sequences(
+        self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
+    ) -> List[int]:
+        """
+        Create a mask from the two sequences passed to be used in a sequence-pair classification task. BARTPho does not
+        make use of token type ids, therefore a list of zeros is returned.
+
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+
+        Returns:
+            `List[int]`: List of zeros.
+
+        """
+
+        sep = [self.sep_token_id]
+        cls = [self.cls_token_id]
+
+        if token_ids_1 is None:
+            return len(cls + token_ids_0 + sep) * [0]
+        return len(cls + token_ids_0 + sep + sep + token_ids_1 + sep) * [0]
+
+    @property
+    def vocab_size(self):
+        return len(self.fairseq_ids_to_tokens)
+
+    def get_vocab(self):
+        vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
+        vocab.update(self.added_tokens_encoder)
+        return vocab
+
+    def _tokenize(self, text: str) -> List[str]:
+        return self.sp_model.encode(text, out_type=str)
+
+    def _convert_token_to_id(self, token):
+        """Converts a token (str) in an id using the vocab."""
+        if token in self.fairseq_tokens_to_ids:
+            return self.fairseq_tokens_to_ids[token]
+        else:
+            return self.unk_token_id
+
+    def _convert_id_to_token(self, index):
+        """Converts an index (integer) in a token (str) using the vocab."""
+        return self.fairseq_ids_to_tokens[index]
+
+    def convert_tokens_to_string(self, tokens):
+        """Converts a sequence of tokens (strings for sub-words) in a single string."""
+        out_string = "".join(tokens).replace(SPIECE_UNDERLINE, " ").strip()
+        return out_string
+
+    def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
+        if not os.path.isdir(save_directory):
+            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
+            return
+        out_vocab_file = os.path.join(
+            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
+        )
+        out_monolingual_vocab_file = os.path.join(
+            save_directory,
+            (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["monolingual_vocab_file"],
+        )
+
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
+            copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
+
+        if os.path.abspath(self.monolingual_vocab_file) != os.path.abspath(
+            out_monolingual_vocab_file
+        ) and os.path.isfile(self.monolingual_vocab_file):
+            copyfile(self.monolingual_vocab_file, out_monolingual_vocab_file)
+        elif not os.path.isfile(self.monolingual_vocab_file):
+            with open(out_monolingual_vocab_file, "w", encoding="utf-8") as fp:
+                for token in self.fairseq_tokens_to_ids:
+                    if token not in self.all_special_tokens:
+                        fp.write(f"{str(token)} \n")
+
+        return out_vocab_file, out_monolingual_vocab_file

venv/lib/python3.10/site-packages/transformers/models/dbrx/__init__.py

@@ -0,0 +1,51 @@
+# Copyright 2024 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from typing import TYPE_CHECKING
+
+from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
+
+
+_import_structure = {
+    "configuration_dbrx": ["DbrxConfig"],
+}
+
+try:
+    if not is_torch_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["modeling_dbrx"] = [
+        "DbrxForCausalLM",
+        "DbrxModel",
+        "DbrxPreTrainedModel",
+    ]
+
+
+if TYPE_CHECKING:
+    from .configuration_dbrx import DbrxConfig
+
+    try:
+        if not is_torch_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .modeling_dbrx import DbrxForCausalLM, DbrxModel, DbrxPreTrainedModel
+
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)

venv/lib/python3.10/site-packages/transformers/models/dbrx/configuration_dbrx.py

@@ -0,0 +1,257 @@
+# coding=utf-8
+# Copyright 2024 Databricks Mosaic Research and The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" DBRX model configuration """
+
+from typing import Any, Optional
+
+from ...configuration_utils import PretrainedConfig
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+
+class DbrxAttentionConfig(PretrainedConfig):
+    """Configuration class for Dbrx Attention.
+
+    [`DbrxAttention`] class. It is used to instantiate attention layers
+    according to the specified arguments, defining the layers architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        attn_pdrop (`float`, *optional*, defaults to 0.0):
+            The dropout probability for the attention layers.
+        clip_qkv (`float`, *optional*):
+            If set, clip the queries, keys, and values in the attention layer to this value.
+        kv_n_heads (`Optional[int]`, defaults to 1): For grouped_query_attention only, allow user to specify number of kv heads.
+        rope_theta (`float`, defaults to 10000.0): The base frequency for rope.
+    """
+
+    def __init__(
+        self,
+        attn_pdrop: float = 0.0,
+        clip_qkv: Optional[float] = None,
+        kv_n_heads: int = 1,
+        rope_theta: float = 10000.0,
+        **kwargs: Any,
+    ):
+        super().__init__(**kwargs)
+        self.attn_pdrop = attn_pdrop
+        self.clip_qkv = clip_qkv
+        self.kv_n_heads = kv_n_heads
+        self.rope_theta = rope_theta
+
+        for k in ["model_type"]:
+            if k in kwargs:
+                kwargs.pop(k)
+        if len(kwargs) != 0:
+            raise ValueError(f"Found unknown {kwargs=}")
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path: str, **kwargs: Any) -> "PretrainedConfig":
+        cls._set_token_in_kwargs(kwargs)
+
+        config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)
+
+        if config_dict.get("model_type") == "dbrx":
+            config_dict = config_dict["attn_config"]
+
+        if "model_type" in config_dict and hasattr(cls, "model_type") and config_dict["model_type"] != cls.model_type:
+            logger.warning(
+                f"You are using a model of type {config_dict['model_type']} to instantiate a model of type "
+                + f"{cls.model_type}. This is not supported for all configurations of models and can yield errors."
+            )
+
+        return cls.from_dict(config_dict, **kwargs)
+
+
+class DbrxFFNConfig(PretrainedConfig):
+    """Configuration class for Dbrx FFN.
+
+    [`DbrxFFN`] class. It is used to instantiate feedforward layers according to
+    the specified arguments, defining the layers architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        ffn_act_fn (`dict`, *optional*, defaults to `None`): A dict specifying activation function for the FFN.
+            The dict should have a key 'name' with the value being the name of the activation function along with
+            any additional keyword arguments. If `None`, then set to `{"name": "silu"}`.
+        ffn_hidden_size (`int`, defaults to 3584): The hidden size of the feedforward network.
+        moe_num_experts (`int`, defaults to 4): The number of experts in the mixture of experts layer.
+        moe_top_k (`int`, defaults to 1): The number of experts to use in the mixture of experts layer.
+        moe_jitter_eps (`float`, *optional*, defaults to `None`): If not `None`, the jitter epsilon for the mixture of experts layer.
+        moe_loss_weight (`float`, defaults to 0.01): The loss weight for the mixture of experts layer.
+        moe_normalize_expert_weights (`float`, *optional*, defaults to 1.0): The normalization factor for the expert weights.
+    """
+
+    def __init__(
+        self,
+        ffn_act_fn: dict = None,
+        ffn_hidden_size: int = 3584,
+        moe_num_experts: int = 4,
+        moe_top_k: int = 1,
+        moe_jitter_eps: Optional[float] = None,
+        moe_loss_weight: float = 0.01,
+        moe_normalize_expert_weights: Optional[float] = 1.0,
+        **kwargs: Any,
+    ):
+        super().__init__()
+        if ffn_act_fn is None:
+            ffn_act_fn = {"name": "silu"}
+        self.ffn_act_fn = ffn_act_fn
+        self.ffn_hidden_size = ffn_hidden_size
+        self.moe_num_experts = moe_num_experts
+        self.moe_top_k = moe_top_k
+        self.moe_jitter_eps = moe_jitter_eps
+        self.moe_loss_weight = moe_loss_weight
+        self.moe_normalize_expert_weights = moe_normalize_expert_weights
+
+        for k in ["model_type"]:
+            if k in kwargs:
+                kwargs.pop(k)
+        if len(kwargs) != 0:
+            raise ValueError(f"Found unknown {kwargs=}")
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path: str, **kwargs: Any) -> "PretrainedConfig":
+        cls._set_token_in_kwargs(kwargs)
+
+        config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)
+
+        if config_dict.get("model_type") == "dbrx":
+            config_dict = config_dict["ffn_config"]
+
+        if "model_type" in config_dict and hasattr(cls, "model_type") and config_dict["model_type"] != cls.model_type:
+            logger.warning(
+                f"You are using a model of type {config_dict['model_type']} to instantiate a model of type "
+                + f"{cls.model_type}. This is not supported for all configurations of models and can yield errors."
+            )
+
+        return cls.from_dict(config_dict, **kwargs)
+
+
+class DbrxConfig(PretrainedConfig):
+    r"""
+
+    This is the configuration class to store the configuration of a [`DbrxModel`]. It is used to instantiate a Dbrx model according to the
+    specified arguments, defining the model architecture. Instantiating a configuration with the
+    defaults will yield a different configuration to that of the [databricks/dbrx-instruct](https://huggingface.co/databricks/dbrx-instruct) architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        d_model (`int`, *optional*, defaults to 2048):
+            Dimensionality of the embeddings and hidden states.
+        n_heads (`int`, *optional*, defaults to 16):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        n_layers (`int`, *optional*, defaults to 24):
+            Number of hidden layers in the Transformer encoder.
+        max_seq_len (`int`, *optional*, defaults to 2048):
+            The maximum sequence length of the model.
+        vocab_size (`int`, *optional*, defaults to 32000):
+            Vocabulary size of the Dbrx model. Defines the maximum number of different tokens that can be represented by
+            the `inputs_ids` passed when calling [`DbrxModel`].
+        resid_pdrop (`float`, *optional*, defaults to 0.0):
+            The dropout probability applied to the attention output before combining with residual.
+        emb_pdrop (`float`, *optional*, defaults to 0.0):
+            The dropout probability for the embedding layer.
+        attn_config (`dict`, *optional*):
+            A dictionary used to configure the model's attention module.
+        ffn_config (`dict`, *optional*):
+            A dictionary used to configure the model's FFN module.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models).
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        output_router_logits (`bool`, *optional*, defaults to `False`):
+            Whether or not the router logits should be returned by the model. Enabling this will also
+            allow the model to output the auxiliary loss. See [here]() for more details.
+
+
+    Example:
+    ```python
+    >>> from transformers import DbrxConfig, DbrxModel
+
+    >>> # Initializing a Dbrx configuration
+    >>> configuration = DbrxConfig(n_layers=2, d_model=256, n_heads=8, vocab_size=128)
+
+    >>> # Initializing a model (with random weights) from the configuration
+    >>> model = DbrxModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```
+    """
+
+    model_type = "dbrx"
+    attribute_map = {
+        "num_attention_heads": "n_heads",
+        "hidden_size": "d_model",
+        "num_hidden_layers": "n_layers",
+        "max_position_embeddings": "max_seq_len",
+    }
+
+    def __init__(
+        self,
+        d_model: int = 2048,
+        n_heads: int = 16,
+        n_layers: int = 24,
+        max_seq_len: int = 2048,
+        vocab_size: int = 32000,
+        resid_pdrop: float = 0.0,
+        emb_pdrop: float = 0.0,
+        attn_config: Optional[DbrxAttentionConfig] = None,
+        ffn_config: Optional[DbrxFFNConfig] = None,
+        use_cache: bool = True,
+        initializer_range: float = 0.02,
+        output_router_logits: bool = False,
+        **kwargs: Any,
+    ):
+        if attn_config is None:
+            self.attn_config = DbrxAttentionConfig()
+        elif isinstance(attn_config, dict):
+            self.attn_config = DbrxAttentionConfig(**attn_config)
+        else:
+            self.attn_config = attn_config
+
+        if ffn_config is None:
+            self.ffn_config = DbrxFFNConfig()
+        elif isinstance(ffn_config, dict):
+            self.ffn_config = DbrxFFNConfig(**ffn_config)
+        else:
+            self.ffn_config = ffn_config
+
+        self.d_model = d_model
+        self.n_heads = n_heads
+        self.n_layers = n_layers
+        self.max_seq_len = max_seq_len
+        self.vocab_size = vocab_size
+        self.resid_pdrop = resid_pdrop
+        self.emb_pdrop = emb_pdrop
+        self.use_cache = use_cache
+        self.initializer_range = initializer_range
+        self.output_router_logits = output_router_logits
+
+        tie_word_embeddings = kwargs.pop("tie_word_embeddings", False)
+        if tie_word_embeddings:
+            raise ValueError("tie_word_embeddings is not supported for DBRX models.")
+
+        super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)

venv/lib/python3.10/site-packages/transformers/models/dbrx/modeling_dbrx.py

@@ -0,0 +1,1523 @@
+# coding=utf-8
+# Copyright 2024 Databricks Mosaic Research and The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" PyTorch DBRX model. """
+
+import math
+from typing import Any, Dict, Optional, Tuple, Union
+
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import nn
+
+from ...activations import ACT2FN
+from ...cache_utils import Cache, DynamicCache, StaticCache
+from ...modeling_attn_mask_utils import AttentionMaskConverter
+from ...modeling_outputs import MoeCausalLMOutputWithPast, MoeModelOutputWithPast
+from ...modeling_utils import PreTrainedModel
+from ...utils import (
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    is_flash_attn_2_available,
+    is_flash_attn_greater_or_equal_2_10,
+    logging,
+    replace_return_docstrings,
+)
+from .configuration_dbrx import DbrxConfig
+
+
+if is_flash_attn_2_available():
+    from flash_attn import flash_attn_func, flash_attn_varlen_func
+    from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input  # noqa
+
+logger = logging.get_logger(__name__)
+
+_CONFIG_FOR_DOC = "DbrxConfig"
+
+
+# Copied from transformers.models.gemma.modeling_gemma.GemmaRotaryEmbedding with Gemma->Dbrx
+class DbrxRotaryEmbedding(nn.Module):
+    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
+        super().__init__()
+
+        self.dim = dim
+        self.max_position_embeddings = max_position_embeddings
+        self.base = base
+        self.register_buffer("inv_freq", None, persistent=False)
+
+    @torch.no_grad()
+    def forward(self, x, position_ids, seq_len=None):
+        # x: [bs, num_attention_heads, seq_len, head_size]
+        if self.inv_freq is None:
+            self.inv_freq = 1.0 / (
+                self.base ** (torch.arange(0, self.dim, 2, dtype=torch.int64, device=x.device).float() / self.dim)
+            )
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
+        position_ids_expanded = position_ids[:, None, :].float()
+        # Force float32 since bfloat16 loses precision on long contexts
+        # See https://github.com/huggingface/transformers/pull/29285
+        device_type = x.device.type
+        device_type = device_type if isinstance(device_type, str) and device_type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+# Copied from transformers.models.llama.modeling_llama.rotate_half
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+# Copied from transformers.models.llama.modeling_llama.apply_rotary_pos_emb
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`, *optional*):
+            Deprecated and unused.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+# Copied from transformers.models.llama.modeling_llama.repeat_kv
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
+    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
+    """
+    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
+    if n_rep == 1:
+        return hidden_states
+    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
+    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
+
+
+def load_balancing_loss_func(
+    gate_logits: torch.Tensor,
+    num_experts: int,
+    top_k: int,
+    attention_mask: Optional[torch.Tensor],
+) -> torch.Tensor:
+    r"""Computes auxiliary load balancing loss as in Switch Transformer - implemented in Pytorch.
+
+    See Switch Transformer (https://arxiv.org/abs/2101.03961) for more details. This function implements the loss
+    function presented in equations (4) - (6) of the paper. It aims at penalizing cases where the routing between
+    experts is too unbalanced.
+
+    Args:
+        gate_logits (Union[`torch.Tensor`, Tuple[torch.Tensor]):
+            Logits from the `gate`, should be a tuple of model.config.num_hidden_layers tensors of
+            shape [batch_size X sequence_length, num_experts].
+        num_experts (`int`):
+            Number of experts.
+        top_k (`int`):
+            The number of experts each token is routed to.
+        attention_mask (`torch.Tensor`, None):
+            The attention_mask used in forward function
+            shape [batch_size X sequence_length] if not None.
+
+    Returns:
+        The auxiliary loss.
+    """
+    if gate_logits is None or not isinstance(gate_logits, tuple):
+        return torch.tensor(0.0)
+
+    if isinstance(gate_logits, tuple):
+        compute_device = gate_logits[0].device
+        concatenated_gate_logits = torch.cat([layer_gate.to(compute_device) for layer_gate in gate_logits], dim=0)
+
+    routing_weights = torch.nn.functional.softmax(concatenated_gate_logits, dim=-1)
+
+    _, selected_experts = torch.topk(routing_weights, top_k, dim=-1)
+
+    expert_mask = torch.nn.functional.one_hot(selected_experts, num_experts)
+
+    if attention_mask is None:
+        # Compute the percentage of tokens routed to each experts
+        tokens_per_expert = torch.mean(expert_mask.float(), dim=0)
+
+        # Compute the average probability of routing to these experts
+        router_prob_per_expert = torch.mean(routing_weights, dim=0)
+    else:
+        batch_size, sequence_length = attention_mask.shape
+        num_hidden_layers = concatenated_gate_logits.shape[0] // (batch_size * sequence_length)
+
+        # Compute the mask that masks all padding tokens as 0 with the same shape of expert_mask
+        expert_attention_mask = (
+            attention_mask[None, :, :, None, None]
+            .expand((num_hidden_layers, batch_size, sequence_length, top_k, num_experts))
+            .reshape(-1, top_k, num_experts)
+            .to(compute_device)
+        )
+
+        # Compute the percentage of tokens routed to each experts
+        tokens_per_expert = torch.sum(expert_mask.float() * expert_attention_mask, dim=0) / torch.sum(
+            expert_attention_mask, dim=0
+        )
+
+        # Compute the mask that masks all padding tokens as 0 with the same shape of tokens_per_expert
+        router_per_expert_attention_mask = (
+            attention_mask[None, :, :, None]
+            .expand((num_hidden_layers, batch_size, sequence_length, num_experts))
+            .reshape(-1, num_experts)
+            .to(compute_device)
+        )
+
+        # Compute the average probability of routing to these experts
+        router_prob_per_expert = torch.sum(routing_weights * router_per_expert_attention_mask, dim=0) / torch.sum(
+            router_per_expert_attention_mask, dim=0
+        )
+
+    overall_loss = torch.sum(tokens_per_expert * router_prob_per_expert.unsqueeze(0))
+    return overall_loss * num_experts
+
+
+# Copied from transformers.models.llama.modeling_llama._get_unpad_data
+def _get_unpad_data(attention_mask):
+    seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
+    indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
+    max_seqlen_in_batch = seqlens_in_batch.max().item()
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
+    return (
+        indices,
+        cu_seqlens,
+        max_seqlen_in_batch,
+    )
+
+
+class DbrxAttention(nn.Module):
+    """Multi-head self attention."""
+
+    def __init__(self, config: DbrxConfig, block_idx: Optional[int] = None):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.d_model
+        self.num_heads = config.n_heads
+        self.head_dim = self.hidden_size // self.num_heads
+        self.max_position_embeddings = config.max_seq_len
+        self.block_idx = block_idx
+        if block_idx is None:
+            logger.warning_once(
+                f"Instantiating {self.__class__.__name__} without passing a `block_idx` is not recommended and will "
+                + "lead to errors during the forward call if caching is used. Please make sure to provide a `block_idx` "
+                + "when creating this class."
+            )
+
+        attn_config = config.attn_config
+        self.attn_pdrop = attn_config.attn_pdrop
+        self.clip_qkv = attn_config.clip_qkv
+        self.num_key_value_heads = attn_config.kv_n_heads
+        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
+        self.rope_theta = attn_config.rope_theta
+        self.is_causal = True
+
+        self.Wqkv = nn.Linear(
+            self.hidden_size, self.hidden_size + 2 * self.num_key_value_heads * self.head_dim, bias=False
+        )
+        self.out_proj = nn.Linear(self.hidden_size, self.hidden_size, bias=False)
+        self.rotary_emb = DbrxRotaryEmbedding(
+            self.head_dim,
+            max_position_embeddings=self.max_position_embeddings,
+            base=self.rope_theta,
+        )
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_ids: torch.LongTensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Any,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Cache]]:
+        bsz, q_len, _ = hidden_states.size()
+
+        qkv_states = self.Wqkv(hidden_states)
+        min_val = -self.clip_qkv if self.clip_qkv is not None else None
+        max_val = self.clip_qkv
+        qkv_states = qkv_states.clamp(min=min_val, max=max_val)
+
+        query_states, key_states, value_states = qkv_states.split(
+            [
+                self.hidden_size,
+                self.num_key_value_heads * self.head_dim,
+                self.num_key_value_heads * self.head_dim,
+            ],
+            dim=2,
+        )
+
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        past_key_value = getattr(self, "past_key_value", past_key_value)
+        cos, sin = self.rotary_emb(value_states, position_ids)
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+        if past_key_value is not None:
+            # sin and cos are specific to RoPE models; position_ids needed for the static cache
+            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+            key_states, value_states = past_key_value.update(key_states, value_states, self.block_idx, cache_kwargs)
+
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+
+        if attention_mask is not None:  # no matter the length, we just slice it
+            causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
+            attn_weights = attn_weights + causal_mask
+
+        # upcast attention to fp32
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
+        attn_weights = nn.functional.dropout(attn_weights, p=self.attn_pdrop, training=self.training)
+        attn_output = torch.matmul(attn_weights, value_states)
+
+        if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
+            raise ValueError(
+                f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
+                + f" {attn_output.size()}"
+            )
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+        attn_output = self.out_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+
+class DbrxFlashAttention2(DbrxAttention):
+    """Dbrx flash attention module.
+
+    This module inherits from `DbrxAttention` as the weights of the module stays
+    untouched. The only required change would be on the forward pass where it
+    calls the public API of flash attention.
+    """
+
+    def __init__(self, *args: Any, **kwargs: Any):
+        super().__init__(*args, **kwargs)
+
+        # TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1.
+        # flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0.
+        # Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left).
+        # From: https://github.com/huggingface/transformers/blob/3b8e2932ce743008f63585aae1e1b8b30dc8b3ac/src/transformers/models/gemma/modeling_gemma.py#L318
+        self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10()
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Any,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        logger.info("Implicitly setting `output_attentions` to False as it is not supported in Flash Attention.")
+        output_attentions = False
+
+        bsz, q_len, _ = hidden_states.size()
+
+        qkv_states = self.Wqkv(hidden_states)
+        if self.clip_qkv is not None:
+            qkv_states = qkv_states.clamp(min=-self.clip_qkv, max=self.clip_qkv)
+
+        query_states, key_states, value_states = qkv_states.split(
+            [
+                self.hidden_size,
+                self.num_key_value_heads * self.head_dim,
+                self.num_key_value_heads * self.head_dim,
+            ],
+            dim=2,
+        )
+
+        # Flash attention requires the input to have the shape
+        # batch_size x seq_length x head_dim x hidden_dim
+        # therefore we just need to keep the original shape
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        cos, sin = self.rotary_emb(value_states, position_ids)
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+        past_key_value = getattr(self, "past_key_value", past_key_value)
+
+        if past_key_value is not None:
+            # sin and cos are specific to RoPE models; cache_position needed for the static cache
+            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+            key_states, value_states = past_key_value.update(key_states, value_states, self.block_idx, cache_kwargs)
+
+        # TODO: These transpose are quite inefficient but Flash Attention requires the layout
+        # [batch_size, sequence_length, num_heads, head_dim]. We would need to refactor the KV cache
+        # to be able to avoid many of these transpose/reshape/view.
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+
+        dropout_rate = self.attn_pdrop if self.training else 0.0
+
+        # In PEFT, usually we cast the layer norms in float32 for training stability reasons
+        # therefore the input hidden states gets silently casted in float32. Hence, we need
+        # cast them back in the correct dtype just to be sure everything works as expected.
+        # This might slowdown training & inference so it is recommended to not cast the LayerNorms
+        # in fp32. (LlamaRMSNorm handles it correctly)
+        input_dtype = query_states.dtype
+        if input_dtype == torch.float32:
+            if torch.is_autocast_enabled():
+                target_dtype = torch.get_autocast_gpu_dtype()
+            # Handle the case where the model is quantized
+            elif hasattr(self.config, "_pre_quantization_dtype"):
+                target_dtype = self.config._pre_quantization_dtype
+            else:
+                target_dtype = query_states.dtype
+
+            logger.warning_once(
+                "The input hidden states seems to be silently casted in float32, this might be "
+                + "related to the fact you have upcasted embedding or layer norm layers in "
+                + f"float32. We will cast back the input in {target_dtype}."
+            )
+
+            query_states = query_states.to(target_dtype)
+            key_states = key_states.to(target_dtype)
+            value_states = value_states.to(target_dtype)
+
+        attn_output = self._flash_attention_forward(
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            q_len,
+            dropout=dropout_rate,
+        )
+
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()
+        attn_output = self.out_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaFlashAttention2._flash_attention_forward
+    def _flash_attention_forward(
+        self, query_states, key_states, value_states, attention_mask, query_length, dropout=0.0, softmax_scale=None
+    ):
+        """
+        Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token
+        first unpad the input, then computes the attention scores and pad the final attention scores.
+
+        Args:
+            query_states (`torch.Tensor`):
+                Input query states to be passed to Flash Attention API
+            key_states (`torch.Tensor`):
+                Input key states to be passed to Flash Attention API
+            value_states (`torch.Tensor`):
+                Input value states to be passed to Flash Attention API
+            attention_mask (`torch.Tensor`):
+                The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the
+                position of padding tokens and 1 for the position of non-padding tokens.
+            dropout (`float`):
+                Attention dropout
+            softmax_scale (`float`, *optional*):
+                The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
+        """
+        if not self._flash_attn_uses_top_left_mask:
+            causal = self.is_causal
+        else:
+            # TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in LlamaFlashAttention2 __init__.
+            causal = self.is_causal and query_length != 1
+
+        # Contains at least one padding token in the sequence
+        if attention_mask is not None:
+            batch_size = query_states.shape[0]
+            query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input(
+                query_states, key_states, value_states, attention_mask, query_length
+            )
+
+            cu_seqlens_q, cu_seqlens_k = cu_seq_lens
+            max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens
+
+            attn_output_unpad = flash_attn_varlen_func(
+                query_states,
+                key_states,
+                value_states,
+                cu_seqlens_q=cu_seqlens_q,
+                cu_seqlens_k=cu_seqlens_k,
+                max_seqlen_q=max_seqlen_in_batch_q,
+                max_seqlen_k=max_seqlen_in_batch_k,
+                dropout_p=dropout,
+                softmax_scale=softmax_scale,
+                causal=causal,
+            )
+
+            attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
+        else:
+            attn_output = flash_attn_func(
+                query_states, key_states, value_states, dropout, softmax_scale=softmax_scale, causal=causal
+            )
+
+        return attn_output
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaFlashAttention2._upad_input
+    def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length):
+        indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask)
+        batch_size, kv_seq_len, num_key_value_heads, head_dim = key_layer.shape
+
+        key_layer = index_first_axis(
+            key_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
+        )
+        value_layer = index_first_axis(
+            value_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
+        )
+        if query_length == kv_seq_len:
+            query_layer = index_first_axis(
+                query_layer.reshape(batch_size * kv_seq_len, self.num_heads, head_dim), indices_k
+            )
+            cu_seqlens_q = cu_seqlens_k
+            max_seqlen_in_batch_q = max_seqlen_in_batch_k
+            indices_q = indices_k
+        elif query_length == 1:
+            max_seqlen_in_batch_q = 1
+            cu_seqlens_q = torch.arange(
+                batch_size + 1, dtype=torch.int32, device=query_layer.device
+            )  # There is a memcpy here, that is very bad.
+            indices_q = cu_seqlens_q[:-1]
+            query_layer = query_layer.squeeze(1)
+        else:
+            # The -q_len: slice assumes left padding.
+            attention_mask = attention_mask[:, -query_length:]
+            query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask)
+
+        return (
+            query_layer,
+            key_layer,
+            value_layer,
+            indices_q,
+            (cu_seqlens_q, cu_seqlens_k),
+            (max_seqlen_in_batch_q, max_seqlen_in_batch_k),
+        )
+
+
+class DbrxSdpaAttention(DbrxAttention):
+    """
+    Dbrx attention module using torch.nn.functional.scaled_dot_product_attention. This module inherits from
+    `DbrxAttention` as the weights of the module stays untouched. The only changes are on the forward pass to adapt to
+    SDPA API.
+    """
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        cache_position: Optional[torch.LongTensor] = None,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        if output_attentions:
+            # TODO: Improve this warning with e.g. `model.config.attn_implementation = "manual"` once this is implemented.
+            logger.warning_once(
+                "DbrxModel is using DbrxSdpaAttention, but `torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to the manual attention implementation, "
+                'but specifying the manual implementation will be required from Transformers version v5.0.0 onwards. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
+            )
+            return super().forward(
+                hidden_states=hidden_states,
+                attention_mask=attention_mask,
+                position_ids=position_ids,
+                past_key_value=past_key_value,
+                output_attentions=output_attentions,
+                use_cache=use_cache,
+                cache_position=cache_position,
+            )
+
+        bsz, q_len, _ = hidden_states.size()
+
+        qkv_states = self.Wqkv(hidden_states)
+        if self.clip_qkv is not None:
+            qkv_states = qkv_states.clamp(min=-self.clip_qkv, max=self.clip_qkv)
+
+        query_states, key_states, value_states = qkv_states.split(
+            [
+                self.hidden_size,
+                self.num_key_value_heads * self.head_dim,
+                self.num_key_value_heads * self.head_dim,
+            ],
+            dim=2,
+        )
+
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        cos, sin = self.rotary_emb(value_states, position_ids, seq_len=None)
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, None)
+
+        past_key_value = getattr(self, "past_key_value", past_key_value)
+
+        if past_key_value is not None:
+            # sin and cos are specific to RoPE models; cache_position needed for the static cache
+            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+            key_states, value_states = past_key_value.update(key_states, value_states, self.block_idx, cache_kwargs)
+
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        causal_mask = attention_mask
+        if attention_mask is not None:
+            causal_mask = causal_mask[:, :, :, : key_states.shape[-2]]
+
+        # SDPA with memory-efficient backend is currently (torch==2.1.2) bugged with non-contiguous inputs with custom attn_mask,
+        # Reference: https://github.com/pytorch/pytorch/issues/112577.
+        if query_states.device.type == "cuda" and causal_mask is not None:
+            query_states = query_states.contiguous()
+            key_states = key_states.contiguous()
+            value_states = value_states.contiguous()
+
+        attn_output = torch.nn.functional.scaled_dot_product_attention(
+            query_states,
+            key_states,
+            value_states,
+            attn_mask=causal_mask,
+            dropout_p=self.attn_pdrop if self.training else 0.0,
+        )
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.view(bsz, q_len, -1)
+
+        attn_output = self.out_proj(attn_output)
+
+        return attn_output, None, past_key_value
+
+
+DBRX_ATTENTION_CLASSES = {
+    "eager": DbrxAttention,
+    "flash_attention_2": DbrxFlashAttention2,
+    "sdpa": DbrxSdpaAttention,
+}
+
+
+class DbrxNormAttentionNorm(nn.Module):
+    def __init__(self, config: DbrxConfig, block_idx: Optional[int] = None):
+        super().__init__()
+        self.block_idx = block_idx
+        self.resid_pdrop = config.resid_pdrop
+        self.norm_1 = nn.LayerNorm(config.d_model, bias=False)
+        self.attn = DBRX_ATTENTION_CLASSES[config._attn_implementation](
+            config=config,
+            block_idx=block_idx,
+        )
+        self.norm_2 = nn.LayerNorm(config.d_model, bias=False)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_ids: torch.LongTensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Any,
+    ) -> Tuple[torch.Tensor, torch.Tensor, Optional[torch.Tensor], Optional[Cache]]:
+        residual_states = hidden_states
+        hidden_states = self.norm_1(hidden_states).to(hidden_states.dtype)
+
+        hidden_states, attn_weights, past_key_value = self.attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+        hidden_states = nn.functional.dropout(hidden_states, p=self.resid_pdrop, training=self.training)
+        hidden_states = hidden_states + residual_states
+
+        residual_states = hidden_states
+        hidden_states = self.norm_2(hidden_states).to(hidden_states.dtype)
+
+        return residual_states, hidden_states, attn_weights, past_key_value
+
+
+class DbrxRouter(nn.Module):
+    def __init__(
+        self,
+        hidden_size: int,
+        moe_num_experts: int,
+        moe_top_k: int,
+        moe_jitter_eps: Optional[float],
+        moe_normalize_expert_weights: Optional[float],
+    ):
+        super().__init__()
+        self.hidden_size = hidden_size
+        self.moe_num_experts = moe_num_experts
+        self.moe_top_k = moe_top_k
+        self.moe_jitter_eps = moe_jitter_eps
+        self.moe_normalize_expert_weights = moe_normalize_expert_weights
+
+        self.layer = nn.Linear(self.hidden_size, self.moe_num_experts, bias=False)
+
+    def forward(self, hidden_states: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.LongTensor]:
+        if self.training and self.moe_jitter_eps is not None:
+            hidden_states *= torch.empty_like(hidden_states).uniform_(
+                1.0 - self.moe_jitter_eps, 1.0 + self.moe_jitter_eps
+            )
+        hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
+        weights = self.layer(hidden_states).softmax(dim=-1, dtype=torch.float32)
+        top_weights, top_experts = torch.topk(weights, self.moe_top_k, dim=-1)
+
+        top_weights_scale = (
+            torch.norm(top_weights, p=self.moe_normalize_expert_weights, dim=-1, keepdim=True)
+            if self.moe_normalize_expert_weights is not None
+            else 1.0
+        )
+        top_weights = top_weights / top_weights_scale
+
+        weights = weights.to(hidden_states.dtype)
+        top_weights = top_weights.to(hidden_states.dtype)
+        return weights, top_weights, top_experts
+
+
+class DbrxExpertGLU(nn.Module):
+    def __init__(self, hidden_size: int, ffn_hidden_size: int, moe_num_experts: int, ffn_act_fn: dict):
+        super().__init__()
+        self.hidden_size = hidden_size
+        self.ffn_hidden_size = ffn_hidden_size
+        self.moe_num_experts = moe_num_experts
+
+        self.w1 = nn.Parameter(torch.empty(moe_num_experts * ffn_hidden_size, hidden_size))
+        self.v1 = nn.Parameter(torch.empty(moe_num_experts * ffn_hidden_size, hidden_size))
+        self.w2 = nn.Parameter(torch.empty(moe_num_experts * ffn_hidden_size, hidden_size))
+
+        act_fn_name = ffn_act_fn.get("name", "silu")
+        self.activation_fn = ACT2FN[act_fn_name]
+
+    def forward(
+        self, x: torch.Tensor, expert_w1: torch.Tensor, expert_v1: torch.Tensor, expert_w2: torch.Tensor
+    ) -> torch.Tensor:
+        gate_proj = x.matmul(expert_w1.t())
+        up_proj = x.matmul(expert_v1.t())
+        gate_proj = self.activation_fn(gate_proj)
+        intermediate_states = gate_proj * up_proj
+        down_proj = intermediate_states.matmul(expert_w2)
+        return down_proj
+
+
+class DbrxExperts(nn.Module):
+    def __init__(self, hidden_size: int, ffn_hidden_size: int, moe_num_experts: int, ffn_act_fn: dict):
+        super().__init__()
+        self.moe_num_experts = moe_num_experts
+        self.mlp = DbrxExpertGLU(
+            hidden_size=hidden_size,
+            ffn_hidden_size=ffn_hidden_size,
+            moe_num_experts=moe_num_experts,
+            ffn_act_fn=ffn_act_fn,
+        )
+
+    def forward(
+        self, x: torch.Tensor, weights: torch.Tensor, top_weights: torch.Tensor, top_experts: torch.LongTensor
+    ) -> torch.Tensor:
+        bsz, q_len, hidden_size = x.shape
+        x = x.view(-1, hidden_size)
+        out = torch.zeros_like(x)
+
+        expert_mask = nn.functional.one_hot(top_experts, num_classes=self.moe_num_experts).permute(2, 1, 0)
+        # Chunk experts at once to avoid storing full parameter multiple times in autograd
+        w1_chunked = self.mlp.w1.view(self.mlp.moe_num_experts, self.mlp.ffn_hidden_size, self.mlp.hidden_size).chunk(
+            self.moe_num_experts, dim=0
+        )
+        v1_chunked = self.mlp.v1.view(self.mlp.moe_num_experts, self.mlp.ffn_hidden_size, self.mlp.hidden_size).chunk(
+            self.moe_num_experts, dim=0
+        )
+        w2_chunked = self.mlp.w2.view(self.mlp.moe_num_experts, self.mlp.ffn_hidden_size, self.mlp.hidden_size).chunk(
+            self.moe_num_experts, dim=0
+        )
+        w1_chunked = [w1.squeeze(dim=0) for w1 in w1_chunked]
+        v1_chunked = [v1.squeeze(dim=0) for v1 in v1_chunked]
+        w2_chunked = [w2.squeeze(dim=0) for w2 in w2_chunked]
+        for expert_idx in range(0, self.moe_num_experts):
+            topk_idx, token_idx = torch.where(expert_mask[expert_idx])
+            if token_idx.shape[0] == 0:
+                continue
+
+            token_list = token_idx
+            topk_list = topk_idx
+
+            expert_tokens = x[None, token_list].reshape(-1, hidden_size)
+            expert_out = (
+                self.mlp(expert_tokens, w1_chunked[expert_idx], v1_chunked[expert_idx], w2_chunked[expert_idx])
+                * top_weights[token_list, topk_list, None]
+            )
+
+            out.index_add_(0, token_idx, expert_out)
+
+        out = out.reshape(bsz, q_len, hidden_size)
+        return out
+
+
+class DbrxFFN(nn.Module):
+    def __init__(self, config: DbrxConfig):
+        super().__init__()
+
+        ffn_config = config.ffn_config
+        self.router = DbrxRouter(
+            hidden_size=config.d_model,
+            moe_num_experts=ffn_config.moe_num_experts,
+            moe_top_k=ffn_config.moe_top_k,
+            moe_jitter_eps=ffn_config.moe_jitter_eps,
+            moe_normalize_expert_weights=ffn_config.moe_normalize_expert_weights,
+        )
+
+        self.experts = DbrxExperts(
+            hidden_size=config.d_model,
+            ffn_hidden_size=ffn_config.ffn_hidden_size,
+            moe_num_experts=ffn_config.moe_num_experts,
+            ffn_act_fn=ffn_config.ffn_act_fn,
+        )
+
+    def forward(self, x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
+        weights, top_weights, top_experts = self.router(x)
+        out = self.experts(x, weights, top_weights, top_experts)
+        return out, weights
+
+
+class DbrxBlock(nn.Module):
+    def __init__(self, config: DbrxConfig, block_idx: int):
+        super().__init__()
+        self.hidden_size = config.d_model
+        self.resid_pdrop = config.resid_pdrop
+        self.block_idx = block_idx
+        self.norm_attn_norm = DbrxNormAttentionNorm(
+            config=config,
+            block_idx=block_idx,
+        )
+        self.ffn = DbrxFFN(config=config)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: torch.LongTensor = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: Optional[bool] = False,
+        output_router_logits: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Any,
+    ) -> Union[
+        Tuple[torch.Tensor],
+        Tuple[torch.Tensor, Optional[torch.Tensor]],
+        Tuple[torch.Tensor, Optional[Cache]],
+        Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Cache]],
+        Tuple[torch.Tensor, Optional[torch.Tensor], Optional[torch.Tensor]],
+        Tuple[torch.Tensor, Optional[Cache], Optional[torch.Tensor]],
+        Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Cache], Optional[torch.Tensor]],
+    ]:
+        """Forward function for DbrxBlock.
+
+        Args:
+            hidden_states (`torch.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
+            position_ids (`torch.LongTensor`): position ids of shape `(batch, seq_len)`
+            attention_mask (`torch.Tensor`, optional): attention mask of size (batch_size, sequence_length)
+                if flash attention is used or (batch_size, 1, query_sequence_length, key_sequence_length)
+                if default attention is used.
+            past_key_value (`Tuple(torch.Tensor)`, optional): cached past key and value projection states
+            output_attentions (`bool`, optional): Whether or not to return the attentions tensors of all
+                attention layers. See `attentions` under returned tensors for more detail.
+            output_router_logits (`bool`, optional): Whether or not to return the router logits.
+            use_cache (`bool`, optional): If set to `True`, `past_key_values` key value states are
+                returned and can be used to speed up decoding (see `past_key_values`).
+            cache_position (`torch.LongTensor`, optional): position ids of the cache
+        """
+
+        # Norm + Attention + Norm
+        resid_states, hidden_states, self_attn_weights, present_key_value = self.norm_attn_norm(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+        # Fully Connected
+        hidden_states, router_logits = self.ffn(hidden_states)
+        hidden_states = nn.functional.dropout(hidden_states, p=self.resid_pdrop, training=self.training)
+        hidden_states = resid_states + hidden_states
+
+        outputs = (hidden_states,)
+
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        if use_cache:
+            outputs += (present_key_value,)
+
+        if output_router_logits:
+            outputs += (router_logits,)
+
+        return outputs
+
+
+DBRX_START_DOCSTRING = r"""
+    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+
+    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
+    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
+    and behavior.
+
+    Parameters:
+        config ([`DbrxConfig`]):
+            Model configuration class with all the parameters of the model. Initializing with a config file does not
+            load the weights associated with the model, only the configuration. Check out the
+            [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+
+
+@add_start_docstrings(
+    "The bare DBRX Model outputting raw hidden-states without any specific head on top.",
+    DBRX_START_DOCSTRING,
+)
+class DbrxPreTrainedModel(PreTrainedModel):
+    config_class = DbrxConfig
+    base_model_prefix = "transformer"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["DbrxBlock"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn_2 = True
+    _supports_sdpa = True
+    _supports_cache_class = True
+
+    def _init_weights(self, module: nn.Module):
+        std = self.config.initializer_range
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+        elif isinstance(module, nn.LayerNorm):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, DbrxExpertGLU):
+            module.w1.data.normal_(mean=0.0, std=std)
+            module.v1.data.normal_(mean=0.0, std=std)
+            module.w2.data.normal_(mean=0.0, std=std)
+
+    def _setup_cache(self, cache_cls: Any, max_batch_size: int, max_cache_len: int):
+        if self.config._attn_implementation == "flash_attention_2" and cache_cls == StaticCache:
+            raise ValueError(
+                "`static` cache implementation is not compatible with "
+                + "`attn_implementation==flash_attention_2`. Make sure to use "
+                + "`spda` in the mean time and open an issue at https://github.com/huggingface/transformers."
+            )
+
+        for block in self.transformer.blocks:
+            device = block.norm_attn_norm.norm_1.weight.device
+            if hasattr(self.config, "_pre_quantization_dtype"):
+                dtype = self.config._pre_quantization_dtype
+            else:
+                dtype = block.norm_attn_norm.attn.out_proj.weight.dtype
+            block.norm_attn_norm.attn.past_key_value = cache_cls(
+                self.config, max_batch_size, max_cache_len, device=device, dtype=dtype
+            )
+
+    def _reset_cache(self):
+        for block in self.transformer.blocks:
+            block.norm_attn_norm.attn.past_key_value = None
+
+
+DBRX_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+            it.
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            [What are attention masks?](../glossary#attention-mask)
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see
+            `past_key_values`).
+
+            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
+            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
+            information on the default strategy.
+
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+        position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.n_positions - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        past_key_values (`Cache` or `tuple(tuple(torch.FloatTensor))`, *optional*):
+            Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
+            blocks) that can be used to speed up sequential decoding. This typically consists in the `past_key_values`
+            returned by the model at a previous stage of decoding, when `use_cache=True` or `config.use_cache=True`.
+
+            Two formats are allowed:
+            - a [`~cache_utils.Cache`] instance;
+            - Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of
+            shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`). This is also known as the legacy
+            cache format.
+
+            The model will output the same cache format that is fed as input. If no `past_key_values` are passed, the
+            legacy cache format will be returned.
+
+            If `past_key_values` are used, the user can optionally input only the last `input_ids` (those that don't
+            have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `input_ids`
+            of shape `(batch_size, sequence_length)`.
+        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
+            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
+            model's internal embedding lookup matrix.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        output_router_logits (`bool`, *optional*):
+            Whether or not to return the logits of all the routers. They are useful for computing the router loss, and
+            should not be returned during inference.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+        cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
+            Indices depicting the position of the input sequence tokens in the sequence. Contrarily to `position_ids`,
+            this tensor is not affected by padding. It is used to update the cache in the correct position and to infer
+            the complete sequence length.
+"""
+
+
+@add_start_docstrings(
+    "The bare DBRX Model outputting raw hidden-states without any specific head on top.",
+    DBRX_START_DOCSTRING,
+)
+class DbrxModel(DbrxPreTrainedModel):
+    """Transformer decoder consisting of *config.num_hidden_layers*. Each layer is a [`DbrxBlock`] layer.
+
+    Args:
+        config ([`DbrxConfig`]): Model configuration class with all parameters of the model.
+            Initializing with a config file does not load the weights associated with the model, only the
+            configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+    """
+
+    def __init__(self, config: DbrxConfig):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+        self.emb_pdrop = config.emb_pdrop
+
+        self.wte = nn.Embedding(config.vocab_size, config.d_model, self.padding_idx)
+        self.blocks = nn.ModuleList([DbrxBlock(config, block_idx) for block_idx in range(config.n_layers)])
+        self.norm_f = nn.LayerNorm(config.d_model, bias=False)
+        self.gradient_checkpointing = False
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self) -> nn.Embedding:
+        return self.wte
+
+    def set_input_embeddings(self, value: nn.Embedding):
+        self.wte = value
+
+    @add_start_docstrings_to_model_forward(DBRX_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        output_router_logits: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+    ) -> Union[Tuple, MoeModelOutputWithPast]:
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        output_router_logits = (
+            output_router_logits if output_router_logits is not None else self.config.output_router_logits
+        )
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if (input_ids is None) ^ (inputs_embeds is not None):
+            raise ValueError(
+                "You cannot specify both input_ids and inputs_embeds at the same time, and must specify either one"
+            )
+
+        if self.gradient_checkpointing and self.training and use_cache:
+            logger.warning_once(
+                "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`."
+            )
+            use_cache = False
+
+        if inputs_embeds is None:
+            inputs_embeds = self.wte(input_ids)
+
+        inputs_embeds = nn.functional.dropout(inputs_embeds, p=self.emb_pdrop, training=self.training)
+
+        past_seen_tokens = 0
+        if use_cache:  # kept for BC (cache positions)
+            if not isinstance(past_key_values, StaticCache):
+                past_key_values = DynamicCache.from_legacy_cache(past_key_values)
+                past_seen_tokens = past_key_values.get_seq_length()
+
+        if cache_position is None:
+            if isinstance(past_key_values, StaticCache):
+                raise ValueError("cache_position is a required argument when using StaticCache.")
+            cache_position = torch.arange(
+                past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
+            )
+
+        if position_ids is None:
+            position_ids = cache_position.unsqueeze(0)
+        causal_mask = self._update_causal_mask(attention_mask, inputs_embeds, cache_position)
+
+        # embed positions
+        hidden_states = inputs_embeds
+
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        all_router_logits = () if output_router_logits else None
+        next_decoder_cache = None
+
+        for block in self.blocks:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+                block_outputs = self._gradient_checkpointing_func(
+                    block.__call__,
+                    hidden_states,
+                    causal_mask,
+                    position_ids,
+                    past_key_values,
+                    output_attentions,
+                    output_router_logits,
+                    use_cache,
+                    cache_position,
+                )
+            else:
+                block_outputs = block(
+                    hidden_states,
+                    attention_mask=causal_mask,
+                    position_ids=position_ids,
+                    past_key_value=past_key_values,
+                    output_attentions=output_attentions,
+                    output_router_logits=output_router_logits,
+                    use_cache=use_cache,
+                    cache_position=cache_position,
+                )
+
+            hidden_states = block_outputs[0]
+
+            if use_cache:
+                next_decoder_cache = block_outputs[2 if output_attentions else 1]
+
+            if output_attentions:
+                all_self_attns += (block_outputs[1],)
+
+            if output_router_logits:
+                all_router_logits += (block_outputs[-1],)
+
+        hidden_states = self.norm_f(hidden_states)
+
+        # add hidden states from the last decoder layer
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+
+        next_cache = None
+        if use_cache:
+            next_cache = (
+                next_decoder_cache.to_legacy_cache() if isinstance(next_decoder_cache, Cache) else next_decoder_cache
+            )
+        if not return_dict:
+            return tuple(
+                v
+                for v in [hidden_states, next_cache, all_hidden_states, all_self_attns, all_router_logits]
+                if v is not None
+            )
+        return MoeModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=next_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+            router_logits=all_router_logits,
+        )
+
+    # TODO: As of torch==2.2.0, the `attention_mask` passed to the model in `generate` is 2D and of dynamic length even when the static
+    # KV cache is used. This is an issue for torch.compile which then recaptures cudagraphs at each decode steps due to the dynamic shapes.
+    # (`recording cudagraph tree for symint key 13`, etc.), which is VERY slow. A workaround is `@torch.compiler.disable`, but this prevents using
+    # `fullgraph=True`. See more context in https://github.com/huggingface/transformers/pull/29114
+    def _update_causal_mask(
+        self, attention_mask: Optional[torch.Tensor], input_tensor: torch.Tensor, cache_position: torch.Tensor
+    ) -> Optional[torch.Tensor]:
+        if self.config._attn_implementation == "flash_attention_2":
+            if attention_mask is not None and 0.0 in attention_mask:
+                return attention_mask
+            return None
+
+        dtype, device = input_tensor.dtype, input_tensor.device
+        min_dtype = torch.finfo(dtype).min
+        sequence_length = input_tensor.shape[1]
+        if hasattr(self.blocks[0].norm_attn_norm.attn, "past_key_value"):  # static cache
+            target_length = self.config.max_position_embeddings
+        else:  # dynamic cache
+            target_length = (
+                attention_mask.shape[-1] if isinstance(attention_mask, torch.Tensor) else cache_position[-1] + 1
+            )
+        target_length = int(target_length)
+
+        causal_mask = torch.full((sequence_length, target_length), fill_value=min_dtype, dtype=dtype, device=device)
+        if sequence_length != 1:
+            causal_mask = torch.triu(causal_mask, diagonal=1)
+        causal_mask *= torch.arange(target_length, device=device) > cache_position.reshape(-1, 1)
+        causal_mask = causal_mask[None, None, :, :].expand(input_tensor.shape[0], 1, -1, -1)
+        if attention_mask is not None:
+            causal_mask = causal_mask.clone()  # copy to contiguous memory for in-place edit
+            if attention_mask.dim() == 2:
+                mask_length = attention_mask.shape[-1]
+                padding_mask = causal_mask[..., :mask_length].eq(0.0) * attention_mask[:, None, None, :].eq(0.0)
+                causal_mask[..., :mask_length] = causal_mask[..., :mask_length].masked_fill(padding_mask, min_dtype)
+            elif attention_mask.dim() == 4:
+                # backwards compatibility: we allow passing a 4D attention mask shorter than the input length with
+                # cache. In that case, the 4D attention mask attends to the newest tokens only.
+                if attention_mask.shape[-2] < cache_position[0] + sequence_length:
+                    offset = cache_position[0]
+                else:
+                    offset = 0
+                mask_shape = attention_mask.shape
+                mask_slice = (attention_mask.eq(0.0)).to(dtype=dtype) * min_dtype
+                causal_mask[
+                    : mask_shape[0], : mask_shape[1], offset : mask_shape[2] + offset, : mask_shape[3]
+                ] = mask_slice
+
+        if (
+            self.config._attn_implementation == "sdpa"
+            and attention_mask is not None
+            and attention_mask.device.type == "cuda"
+        ):
+            # TODO: For dynamo, rather use a check on fullgraph=True once this is possible (https://github.com/pytorch/pytorch/pull/120400).
+            is_tracing = (
+                torch.jit.is_tracing()
+                or isinstance(input_tensor, torch.fx.Proxy)
+                or (hasattr(torch, "_dynamo") and torch._dynamo.is_compiling())
+            )
+            if not is_tracing and torch.any(attention_mask != 1):
+                # Attend to all tokens in fully masked rows in the causal_mask, for example the relevant first rows when
+                # using left padding. This is required by F.scaled_dot_product_attention memory-efficient attention path.
+                # Details: https://github.com/pytorch/pytorch/issues/110213
+                causal_mask = AttentionMaskConverter._unmask_unattended(causal_mask, min_dtype)
+
+        return causal_mask
+
+
+@add_start_docstrings("The DBRX Model transformer for causal language modeling.", DBRX_START_DOCSTRING)
+class DbrxForCausalLM(DbrxPreTrainedModel):
+    def __init__(self, config: DbrxConfig):
+        super().__init__(config)
+        self.transformer = DbrxModel(config)
+        self.vocab_size = config.vocab_size
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+        self.moe_loss_weight = config.ffn_config.moe_loss_weight
+        self.num_experts = config.ffn_config.moe_num_experts
+        self.num_experts_per_tok = config.ffn_config.moe_top_k
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self) -> nn.Embedding:
+        return self.transformer.get_input_embeddings()
+
+    def set_input_embeddings(self, value: nn.Embedding):
+        self.transformer.set_input_embeddings(value)
+
+    def get_output_embeddings(self) -> nn.Linear:
+        return self.lm_head
+
+    def set_output_embeddings(self, new_embeddings: nn.Linear):
+        self.lm_head = new_embeddings
+
+    def set_decoder(self, decoder: DbrxModel):
+        self.transformer = decoder
+
+    def get_decoder(self) -> DbrxModel:
+        return self.transformer
+
+    @add_start_docstrings_to_model_forward(DBRX_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=MoeCausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC)
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        output_router_logits: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+    ) -> Union[Tuple, MoeCausalLMOutputWithPast]:
+        r"""Forward function for causal language modeling.
+
+        Args:
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+
+        Returns:
+
+        Example:
+
+        ```python
+        >> from transformers import AutoTokenizer, DbrxForCausalLM
+
+        >> model = DbrxForCausalLM.from_pretrained("databricks/dbrx-instruct")
+        >> tokenizer = AutoTokenizer.from_pretrained("databricks/dbrx-instruct")
+
+        >> prompt = "Hey, are you conscious? Can you talk to me?"
+        >> inputs = tokenizer(prompt, return_tensors="pt")
+
+        >> # Generate
+        >> generate_ids = model.generate(inputs.input_ids, max_length=30)
+        >> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
+        ```
+        """
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        output_router_logits = (
+            output_router_logits if output_router_logits is not None else self.config.output_router_logits
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs = self.transformer(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            output_router_logits=output_router_logits,
+            return_dict=return_dict,
+            cache_position=cache_position,
+        )
+
+        hidden_states = outputs[0]
+        logits = self.lm_head(hidden_states)
+
+        loss = None
+        if labels is not None:
+            # Shift so that tokens < n predict n
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            # Flatten the tokens
+            loss_fct = nn.CrossEntropyLoss()
+            shift_logits = shift_logits.view(-1, self.config.vocab_size)
+            shift_labels = shift_labels.view(-1)
+            # Enable model parallelism
+            shift_labels = shift_labels.to(shift_logits.device)
+            loss = loss_fct(shift_logits, shift_labels)
+
+        aux_loss = None
+        if output_router_logits:
+            aux_loss = load_balancing_loss_func(
+                outputs.router_logits if return_dict else outputs[-1],
+                self.num_experts,
+                self.num_experts_per_tok,
+                attention_mask,
+            )
+            if labels is not None and loss is not None:
+                loss += self.moe_loss_weight * aux_loss.to(loss.device)  # make sure to reside in the same device
+
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            if output_router_logits:
+                output = (aux_loss,) + output
+            return (loss,) + output if loss is not None else output
+
+        return MoeCausalLMOutputWithPast(
+            loss=loss,
+            aux_loss=aux_loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+            router_logits=outputs.router_logits,
+        )
+
+    def prepare_inputs_for_generation(
+        self,
+        input_ids: torch.Tensor,
+        past_key_values: Optional[Cache] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        **kwargs: Any,
+    ) -> Dict[str, Any]:
+        past_length = 0
+        if past_key_values is not None:
+            if isinstance(past_key_values, Cache):
+                cache_length = past_key_values.get_seq_length()
+                past_length = past_key_values.seen_tokens
+                max_cache_length = past_key_values.get_max_length()
+            else:
+                cache_length = past_length = past_key_values[0][0].shape[2]
+                max_cache_length = None
+
+            # Keep only the unprocessed tokens:
+            # 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where
+            # some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as
+            # input)
+            if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
+                input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :]
+            # 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
+            # input_ids based on the past_length.
+            elif past_length < input_ids.shape[1]:
+                input_ids = input_ids[:, past_length:]
+            # 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens.
+
+            # If we are about to go beyond the maximum cache length, we need to crop the input attention mask.
+            if (
+                max_cache_length is not None
+                and attention_mask is not None
+                and cache_length + input_ids.shape[1] > max_cache_length
+            ):
+                attention_mask = attention_mask[:, -max_cache_length:]
+
+        position_ids = kwargs.get("position_ids", None)
+        if attention_mask is not None and position_ids is None:
+            # create position_ids on the fly for batch generation
+            position_ids = attention_mask.long().cumsum(-1) - 1
+            position_ids.masked_fill_(attention_mask == 0, 1)
+            if past_key_values:
+                position_ids = position_ids[:, -input_ids.shape[1] :]
+
+        if self.generation_config.cache_implementation == "static":
+            # generation with static cache
+            cache_position = kwargs.get("cache_position", None)
+            if cache_position is None:
+                past_length = 0
+            else:
+                past_length = cache_position[-1] + 1
+            input_ids = input_ids[:, past_length:]
+            position_ids = position_ids[:, past_length:] if position_ids is not None else None
+
+        # TODO @gante we should only keep a `cache_position` in generate, and do +=1.
+        # same goes for position ids. Could also help with continued generation.
+        input_length = position_ids.shape[-1] if position_ids is not None else input_ids.shape[-1]
+        cache_position = torch.arange(past_length, past_length + input_length, device=input_ids.device)
+        position_ids = position_ids.contiguous() if position_ids is not None else None
+
+        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
+        if inputs_embeds is not None and past_key_values is None:
+            model_inputs = {"inputs_embeds": inputs_embeds}
+        else:
+            # The `contiguous()` here is necessary to have a static stride during decoding. torchdynamo otherwise
+            # recompiles graphs as the stride of the inputs is a guard. Ref: https://github.com/huggingface/transformers/pull/29114
+            # TODO: use `next_tokens` directly instead.
+            model_inputs = {"input_ids": input_ids.contiguous()}
+
+        model_inputs.update(
+            {
+                "position_ids": position_ids,
+                "cache_position": cache_position,
+                "past_key_values": past_key_values,
+                "use_cache": kwargs.get("use_cache"),
+                "attention_mask": attention_mask,
+            }
+        )
+        return model_inputs
+
+    @staticmethod
+    def _reorder_cache(past_key_values: Cache, beam_idx: torch.LongTensor):
+        reordered_past = ()
+        for layer_past in past_key_values:
+            reordered_past += (
+                tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past),
+            )
+        return reordered_past

venv/lib/python3.10/site-packages/transformers/models/esm/__init__.py

@@ -0,0 +1,94 @@
+# Copyright 2022 Facebook and The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from typing import TYPE_CHECKING
+
+from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_tf_available, is_torch_available
+
+
+_import_structure = {
+    "configuration_esm": ["ESM_PRETRAINED_CONFIG_ARCHIVE_MAP", "EsmConfig"],
+    "tokenization_esm": ["EsmTokenizer"],
+}
+
+try:
+    if not is_torch_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["modeling_esm"] = [
+        "ESM_PRETRAINED_MODEL_ARCHIVE_LIST",
+        "EsmForMaskedLM",
+        "EsmForSequenceClassification",
+        "EsmForTokenClassification",
+        "EsmModel",
+        "EsmPreTrainedModel",
+    ]
+    _import_structure["modeling_esmfold"] = ["EsmForProteinFolding", "EsmFoldPreTrainedModel"]
+
+try:
+    if not is_tf_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["modeling_tf_esm"] = [
+        "TF_ESM_PRETRAINED_MODEL_ARCHIVE_LIST",
+        "TFEsmForMaskedLM",
+        "TFEsmForSequenceClassification",
+        "TFEsmForTokenClassification",
+        "TFEsmModel",
+        "TFEsmPreTrainedModel",
+    ]
+
+if TYPE_CHECKING:
+    from .configuration_esm import ESM_PRETRAINED_CONFIG_ARCHIVE_MAP, EsmConfig
+    from .tokenization_esm import EsmTokenizer
+
+    try:
+        if not is_torch_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .modeling_esm import (
+            ESM_PRETRAINED_MODEL_ARCHIVE_LIST,
+            EsmForMaskedLM,
+            EsmForSequenceClassification,
+            EsmForTokenClassification,
+            EsmModel,
+            EsmPreTrainedModel,
+        )
+        from .modeling_esmfold import EsmFoldPreTrainedModel, EsmForProteinFolding
+
+    try:
+        if not is_tf_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .modeling_tf_esm import (
+            TF_ESM_PRETRAINED_MODEL_ARCHIVE_LIST,
+            TFEsmForMaskedLM,
+            TFEsmForSequenceClassification,
+            TFEsmForTokenClassification,
+            TFEsmModel,
+            TFEsmPreTrainedModel,
+        )
+
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure)

venv/lib/python3.10/site-packages/transformers/models/esm/configuration_esm.py

@@ -0,0 +1,361 @@
+# coding=utf-8
+# Copyright 2022 Meta and The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" ESM model configuration"""
+
+from dataclasses import asdict, dataclass
+from typing import Optional
+
+from ...configuration_utils import PretrainedConfig
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+# TODO Update this
+
+from ..deprecated._archive_maps import ESM_PRETRAINED_CONFIG_ARCHIVE_MAP  # noqa: F401, E402
+
+
+class EsmConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`ESMModel`]. It is used to instantiate a ESM model
+    according to the specified arguments, defining the model architecture. Instantiating a configuration with the
+    defaults will yield a similar configuration to that of the ESM
+    [facebook/esm-1b](https://huggingface.co/facebook/esm-1b) architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*):
+            Vocabulary size of the ESM model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`ESMModel`].
+        mask_token_id (`int`, *optional*):
+            The index of the mask token in the vocabulary. This must be included in the config because of the
+            "mask-dropout" scaling trick, which will scale the inputs depending on the number of masked tokens.
+        pad_token_id (`int`, *optional*):
+            The index of the padding token in the vocabulary. This must be included in the config because certain parts
+            of the ESM code use this instead of the attention mask.
+        hidden_size (`int`, *optional*, defaults to 768):
+            Dimensionality of the encoder layers and the pooler layer.
+        num_hidden_layers (`int`, *optional*, defaults to 12):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 12):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        intermediate_size (`int`, *optional*, defaults to 3072):
+            Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
+        hidden_dropout_prob (`float`, *optional*, defaults to 0.1):
+            The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
+        attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1):
+            The dropout ratio for the attention probabilities.
+        max_position_embeddings (`int`, *optional*, defaults to 1026):
+            The maximum sequence length that this model might ever be used with. Typically set this to something large
+            just in case (e.g., 512 or 1024 or 2048).
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        layer_norm_eps (`float`, *optional*, defaults to 1e-12):
+            The epsilon used by the layer normalization layers.
+        position_embedding_type (`str`, *optional*, defaults to `"absolute"`):
+            Type of position embedding. Choose one of `"absolute"`, `"relative_key"`, `"relative_key_query", "rotary"`.
+            For positional embeddings use `"absolute"`. For more information on `"relative_key"`, please refer to
+            [Self-Attention with Relative Position Representations (Shaw et al.)](https://arxiv.org/abs/1803.02155).
+            For more information on `"relative_key_query"`, please refer to *Method 4* in [Improve Transformer Models
+            with Better Relative Position Embeddings (Huang et al.)](https://arxiv.org/abs/2009.13658).
+        is_decoder (`bool`, *optional*, defaults to `False`):
+            Whether the model is used as a decoder or not. If `False`, the model is used as an encoder.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        emb_layer_norm_before (`bool`, *optional*):
+            Whether to apply layer normalization after embeddings but before the main stem of the network.
+        token_dropout (`bool`, defaults to `False`):
+            When this is enabled, masked tokens are treated as if they had been dropped out by input dropout.
+
+    Examples:
+
+    ```python
+    >>> from transformers import EsmModel, EsmConfig
+
+    >>> # Initializing a ESM facebook/esm-1b style configuration >>> configuration = EsmConfig()
+
+    >>> # Initializing a model from the configuration >>> model = ESMModel(configuration)
+
+    >>> # Accessing the model configuration >>> configuration = model.config
+    ```"""
+
+    model_type = "esm"
+
+    def __init__(
+        self,
+        vocab_size=None,
+        mask_token_id=None,
+        pad_token_id=None,
+        hidden_size=768,
+        num_hidden_layers=12,
+        num_attention_heads=12,
+        intermediate_size=3072,
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        max_position_embeddings=1026,
+        initializer_range=0.02,
+        layer_norm_eps=1e-12,
+        position_embedding_type="absolute",
+        use_cache=True,
+        emb_layer_norm_before=None,
+        token_dropout=False,
+        is_folding_model=False,
+        esmfold_config=None,
+        vocab_list=None,
+        **kwargs,
+    ):
+        super().__init__(pad_token_id=pad_token_id, mask_token_id=mask_token_id, **kwargs)
+
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.intermediate_size = intermediate_size
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.max_position_embeddings = max_position_embeddings
+        self.initializer_range = initializer_range
+        self.layer_norm_eps = layer_norm_eps
+        self.position_embedding_type = position_embedding_type
+        self.use_cache = use_cache
+        self.emb_layer_norm_before = emb_layer_norm_before
+        self.token_dropout = token_dropout
+        self.is_folding_model = is_folding_model
+        if is_folding_model:
+            if esmfold_config is None:
+                logger.info("No esmfold_config supplied for folding model, using default values.")
+                esmfold_config = EsmFoldConfig()
+            elif isinstance(esmfold_config, dict):
+                esmfold_config = EsmFoldConfig(**esmfold_config)
+            self.esmfold_config = esmfold_config
+            if vocab_list is None:
+                logger.warning("No vocab_list supplied for folding model, assuming the ESM-2 vocabulary!")
+                self.vocab_list = get_default_vocab_list()
+            else:
+                self.vocab_list = vocab_list
+        else:
+            self.esmfold_config = None
+            self.vocab_list = None
+        if self.esmfold_config is not None and getattr(self.esmfold_config, "use_esm_attn_map", False):
+            raise ValueError("The HuggingFace port of ESMFold does not support use_esm_attn_map at this time!")
+
+    def to_dict(self):
+        """
+        Serializes this instance to a Python dictionary. Override the default [`~PretrainedConfig.to_dict`].
+
+        Returns:
+            `Dict[str, any]`: Dictionary of all the attributes that make up this configuration instance,
+        """
+        output = super().to_dict()
+        if isinstance(self.esmfold_config, EsmFoldConfig):
+            output["esmfold_config"] = self.esmfold_config.to_dict()
+        return output
+
+
+@dataclass
+class EsmFoldConfig:
+    esm_type: str = None
+    fp16_esm: bool = True
+    use_esm_attn_map: bool = False
+    esm_ablate_pairwise: bool = False
+    esm_ablate_sequence: bool = False
+    esm_input_dropout: float = 0
+
+    embed_aa: bool = True
+    bypass_lm: bool = False
+
+    lddt_head_hid_dim: int = 128
+    trunk: "TrunkConfig" = None
+
+    def __post_init__(self):
+        if self.trunk is None:
+            self.trunk = TrunkConfig()
+        elif isinstance(self.trunk, dict):
+            self.trunk = TrunkConfig(**self.trunk)
+
+    def to_dict(self):
+        """
+        Serializes this instance to a Python dictionary. Override the default [`~PretrainedConfig.to_dict`].
+
+        Returns:
+            `Dict[str, any]`: Dictionary of all the attributes that make up this configuration instance,
+        """
+        output = asdict(self)
+        output["trunk"] = self.trunk.to_dict()
+        return output
+
+
+@dataclass
+class TrunkConfig:
+    num_blocks: int = 48
+    sequence_state_dim: int = 1024
+    pairwise_state_dim: int = 128
+    sequence_head_width: int = 32
+    pairwise_head_width: int = 32
+    position_bins: int = 32
+    dropout: float = 0
+    layer_drop: float = 0
+    cpu_grad_checkpoint: bool = False
+    max_recycles: int = 4
+    chunk_size: Optional[int] = 128
+    structure_module: "StructureModuleConfig" = None
+
+    def __post_init__(self):
+        if self.structure_module is None:
+            self.structure_module = StructureModuleConfig()
+        elif isinstance(self.structure_module, dict):
+            self.structure_module = StructureModuleConfig(**self.structure_module)
+
+        if self.max_recycles <= 0:
+            raise ValueError(f"`max_recycles` should be positive, got {self.max_recycles}.")
+        if self.sequence_state_dim % self.sequence_state_dim != 0:
+            raise ValueError(
+                "`sequence_state_dim` should be a round multiple of `sequence_state_dim`, got"
+                f" {self.sequence_state_dim} and {self.sequence_state_dim}."
+            )
+        if self.pairwise_state_dim % self.pairwise_state_dim != 0:
+            raise ValueError(
+                "`pairwise_state_dim` should be a round multiple of `pairwise_state_dim`, got"
+                f" {self.pairwise_state_dim} and {self.pairwise_state_dim}."
+            )
+
+        sequence_num_heads = self.sequence_state_dim // self.sequence_head_width
+        pairwise_num_heads = self.pairwise_state_dim // self.pairwise_head_width
+
+        if self.sequence_state_dim != sequence_num_heads * self.sequence_head_width:
+            raise ValueError(
+                "`sequence_state_dim` should be equal to `sequence_num_heads * sequence_head_width, got"
+                f" {self.sequence_state_dim} != {sequence_num_heads} * {self.sequence_head_width}."
+            )
+        if self.pairwise_state_dim != pairwise_num_heads * self.pairwise_head_width:
+            raise ValueError(
+                "`pairwise_state_dim` should be equal to `pairwise_num_heads * pairwise_head_width, got"
+                f" {self.pairwise_state_dim} != {pairwise_num_heads} * {self.pairwise_head_width}."
+            )
+        if self.pairwise_state_dim % 2 != 0:
+            raise ValueError(f"`pairwise_state_dim` should be even, got {self.pairwise_state_dim}.")
+
+        if self.dropout >= 0.4:
+            raise ValueError(f"`dropout` should not be greater than 0.4, got {self.dropout}.")
+
+    def to_dict(self):
+        """
+        Serializes this instance to a Python dictionary. Override the default [`~PretrainedConfig.to_dict`].
+
+        Returns:
+            `Dict[str, any]`: Dictionary of all the attributes that make up this configuration instance,
+        """
+        output = asdict(self)
+        output["structure_module"] = self.structure_module.to_dict()
+        return output
+
+
+@dataclass
+class StructureModuleConfig:
+    """
+    Args:
+        sequence_dim:
+            Single representation channel dimension
+        pairwise_dim:
+            Pair representation channel dimension
+        ipa_dim:
+            IPA hidden channel dimension
+        resnet_dim:
+            Angle resnet (Alg. 23 lines 11-14) hidden channel dimension
+        num_heads_ipa:
+            Number of IPA heads
+        num_qk_points:
+            Number of query/key points to generate during IPA
+        num_v_points:
+            Number of value points to generate during IPA
+        dropout_rate:
+            Dropout rate used throughout the layer
+        num_blocks:
+            Number of structure module blocks
+        num_transition_layers:
+            Number of layers in the single representation transition (Alg. 23 lines 8-9)
+        num_resnet_blocks:
+            Number of blocks in the angle resnet
+        num_angles:
+            Number of angles to generate in the angle resnet
+        trans_scale_factor:
+            Scale of single representation transition hidden dimension
+        epsilon:
+            Small number used in angle resnet normalization
+        inf:
+            Large number used for attention masking
+    """
+
+    sequence_dim: int = 384
+    pairwise_dim: int = 128
+    ipa_dim: int = 16
+    resnet_dim: int = 128
+    num_heads_ipa: int = 12
+    num_qk_points: int = 4
+    num_v_points: int = 8
+    dropout_rate: float = 0.1
+    num_blocks: int = 8
+    num_transition_layers: int = 1
+    num_resnet_blocks: int = 2
+    num_angles: int = 7
+    trans_scale_factor: int = 10
+    epsilon: float = 1e-8
+    inf: float = 1e5
+
+    def to_dict(self):
+        return asdict(self)
+
+
+def get_default_vocab_list():
+    return (
+        "<cls>",
+        "<pad>",
+        "<eos>",
+        "<unk>",
+        "L",
+        "A",
+        "G",
+        "V",
+        "S",
+        "E",
+        "R",
+        "T",
+        "I",
+        "D",
+        "P",
+        "K",
+        "Q",
+        "N",
+        "F",
+        "Y",
+        "M",
+        "H",
+        "W",
+        "C",
+        "X",
+        "B",
+        "U",
+        "Z",
+        "O",
+        ".",
+        "-",
+        "<null_1>",
+        "<mask>",
+    )

venv/lib/python3.10/site-packages/transformers/models/esm/convert_esm.py

@@ -0,0 +1,400 @@
+# coding=utf-8
+# Copyright 2022 The HuggingFace Inc. team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Convert ESM checkpoint."""
+
+
+import argparse
+import pathlib
+from pathlib import Path
+from tempfile import TemporaryDirectory
+
+import esm as esm_module
+import torch
+from esm.esmfold.v1.misc import batch_encode_sequences as esmfold_encode_sequences
+from esm.esmfold.v1.pretrained import esmfold_v1
+
+from transformers.models.esm.configuration_esm import EsmConfig, EsmFoldConfig
+from transformers.models.esm.modeling_esm import (
+    EsmForMaskedLM,
+    EsmForSequenceClassification,
+    EsmIntermediate,
+    EsmLayer,
+    EsmOutput,
+    EsmSelfAttention,
+    EsmSelfOutput,
+)
+from transformers.models.esm.modeling_esmfold import EsmForProteinFolding
+from transformers.models.esm.tokenization_esm import EsmTokenizer
+from transformers.utils import logging
+
+
+logging.set_verbosity_info()
+logger = logging.get_logger(__name__)
+
+SAMPLE_DATA = [
+    (
+        "protein1",
+        "MNGTEGPNFYVPFSNATGVVRSPFEYPQYYLAEPWQFSMLAAYMFLLIVLGFPINFLTLYVTVQHKKLRTPLNYILLNLAVADLFMVLGGFTSTLYTSLHGYFVFGPTGCNLEGFFATLGGEIALWSLVVLAIERYVVVCKPMSNFRFGENHAIMGVAFTWVMALACAAPPLAGWSRYIPEGLQCSCGIDYYTLKPEVNNESFVIYMFVVHFTIPMIIIFFCYGQLVFTVKEAAAQQQESATTQKAEKEVTRMVIIMVIAFLICWVPYASVAFYIFTHQGSNFGPIFMTIPAFFAKSAAIYNPVIYIMMNKQFRNCMLTTICCGKNPLGDDEASATVSKTETSQVAPA",
+    ),
+    ("protein2", "MKTVRQERLKSIVRILERSKEPVSGAQLAEELSVSRQVIVQDIAYLRSLGYNIVATPRGYVLA"),
+    ("protein3", "MKTVRQERLKSI<mask>RILERSKEPVSGAQLAEELS<mask>SRQVIVQDIAYLRSLGYN<mask>VATPRGYVLAGG"),
+    ("protein4", "MKTVRQERLKSI<mask>RILERSKEPVSGAQLAEELS<mask>SRQVIVQDIAYLRSLGYN<mask>VATPRGYVLA"),
+]
+
+MODEL_MAPPING = {
+    "esm1b_t33_650M_UR50S": esm_module.pretrained.esm1b_t33_650M_UR50S,
+    "esm1v_t33_650M_UR90S_1": esm_module.pretrained.esm1v_t33_650M_UR90S_1,
+    "esm1v_t33_650M_UR90S_2": esm_module.pretrained.esm1v_t33_650M_UR90S_2,
+    "esm1v_t33_650M_UR90S_3": esm_module.pretrained.esm1v_t33_650M_UR90S_3,
+    "esm1v_t33_650M_UR90S_4": esm_module.pretrained.esm1v_t33_650M_UR90S_4,
+    "esm1v_t33_650M_UR90S_5": esm_module.pretrained.esm1v_t33_650M_UR90S_5,
+    "esm2_t48_15B_UR50D": esm_module.pretrained.esm2_t48_15B_UR50D,
+    "esm2_t36_3B_UR50D": esm_module.pretrained.esm2_t36_3B_UR50D,
+    "esm2_t33_650M_UR50D": esm_module.pretrained.esm2_t33_650M_UR50D,
+    "esm2_t30_150M_UR50D": esm_module.pretrained.esm2_t30_150M_UR50D,
+    "esm2_t12_35M_UR50D": esm_module.pretrained.esm2_t12_35M_UR50D,
+    "esm2_t6_8M_UR50D": esm_module.pretrained.esm2_t6_8M_UR50D,
+    "esmfold_v1": esmfold_v1,
+}
+
+restypes = list("ARNDCQEGHILKMFPSTWYV")
+
+restypes_with_x = restypes + ["X"]
+restypes_with_extras = restypes_with_x + ["<pad>", "<mask>", "<cls>", "<sep>", "<eos>"]
+
+
+def get_esmfold_tokenizer():
+    with TemporaryDirectory() as tempdir:
+        vocab = "\n".join(restypes_with_extras)
+        vocab_file = Path(tempdir) / "vocab.txt"
+        vocab_file.write_text(vocab)
+        hf_tokenizer = EsmTokenizer(vocab_file=str(vocab_file))
+    hf_tokenizer.pad_token_id = 0  # Overlaps with 'A' but that seems to be what they want
+    return hf_tokenizer
+
+
+def transfer_and_check_weights(original_module, our_module):
+    status = our_module.load_state_dict(original_module.state_dict())
+    if status.missing_keys:
+        raise ValueError(f"Missing keys: {status.missing_keys}")
+    if status.unexpected_keys:
+        raise ValueError(f"Unexpected keys: {status.unexpected_keys}")
+
+
+def convert_esm_checkpoint_to_pytorch(
+    model: str, pytorch_dump_folder_path: str, classification_head: bool, push_to_repo: str, auth_token: str
+):
+    """
+    Copy/paste/tweak esm's weights to our BERT structure.
+    """
+    if model.startswith("esmfold"):
+        esm = MODEL_MAPPING[model]()
+    else:
+        esm, alphabet = MODEL_MAPPING[model]()
+    esm.eval()  # disable dropout
+
+    if model.startswith("esmfold"):
+        embed_dim = esm.esm.embed_dim
+        num_layers = esm.esm.num_layers
+        num_attention_heads = esm.esm.attention_heads
+        intermediate_size = 4 * embed_dim
+        token_dropout = esm.esm.token_dropout
+        emb_layer_norm_before = False  # This code path does not exist in ESM-2
+        position_embedding_type = "rotary"
+        is_folding_model = True
+        esmfold_config = EsmFoldConfig()
+        for key, val in esm.cfg.items():
+            if hasattr(esmfold_config, key) and key != "trunk":
+                setattr(esmfold_config, key, val)
+        for key, val in esm.cfg.trunk.items():
+            if hasattr(esmfold_config.trunk, key) and key != "structure_module":
+                setattr(esmfold_config.trunk, key, val)
+        for key, val in esm.cfg.trunk.structure_module.items():
+            if hasattr(esmfold_config.trunk.structure_module, key):
+                setattr(esmfold_config.trunk.structure_module, key, val)
+    elif hasattr(esm, "args"):
+        # Indicates an ESM-1b or ESM-1v model
+        embed_dim = esm.args.embed_dim
+        num_layers = esm.args.layers
+        num_attention_heads = esm.args.attention_heads
+        intermediate_size = esm.args.ffn_embed_dim
+        token_dropout = esm.args.token_dropout
+        emb_layer_norm_before = True if esm.emb_layer_norm_before else False
+        position_embedding_type = "absolute"
+        is_folding_model = False
+        esmfold_config = None
+    else:
+        # Indicates an ESM-2 model
+        embed_dim = esm.embed_dim
+        num_layers = esm.num_layers
+        num_attention_heads = esm.attention_heads
+        intermediate_size = 4 * embed_dim  # This is hardcoded in ESM-2
+        token_dropout = esm.token_dropout
+        emb_layer_norm_before = False  # This code path does not exist in ESM-2
+        position_embedding_type = "rotary"
+        is_folding_model = False
+        esmfold_config = None
+
+    if is_folding_model:
+        alphabet = esm.esm.alphabet
+    vocab_list = tuple(alphabet.all_toks)
+    mask_token_id = alphabet.mask_idx
+    pad_token_id = alphabet.padding_idx
+
+    if is_folding_model:
+        original_esm_model = esm.esm
+    else:
+        original_esm_model = esm
+
+    config = EsmConfig(
+        vocab_size=original_esm_model.embed_tokens.num_embeddings,
+        mask_token_id=mask_token_id,
+        hidden_size=embed_dim,
+        num_hidden_layers=num_layers,
+        num_attention_heads=num_attention_heads,
+        intermediate_size=intermediate_size,
+        max_position_embeddings=1026,
+        layer_norm_eps=1e-5,  # PyTorch default used in fairseq
+        attention_probs_dropout_prob=0.0,
+        hidden_dropout_prob=0.0,
+        pad_token_id=pad_token_id,
+        emb_layer_norm_before=emb_layer_norm_before,
+        token_dropout=token_dropout,
+        position_embedding_type=position_embedding_type,
+        is_folding_model=is_folding_model,
+        esmfold_config=esmfold_config,
+        vocab_list=vocab_list,
+    )
+    if classification_head:
+        config.num_labels = esm.classification_heads["mnli"].out_proj.weight.shape[0]
+    print("Our ESM config:", config)
+
+    if model.startswith("esmfold"):
+        model_class = EsmForProteinFolding
+    elif classification_head:
+        model_class = EsmForSequenceClassification
+    else:
+        model_class = EsmForMaskedLM
+    model = model_class(config)
+    model.eval()
+
+    # Now let's copy all the weights.
+    # Embeddings
+    model.esm.embeddings.word_embeddings.weight = original_esm_model.embed_tokens.weight
+    if position_embedding_type == "absolute":
+        model.esm.embeddings.position_embeddings.weight = original_esm_model.embed_positions.weight
+
+    if config.emb_layer_norm_before:
+        model.esm.embeddings.layer_norm.weight = original_esm_model.emb_layer_norm_before.weight
+        model.esm.embeddings.layer_norm.bias = original_esm_model.emb_layer_norm_before.bias
+
+    model.esm.encoder.emb_layer_norm_after.weight = original_esm_model.emb_layer_norm_after.weight
+    model.esm.encoder.emb_layer_norm_after.bias = original_esm_model.emb_layer_norm_after.bias
+
+    for i in range(config.num_hidden_layers):
+        # Encoder: start of layer
+        layer: EsmLayer = model.esm.encoder.layer[i]
+        # esm_layer: TransformerSentenceEncoderLayer = original_esm_model.layers[i]
+        esm_layer = original_esm_model.layers[i]
+
+        # self attention
+        self_attn: EsmSelfAttention = layer.attention.self
+        assert (
+            esm_layer.self_attn.k_proj.weight.data.shape
+            == esm_layer.self_attn.q_proj.weight.data.shape
+            == esm_layer.self_attn.v_proj.weight.data.shape
+            == torch.Size((config.hidden_size, config.hidden_size))
+        )
+
+        self_attn.query.weight.data = esm_layer.self_attn.q_proj.weight
+        self_attn.query.bias.data = esm_layer.self_attn.q_proj.bias
+        self_attn.key.weight.data = esm_layer.self_attn.k_proj.weight
+        self_attn.key.bias.data = esm_layer.self_attn.k_proj.bias
+        self_attn.value.weight.data = esm_layer.self_attn.v_proj.weight
+        self_attn.value.bias.data = esm_layer.self_attn.v_proj.bias
+
+        if getattr(esm_layer.self_attn, "rot_emb", None) is not None:
+            # Matt: Although inv_freq is not a trainable weight, it is computed at model init and cached.
+            # During the training of ESM-2 the model was converted to float16 precision, which also converts
+            # the inv_freq tensor, and the loss of precision remains even if the model is loaded later as float32.
+            # If we recompute inv_freq without this loss of precision then we will get subtly different rotary
+            # embeddings, which are enough to cause significant discrepancies in model outputs. To avoid this,
+            # we make sure the new model copies the data from the old inv_freq.
+            self_attn.rotary_embeddings.inv_freq.data = esm_layer.self_attn.rot_emb.inv_freq
+
+        # LayerNorm changes for pre-activation
+        layer.attention.LayerNorm.weight = esm_layer.self_attn_layer_norm.weight
+        layer.attention.LayerNorm.bias = esm_layer.self_attn_layer_norm.bias
+        layer.LayerNorm.weight = esm_layer.final_layer_norm.weight
+        layer.LayerNorm.bias = esm_layer.final_layer_norm.bias
+
+        # self-attention output
+        self_output: EsmSelfOutput = layer.attention.output
+        assert self_output.dense.weight.shape == esm_layer.self_attn.out_proj.weight.shape
+        self_output.dense.weight = esm_layer.self_attn.out_proj.weight
+        self_output.dense.bias = esm_layer.self_attn.out_proj.bias
+
+        # intermediate
+        intermediate: EsmIntermediate = layer.intermediate
+        assert intermediate.dense.weight.shape == esm_layer.fc1.weight.shape
+        intermediate.dense.weight = esm_layer.fc1.weight
+        intermediate.dense.bias = esm_layer.fc1.bias
+
+        # output
+        bert_output: EsmOutput = layer.output
+        assert bert_output.dense.weight.shape == esm_layer.fc2.weight.shape
+        bert_output.dense.weight = esm_layer.fc2.weight
+        bert_output.dense.bias = esm_layer.fc2.bias
+        # end of layer
+
+    if is_folding_model:
+        model.esm_s_combine.data = esm.esm_s_combine.data
+        model.af2_to_esm.data = esm.af2_to_esm.data
+        transfer_and_check_weights(esm.embedding, model.embedding)
+        transfer_and_check_weights(esm.esm_s_mlp, model.esm_s_mlp)
+        transfer_and_check_weights(esm.trunk, model.trunk)
+        transfer_and_check_weights(esm.distogram_head, model.distogram_head)
+        transfer_and_check_weights(esm.ptm_head, model.ptm_head)
+        transfer_and_check_weights(esm.lm_head, model.lm_head)
+        transfer_and_check_weights(esm.lddt_head, model.lddt_head)
+
+    elif classification_head:
+        model.classifier.dense.weight = esm.esm.classification_heads["mnli"].dense.weight
+        model.classifier.dense.bias = esm.classification_heads["mnli"].dense.bias
+        model.classifier.out_proj.weight = esm.classification_heads["mnli"].out_proj.weight
+        model.classifier.out_proj.bias = esm.classification_heads["mnli"].out_proj.bias
+    else:
+        # LM Head
+        model.lm_head.dense.weight = esm.lm_head.dense.weight
+        model.lm_head.dense.bias = esm.lm_head.dense.bias
+        model.lm_head.layer_norm.weight = esm.lm_head.layer_norm.weight
+        model.lm_head.layer_norm.bias = esm.lm_head.layer_norm.bias
+        model.lm_head.decoder.weight = esm.lm_head.weight
+        model.lm_head.bias = esm.lm_head.bias
+
+    # Contact prediction head
+    transfer_and_check_weights(esm.contact_head, model.esm.contact_head)
+
+    # Prepare data (first 2 sequences from ESMStructuralSplitDataset superfamily / 4)
+    if is_folding_model:
+        # Folding models aren't trained on masked inputs and don't like mask tokens.
+        sample_data = SAMPLE_DATA[:2]
+    else:
+        sample_data = SAMPLE_DATA
+
+    if is_folding_model:
+        hf_tokenizer = get_esmfold_tokenizer()
+        hf_tokens = hf_tokenizer(
+            [row[1] for row in sample_data], return_tensors="pt", padding=True, add_special_tokens=False
+        )
+        esmfold_aas, esmfold_mask, _, _, _ = esmfold_encode_sequences([row[1] for row in sample_data])
+        success = torch.all(hf_tokens["input_ids"] == esmfold_aas) and torch.all(
+            hf_tokens["attention_mask"] == esmfold_mask
+        )
+    else:
+        # Let's check that we get the same results.
+        batch_converter = alphabet.get_batch_converter()
+        batch_labels, batch_strs, batch_tokens = batch_converter(sample_data)
+        # Prepare tokenizer and make sure it matches
+        with TemporaryDirectory() as tempdir:
+            vocab = "\n".join(alphabet.all_toks)
+            vocab_file = Path(tempdir) / "vocab.txt"
+            vocab_file.write_text(vocab)
+            hf_tokenizer = EsmTokenizer(vocab_file=str(vocab_file))
+
+        hf_tokens = hf_tokenizer([row[1] for row in sample_data], return_tensors="pt", padding=True)
+        success = torch.all(hf_tokens["input_ids"] == batch_tokens)
+
+    print("Do both models tokenizers output the same tokens?", "🔥" if success else "💩")
+    if not success:
+        raise Exception("Tokenization does not match!")
+
+    with torch.no_grad():
+        if is_folding_model:
+            # Let's test the model in parts
+            # ESMFold always converts the ESM stem to float16, which requires float16 ops
+            # that don't exist on CPU. Therefore, to test it we need to run it on GPU. However,
+            # ESMFold is what we in the community call a "big boy" and so we desperately avoid putting both the
+            # original and the converted model on the GPU at the same time.
+            their_output = esm.cuda().infer([row[1] for row in sample_data])
+            our_output = model.cuda()(
+                input_ids=hf_tokens["input_ids"].cuda(), attention_mask=hf_tokens["attention_mask"].cuda()
+            )
+        else:
+            our_output = model(**hf_tokens, output_hidden_states=True)
+            our_output = our_output["logits"]
+            if classification_head:
+                their_output = esm.model.classification_heads["mnli"](esm.extract_features(batch_tokens))
+            else:
+                their_output = esm(hf_tokens["input_ids"], repr_layers=list(range(999)))
+                their_output = their_output["logits"]
+
+        if is_folding_model:
+            max_absolute_diff = torch.max(torch.abs(our_output["positions"] - their_output["positions"])).item()
+            success = torch.allclose(our_output["positions"], their_output["positions"], atol=1e-5)
+        else:
+            max_absolute_diff = torch.max(torch.abs(our_output - their_output)).item()
+            success = torch.allclose(our_output, their_output, atol=1e-5)
+
+        print(f"max_absolute_diff = {max_absolute_diff}")  # ~ 1e-5
+        print("Do both models output the same tensors?", "🔥" if success else "💩")
+
+        if not success:
+            raise Exception("Something went wRoNg")
+
+        if not is_folding_model:
+            # Let's check contact prediction too
+            our_output = model.predict_contacts(hf_tokens["input_ids"], hf_tokens["attention_mask"])
+            their_output = esm.predict_contacts(hf_tokens["input_ids"])
+            max_absolute_diff = torch.max(torch.abs(our_output - their_output)).item()
+            success = torch.allclose(our_output, their_output, atol=1e-5)
+
+            print("Contact prediction testing:")
+            print(f"max_absolute_diff = {max_absolute_diff}")  # ~ 1e-5
+            print("Do both models output the same tensors?", "🔥" if success else "💩")
+
+            if not success:
+                raise Exception("Something went wRoNg")
+
+        pathlib.Path(pytorch_dump_folder_path).mkdir(parents=True, exist_ok=True)
+        print(f"Saving model to {pytorch_dump_folder_path}")
+        model.save_pretrained(pytorch_dump_folder_path)
+
+        del esm  # Free up some memory before continuing
+
+    print(f"Saving tokenizer to {pytorch_dump_folder_path}")
+    hf_tokenizer.save_pretrained(pytorch_dump_folder_path)
+
+    if push_to_repo:
+        model.push_to_hub(repo_id=push_to_repo, token_token=auth_token)
+        hf_tokenizer.push_to_hub(repo_id=push_to_repo, token_token=auth_token)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    # Required parameters
+    parser.add_argument(
+        "--pytorch_dump_folder_path", type=str, required=True, help="Path to the output PyTorch model."
+    )
+    parser.add_argument(
+        "--classification_head", action="store_true", help="Whether to convert a final classification head."
+    )
+    parser.add_argument("--model", default=None, type=str, required=True, help="Name of model to convert.")
+    parser.add_argument("--push_to_repo", type=str, help="Repo to upload to (including username!).")
+    parser.add_argument("--auth_token", type=str, help="HuggingFace auth token.")
+    args = parser.parse_args()
+    convert_esm_checkpoint_to_pytorch(
+        args.model, args.pytorch_dump_folder_path, args.classification_head, args.push_to_repo, args.auth_token
+    )

venv/lib/python3.10/site-packages/transformers/models/esm/modeling_esm.py

@@ -0,0 +1,1265 @@
+# coding=utf-8
+# Copyright 2022 Meta and The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" PyTorch ESM model."""
+
+import math
+from typing import List, Optional, Tuple, Union
+
+import torch
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+
+from ...file_utils import add_code_sample_docstrings, add_start_docstrings, add_start_docstrings_to_model_forward
+from ...modeling_outputs import (
+    BaseModelOutputWithPastAndCrossAttentions,
+    BaseModelOutputWithPoolingAndCrossAttentions,
+    MaskedLMOutput,
+    SequenceClassifierOutput,
+    TokenClassifierOutput,
+)
+from ...modeling_utils import PreTrainedModel, find_pruneable_heads_and_indices, prune_linear_layer
+from ...utils import logging
+from .configuration_esm import EsmConfig
+
+
+logger = logging.get_logger(__name__)
+
+_CHECKPOINT_FOR_DOC = "facebook/esm2_t6_8M_UR50D"
+_CONFIG_FOR_DOC = "EsmConfig"
+
+
+from ..deprecated._archive_maps import ESM_PRETRAINED_MODEL_ARCHIVE_LIST  # noqa: F401, E402
+
+
+def rotate_half(x):
+    x1, x2 = x.chunk(2, dim=-1)
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(x, cos, sin):
+    cos = cos[:, :, : x.shape[-2], :]
+    sin = sin[:, :, : x.shape[-2], :]
+
+    return (x * cos) + (rotate_half(x) * sin)
+
+
+def gelu(x):
+    """
+    This is the gelu implementation from the original ESM repo. Using F.gelu yields subtly wrong results.
+    """
+    return x * 0.5 * (1.0 + torch.erf(x / math.sqrt(2.0)))
+
+
+def symmetrize(x):
+    "Make layer symmetric in final two dimensions, used for contact prediction."
+    return x + x.transpose(-1, -2)
+
+
+def average_product_correct(x):
+    "Perform average product correct, used for contact prediction."
+    a1 = x.sum(-1, keepdims=True)
+    a2 = x.sum(-2, keepdims=True)
+    a12 = x.sum((-1, -2), keepdims=True)
+
+    avg = a1 * a2
+    avg.div_(a12)  # in-place to reduce memory
+    normalized = x - avg
+    return normalized
+
+
+class RotaryEmbedding(torch.nn.Module):
+    """
+    Rotary position embeddings based on those in
+    [RoFormer](https://huggingface.co/docs/transformers/model_doc/roformer). Query and keys are transformed by rotation
+    matrices which depend on their relative positions.
+    """
+
+    def __init__(self, dim: int):
+        super().__init__()
+        # Generate and save the inverse frequency buffer (non trainable)
+        inv_freq = 1.0 / (10000 ** (torch.arange(0, dim, 2, dtype=torch.int64).float() / dim))
+        inv_freq = inv_freq
+        self.register_buffer("inv_freq", inv_freq)
+
+        self._seq_len_cached = None
+        self._cos_cached = None
+        self._sin_cached = None
+
+    def _update_cos_sin_tables(self, x, seq_dimension=2):
+        seq_len = x.shape[seq_dimension]
+
+        # Reset the tables if the sequence length has changed,
+        # or if we're on a new device (possibly due to tracing for instance)
+        if seq_len != self._seq_len_cached or self._cos_cached.device != x.device:
+            self._seq_len_cached = seq_len
+            t = torch.arange(x.shape[seq_dimension], device=x.device).type_as(self.inv_freq)
+            freqs = torch.outer(t, self.inv_freq)
+            emb = torch.cat((freqs, freqs), dim=-1).to(x.device)
+
+            self._cos_cached = emb.cos()[None, None, :, :]
+            self._sin_cached = emb.sin()[None, None, :, :]
+
+        return self._cos_cached, self._sin_cached
+
+    def forward(self, q: torch.Tensor, k: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
+        self._cos_cached, self._sin_cached = self._update_cos_sin_tables(k, seq_dimension=-2)
+
+        return (
+            apply_rotary_pos_emb(q, self._cos_cached, self._sin_cached),
+            apply_rotary_pos_emb(k, self._cos_cached, self._sin_cached),
+        )
+
+
+class EsmContactPredictionHead(nn.Module):
+    """Performs symmetrization, apc, and computes a logistic regression on the output features"""
+
+    def __init__(
+        self,
+        in_features: int,
+        bias=True,
+        eos_idx: int = 2,
+    ):
+        super().__init__()
+        self.in_features = in_features
+        self.eos_idx = eos_idx
+        self.regression = nn.Linear(in_features, 1, bias)
+        self.activation = nn.Sigmoid()
+
+    def forward(self, tokens, attentions):
+        # remove eos token attentions
+        eos_mask = tokens.ne(self.eos_idx).to(attentions)
+        eos_mask = eos_mask.unsqueeze(1) * eos_mask.unsqueeze(2)
+        attentions = attentions * eos_mask[:, None, None, :, :]
+        attentions = attentions[..., :-1, :-1]
+        # remove cls token attentions
+        attentions = attentions[..., 1:, 1:]
+        batch_size, layers, heads, seqlen, _ = attentions.size()
+        attentions = attentions.view(batch_size, layers * heads, seqlen, seqlen)
+
+        # features: batch x channels x tokens x tokens (symmetric)
+        attentions = attentions.to(
+            self.regression.weight.device
+        )  # attentions always float32, may need to convert to float16
+        attentions = average_product_correct(symmetrize(attentions))
+        attentions = attentions.permute(0, 2, 3, 1)
+        return self.activation(self.regression(attentions).squeeze(3))
+
+
+class EsmEmbeddings(nn.Module):
+    """
+    Same as BertEmbeddings with a tiny tweak for positional embeddings indexing.
+    """
+
+    def __init__(self, config):
+        super().__init__()
+        self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
+
+        if config.emb_layer_norm_before:
+            self.layer_norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        else:
+            self.layer_norm = None
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+        # position_ids (1, len position emb) is contiguous in memory and exported when serialized
+        self.position_embedding_type = getattr(config, "position_embedding_type", "absolute")
+        self.register_buffer(
+            "position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)), persistent=False
+        )
+
+        self.padding_idx = config.pad_token_id
+        self.position_embeddings = nn.Embedding(
+            config.max_position_embeddings, config.hidden_size, padding_idx=self.padding_idx
+        )
+        self.token_dropout = config.token_dropout
+        self.mask_token_id = config.mask_token_id
+
+    def forward(
+        self, input_ids=None, attention_mask=None, position_ids=None, inputs_embeds=None, past_key_values_length=0
+    ):
+        if position_ids is None:
+            if input_ids is not None:
+                # Create the position ids from the input token ids. Any padded tokens remain padded.
+                position_ids = create_position_ids_from_input_ids(input_ids, self.padding_idx, past_key_values_length)
+            else:
+                position_ids = self.create_position_ids_from_inputs_embeds(inputs_embeds)
+
+        if inputs_embeds is None:
+            inputs_embeds = self.word_embeddings(input_ids)
+
+        # Note that if we want to support ESM-1 (not 1b!) in future then we need to support an
+        # embedding_scale factor here.
+        embeddings = inputs_embeds
+
+        # Matt: ESM has the option to handle masking in MLM in a slightly unusual way. If the token_dropout
+        # flag is False then it is handled in the same was as BERT/RoBERTa. If it is set to True, however,
+        # masked tokens are treated as if they were selected for input dropout and zeroed out.
+        # This "mask-dropout" is compensated for when masked tokens are not present, by scaling embeddings by
+        # a factor of (fraction of unmasked tokens during training) / (fraction of unmasked tokens in sample).
+        # This is analogous to the way that dropout layers scale down outputs during evaluation when not
+        # actually dropping out values (or, equivalently, scale up their un-dropped outputs in training).
+        if self.token_dropout:
+            embeddings = embeddings.masked_fill((input_ids == self.mask_token_id).unsqueeze(-1), 0.0)
+            mask_ratio_train = 0.15 * 0.8  # Hardcoded as the ratio used in all ESM model training runs
+            src_lengths = attention_mask.sum(-1)
+            mask_ratio_observed = (input_ids == self.mask_token_id).sum(-1).float() / src_lengths
+            embeddings = (embeddings * (1 - mask_ratio_train) / (1 - mask_ratio_observed)[:, None, None]).to(
+                embeddings.dtype
+            )
+
+        if self.position_embedding_type == "absolute":
+            position_embeddings = self.position_embeddings(position_ids)
+            embeddings = embeddings + position_embeddings
+
+        if self.layer_norm is not None:
+            embeddings = self.layer_norm(embeddings)
+        if attention_mask is not None:
+            embeddings = (embeddings * attention_mask.unsqueeze(-1)).to(embeddings.dtype)
+        # Matt: I think this line was copied incorrectly from BERT, disabling it for now.
+        # embeddings = self.dropout(embeddings)
+        return embeddings
+
+    def create_position_ids_from_inputs_embeds(self, inputs_embeds):
+        """
+        We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids.
+
+        Args:
+            inputs_embeds: torch.Tensor
+
+        Returns: torch.Tensor
+        """
+        input_shape = inputs_embeds.size()[:-1]
+        sequence_length = input_shape[1]
+
+        position_ids = torch.arange(
+            self.padding_idx + 1, sequence_length + self.padding_idx + 1, dtype=torch.long, device=inputs_embeds.device
+        )
+        return position_ids.unsqueeze(0).expand(input_shape)
+
+
+class EsmSelfAttention(nn.Module):
+    def __init__(self, config, position_embedding_type=None):
+        super().__init__()
+        if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"):
+            raise ValueError(
+                f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
+                f"heads ({config.num_attention_heads})"
+            )
+
+        self.num_attention_heads = config.num_attention_heads
+        self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
+        self.all_head_size = self.num_attention_heads * self.attention_head_size
+
+        self.query = nn.Linear(config.hidden_size, self.all_head_size)
+        self.key = nn.Linear(config.hidden_size, self.all_head_size)
+        self.value = nn.Linear(config.hidden_size, self.all_head_size)
+
+        self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
+        self.position_embedding_type = position_embedding_type or getattr(
+            config, "position_embedding_type", "absolute"
+        )
+        self.rotary_embeddings = None
+        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
+            self.max_position_embeddings = config.max_position_embeddings
+            self.distance_embedding = nn.Embedding(2 * config.max_position_embeddings - 1, self.attention_head_size)
+        elif self.position_embedding_type == "rotary":
+            self.rotary_embeddings = RotaryEmbedding(dim=self.attention_head_size)
+
+        self.is_decoder = config.is_decoder
+
+    def transpose_for_scores(self, x: torch.Tensor) -> torch.Tensor:
+        new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size)
+        x = x.view(new_x_shape)
+        return x.permute(0, 2, 1, 3)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
+        output_attentions: Optional[bool] = False,
+    ) -> Tuple[torch.Tensor]:
+        mixed_query_layer = self.query(hidden_states)
+
+        # If this is instantiated as a cross-attention module, the keys
+        # and values come from an encoder; the attention mask needs to be
+        # such that the encoder's padding tokens are not attended to.
+        is_cross_attention = encoder_hidden_states is not None
+
+        if is_cross_attention and past_key_value is not None:
+            # reuse k,v, cross_attentions
+            key_layer = past_key_value[0]
+            value_layer = past_key_value[1]
+            attention_mask = encoder_attention_mask
+        elif is_cross_attention:
+            key_layer = self.transpose_for_scores(self.key(encoder_hidden_states))
+            value_layer = self.transpose_for_scores(self.value(encoder_hidden_states))
+            attention_mask = encoder_attention_mask
+        elif past_key_value is not None:
+            key_layer = self.transpose_for_scores(self.key(hidden_states))
+            value_layer = self.transpose_for_scores(self.value(hidden_states))
+            key_layer = torch.cat([past_key_value[0], key_layer], dim=2)
+            value_layer = torch.cat([past_key_value[1], value_layer], dim=2)
+        else:
+            key_layer = self.transpose_for_scores(self.key(hidden_states))
+            value_layer = self.transpose_for_scores(self.value(hidden_states))
+
+        query_layer = self.transpose_for_scores(mixed_query_layer)
+
+        # Matt: Our BERT model (which this code was derived from) scales attention logits down by sqrt(head_dim).
+        # ESM scales the query down by the same factor instead. Modulo numerical stability these are equivalent,
+        # but not when rotary embeddings get involved. Therefore, we scale the query here to match the original
+        # ESM code and fix rotary embeddings.
+        query_layer = query_layer * self.attention_head_size**-0.5
+
+        if self.is_decoder:
+            # if cross_attention save Tuple(torch.Tensor, torch.Tensor) of all cross attention key/value_states.
+            # Further calls to cross_attention layer can then reuse all cross-attention
+            # key/value_states (first "if" case)
+            # if uni-directional self-attention (decoder) save Tuple(torch.Tensor, torch.Tensor) of
+            # all previous decoder key/value_states. Further calls to uni-directional self-attention
+            # can concat previous decoder key/value_states to current projected key/value_states (third "elif" case)
+            # if encoder bi-directional self-attention `past_key_value` is always `None`
+            past_key_value = (key_layer, value_layer)
+
+        if self.position_embedding_type == "rotary":
+            query_layer, key_layer = self.rotary_embeddings(query_layer, key_layer)
+
+        # Take the dot product between "query" and "key" to get the raw attention scores.
+        attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2))
+
+        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
+            seq_length = hidden_states.size()[1]
+            position_ids_l = torch.arange(seq_length, dtype=torch.long, device=hidden_states.device).view(-1, 1)
+            position_ids_r = torch.arange(seq_length, dtype=torch.long, device=hidden_states.device).view(1, -1)
+            distance = position_ids_l - position_ids_r
+            positional_embedding = self.distance_embedding(distance + self.max_position_embeddings - 1)
+            positional_embedding = positional_embedding.to(dtype=query_layer.dtype)  # fp16 compatibility
+
+            if self.position_embedding_type == "relative_key":
+                relative_position_scores = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
+                attention_scores = attention_scores + relative_position_scores
+            elif self.position_embedding_type == "relative_key_query":
+                relative_position_scores_query = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
+                relative_position_scores_key = torch.einsum("bhrd,lrd->bhlr", key_layer, positional_embedding)
+                attention_scores = attention_scores + relative_position_scores_query + relative_position_scores_key
+
+        if attention_mask is not None:
+            # Apply the attention mask is (precomputed for all layers in EsmModel forward() function)
+            attention_scores = attention_scores + attention_mask
+
+        # Normalize the attention scores to probabilities.
+        attention_probs = nn.functional.softmax(attention_scores, dim=-1)
+
+        # This is actually dropping out entire tokens to attend to, which might
+        # seem a bit unusual, but is taken from the original Transformer paper.
+        attention_probs = self.dropout(attention_probs)
+
+        # Mask heads if we want to
+        if head_mask is not None:
+            attention_probs = attention_probs * head_mask
+
+        context_layer = torch.matmul(attention_probs.to(value_layer.dtype), value_layer)
+
+        context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
+        new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
+        context_layer = context_layer.view(new_context_layer_shape)
+
+        outputs = (context_layer, attention_probs) if output_attentions else (context_layer,)
+
+        if self.is_decoder:
+            outputs = outputs + (past_key_value,)
+        return outputs
+
+
+class EsmSelfOutput(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+
+    def forward(self, hidden_states, input_tensor):
+        hidden_states = self.dense(hidden_states)
+        hidden_states = self.dropout(hidden_states)
+        hidden_states = hidden_states + input_tensor
+        return hidden_states
+
+
+class EsmAttention(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.self = EsmSelfAttention(config)
+        self.output = EsmSelfOutput(config)
+        self.pruned_heads = set()
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+
+    def prune_heads(self, heads):
+        if len(heads) == 0:
+            return
+        heads, index = find_pruneable_heads_and_indices(
+            heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads
+        )
+
+        # Prune linear layers
+        self.self.query = prune_linear_layer(self.self.query, index)
+        self.self.key = prune_linear_layer(self.self.key, index)
+        self.self.value = prune_linear_layer(self.self.value, index)
+        self.output.dense = prune_linear_layer(self.output.dense, index, dim=1)
+
+        # Update hyper params and store pruned heads
+        self.self.num_attention_heads = self.self.num_attention_heads - len(heads)
+        self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads
+        self.pruned_heads = self.pruned_heads.union(heads)
+
+    def forward(
+        self,
+        hidden_states,
+        attention_mask=None,
+        head_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_value=None,
+        output_attentions=False,
+    ):
+        hidden_states_ln = self.LayerNorm(hidden_states)
+        self_outputs = self.self(
+            hidden_states_ln,
+            attention_mask,
+            head_mask,
+            encoder_hidden_states,
+            encoder_attention_mask,
+            past_key_value,
+            output_attentions,
+        )
+        attention_output = self.output(self_outputs[0], hidden_states)
+        outputs = (attention_output,) + self_outputs[1:]  # add attentions if we output them
+        return outputs
+
+
+class EsmIntermediate(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.intermediate_size)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.dense(hidden_states)
+        hidden_states = gelu(hidden_states)
+        return hidden_states
+
+
+class EsmOutput(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.intermediate_size, config.hidden_size)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+
+    def forward(self, hidden_states, input_tensor):
+        hidden_states = self.dense(hidden_states)
+        hidden_states = self.dropout(hidden_states)
+        hidden_states = hidden_states + input_tensor
+        return hidden_states
+
+
+class EsmLayer(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.chunk_size_feed_forward = config.chunk_size_feed_forward
+        self.seq_len_dim = 1
+        self.attention = EsmAttention(config)
+        self.is_decoder = config.is_decoder
+        self.add_cross_attention = config.add_cross_attention
+        if self.add_cross_attention:
+            if not self.is_decoder:
+                raise RuntimeError(f"{self} should be used as a decoder model if cross attention is added")
+            self.crossattention = EsmAttention(config)
+        self.intermediate = EsmIntermediate(config)
+        self.output = EsmOutput(config)
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+
+    def forward(
+        self,
+        hidden_states,
+        attention_mask=None,
+        head_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_value=None,
+        output_attentions=False,
+    ):
+        # decoder uni-directional self-attention cached key/values tuple is at positions 1,2
+        self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None
+        self_attention_outputs = self.attention(
+            hidden_states,
+            attention_mask,
+            head_mask,
+            output_attentions=output_attentions,
+            past_key_value=self_attn_past_key_value,
+        )
+        attention_output = self_attention_outputs[0]
+
+        # if decoder, the last output is tuple of self-attn cache
+        if self.is_decoder:
+            outputs = self_attention_outputs[1:-1]
+            present_key_value = self_attention_outputs[-1]
+        else:
+            outputs = self_attention_outputs[1:]  # add self attentions if we output attention weights
+
+        cross_attn_present_key_value = None
+        if self.is_decoder and encoder_hidden_states is not None:
+            if not hasattr(self, "crossattention"):
+                raise AttributeError(
+                    f"If `encoder_hidden_states` are passed, {self} has to be instantiated"
+                    " with cross-attention layers by setting `config.add_cross_attention=True`"
+                )
+
+            # cross_attn cached key/values tuple is at positions 3,4 of past_key_value tuple
+            cross_attn_past_key_value = past_key_value[-2:] if past_key_value is not None else None
+            cross_attention_outputs = self.crossattention(
+                attention_output,
+                attention_mask,
+                head_mask,
+                encoder_hidden_states,
+                encoder_attention_mask,
+                cross_attn_past_key_value,
+                output_attentions,
+            )
+            attention_output = cross_attention_outputs[0]
+            outputs = outputs + cross_attention_outputs[1:-1]  # add cross attentions if we output attention weights
+
+            # add cross-attn cache to positions 3,4 of present_key_value tuple
+            cross_attn_present_key_value = cross_attention_outputs[-1]
+            present_key_value = present_key_value + cross_attn_present_key_value
+
+        layer_output = self.feed_forward_chunk(attention_output)
+
+        outputs = (layer_output,) + outputs
+
+        # if decoder, return the attn key/values as the last output
+        if self.is_decoder:
+            outputs = outputs + (present_key_value,)
+        return outputs
+
+    def feed_forward_chunk(self, attention_output):
+        attention_output_ln = self.LayerNorm(attention_output)
+        intermediate_output = self.intermediate(attention_output_ln)
+        layer_output = self.output(intermediate_output, attention_output)
+        return layer_output
+
+
+class EsmEncoder(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.layer = nn.ModuleList([EsmLayer(config) for _ in range(config.num_hidden_layers)])
+        self.emb_layer_norm_after = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.gradient_checkpointing = False
+
+    def forward(
+        self,
+        hidden_states,
+        attention_mask=None,
+        head_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_values=None,
+        use_cache=None,
+        output_attentions=False,
+        output_hidden_states=False,
+        return_dict=True,
+    ):
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with `config.gradient_checkpointing=True`. Setting "
+                    "`use_cache=False`..."
+                )
+                use_cache = False
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attentions = () if output_attentions else None
+        all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None
+
+        next_decoder_cache = () if use_cache else None
+        for i, layer_module in enumerate(self.layer):
+            if output_hidden_states:
+                all_hidden_states = all_hidden_states + (hidden_states,)
+
+            layer_head_mask = head_mask[i] if head_mask is not None else None
+            past_key_value = past_key_values[i] if past_key_values is not None else None
+
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = self._gradient_checkpointing_func(
+                    layer_module.__call__,
+                    hidden_states,
+                    attention_mask,
+                    layer_head_mask,
+                    encoder_hidden_states,
+                    encoder_attention_mask,
+                    past_key_value,
+                    output_attentions,
+                )
+            else:
+                layer_outputs = layer_module(
+                    hidden_states,
+                    attention_mask,
+                    layer_head_mask,
+                    encoder_hidden_states,
+                    encoder_attention_mask,
+                    past_key_value,
+                    output_attentions,
+                )
+
+            hidden_states = layer_outputs[0]
+            if use_cache:
+                next_decoder_cache = next_decoder_cache + (layer_outputs[-1],)
+            if output_attentions:
+                all_self_attentions = all_self_attentions + (layer_outputs[1],)
+                if self.config.add_cross_attention:
+                    all_cross_attentions = all_cross_attentions + (layer_outputs[2],)
+
+        if self.emb_layer_norm_after:
+            hidden_states = self.emb_layer_norm_after(hidden_states)
+
+        if output_hidden_states:
+            all_hidden_states = all_hidden_states + (hidden_states,)
+
+        if not return_dict:
+            return tuple(
+                v
+                for v in [
+                    hidden_states,
+                    next_decoder_cache,
+                    all_hidden_states,
+                    all_self_attentions,
+                    all_cross_attentions,
+                ]
+                if v is not None
+            )
+        return BaseModelOutputWithPastAndCrossAttentions(
+            last_hidden_state=hidden_states,
+            past_key_values=next_decoder_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attentions,
+            cross_attentions=all_cross_attentions,
+        )
+
+
+# Copied from transformers.models.bert.modeling_bert.BertPooler
+class EsmPooler(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        self.activation = nn.Tanh()
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        # We "pool" the model by simply taking the hidden state corresponding
+        # to the first token.
+        first_token_tensor = hidden_states[:, 0]
+        pooled_output = self.dense(first_token_tensor)
+        pooled_output = self.activation(pooled_output)
+        return pooled_output
+
+
+class EsmPreTrainedModel(PreTrainedModel):
+    """
+    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
+    models.
+    """
+
+    config_class = EsmConfig
+    base_model_prefix = "esm"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["EsmLayer", "EsmFoldTriangularSelfAttentionBlock", "EsmEmbeddings"]
+
+    # Copied from transformers.models.bert.modeling_bert.BertPreTrainedModel._init_weights
+    def _init_weights(self, module):
+        """Initialize the weights"""
+        if isinstance(module, nn.Linear):
+            # Slightly different from the TF version which uses truncated_normal for initialization
+            # cf https://github.com/pytorch/pytorch/pull/5617
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+        elif isinstance(module, nn.LayerNorm):
+            module.bias.data.zero_()
+            module.weight.data.fill_(1.0)
+
+
+ESM_START_DOCSTRING = r"""
+
+    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+
+    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
+    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
+    and behavior.
+
+    Parameters:
+        config ([`EsmConfig`]): Model configuration class with all the parameters of the
+            model. Initializing with a config file does not load the weights associated with the model, only the
+            configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+
+ESM_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`torch.LongTensor` of shape `({0})`):
+            Indices of input sequence tokens in the vocabulary.
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`torch.FloatTensor` of shape `({0})`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            [What are attention masks?](../glossary#attention-mask)
+        position_ids (`torch.LongTensor` of shape `({0})`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.max_position_embeddings - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        head_mask (`torch.FloatTensor` of shape `(num_heads,)` or `(num_layers, num_heads)`, *optional*):
+            Mask to nullify selected heads of the self-attention modules. Mask values selected in `[0, 1]`:
+
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+
+        inputs_embeds (`torch.FloatTensor` of shape `({0}, hidden_size)`, *optional*):
+            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
+            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
+            model's internal embedding lookup matrix.
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~file_utils.ModelOutput`] instead of a plain tuple.
+"""
+
+
+@add_start_docstrings(
+    "The bare ESM Model transformer outputting raw hidden-states without any specific head on top.",
+    ESM_START_DOCSTRING,
+)
+class EsmModel(EsmPreTrainedModel):
+    """
+
+    The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of
+    cross-attention is added between the self-attention layers, following the architecture described in [Attention is
+    all you need](https://arxiv.org/abs/1706.03762) by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit,
+    Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin.
+
+    To behave as an decoder the model needs to be initialized with the `is_decoder` argument of the configuration set
+    to `True`. To be used in a Seq2Seq model, the model needs to initialized with both `is_decoder` argument and
+    `add_cross_attention` set to `True`; an `encoder_hidden_states` is then expected as an input to the forward pass.
+    """
+
+    def __init__(self, config, add_pooling_layer=True):
+        super().__init__(config)
+        self.config = config
+
+        self.embeddings = EsmEmbeddings(config)
+        self.encoder = EsmEncoder(config)
+
+        self.pooler = EsmPooler(config) if add_pooling_layer else None
+
+        self.contact_head = EsmContactPredictionHead(
+            in_features=config.num_hidden_layers * config.num_attention_heads, bias=True
+        )
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.embeddings.word_embeddings
+
+    def set_input_embeddings(self, value):
+        self.embeddings.word_embeddings = value
+
+    def _prune_heads(self, heads_to_prune):
+        """
+        Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base
+        class PreTrainedModel
+        """
+        for layer, heads in heads_to_prune.items():
+            self.encoder.layer[layer].attention.prune_heads(heads)
+
+    @add_start_docstrings_to_model_forward(ESM_INPUTS_DOCSTRING.format("(batch_size, sequence_length)"))
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=BaseModelOutputWithPoolingAndCrossAttentions,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]:
+        r"""
+        encoder_hidden_states  (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if
+            the model is configured as a decoder.
+        encoder_attention_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in
+            the cross-attention if the model is configured as a decoder. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+        past_key_values (`tuple(tuple(torch.FloatTensor))` of length `config.n_layers` with each tuple having 4 tensors of shape `(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
+            Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding.
+
+            If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
+            don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
+            `decoder_input_ids` of shape `(batch_size, sequence_length)`.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+        """
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if self.config.is_decoder:
+            use_cache = use_cache if use_cache is not None else self.config.use_cache
+        else:
+            use_cache = False
+
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
+        elif input_ids is not None:
+            self.warn_if_padding_and_no_attention_mask(input_ids, attention_mask)
+            input_shape = input_ids.size()
+        elif inputs_embeds is not None:
+            input_shape = inputs_embeds.size()[:-1]
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        batch_size, seq_length = input_shape
+        device = input_ids.device if input_ids is not None else inputs_embeds.device
+
+        # past_key_values_length
+        past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0
+
+        if attention_mask is None:
+            attention_mask = torch.ones(((batch_size, seq_length + past_key_values_length)), device=device)
+
+        # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
+        # ourselves in which case we just need to make it broadcastable to all heads.
+        extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, input_shape)
+
+        # If a 2D or 3D attention mask is provided for the cross-attention
+        # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
+        if self.config.is_decoder and encoder_hidden_states is not None:
+            encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
+            encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
+            if encoder_attention_mask is None:
+                encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
+            encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask)
+        else:
+            encoder_extended_attention_mask = None
+
+        # Prepare head mask if needed
+        # 1.0 in head_mask indicate we keep the head
+        # attention_probs has shape bsz x n_heads x N x N
+        # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
+        # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
+        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
+
+        embedding_output = self.embeddings(
+            input_ids=input_ids,
+            position_ids=position_ids,
+            attention_mask=attention_mask,
+            inputs_embeds=inputs_embeds,
+            past_key_values_length=past_key_values_length,
+        )
+        encoder_outputs = self.encoder(
+            embedding_output,
+            attention_mask=extended_attention_mask,
+            head_mask=head_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_extended_attention_mask,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        sequence_output = encoder_outputs[0]
+        pooled_output = self.pooler(sequence_output) if self.pooler is not None else None
+
+        if not return_dict:
+            return (sequence_output, pooled_output) + encoder_outputs[1:]
+
+        return BaseModelOutputWithPoolingAndCrossAttentions(
+            last_hidden_state=sequence_output,
+            pooler_output=pooled_output,
+            past_key_values=encoder_outputs.past_key_values,
+            hidden_states=encoder_outputs.hidden_states,
+            attentions=encoder_outputs.attentions,
+            cross_attentions=encoder_outputs.cross_attentions,
+        )
+
+    def predict_contacts(self, tokens, attention_mask):
+        attns = self(tokens, attention_mask=attention_mask, return_dict=True, output_attentions=True).attentions
+        attns = torch.stack(attns, dim=1)  # Matches the original model layout
+        # In the original model, attentions for padding tokens are completely zeroed out.
+        # This makes no difference most of the time because the other tokens won't attend to them,
+        # but it does for the contact prediction task, which takes attentions as input,
+        # so we have to mimic that here.
+        attns *= attention_mask.unsqueeze(1).unsqueeze(2).unsqueeze(3)
+        attns *= attention_mask.unsqueeze(1).unsqueeze(2).unsqueeze(4)
+        return self.contact_head(tokens, attns)
+
+
+@add_start_docstrings("""ESM Model with a `language modeling` head on top.""", ESM_START_DOCSTRING)
+class EsmForMaskedLM(EsmPreTrainedModel):
+    _tied_weights_keys = ["lm_head.decoder.weight"]
+
+    def __init__(self, config):
+        super().__init__(config)
+
+        if config.is_decoder:
+            logger.warning(
+                "If you want to use `EsmForMaskedLM` make sure `config.is_decoder=False` for "
+                "bi-directional self-attention."
+            )
+
+        self.esm = EsmModel(config, add_pooling_layer=False)
+        self.lm_head = EsmLMHead(config)
+
+        self.init_weights()
+
+    def get_output_embeddings(self):
+        return self.lm_head.decoder
+
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head.decoder = new_embeddings
+
+    @add_start_docstrings_to_model_forward(ESM_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=MaskedLMOutput,
+        config_class=_CONFIG_FOR_DOC,
+        mask="<mask>",
+    )
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, MaskedLMOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
+            config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
+            loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
+        kwargs (`Dict[str, any]`, optional, defaults to *{}*):
+            Used to hide legacy arguments that have been deprecated.
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.esm(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        sequence_output = outputs[0]
+        prediction_scores = self.lm_head(sequence_output)
+
+        masked_lm_loss = None
+        if labels is not None:
+            loss_fct = CrossEntropyLoss()
+
+            labels = labels.to(prediction_scores.device)
+            masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))
+
+        if not return_dict:
+            output = (prediction_scores,) + outputs[2:]
+            return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output
+
+        return MaskedLMOutput(
+            loss=masked_lm_loss,
+            logits=prediction_scores,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    def predict_contacts(self, tokens, attention_mask):
+        return self.esm.predict_contacts(tokens, attention_mask=attention_mask)
+
+
+class EsmLMHead(nn.Module):
+    """ESM Head for masked language modeling."""
+
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        self.layer_norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+
+        self.decoder = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+        self.bias = nn.Parameter(torch.zeros(config.vocab_size))
+
+    def forward(self, features, **kwargs):
+        x = self.dense(features)
+        x = gelu(x)
+        x = self.layer_norm(x)
+
+        # project back to size of vocabulary with bias
+        x = self.decoder(x) + self.bias
+        return x
+
+
+@add_start_docstrings(
+    """
+    ESM Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled
+    output) e.g. for GLUE tasks.
+    """,
+    ESM_START_DOCSTRING,
+)
+class EsmForSequenceClassification(EsmPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.config = config
+
+        self.esm = EsmModel(config, add_pooling_layer=False)
+        self.classifier = EsmClassificationHead(config)
+
+        self.init_weights()
+
+    @add_start_docstrings_to_model_forward(ESM_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=SequenceClassifierOutput,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, SequenceClassifierOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.esm(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        sequence_output = outputs[0]
+        logits = self.classifier(sequence_output)
+
+        loss = None
+        if labels is not None:
+            labels = labels.to(logits.device)
+
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = "regression"
+                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
+                    self.config.problem_type = "single_label_classification"
+                else:
+                    self.config.problem_type = "multi_label_classification"
+
+            if self.config.problem_type == "regression":
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(logits, labels)
+            elif self.config.problem_type == "single_label_classification":
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+            elif self.config.problem_type == "multi_label_classification":
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(logits, labels)
+
+        if not return_dict:
+            output = (logits,) + outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+@add_start_docstrings(
+    """
+    ESM Model with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for
+    Named-Entity-Recognition (NER) tasks.
+    """,
+    ESM_START_DOCSTRING,
+)
+class EsmForTokenClassification(EsmPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.esm = EsmModel(config, add_pooling_layer=False)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+
+        self.init_weights()
+
+    @add_start_docstrings_to_model_forward(ESM_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=TokenClassifierOutput,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, TokenClassifierOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.esm(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        sequence_output = outputs[0]
+
+        sequence_output = self.dropout(sequence_output)
+        logits = self.classifier(sequence_output)
+
+        loss = None
+        if labels is not None:
+            loss_fct = CrossEntropyLoss()
+
+            labels = labels.to(logits.device)
+            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+
+        if not return_dict:
+            output = (logits,) + outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return TokenClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+class EsmClassificationHead(nn.Module):
+    """Head for sentence-level classification tasks."""
+
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+        self.out_proj = nn.Linear(config.hidden_size, config.num_labels)
+
+    def forward(self, features, **kwargs):
+        x = features[:, 0, :]  # take <s> token (equiv. to [CLS])
+        x = self.dropout(x)
+        x = self.dense(x)
+        x = torch.tanh(x)
+        x = self.dropout(x)
+        x = self.out_proj(x)
+        return x
+
+
+def create_position_ids_from_input_ids(input_ids, padding_idx, past_key_values_length=0):
+    """
+    Replace non-padding symbols with their position numbers. Position numbers begin at padding_idx+1. Padding symbols
+    are ignored. This is modified from fairseq's `utils.make_positions`.
+
+    Args:
+        x: torch.Tensor x:
+
+    Returns: torch.Tensor
+    """
+    # The series of casts and type-conversions here are carefully balanced to both work with ONNX export and XLA.
+    mask = input_ids.ne(padding_idx).int()
+    incremental_indices = (torch.cumsum(mask, dim=1).type_as(mask) + past_key_values_length) * mask
+    return incremental_indices.long() + padding_idx

venv/lib/python3.10/site-packages/transformers/models/esm/modeling_esmfold.py

@@ -0,0 +1,2322 @@
+# coding=utf-8
+# Copyright 2022 Meta and The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import math
+import sys
+from dataclasses import dataclass
+from functools import partial
+from typing import Callable, Dict, List, Optional, Sequence, Tuple, Union
+
+import numpy as np
+import torch
+import torch.nn as nn
+from torch.nn import LayerNorm
+
+from ...integrations.deepspeed import is_deepspeed_available
+from ...modeling_outputs import ModelOutput
+from ...utils import (
+    ContextManagers,
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    is_scipy_available,
+    logging,
+    replace_return_docstrings,
+)
+from .configuration_esm import EsmConfig
+from .modeling_esm import ESM_START_DOCSTRING, EsmModel, EsmPreTrainedModel
+from .openfold_utils import (
+    OFProtein,
+    Rigid,
+    Rotation,
+    atom14_to_atom37,
+    chunk_layer,
+    compute_predicted_aligned_error,
+    compute_tm,
+    frames_and_literature_positions_to_atom14_pos,
+    make_atom14_masks,
+    residue_constants,
+    to_pdb,
+    torsion_angles_to_frames,
+)
+
+
+logger = logging.get_logger(__name__)
+_CHECKPOINT_FOR_DOC = "facebook/esmfold_v1"
+_CONFIG_FOR_DOC = "EsmConfig"
+
+
+@dataclass
+class EsmForProteinFoldingOutput(ModelOutput):
+    """
+    Output type of [`EsmForProteinFoldingOutput`].
+
+    Args:
+        frames (`torch.FloatTensor`):
+            Output frames.
+        sidechain_frames (`torch.FloatTensor`):
+            Output sidechain frames.
+        unnormalized_angles (`torch.FloatTensor`):
+            Predicted unnormalized backbone and side chain torsion angles.
+        angles (`torch.FloatTensor`):
+            Predicted backbone and side chain torsion angles.
+        positions (`torch.FloatTensor`):
+            Predicted positions of the backbone and side chain atoms.
+        states (`torch.FloatTensor`):
+            Hidden states from the protein folding trunk.
+        s_s (`torch.FloatTensor`):
+            Per-residue embeddings derived by concatenating the hidden states of each layer of the ESM-2 LM stem.
+        s_z (`torch.FloatTensor`):
+            Pairwise residue embeddings.
+        distogram_logits (`torch.FloatTensor`):
+            Input logits to the distogram used to compute residue distances.
+        lm_logits (`torch.FloatTensor`):
+            Logits output by the ESM-2 protein language model stem.
+        aatype (`torch.FloatTensor`):
+            Input amino acids (AlphaFold2 indices).
+        atom14_atom_exists (`torch.FloatTensor`):
+            Whether each atom exists in the atom14 representation.
+        residx_atom14_to_atom37 (`torch.FloatTensor`):
+            Mapping between atoms in the atom14 and atom37 representations.
+        residx_atom37_to_atom14 (`torch.FloatTensor`):
+            Mapping between atoms in the atom37 and atom14 representations.
+        atom37_atom_exists (`torch.FloatTensor`):
+            Whether each atom exists in the atom37 representation.
+        residue_index (`torch.FloatTensor`):
+            The index of each residue in the protein chain. Unless internal padding tokens are used, this will just be
+            a sequence of integers from 0 to `sequence_length`.
+        lddt_head (`torch.FloatTensor`):
+            Raw outputs from the lddt head used to compute plddt.
+        plddt (`torch.FloatTensor`):
+            Per-residue confidence scores. Regions of low confidence may indicate areas where the model's prediction is
+            uncertain, or where the protein structure is disordered.
+        ptm_logits (`torch.FloatTensor`):
+            Raw logits used for computing ptm.
+        ptm (`torch.FloatTensor`):
+            TM-score output representing the model's high-level confidence in the overall structure.
+        aligned_confidence_probs (`torch.FloatTensor`):
+            Per-residue confidence scores for the aligned structure.
+        predicted_aligned_error (`torch.FloatTensor`):
+            Predicted error between the model's prediction and the ground truth.
+        max_predicted_aligned_error (`torch.FloatTensor`):
+            Per-sample maximum predicted error.
+    """
+
+    frames: torch.FloatTensor = None
+    sidechain_frames: torch.FloatTensor = None
+    unnormalized_angles: torch.FloatTensor = None
+    angles: torch.FloatTensor = None
+    positions: torch.FloatTensor = None
+    states: torch.FloatTensor = None
+    s_s: torch.FloatTensor = None
+    s_z: torch.FloatTensor = None
+    distogram_logits: torch.FloatTensor = None
+    lm_logits: torch.FloatTensor = None
+    aatype: torch.FloatTensor = None
+    atom14_atom_exists: torch.FloatTensor = None
+    residx_atom14_to_atom37: torch.FloatTensor = None
+    residx_atom37_to_atom14: torch.FloatTensor = None
+    atom37_atom_exists: torch.FloatTensor = None
+    residue_index: torch.FloatTensor = None
+    lddt_head: torch.FloatTensor = None
+    plddt: torch.FloatTensor = None
+    ptm_logits: torch.FloatTensor = None
+    ptm: torch.FloatTensor = None
+    aligned_confidence_probs: torch.FloatTensor = None
+    predicted_aligned_error: torch.FloatTensor = None
+    max_predicted_aligned_error: torch.FloatTensor = None
+
+
+ESMFOLD_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`torch.LongTensor` of shape `({0})`):
+            Indices of input sequence tokens in the vocabulary.
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`torch.FloatTensor` of shape `({0})`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            [What are attention masks?](../glossary#attention-mask)
+        position_ids (`torch.LongTensor` of shape `({0})`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.max_position_embeddings - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        masking_pattern (`torch.LongTensor` of shape `({0})`, *optional*):
+            Locations of tokens to mask during training as a form of regularization. Mask values selected in `[0, 1]`.
+        num_recycles (`int`, *optional*, defaults to `None`):
+            Number of times to recycle the input sequence. If `None`, defaults to `config.num_recycles`. "Recycling"
+            consists of passing the output of the folding trunk back in as input to the trunk. During training, the
+            number of recycles should vary with each batch, to ensure that the model learns to output valid predictions
+            after each recycle. During inference, num_recycles should be set to the highest value that the model was
+            trained with for maximum accuracy. Accordingly, when this value is set to `None`, config.max_recycles is
+            used.
+"""
+
+
+def is_fp16_enabled():
+    # Autocast world
+    fp16_enabled = torch.get_autocast_gpu_dtype() == torch.float16
+    fp16_enabled = fp16_enabled and torch.is_autocast_enabled()
+
+    return fp16_enabled
+
+
+def is_deepspeed_initialized():
+    if is_deepspeed_available():
+        return False
+    else:
+        try:
+            import deepspeed
+
+            # This is not available in all DeepSpeed versions.
+            return deepspeed.utils.is_initialized()
+        except Exception:
+            return False
+
+
+def collate_dense_tensors(samples: List[torch.Tensor], pad_v: float = 0) -> torch.Tensor:
+    """
+    Takes a list of tensors with the following dimensions:
+        [(d_11, ..., d_1K),
+         (d_21, ..., d_2K), ..., (d_N1, ..., d_NK)]
+    and stack + pads them into a single tensor of:
+    (N, max_i=1,N { d_i1 }, ..., max_i=1,N {diK})
+    """
+    if len(samples) == 0:
+        return torch.Tensor()
+    if len({x.dim() for x in samples}) != 1:
+        raise RuntimeError(f"Samples has varying dimensions: {[x.dim() for x in samples]}")
+    (device,) = tuple({x.device for x in samples})  # assumes all on same device
+    max_shape = [max(lst) for lst in zip(*[x.shape for x in samples])]
+    result = torch.empty(len(samples), *max_shape, dtype=samples[0].dtype, device=device)
+    result.fill_(pad_v)
+    for i in range(len(samples)):
+        result_i = result[i]
+        t = samples[i]
+        result_i[tuple(slice(0, k) for k in t.shape)] = t
+    return result
+
+
+def flatten_final_dims(t: torch.Tensor, no_dims: int):
+    return t.reshape(t.shape[:-no_dims] + (-1,))
+
+
+def permute_final_dims(tensor: torch.Tensor, inds: List[int]):
+    zero_index = -1 * len(inds)
+    first_inds = list(range(len(tensor.shape[:zero_index])))
+    return tensor.permute(first_inds + [zero_index + i for i in inds])
+
+
+def dict_multimap(fn, dicts):
+    first = dicts[0]
+    new_dict = {}
+    for k, v in first.items():
+        all_v = [d[k] for d in dicts]
+        if isinstance(v, dict):
+            new_dict[k] = dict_multimap(fn, all_v)
+        else:
+            new_dict[k] = fn(all_v)
+
+    return new_dict
+
+
+def trunc_normal_init_(weights, scale=1.0, fan="fan_in"):
+    shape = weights.shape
+    scale = scale / max(1, shape[1])
+
+    if not is_scipy_available():
+        logger.warning(
+            "This init requires scipy, but scipy was not found, default to an approximation that might not be"
+            " equivalent."
+        )
+        std = math.sqrt(scale)
+        torch.nn.init.normal_(weights, std=std).clamp(min=0.0, max=2.0 * std)
+
+    else:
+        from scipy.stats import truncnorm
+
+        std = math.sqrt(scale) / truncnorm.std(a=-2, b=2, loc=0, scale=1)
+        samples = truncnorm.rvs(a=-2, b=2, loc=0, scale=std, size=weights.numel())
+        samples = np.reshape(samples, shape)
+        weights.copy_(torch.tensor(samples, device=weights.device))
+
+
+def ipa_point_weights_init_(weights):
+    with torch.no_grad():
+        softplus_inverse_1 = 0.541324854612918
+        weights.fill_(softplus_inverse_1)
+
+
+class EsmFoldLinear(nn.Linear):
+    """
+    A Linear layer with built-in nonstandard initializations. Called just like torch.nn.Linear.
+
+    Implements the initializers in 1.11.4, plus some additional ones found in the code.
+    """
+
+    def __init__(
+        self,
+        in_dim: int,
+        out_dim: int,
+        bias: bool = True,
+        init: str = "default",
+        init_fn: Optional[Callable[[torch.Tensor, torch.Tensor], None]] = None,
+    ):
+        """
+        Args:
+            in_dim:
+                The final dimension of inputs to the layer
+            out_dim:
+                The final dimension of layer outputs
+            bias:
+                Whether to learn an additive bias. True by default
+            init:
+                The initializer to use. Choose from:
+
+                "default": LeCun fan-in truncated normal initialization "relu": He initialization w/ truncated normal
+                distribution "glorot": Fan-average Glorot uniform initialization "gating": Weights=0, Bias=1 "normal":
+                Normal initialization with std=1/sqrt(fan_in) "final": Weights=0, Bias=0
+
+                Overridden by init_fn if the latter is not None.
+            init_fn:
+                A custom initializer taking weight and bias as inputs. Overrides init if not None.
+        """
+        super().__init__(in_dim, out_dim, bias=bias)
+
+        if bias:
+            with torch.no_grad():
+                self.bias.fill_(0)
+        self.init = init
+        self.init_fn = init_fn
+
+        if init not in ["default", "relu", "glorot", "gating", "normal", "final"]:
+            raise ValueError("Invalid init string.")
+
+
+class EsmFoldLayerNorm(nn.Module):
+    def __init__(self, c_in, eps=1e-5):
+        super().__init__()
+
+        self.c_in = (c_in,)
+        self.eps = eps
+
+        self.weight = nn.Parameter(torch.ones(c_in))
+        self.bias = nn.Parameter(torch.zeros(c_in))
+
+    def forward(self, x):
+        d = x.dtype
+        if d is torch.bfloat16 and not is_deepspeed_initialized():
+            with torch.cuda.amp.autocast(enabled=False):
+                out = nn.functional.layer_norm(x, self.c_in, self.weight.to(dtype=d), self.bias.to(dtype=d), self.eps)
+        else:
+            out = nn.functional.layer_norm(x, self.c_in, self.weight, self.bias, self.eps)
+
+        return out
+
+
+@torch.jit.ignore
+def softmax_no_cast(t: torch.Tensor, dim: int = -1) -> torch.Tensor:
+    """
+    Softmax, but without automatic casting to fp32 when the input is of type bfloat16
+    """
+    d = t.dtype
+    if d is torch.bfloat16 and not is_deepspeed_initialized():
+        with torch.cuda.amp.autocast(enabled=False):
+            s = torch.nn.functional.softmax(t, dim=dim)
+    else:
+        s = torch.nn.functional.softmax(t, dim=dim)
+
+    return s
+
+
+class EsmFoldAttention(nn.Module):
+    """
+    Standard multi-head attention using AlphaFold's default layer initialization. Allows multiple bias vectors.
+    """
+
+    def __init__(
+        self,
+        c_q: int,
+        c_k: int,
+        c_v: int,
+        c_hidden: int,
+        no_heads: int,
+        gating: bool = True,
+    ):
+        """
+        Args:
+            c_q:
+                Input dimension of query data
+            c_k:
+                Input dimension of key data
+            c_v:
+                Input dimension of value data
+            c_hidden:
+                Per-head hidden dimension
+            no_heads:
+                Number of attention heads
+            gating:
+                Whether the output should be gated using query data
+        """
+        super().__init__()
+
+        self.c_q = c_q
+        self.c_k = c_k
+        self.c_v = c_v
+        self.c_hidden = c_hidden
+        self.no_heads = no_heads
+        self.gating = gating
+
+        # DISCREPANCY: c_hidden is not the per-head channel dimension, as
+        # stated in the supplement, but the overall channel dimension.
+
+        self.linear_q = EsmFoldLinear(self.c_q, self.c_hidden * self.no_heads, bias=False, init="glorot")
+        self.linear_k = EsmFoldLinear(self.c_k, self.c_hidden * self.no_heads, bias=False, init="glorot")
+        self.linear_v = EsmFoldLinear(self.c_v, self.c_hidden * self.no_heads, bias=False, init="glorot")
+        self.linear_o = EsmFoldLinear(self.c_hidden * self.no_heads, self.c_q, init="final")
+
+        self.linear_g = None
+        if self.gating:
+            self.linear_g = EsmFoldLinear(self.c_q, self.c_hidden * self.no_heads, init="gating")
+
+        self.sigmoid = nn.Sigmoid()
+
+    def _prep_qkv(self, q_x: torch.Tensor, kv_x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
+        # [*, Q/K/V, H * C_hidden]
+        q = self.linear_q(q_x)
+        k = self.linear_k(kv_x)
+        v = self.linear_v(kv_x)
+
+        # [*, Q/K, H, C_hidden]
+        q = q.view(q.shape[:-1] + (self.no_heads, -1))
+        k = k.view(k.shape[:-1] + (self.no_heads, -1))
+        v = v.view(v.shape[:-1] + (self.no_heads, -1))
+
+        # [*, H, Q/K, C_hidden]
+        q = q.transpose(-2, -3)
+        k = k.transpose(-2, -3)
+        v = v.transpose(-2, -3)
+
+        q /= math.sqrt(self.c_hidden)
+
+        return q, k, v
+
+    def _wrap_up(self, o: torch.Tensor, q_x: torch.Tensor) -> torch.Tensor:
+        if self.linear_g is not None:
+            g = self.sigmoid(self.linear_g(q_x))
+
+            # [*, Q, H, C_hidden]
+            g = g.view(g.shape[:-1] + (self.no_heads, -1))
+            o = o * g
+
+        # [*, Q, H * C_hidden]
+        o = flatten_final_dims(o, 2)
+
+        # [*, Q, C_q]
+        o = self.linear_o(o)
+
+        return o
+
+    def forward(
+        self,
+        q_x: torch.Tensor,
+        kv_x: torch.Tensor,
+        biases: Optional[List[torch.Tensor]] = None,
+        use_memory_efficient_kernel: bool = False,
+        use_lma: bool = False,
+        lma_q_chunk_size: int = 1024,
+        lma_kv_chunk_size: int = 4096,
+        use_flash: bool = False,
+        flash_mask: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        """
+        Args:
+            q_x:
+                [*, Q, C_q] query data
+            kv_x:
+                [*, K, C_k] key data
+            biases:
+                List of biases that broadcast to [*, H, Q, K]
+            use_memory_efficient_kernel:
+                Whether to use a custom memory-efficient attention kernel. This should be the default choice for most.
+                If none of the "use_<...>" flags are True, a stock PyTorch implementation is used instead
+            use_lma:
+                Whether to use low-memory attention (Staats & Rabe 2021). If none of the "use_<...>" flags are True, a
+                stock PyTorch implementation is used instead
+            lma_q_chunk_size:
+                Query chunk size (for LMA)
+            lma_kv_chunk_size:
+                Key/Value chunk size (for LMA)
+        Returns
+            [*, Q, C_q] attention update
+        """
+        if use_lma and (lma_q_chunk_size is None or lma_kv_chunk_size is None):
+            raise ValueError("If use_lma is specified, lma_q_chunk_size and lma_kv_chunk_size must be provided")
+
+        if use_flash and biases is not None:
+            raise ValueError("use_flash is incompatible with the bias option. For masking, use flash_mask instead")
+
+        attn_options = [use_memory_efficient_kernel, use_lma, use_flash]
+        if sum(attn_options) > 1:
+            raise ValueError("Choose at most one alternative attention algorithm")
+
+        if biases is None:
+            biases = []
+
+        # [*, H, Q/K, C_hidden]
+        query, key, value = self._prep_qkv(q_x, kv_x)
+        key = permute_final_dims(key, (1, 0))
+
+        # [*, H, Q, K]
+        output = torch.matmul(query, key)
+        for b in biases:
+            output += b
+        output = softmax_no_cast(output, -1)
+
+        # [*, H, Q, C_hidden]
+        output = torch.matmul(output, value)
+        output = output.transpose(-2, -3)
+        output = self._wrap_up(output, q_x)
+
+        return output
+
+
+class EsmFoldTriangleAttention(nn.Module):
+    def __init__(self, c_in, c_hidden, no_heads, starting=True, inf=1e9):
+        """
+        Args:
+            c_in:
+                Input channel dimension
+            c_hidden:
+                Overall hidden channel dimension (not per-head)
+            no_heads:
+                Number of attention heads
+        """
+        super().__init__()
+
+        self.c_in = c_in
+        self.c_hidden = c_hidden
+        self.no_heads = no_heads
+        self.starting = starting
+        self.inf = inf
+
+        self.layer_norm = LayerNorm(self.c_in)
+
+        self.linear = EsmFoldLinear(c_in, self.no_heads, bias=False, init="normal")
+
+        self.mha = EsmFoldAttention(self.c_in, self.c_in, self.c_in, self.c_hidden, self.no_heads)
+
+    @torch.jit.ignore
+    def _chunk(
+        self,
+        x: torch.Tensor,
+        biases: List[torch.Tensor],
+        chunk_size: int,
+        use_memory_efficient_kernel: bool = False,
+        use_lma: bool = False,
+        inplace_safe: bool = False,
+    ) -> torch.Tensor:
+        "triangle! triangle!"
+        mha_inputs = {
+            "q_x": x,
+            "kv_x": x,
+            "biases": biases,
+        }
+
+        return chunk_layer(
+            partial(self.mha, use_memory_efficient_kernel=use_memory_efficient_kernel, use_lma=use_lma),
+            mha_inputs,
+            chunk_size=chunk_size,
+            no_batch_dims=len(x.shape[:-2]),
+            _out=x if inplace_safe else None,
+        )
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        mask: Optional[torch.Tensor] = None,
+        chunk_size: Optional[int] = None,
+        use_memory_efficient_kernel: bool = False,
+        use_lma: bool = False,
+        inplace_safe: bool = False,
+    ) -> torch.Tensor:
+        """
+        Args:
+            x:
+                [*, I, J, C_in] input tensor (e.g. the pair representation)
+        Returns:
+            [*, I, J, C_in] output tensor
+        """
+        if mask is None:
+            # [*, I, J]
+            mask = x.new_ones(
+                x.shape[:-1],
+            )
+
+        if not self.starting:
+            x = x.transpose(-2, -3)
+            mask = mask.transpose(-1, -2)
+
+        # [*, I, J, C_in]
+        x = self.layer_norm(x)
+
+        # [*, I, 1, 1, J]
+        mask_bias = (self.inf * (mask - 1))[..., :, None, None, :]
+
+        # [*, H, I, J]
+        triangle_bias = permute_final_dims(self.linear(x), (2, 0, 1))
+
+        # [*, 1, H, I, J]
+        triangle_bias = triangle_bias.unsqueeze(-4)
+
+        biases = [mask_bias, triangle_bias]
+
+        if chunk_size is not None:
+            x = self._chunk(
+                x,
+                biases,
+                chunk_size,
+                use_memory_efficient_kernel=use_memory_efficient_kernel,
+                use_lma=use_lma,
+                inplace_safe=inplace_safe,
+            )
+        else:
+            x = self.mha(
+                q_x=x, kv_x=x, biases=biases, use_memory_efficient_kernel=use_memory_efficient_kernel, use_lma=use_lma
+            )
+
+        if not self.starting:
+            x = x.transpose(-2, -3)
+
+        return x
+
+
+class EsmFoldTriangleMultiplicativeUpdate(nn.Module):
+    """
+    Implements Algorithms 11 and 12.
+    """
+
+    def __init__(self, config, _outgoing=True):
+        super().__init__()
+        c_hidden = config.pairwise_state_dim
+        self._outgoing = _outgoing
+
+        self.linear_a_p = EsmFoldLinear(c_hidden, c_hidden)
+        self.linear_a_g = EsmFoldLinear(c_hidden, c_hidden, init="gating")
+        self.linear_b_p = EsmFoldLinear(c_hidden, c_hidden)
+        self.linear_b_g = EsmFoldLinear(c_hidden, c_hidden, init="gating")
+        self.linear_g = EsmFoldLinear(c_hidden, c_hidden, init="gating")
+        self.linear_z = EsmFoldLinear(c_hidden, c_hidden, init="final")
+
+        self.layer_norm_in = LayerNorm(c_hidden)
+        self.layer_norm_out = LayerNorm(c_hidden)
+
+        self.sigmoid = nn.Sigmoid()
+
+    def _combine_projections(
+        self, a: torch.Tensor, b: torch.Tensor, _inplace_chunk_size: Optional[int] = None
+    ) -> torch.Tensor:
+        if self._outgoing:
+            a = permute_final_dims(a, (2, 0, 1))
+            b = permute_final_dims(b, (2, 1, 0))
+        else:
+            a = permute_final_dims(a, (2, 1, 0))
+            b = permute_final_dims(b, (2, 0, 1))
+
+        if _inplace_chunk_size is not None:
+            # To be replaced by torch vmap
+            for i in range(0, a.shape[-3], _inplace_chunk_size):
+                a_chunk = a[..., i : i + _inplace_chunk_size, :, :]
+                b_chunk = b[..., i : i + _inplace_chunk_size, :, :]
+                a[..., i : i + _inplace_chunk_size, :, :] = torch.matmul(
+                    a_chunk,
+                    b_chunk,
+                )
+
+            p = a
+        else:
+            p = torch.matmul(a, b)
+
+        return permute_final_dims(p, (1, 2, 0))
+
+    def _inference_forward(
+        self,
+        z: torch.Tensor,
+        mask: Optional[torch.Tensor] = None,
+        inplace_chunk_size: Optional[int] = None,
+        with_add: bool = True,
+    ):
+        """
+        Args:
+            z:
+                A [*, N, N, C_z] pair representation
+            mask:
+                A [*, N, N] pair mask
+            inplace_chunk_size:
+                Size of chunks used in the main computation. Increase to trade memory for speed.
+            with_add:
+                If True, z is overwritten with (z + update). Otherwise, it is overwritten with (update).
+        Returns:
+            A reference to the overwritten z
+
+        More memory-efficient, inference-only version of the forward function. Uses in-place operations, fusion of the
+        addition that happens after this module in the Evoformer, a smidge of recomputation, and a cache of overwritten
+        values to lower peak memory consumption of this module from 5x the size of the input tensor z to 2.5x its size.
+        Useful for inference on extremely long sequences.
+
+        It works as follows. We will make reference to variables used in the default forward implementation below.
+        Naively, triangle multiplication attention requires the manifestation of 5 tensors the size of z: 1) z, the
+        "square" input tensor, 2) a, the first projection of z, 3) b, the second projection of b, 4) g, a z-sized mask,
+        and 5) a z-sized tensor for intermediate computations. For large N, this is prohibitively expensive; for
+        N=4000, for example, z is more than 8GB alone. To avoid this problem, we compute b, g, and all intermediate
+        tensors in small chunks, noting that the chunks required to compute a chunk of the output depend only on the
+        tensor a and corresponding vertical and horizontal chunks of z. This suggests an algorithm that loops over
+        pairs of chunks of z: hereafter "columns" and "rows" of z, even though each "column" and "row" in fact contains
+        inplace_chunk_size contiguous true columns and rows of z. Writing output chunks to a new tensor would bring
+        total memory consumption down to 3x the size of z. However, more memory can be saved by writing output chunks
+        directly to z in-place. WLOG, we choose to write output chunks vertically, overwriting the ith "column" of z at
+        the end of the ith iteration of the main loop. Despite this overwriting, the ith column is always one column
+        ahead of previously overwritten columns and can be recovered directly from z. After the first iteration,
+        however, the ith row of z is always at least partially overwritten. For this reason, we introduce the z-cache,
+        a tensor one-half the size of z. The z-cache initially contains the left half (2nd and 3rd quadrants) of z. For
+        0 < i < N/2, the missing left part of the ith row of z is recovered from this cache at the beginning of the ith
+        iteration. Once i exceeds n/2, the cache is "reoriented" to encompass the 3rd and 4th quadrants of z instead.
+        Though the 3rd quadrant of the original z is entirely overwritten at this point, it can be recovered from the
+        z-cache itself. Thereafter, the ith row of z can be recovered in its entirety from the reoriented z-cache.
+        After the final iteration, z has been completely overwritten and contains the triangular multiplicative update.
+        If with_add is True, it instead contains the sum of z and the triangular multiplicative update. In either case,
+        peak memory consumption is just 2.5x the size of z, disregarding memory used for chunks and other small
+        variables.
+        """
+        if mask is None:
+            mask = z.new_ones(z.shape[:-1])
+
+        mask = mask.unsqueeze(-1)
+
+        def compute_projection_helper(pair, mask, a=True):
+            if a:
+                linear_g = self.linear_a_g
+                linear_p = self.linear_a_p
+            else:
+                linear_g = self.linear_b_g
+                linear_p = self.linear_b_p
+
+            pair = self.layer_norm_in(pair)
+            p = linear_g(pair)
+            p.sigmoid_()
+            p *= linear_p(pair)
+            p *= mask
+            p = permute_final_dims(p, (2, 0, 1))
+            return p
+
+        def compute_projection(pair, mask, a=True, chunked=True):
+            need_transpose = self._outgoing ^ a
+            if not chunked:
+                p = compute_projection_helper(pair, mask, a)
+                if need_transpose:
+                    p = p.transpose(-1, -2)
+            else:
+                # This computation is chunked so as not to exceed our 2.5x
+                # budget with a large intermediate tensor
+                linear_g = self.linear_a_g if a else self.linear_b_g
+                c = linear_g.bias.shape[-1]
+                out_shape = pair.shape[:-3] + (c,) + pair.shape[-3:-1]
+                p = pair.new_zeros(out_shape)
+                for i in range(0, pair.shape[-3], inplace_chunk_size):
+                    pair_chunk = pair[..., i : i + inplace_chunk_size, :, :]
+                    pair_chunk = compute_projection_helper(
+                        pair[..., i : i + inplace_chunk_size, :, :],
+                        mask[..., i : i + inplace_chunk_size, :, :],
+                        a,
+                    )
+                    if need_transpose:
+                        pair_chunk = pair_chunk.transpose(-1, -2)
+                        p[..., i : i + inplace_chunk_size] = pair_chunk
+                    else:
+                        p[..., i : i + inplace_chunk_size, :] = pair_chunk
+
+                    del pair_chunk
+
+            return p
+
+        # We start by fully manifesting a. In addition to the input, this
+        # brings total memory consumption to 2x z (disregarding size of chunks)
+        # [*, N, N, c]
+        a = compute_projection(z, mask, True, chunked=True)
+
+        if inplace_chunk_size is not None:
+            n = a.shape[-1]
+            half_n = n // 2 + n % 2
+            row_dim = -3
+            col_dim = -2
+            b_chunk_dim = row_dim if self._outgoing else col_dim
+
+            def empty_slicer(t):
+                return [slice(None) for _ in t.shape]
+
+            def slice_tensor(t, start, end, dim):
+                # Slices start:end from the dim dimension of t
+                s = empty_slicer(t)
+                s[dim] = slice(start, end)
+                return t[s]
+
+            def flip_z_cache_(z_cache, z):
+                # "Reorient" the z_cache (see below), filling it with quadrants
+                # 3---recovered from the z_cache---and 4---recovered from z---
+                # of the input tensor z.
+                quadrant_3 = slice_tensor(z_cache, half_n, None, row_dim)
+                z_cache = z_cache.transpose(row_dim, col_dim)
+
+                # If n is odd, we need to shrink the z_cache by one row
+                z_cache = z_cache[..., : (n // 2), :, :]
+
+                # Move the 3rd quadrant of z into the
+                first_half_slicer = empty_slicer(z_cache)
+                first_half_slicer[col_dim] = slice(0, half_n)
+                z_cache[first_half_slicer] = quadrant_3
+
+                # Get the fourth quadrant of z
+                quadrant_4 = slice_tensor(z, half_n, None, row_dim)
+                quadrant_4 = slice_tensor(quadrant_4, half_n, None, col_dim)
+
+                # Insert said quadrant into the rotated z-cache
+                quadrant_3_slicer = empty_slicer(z_cache)
+                quadrant_3_slicer[col_dim] = slice(half_n, None)
+
+                z_cache[quadrant_3_slicer] = quadrant_4
+
+                return z_cache
+
+            # Initialize the z cache to the left half of z.
+            z_cache_shape = list(z.shape)
+            z_cache_shape[col_dim] = half_n
+            z_cache = z.new_zeros(z_cache_shape)
+            z_cache_slicer = empty_slicer(z_cache)
+            z_cache_slicer[col_dim] = slice(0, half_n)
+            z_cache.copy_(z[z_cache_slicer])
+            z_cache_rotated = False
+
+            # We need to reorient the z-cache at the halfway point, and we
+            # don't want a single chunk to straddle that point. We contract one
+            # of the chunks in the middle to address that problem.
+            i_range = list(range(0, half_n, inplace_chunk_size))
+            initial_offsets = [i_2 - i_1 for i_1, i_2 in zip(i_range, i_range[1:] + [half_n])]
+            after_half = list(range(half_n, n, inplace_chunk_size))
+            after_half_offsets = [inplace_chunk_size for _ in after_half]
+            combined_range_with_offsets = zip(i_range + after_half, initial_offsets + after_half_offsets)
+            for i, offset in combined_range_with_offsets:
+                if not z_cache_rotated and i >= half_n:
+                    z_cache = flip_z_cache_(z_cache, z)
+                    z_cache_rotated = True
+
+                z_chunk_b = slice_tensor(z, i, i + offset, b_chunk_dim)
+                mask_chunk = slice_tensor(mask, i, i + offset, b_chunk_dim)
+
+                z_chunk_b = z_chunk_b.clone()
+                if b_chunk_dim == col_dim:
+                    z_chunk_b = slice_tensor(z, i, i + offset, col_dim)
+                else:  # b_chunk_dim == row_dim
+                    # In this case, the b-dimension (b_chunk_dim) is partially
+                    # overwritten at the end of each iteration. We need to
+                    # restore the missing component from the z-cache.
+                    if not z_cache_rotated:
+                        z_chunk_slicer = empty_slicer(z_chunk_b)
+                        z_chunk_slicer[col_dim] = slice(0, half_n)
+                        z_chunk_b[z_chunk_slicer] = slice_tensor(z_cache, i, i + offset, row_dim)
+                    else:
+                        z_cache_offset = i - half_n
+                        z_chunk_b = slice_tensor(z_cache, z_cache_offset, z_cache_offset + offset, row_dim)
+
+                b_chunk = compute_projection(z_chunk_b, mask_chunk, a=False, chunked=False)
+                del z_chunk_b
+
+                x_chunk = torch.matmul(a, b_chunk)
+                x_chunk = permute_final_dims(x_chunk, (1, 2, 0))
+                x_chunk = self.layer_norm_out(x_chunk)
+                x_chunk = self.linear_z(x_chunk)
+
+                # The g dimension (col_dim) is parallel to and ahead of the
+                # overwrites in z. We can extract the g chunk normally.
+                z_chunk_g = slice_tensor(z, i, i + offset, col_dim)
+                g_chunk = self.linear_g(self.layer_norm_in(z_chunk_g))
+                g_chunk.sigmoid_()
+                del z_chunk_g
+
+                x_chunk *= g_chunk
+
+                # Write the columns into z in-place
+                z_slicer = empty_slicer(z)
+                z_slicer[col_dim] = slice(i, i + offset)
+                if with_add:
+                    z[z_slicer] += x_chunk
+                else:
+                    z[z_slicer] = x_chunk
+        else:
+            b = compute_projection(z, mask, False, False)
+            x = torch.matmul(a, b)
+            x = self.layer_norm_out(x)
+            x = self.linear_z(x)
+            g = self.linear_g(z)
+            g.sigmoid_()
+            x *= g
+            if with_add:
+                z += x
+            else:
+                z = x
+
+        return z
+
+    def forward(
+        self,
+        z: torch.Tensor,
+        mask: Optional[torch.Tensor] = None,
+        inplace_safe: bool = False,
+        _add_with_inplace: bool = False,
+        _inplace_chunk_size: Optional[int] = 256,
+    ) -> torch.Tensor:
+        """
+        Args:
+            x:
+                [*, N_res, N_res, C_z] input tensor
+            mask:
+                [*, N_res, N_res] input mask
+        Returns:
+            [*, N_res, N_res, C_z] output tensor
+        """
+        if inplace_safe:
+            x = self._inference_forward(
+                z,
+                mask,
+                inplace_chunk_size=_inplace_chunk_size,
+                with_add=_add_with_inplace,
+            )
+            return x
+
+        if mask is None:
+            mask = z.new_ones(z.shape[:-1])
+
+        mask = mask.unsqueeze(-1)
+
+        z = self.layer_norm_in(z)
+        a = mask
+        a = a * self.sigmoid(self.linear_a_g(z))
+        a = a * self.linear_a_p(z)
+        b = mask
+        b = b * self.sigmoid(self.linear_b_g(z))
+        b = b * self.linear_b_p(z)
+
+        if is_fp16_enabled():
+            with torch.cuda.amp.autocast(enabled=False):
+                x = self._combine_projections(a.float(), b.float())
+        else:
+            x = self._combine_projections(a, b)
+
+        del a, b
+        x = self.layer_norm_out(x)
+        x = self.linear_z(x)
+        g = self.sigmoid(self.linear_g(z))
+        x = x * g
+
+        return x
+
+
+class EsmFoldPreTrainedModel(EsmPreTrainedModel):
+    """
+    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
+    models.
+    """
+
+    # Subclass `EsMPreTrainedModel` to deal with special init
+    def _init_weights(self, module):
+        """Initialize the weights"""
+        if isinstance(module, EsmFoldLinear):
+            with torch.no_grad():
+                if module.init_fn is not None:
+                    module.init_fn(module.weight, module.bias)
+                elif module.init == "default":
+                    trunc_normal_init_(module.weight, scale=1.0)
+                elif module.init == "relu":
+                    trunc_normal_init_(module.weight, scale=2.0)
+                elif module.init == "glorot":
+                    nn.init.xavier_uniform_(module.weight, gain=1)
+                elif module.init == "gating":
+                    module.weight.fill_(0.0)
+                    if module.bias:
+                        module.bias.fill_(1.0)
+                elif module.init == "normal":
+                    torch.nn.init.kaiming_normal_(module.weight, nonlinearity="linear")
+                elif module.init == "final":
+                    module.weight.fill_(0.0)
+        elif isinstance(module, EsmFoldInvariantPointAttention):
+            ipa_point_weights_init_(module.head_weights)
+        elif isinstance(module, EsmFoldTriangularSelfAttentionBlock):
+            torch.nn.init.zeros_(module.tri_mul_in.linear_z.weight)
+            torch.nn.init.zeros_(module.tri_mul_in.linear_z.bias)
+            torch.nn.init.zeros_(module.tri_mul_out.linear_z.weight)
+            torch.nn.init.zeros_(module.tri_mul_out.linear_z.bias)
+            torch.nn.init.zeros_(module.tri_att_start.mha.linear_o.weight)
+            torch.nn.init.zeros_(module.tri_att_start.mha.linear_o.bias)
+            torch.nn.init.zeros_(module.tri_att_end.mha.linear_o.weight)
+            torch.nn.init.zeros_(module.tri_att_end.mha.linear_o.bias)
+
+            torch.nn.init.zeros_(module.sequence_to_pair.o_proj.weight)
+            torch.nn.init.zeros_(module.sequence_to_pair.o_proj.bias)
+            torch.nn.init.zeros_(module.pair_to_sequence.linear.weight)
+            torch.nn.init.zeros_(module.seq_attention.o_proj.weight)
+            torch.nn.init.zeros_(module.seq_attention.o_proj.bias)
+            torch.nn.init.zeros_(module.mlp_seq.mlp[-2].weight)
+            torch.nn.init.zeros_(module.mlp_seq.mlp[-2].bias)
+            torch.nn.init.zeros_(module.mlp_pair.mlp[-2].weight)
+            torch.nn.init.zeros_(module.mlp_pair.mlp[-2].bias)
+        else:
+            super()._init_weights(module)
+
+
+class EsmFoldSelfAttention(nn.Module):
+    def __init__(self, embed_dim, num_heads, head_width, gated=False):
+        super().__init__()
+        assert embed_dim == num_heads * head_width
+
+        self.embed_dim = embed_dim
+        self.num_heads = num_heads
+        self.head_width = head_width
+
+        self.proj = nn.Linear(embed_dim, embed_dim * 3, bias=False)
+        self.o_proj = nn.Linear(embed_dim, embed_dim, bias=True)
+        self.gated = gated
+        if gated:
+            self.g_proj = nn.Linear(embed_dim, embed_dim)
+            torch.nn.init.zeros_(self.g_proj.weight)
+            torch.nn.init.ones_(self.g_proj.bias)
+
+        self.rescale_factor = self.head_width**-0.5
+
+        torch.nn.init.zeros_(self.o_proj.bias)
+
+    def forward(self, x, mask=None, bias=None, indices=None):
+        """
+        Basic self attention with optional mask and external pairwise bias. To handle sequences of different lengths,
+        use mask.
+
+        Inputs:
+            x: batch of input sequneces (.. x L x C) mask: batch of boolean masks where 1=valid, 0=padding position (..
+            x L_k) bias: batch of scalar pairwise attention biases (.. x Lq x Lk x num_heads)
+
+        Outputs:
+          sequence projection (B x L x embed_dim), attention maps (B x L x L x num_heads)
+        """
+
+        t = self.proj(x).view(*x.shape[:2], self.num_heads, -1)
+        t = t.permute(0, 2, 1, 3)
+        q, k, v = t.chunk(3, dim=-1)
+
+        q = self.rescale_factor * q
+        a = torch.einsum("...qc,...kc->...qk", q, k)
+
+        # Add external attention bias.
+        if bias is not None:
+            a = a + bias.permute(0, 3, 1, 2)
+
+        # Do not attend to padding tokens.
+        if mask is not None:
+            mask = mask[:, None, None]
+            a = a.masked_fill(mask == False, -np.inf)  # noqa: E712
+
+        a = nn.functional.softmax(a, dim=-1)
+
+        y = torch.einsum("...hqk,...hkc->...qhc", a, v)
+        y = y.reshape(*y.shape[:2], -1)
+
+        if self.gated:
+            y = self.g_proj(x).sigmoid() * y
+        y = self.o_proj(y)
+
+        return y, a.permute(0, 3, 1, 2)
+
+
+class EsmFoldDropout(nn.Module):
+    """
+    Implementation of dropout with the ability to share the dropout mask along a particular dimension.
+    """
+
+    def __init__(self, r: float, batch_dim: Union[int, List[int]]):
+        super().__init__()
+
+        self.r = r
+        if isinstance(batch_dim, int):
+            batch_dim = [batch_dim]
+        self.batch_dim = batch_dim
+        self.dropout = nn.Dropout(self.r)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        shape = list(x.shape)
+        if self.batch_dim is not None:
+            for bd in self.batch_dim:
+                shape[bd] = 1
+        return x * self.dropout(x.new_ones(shape))
+
+
+class EsmFoldSequenceToPair(nn.Module):
+    def __init__(self, sequence_state_dim, inner_dim, pairwise_state_dim):
+        super().__init__()
+
+        self.layernorm = nn.LayerNorm(sequence_state_dim)
+        self.proj = nn.Linear(sequence_state_dim, inner_dim * 2, bias=True)
+        self.o_proj = nn.Linear(2 * inner_dim, pairwise_state_dim, bias=True)
+
+        torch.nn.init.zeros_(self.proj.bias)
+        torch.nn.init.zeros_(self.o_proj.bias)
+
+    def forward(self, sequence_state):
+        """
+        Inputs:
+          sequence_state: B x L x sequence_state_dim
+
+        Output:
+          pairwise_state: B x L x L x pairwise_state_dim
+
+        Intermediate state:
+          B x L x L x 2*inner_dim
+        """
+
+        assert len(sequence_state.shape) == 3
+
+        s = self.layernorm(sequence_state)
+        s = self.proj(s)
+        q, k = s.chunk(2, dim=-1)
+
+        prod = q[:, None, :, :] * k[:, :, None, :]
+        diff = q[:, None, :, :] - k[:, :, None, :]
+
+        x = torch.cat([prod, diff], dim=-1)
+        x = self.o_proj(x)
+
+        return x
+
+
+class EsmFoldPairToSequence(nn.Module):
+    def __init__(self, pairwise_state_dim, num_heads):
+        super().__init__()
+
+        self.layernorm = nn.LayerNorm(pairwise_state_dim)
+        self.linear = nn.Linear(pairwise_state_dim, num_heads, bias=False)
+
+    def forward(self, pairwise_state):
+        """
+        Inputs:
+          pairwise_state: B x L x L x pairwise_state_dim
+
+        Output:
+          pairwise_bias: B x L x L x num_heads
+        """
+        assert len(pairwise_state.shape) == 4
+        z = self.layernorm(pairwise_state)
+        pairwise_bias = self.linear(z)
+        return pairwise_bias
+
+
+class EsmFoldResidueMLP(nn.Module):
+    def __init__(self, embed_dim, inner_dim, dropout=0):
+        super().__init__()
+
+        self.mlp = nn.Sequential(
+            nn.LayerNorm(embed_dim),
+            nn.Linear(embed_dim, inner_dim),
+            nn.ReLU(),
+            nn.Linear(inner_dim, embed_dim),
+            nn.Dropout(dropout),
+        )
+
+    def forward(self, x):
+        return x + self.mlp(x)
+
+
+class EsmFoldTriangularSelfAttentionBlock(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+
+        sequence_state_dim = config.sequence_state_dim
+        pairwise_state_dim = config.pairwise_state_dim
+        sequence_num_heads = sequence_state_dim // config.sequence_head_width
+        pairwise_num_heads = pairwise_state_dim // config.pairwise_head_width
+
+        self.layernorm_1 = nn.LayerNorm(sequence_state_dim)
+
+        self.sequence_to_pair = EsmFoldSequenceToPair(sequence_state_dim, pairwise_state_dim // 2, pairwise_state_dim)
+        self.pair_to_sequence = EsmFoldPairToSequence(pairwise_state_dim, sequence_num_heads)
+
+        self.seq_attention = EsmFoldSelfAttention(
+            sequence_state_dim, sequence_num_heads, config.sequence_head_width, gated=True
+        )
+        self.tri_mul_out = EsmFoldTriangleMultiplicativeUpdate(config, _outgoing=True)
+        self.tri_mul_in = EsmFoldTriangleMultiplicativeUpdate(config, _outgoing=False)
+
+        self.tri_att_start = EsmFoldTriangleAttention(
+            pairwise_state_dim, config.pairwise_head_width, pairwise_num_heads, inf=1e9, starting=True
+        )
+        self.tri_att_end = EsmFoldTriangleAttention(
+            pairwise_state_dim, config.pairwise_head_width, pairwise_num_heads, inf=1e9, starting=False
+        )
+
+        self.mlp_seq = EsmFoldResidueMLP(sequence_state_dim, 4 * sequence_state_dim, dropout=config.dropout)
+        self.mlp_pair = EsmFoldResidueMLP(pairwise_state_dim, 4 * pairwise_state_dim, dropout=config.dropout)
+
+        self.drop = nn.Dropout(config.dropout)
+        self.row_drop = EsmFoldDropout(config.dropout * 2, 2)
+        self.col_drop = EsmFoldDropout(config.dropout * 2, 1)
+
+    def forward(self, sequence_state, pairwise_state, mask=None, chunk_size=None, **__kwargs):
+        """
+        Inputs:
+          sequence_state: B x L x sequence_state_dim pairwise_state: B x L x L x pairwise_state_dim mask: B x L boolean
+          tensor of valid positions
+
+        Output:
+          sequence_state: B x L x sequence_state_dim pairwise_state: B x L x L x pairwise_state_dim
+        """
+        if len(sequence_state.shape) != 3:
+            raise ValueError(f"`sequence_state` should be a 3d-tensor, got {len(sequence_state.shape)} dims.")
+        if len(pairwise_state.shape) != 4:
+            raise ValueError(f"`pairwise_state` should be a 4d-tensor, got {len(pairwise_state.shape)} dims.")
+        if mask is not None and len(mask.shape) != 2:
+            raise ValueError(f"`mask` should be a 2d-tensor, got {len(mask.shape)} dims.")
+
+        batch_dim, seq_dim, sequence_state_dim = sequence_state.shape
+        pairwise_state_dim = pairwise_state.shape[3]
+
+        if sequence_state_dim != self.config.sequence_state_dim:
+            raise ValueError(
+                "`sequence_state` last dimension should be equal to `self.sequence_state_dim`. Got "
+                f"{sequence_state_dim} != {self.config.sequence_state_dim}."
+            )
+        if pairwise_state_dim != self.config.pairwise_state_dim:
+            raise ValueError(
+                "`pairwise_state` last dimension should be equal to `self.pairwise_state_dim`. Got "
+                f"{pairwise_state_dim} != {self.config.pairwise_state_dim}."
+            )
+        if batch_dim != pairwise_state.shape[0]:
+            raise ValueError(
+                f"`sequence_state` and `pairwise_state` have inconsistent batch size: {batch_dim} != "
+                f"{pairwise_state.shape[0]}."
+            )
+        if seq_dim != pairwise_state.shape[1] or seq_dim != pairwise_state.shape[2]:
+            raise ValueError(
+                f"`sequence_state` and `pairwise_state` have inconsistent sequence length: {seq_dim} != "
+                f"{pairwise_state.shape[1]} or {pairwise_state.shape[2]}."
+            )
+
+        # Update sequence state
+        bias = self.pair_to_sequence(pairwise_state)
+
+        # Self attention with bias + mlp.
+        y = self.layernorm_1(sequence_state)
+        y, _ = self.seq_attention(y, mask=mask, bias=bias)
+        sequence_state = sequence_state + self.drop(y)
+        sequence_state = self.mlp_seq(sequence_state)
+
+        # Update pairwise state
+        pairwise_state = pairwise_state + self.sequence_to_pair(sequence_state)
+
+        # Axial attention with triangular bias.
+        tri_mask = mask.unsqueeze(2) * mask.unsqueeze(1) if mask is not None else None
+        pairwise_state = pairwise_state + self.row_drop(self.tri_mul_out(pairwise_state, mask=tri_mask))
+        pairwise_state = pairwise_state + self.col_drop(self.tri_mul_in(pairwise_state, mask=tri_mask))
+        pairwise_state = pairwise_state + self.row_drop(
+            self.tri_att_start(pairwise_state, mask=tri_mask, chunk_size=chunk_size)
+        )
+        pairwise_state = pairwise_state + self.col_drop(
+            self.tri_att_end(pairwise_state, mask=tri_mask, chunk_size=chunk_size)
+        )
+
+        # MLP over pairs.
+        pairwise_state = self.mlp_pair(pairwise_state)
+
+        return sequence_state, pairwise_state
+
+
+class EsmCategoricalMixture:
+    def __init__(self, param, bins=50, start=0, end=1):
+        # All tensors are of shape ..., bins.
+        self.logits = param
+        bins = torch.linspace(start, end, bins + 1, device=self.logits.device, dtype=self.logits.dtype)
+        self.v_bins = (bins[:-1] + bins[1:]) / 2
+
+    def log_prob(self, true):
+        # Shapes are:
+        #     self.probs: ... x bins
+        #     true      : ...
+        true_index = (true.unsqueeze(-1) - self.v_bins[[None] * true.ndim]).abs().argmin(-1)
+        nll = self.logits.log_softmax(-1)
+        return torch.take_along_dim(nll, true_index.unsqueeze(-1), dim=-1).squeeze(-1)
+
+    def mean(self):
+        return (self.logits.softmax(-1) @ self.v_bins.unsqueeze(1)).squeeze(-1)
+
+
+def categorical_lddt(logits, bins=50):
+    # Logits are ..., 37, bins.
+    return EsmCategoricalMixture(logits, bins=bins).mean()
+
+
+def get_axial_mask(mask):
+    """
+    Helper to convert B x L mask of valid positions to axial mask used in row column attentions.
+
+    Input:
+      mask: B x L tensor of booleans
+
+    Output:
+      mask: B x L x L tensor of booleans
+    """
+
+    if mask is None:
+        return None
+
+    if len(mask.shape) != 2:
+        raise ValueError(f"`mask` should be a 2d-tensor, got {len(mask.shape)} dims.")
+    batch_dim, seq_dim = mask.shape
+    m = mask.unsqueeze(1).expand(batch_dim, seq_dim, seq_dim)
+    m = m.reshape(batch_dim * seq_dim, seq_dim)
+    return m
+
+
+class EsmFoldRelativePosition(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.bins = config.position_bins
+
+        # Note an additional offset is used so that the 0th position
+        # is reserved for masked pairs.
+        self.embedding = torch.nn.Embedding(2 * self.bins + 2, config.pairwise_state_dim)
+
+    def forward(self, residue_index, mask=None):
+        """
+        Input:
+          residue_index: B x L tensor of indices (dytpe=torch.long) mask: B x L tensor of booleans
+
+        Output:
+          pairwise_state: B x L x L x pairwise_state_dim tensor of embeddings
+        """
+        if residue_index.dtype != torch.long:
+            raise ValueError(f"`residue_index` has dtype {residue_index.dtype}, it should be `torch.long`.")
+        if mask is not None and residue_index.shape != mask.shape:
+            raise ValueError(
+                f"`residue_index` and `mask` have inconsistent shapes: {residue_index.shape} != {mask.shape}."
+            )
+
+        diff = residue_index[:, None, :] - residue_index[:, :, None]
+        diff = diff.clamp(-self.bins, self.bins)
+        diff = diff + self.bins + 1  # Add 1 to adjust for padding index.
+
+        if mask is not None:
+            mask = mask[:, None, :] * mask[:, :, None]
+            diff[mask == False] = 0  # noqa: E712
+
+        output = self.embedding(diff)
+        return output
+
+
+class EsmFoldAngleResnetBlock(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+
+        self.linear_1 = EsmFoldLinear(config.resnet_dim, config.resnet_dim, init="relu")
+        self.linear_2 = EsmFoldLinear(config.resnet_dim, config.resnet_dim, init="final")
+
+        self.relu = nn.ReLU()
+
+    def forward(self, a: torch.Tensor) -> torch.Tensor:
+        s_initial = a
+
+        a = self.relu(a)
+        a = self.linear_1(a)
+        a = self.relu(a)
+        a = self.linear_2(a)
+
+        return a + s_initial
+
+
+class EsmFoldAngleResnet(nn.Module):
+    """
+    Implements Algorithm 20, lines 11-14
+    """
+
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+
+        self.linear_in = EsmFoldLinear(config.sequence_dim, config.resnet_dim)
+        self.linear_initial = EsmFoldLinear(config.sequence_dim, config.resnet_dim)
+
+        self.layers = nn.ModuleList()
+        for _ in range(config.num_resnet_blocks):
+            layer = EsmFoldAngleResnetBlock(config)
+            self.layers.append(layer)
+
+        self.linear_out = EsmFoldLinear(config.resnet_dim, config.num_angles * 2)
+
+        self.relu = nn.ReLU()
+
+    def forward(self, s: torch.Tensor, s_initial: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Args:
+            s:
+                [*, C_hidden] single embedding
+            s_initial:
+                [*, C_hidden] single embedding as of the start of the StructureModule
+        Returns:
+            [*, no_angles, 2] predicted angles
+        """
+        # NOTE: The ReLU's applied to the inputs are absent from the supplement
+        # pseudocode but present in the source. For maximal compatibility with
+        # the pretrained weights, I'm going with the source.
+
+        # [*, C_hidden]
+        s_initial = self.relu(s_initial)
+        s_initial = self.linear_initial(s_initial)
+        s = self.relu(s)
+        s = self.linear_in(s)
+        s = s + s_initial
+
+        for l in self.layers:
+            s = l(s)
+
+        s = self.relu(s)
+
+        # [*, no_angles * 2]
+        s = self.linear_out(s)
+
+        # [*, no_angles, 2]
+        s = s.view(s.shape[:-1] + (-1, 2))
+
+        unnormalized_s = s
+        norm_denom = torch.sqrt(
+            torch.clamp(
+                torch.sum(s**2, dim=-1, keepdim=True),
+                min=self.config.epsilon,
+            )
+        )
+        s = s / norm_denom
+
+        return unnormalized_s, s
+
+
+class EsmFoldInvariantPointAttention(nn.Module):
+    """
+    Implements Algorithm 22.
+    """
+
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+
+        c_s = config.sequence_dim
+        c_z = config.pairwise_dim
+        self.hidden_dim = config.ipa_dim
+        self.num_heads = config.num_heads_ipa
+        self.num_qk_points = config.num_qk_points
+        self.num_v_points = config.num_v_points
+
+        # These linear layers differ from their specifications in the
+        # supplement. There, they lack bias and use Glorot initialization.
+        # Here as in the official source, they have bias and use the default
+        # Lecun initialization.
+        hc = config.ipa_dim * config.num_heads_ipa
+        self.linear_q = EsmFoldLinear(c_s, hc)
+        self.linear_kv = EsmFoldLinear(c_s, 2 * hc)
+
+        hpq = config.num_heads_ipa * config.num_qk_points * 3
+        self.linear_q_points = EsmFoldLinear(c_s, hpq)
+
+        hpkv = config.num_heads_ipa * (config.num_qk_points + config.num_v_points) * 3
+        self.linear_kv_points = EsmFoldLinear(c_s, hpkv)
+
+        self.linear_b = EsmFoldLinear(c_z, config.num_heads_ipa)
+
+        self.head_weights = nn.Parameter(torch.zeros((config.num_heads_ipa)))
+
+        concat_out_dim = config.num_heads_ipa * (c_z + config.ipa_dim + config.num_v_points * 4)
+        self.linear_out = EsmFoldLinear(concat_out_dim, c_s, init="final")
+
+        self.softmax = nn.Softmax(dim=-1)
+        self.softplus = nn.Softplus()
+
+    def forward(
+        self,
+        s: torch.Tensor,
+        z: Optional[torch.Tensor],
+        r: Rigid,
+        mask: torch.Tensor,
+        _offload_inference: bool = False,
+        _z_reference_list: Optional[Sequence[torch.Tensor]] = None,
+    ) -> torch.Tensor:
+        """
+        Args:
+            s:
+                [*, N_res, C_s] single representation
+            z:
+                [*, N_res, N_res, C_z] pair representation
+            r:
+                [*, N_res] transformation object
+            mask:
+                [*, N_res] mask
+        Returns:
+            [*, N_res, C_s] single representation update
+        """
+        z = [z]
+
+        #######################################
+        # Generate scalar and point activations
+        #######################################
+        # [*, N_res, H * C_hidden]
+        q = self.linear_q(s)
+        kv = self.linear_kv(s)
+
+        # [*, N_res, H, C_hidden]
+        q = q.view(q.shape[:-1] + (self.num_heads, -1))
+
+        # [*, N_res, H, 2 * C_hidden]
+        kv = kv.view(kv.shape[:-1] + (self.num_heads, -1))
+
+        # [*, N_res, H, C_hidden]
+        k, v = torch.split(kv, self.hidden_dim, dim=-1)
+
+        # [*, N_res, H * P_q * 3]
+        q_pts = self.linear_q_points(s)
+
+        # This is kind of clunky, but it's how the original does it
+        # [*, N_res, H * P_q, 3]
+        q_pts = torch.split(q_pts, q_pts.shape[-1] // 3, dim=-1)
+        q_pts = torch.stack(q_pts, dim=-1)
+        q_pts = r[..., None].apply(q_pts)
+
+        # [*, N_res, H, P_q, 3]
+        q_pts = q_pts.view(q_pts.shape[:-2] + (self.num_heads, self.num_qk_points, 3))
+
+        # [*, N_res, H * (P_q + P_v) * 3]
+        kv_pts = self.linear_kv_points(s)
+
+        # [*, N_res, H * (P_q + P_v), 3]
+        kv_pts = torch.split(kv_pts, kv_pts.shape[-1] // 3, dim=-1)
+        kv_pts = torch.stack(kv_pts, dim=-1)
+        kv_pts = r[..., None].apply(kv_pts)
+
+        # [*, N_res, H, (P_q + P_v), 3]
+        kv_pts = kv_pts.view(kv_pts.shape[:-2] + (self.num_heads, -1, 3))
+
+        # [*, N_res, H, P_q/P_v, 3]
+        k_pts, v_pts = torch.split(kv_pts, [self.num_qk_points, self.num_v_points], dim=-2)
+
+        ##########################
+        # Compute attention scores
+        ##########################
+        # [*, N_res, N_res, H]
+        b = self.linear_b(z[0])
+
+        if _offload_inference:
+            assert sys.getrefcount(z[0]) == 2
+            z[0] = z[0].cpu()
+
+        # [*, H, N_res, N_res]
+        if is_fp16_enabled():
+            with torch.cuda.amp.autocast(enabled=False):
+                a = torch.matmul(
+                    permute_final_dims(q.float(), (1, 0, 2)),  # [*, H, N_res, C_hidden]
+                    permute_final_dims(k.float(), (1, 2, 0)),  # [*, H, C_hidden, N_res]
+                )
+        else:
+            a = torch.matmul(
+                permute_final_dims(q, (1, 0, 2)),  # [*, H, N_res, C_hidden]
+                permute_final_dims(k, (1, 2, 0)),  # [*, H, C_hidden, N_res]
+            )
+
+        a *= math.sqrt(1.0 / (3 * self.hidden_dim))
+        a += math.sqrt(1.0 / 3) * permute_final_dims(b, (2, 0, 1))
+
+        # [*, N_res, N_res, H, P_q, 3]
+        pt_att = q_pts.unsqueeze(-4) - k_pts.unsqueeze(-5)
+        pt_att = pt_att**2
+
+        # [*, N_res, N_res, H, P_q]
+        pt_att = sum(torch.unbind(pt_att, dim=-1))
+        head_weights = self.softplus(self.head_weights).view(*((1,) * len(pt_att.shape[:-2]) + (-1, 1)))
+        head_weights = head_weights * math.sqrt(1.0 / (3 * (self.num_qk_points * 9.0 / 2)))
+        pt_att = pt_att * head_weights
+
+        # [*, N_res, N_res, H]
+        pt_att = torch.sum(pt_att, dim=-1) * (-0.5)
+        # [*, N_res, N_res]
+        square_mask = mask.unsqueeze(-1) * mask.unsqueeze(-2)
+        square_mask = self.config.inf * (square_mask - 1)
+
+        # [*, H, N_res, N_res]
+        pt_att = permute_final_dims(pt_att, (2, 0, 1))
+
+        a = a + pt_att
+        a = a + square_mask.unsqueeze(-3)
+        a = self.softmax(a)
+
+        ################
+        # Compute output
+        ################
+        # [*, N_res, H, C_hidden]
+        o = torch.matmul(a, v.transpose(-2, -3).to(dtype=a.dtype)).transpose(-2, -3)
+
+        # [*, N_res, H * C_hidden]
+        o = flatten_final_dims(o, 2)
+
+        # [*, H, 3, N_res, P_v]
+        o_pt = torch.sum(
+            (a[..., None, :, :, None] * permute_final_dims(v_pts, (1, 3, 0, 2))[..., None, :, :]),
+            dim=-2,
+        )
+
+        # [*, N_res, H, P_v, 3]
+        o_pt = permute_final_dims(o_pt, (2, 0, 3, 1))
+        o_pt = r[..., None, None].invert_apply(o_pt)
+
+        # [*, N_res, H * P_v]
+        o_pt_norm = flatten_final_dims(torch.sqrt(torch.sum(o_pt**2, dim=-1) + self.config.epsilon), 2)
+
+        # [*, N_res, H * P_v, 3]
+        o_pt = o_pt.reshape(*o_pt.shape[:-3], -1, 3)
+
+        if _offload_inference:
+            z[0] = z[0].to(o_pt.device)
+
+        # [*, N_res, H, C_z]
+        o_pair = torch.matmul(a.transpose(-2, -3), z[0].to(dtype=a.dtype))
+
+        # [*, N_res, H * C_z]
+        o_pair = flatten_final_dims(o_pair, 2)
+
+        # [*, N_res, C_s]
+        s = self.linear_out(
+            torch.cat((o, *torch.unbind(o_pt, dim=-1), o_pt_norm, o_pair), dim=-1).to(dtype=z[0].dtype)
+        )
+
+        return s
+
+
+class EsmFoldBackboneUpdate(nn.Module):
+    """
+    Implements part of Algorithm 23.
+    """
+
+    def __init__(self, config):
+        super().__init__()
+
+        self.linear = EsmFoldLinear(config.sequence_dim, 6, init="final")
+
+    def forward(self, s: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Args:
+            [*, N_res, C_s] single representation
+        Returns:
+            [*, N_res, 6] update vector
+        """
+        # [*, 6]
+        update = self.linear(s)
+
+        return update
+
+
+class EsmFoldStructureModuleTransitionLayer(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+
+        self.linear_1 = EsmFoldLinear(config.sequence_dim, config.sequence_dim, init="relu")
+        self.linear_2 = EsmFoldLinear(config.sequence_dim, config.sequence_dim, init="relu")
+        self.linear_3 = EsmFoldLinear(config.sequence_dim, config.sequence_dim, init="final")
+
+        self.relu = nn.ReLU()
+
+    def forward(self, s):
+        s_initial = s
+        s = self.linear_1(s)
+        s = self.relu(s)
+        s = self.linear_2(s)
+        s = self.relu(s)
+        s = self.linear_3(s)
+
+        s = s + s_initial
+
+        return s
+
+
+class EsmFoldStructureModuleTransition(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+
+        self.layers = nn.ModuleList()
+        for _ in range(config.num_transition_layers):
+            l = EsmFoldStructureModuleTransitionLayer(config)
+            self.layers.append(l)
+
+        self.dropout = nn.Dropout(config.dropout_rate)
+        self.layer_norm = LayerNorm(config.sequence_dim)
+
+    def forward(self, s):
+        for l in self.layers:
+            s = l(s)
+
+        s = self.dropout(s)
+        s = self.layer_norm(s)
+
+        return s
+
+
+class EsmFoldStructureModule(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+
+        # Buffers to be lazily initialized later
+        # self.default_frames
+        # self.group_idx
+        # self.atom_mask
+        # self.lit_positions
+
+        self.layer_norm_s = LayerNorm(config.sequence_dim)
+        self.layer_norm_z = LayerNorm(config.pairwise_dim)
+
+        self.linear_in = EsmFoldLinear(config.sequence_dim, config.sequence_dim)
+
+        self.ipa = EsmFoldInvariantPointAttention(config)
+
+        self.ipa_dropout = nn.Dropout(config.dropout_rate)
+        self.layer_norm_ipa = LayerNorm(config.sequence_dim)
+
+        self.transition = EsmFoldStructureModuleTransition(config)
+        self.bb_update = EsmFoldBackboneUpdate(config)
+        self.angle_resnet = EsmFoldAngleResnet(config)
+
+    def forward(
+        self,
+        evoformer_output_dict,
+        aatype,
+        mask=None,
+        _offload_inference=False,
+    ):
+        """
+        Args:
+            evoformer_output_dict:
+                Dictionary containing:
+                    "single":
+                        [*, N_res, C_s] single representation
+                    "pair":
+                        [*, N_res, N_res, C_z] pair representation
+            aatype:
+                [*, N_res] amino acid indices
+            mask:
+                Optional [*, N_res] sequence mask
+        Returns:
+            A dictionary of outputs
+        """
+        s = evoformer_output_dict["single"]
+
+        if mask is None:
+            # [*, N]
+            mask = s.new_ones(s.shape[:-1])
+
+        # [*, N, C_s]
+        s = self.layer_norm_s(s)
+
+        # [*, N, N, C_z]
+        z = self.layer_norm_z(evoformer_output_dict["pair"])
+
+        z_reference_list = None
+        if _offload_inference:
+            assert sys.getrefcount(evoformer_output_dict["pair"]) == 2
+            evoformer_output_dict["pair"] = evoformer_output_dict["pair"].cpu()
+            z_reference_list = [z]
+            z = None
+
+        # [*, N, C_s]
+        s_initial = s
+        s = self.linear_in(s)
+
+        # [*, N]
+        rigids = Rigid.identity(
+            s.shape[:-1],
+            s.dtype,
+            s.device,
+            self.training,
+            fmt="quat",
+        )
+        outputs = []
+        for i in range(self.config.num_blocks):
+            # [*, N, C_s]
+            s = s + self.ipa(
+                s,
+                z,
+                rigids,
+                mask,
+                _offload_inference=_offload_inference,
+                _z_reference_list=z_reference_list,
+            )
+            s = self.ipa_dropout(s)
+            s = self.layer_norm_ipa(s)
+            s = self.transition(s)
+
+            # [*, N]
+            rigids = rigids.compose_q_update_vec(self.bb_update(s))
+
+            # To hew as closely as possible to AlphaFold, we convert our
+            # quaternion-based transformations to rotation-matrix ones
+            # here
+            backb_to_global = Rigid(
+                Rotation(rot_mats=rigids.get_rots().get_rot_mats(), quats=None),
+                rigids.get_trans(),
+            )
+
+            backb_to_global = backb_to_global.scale_translation(self.config.trans_scale_factor)
+
+            # [*, N, 7, 2]
+            unnormalized_angles, angles = self.angle_resnet(s, s_initial)
+
+            all_frames_to_global = self.torsion_angles_to_frames(backb_to_global, angles, aatype)
+
+            pred_xyz = self.frames_and_literature_positions_to_atom14_pos(all_frames_to_global, aatype)
+
+            scaled_rigids = rigids.scale_translation(self.config.trans_scale_factor)
+
+            preds = {
+                "frames": scaled_rigids.to_tensor_7(),
+                "sidechain_frames": all_frames_to_global.to_tensor_4x4(),
+                "unnormalized_angles": unnormalized_angles,
+                "angles": angles,
+                "positions": pred_xyz,
+                "states": s,
+            }
+
+            outputs.append(preds)
+
+            rigids = rigids.stop_rot_gradient()
+
+        del z, z_reference_list
+
+        if _offload_inference:
+            evoformer_output_dict["pair"] = evoformer_output_dict["pair"].to(s.device)
+
+        outputs = dict_multimap(torch.stack, outputs)
+        outputs["single"] = s
+
+        return outputs
+
+    def _init_residue_constants(self, float_dtype, device):
+        if not hasattr(self, "default_frames"):
+            self.register_buffer(
+                "default_frames",
+                torch.tensor(
+                    residue_constants.restype_rigid_group_default_frame,
+                    dtype=float_dtype,
+                    device=device,
+                    requires_grad=False,
+                ),
+                persistent=False,
+            )
+        if not hasattr(self, "group_idx"):
+            self.register_buffer(
+                "group_idx",
+                torch.tensor(
+                    residue_constants.restype_atom14_to_rigid_group,
+                    device=device,
+                    requires_grad=False,
+                ),
+                persistent=False,
+            )
+        if not hasattr(self, "atom_mask"):
+            self.register_buffer(
+                "atom_mask",
+                torch.tensor(
+                    residue_constants.restype_atom14_mask,
+                    dtype=float_dtype,
+                    device=device,
+                    requires_grad=False,
+                ),
+                persistent=False,
+            )
+        if not hasattr(self, "lit_positions"):
+            self.register_buffer(
+                "lit_positions",
+                torch.tensor(
+                    residue_constants.restype_atom14_rigid_group_positions,
+                    dtype=float_dtype,
+                    device=device,
+                    requires_grad=False,
+                ),
+                persistent=False,
+            )
+
+    def torsion_angles_to_frames(self, r, alpha, f):
+        # Lazily initialize the residue constants on the correct device
+        self._init_residue_constants(alpha.dtype, alpha.device)
+        # Separated purely to make testing less annoying
+        return torsion_angles_to_frames(r, alpha, f, self.default_frames)
+
+    def frames_and_literature_positions_to_atom14_pos(self, r, f):  # [*, N, 8]  # [*, N]
+        # Lazily initialize the residue constants on the correct device
+        self._init_residue_constants(r.get_rots().dtype, r.get_rots().device)
+        return frames_and_literature_positions_to_atom14_pos(
+            r,
+            f,
+            self.default_frames,
+            self.group_idx,
+            self.atom_mask,
+            self.lit_positions,
+        )
+
+
+class EsmFoldingTrunk(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+
+        c_s = config.sequence_state_dim
+        c_z = config.pairwise_state_dim
+
+        self.pairwise_positional_embedding = EsmFoldRelativePosition(config)
+
+        self.blocks = nn.ModuleList([EsmFoldTriangularSelfAttentionBlock(config) for _ in range(config.num_blocks)])
+
+        self.recycle_bins = 15
+        self.recycle_s_norm = nn.LayerNorm(c_s)
+        self.recycle_z_norm = nn.LayerNorm(c_z)
+        self.recycle_disto = nn.Embedding(self.recycle_bins, c_z)
+        self.recycle_disto.weight[0].detach().zero_()
+
+        self.structure_module = EsmFoldStructureModule(config.structure_module)
+        self.trunk2sm_s = nn.Linear(c_s, config.structure_module.sequence_dim)
+        self.trunk2sm_z = nn.Linear(c_z, config.structure_module.pairwise_dim)
+
+        self.chunk_size = config.chunk_size
+
+    def set_chunk_size(self, chunk_size):
+        # This parameter means the axial attention will be computed
+        # in a chunked manner. This should make the memory used more or less O(L) instead of O(L^2).
+        # It's equivalent to running a for loop over chunks of the dimension we're iterative over,
+        # where the chunk_size is the size of the chunks, so 128 would mean to parse 128-length chunks.
+        self.chunk_size = chunk_size
+
+    def forward(self, seq_feats, pair_feats, true_aa, residx, mask, no_recycles):
+        """
+        Inputs:
+          seq_feats: B x L x C tensor of sequence features pair_feats: B x L x L x C tensor of pair features residx: B
+          x L long tensor giving the position in the sequence mask: B x L boolean tensor indicating valid residues
+
+        Output:
+          predicted_structure: B x L x (num_atoms_per_residue * 3) tensor wrapped in a Coordinates object
+        """
+
+        device = seq_feats.device
+        s_s_0 = seq_feats
+        s_z_0 = pair_feats
+
+        if no_recycles is None:
+            no_recycles = self.config.max_recycles
+        else:
+            if no_recycles < 0:
+                raise ValueError("Number of recycles must not be negative.")
+            no_recycles += 1  # First 'recycle' is just the standard forward pass through the model.
+
+        def trunk_iter(s, z, residx, mask):
+            z = z + self.pairwise_positional_embedding(residx, mask=mask)
+
+            for block in self.blocks:
+                s, z = block(s, z, mask=mask, residue_index=residx, chunk_size=self.chunk_size)
+            return s, z
+
+        s_s = s_s_0
+        s_z = s_z_0
+        recycle_s = torch.zeros_like(s_s)
+        recycle_z = torch.zeros_like(s_z)
+        recycle_bins = torch.zeros(*s_z.shape[:-1], device=device, dtype=torch.int64)
+
+        for recycle_idx in range(no_recycles):
+            with ContextManagers([] if recycle_idx == no_recycles - 1 else [torch.no_grad()]):
+                # === Recycling ===
+                recycle_s = self.recycle_s_norm(recycle_s.detach()).to(device)
+                recycle_z = self.recycle_z_norm(recycle_z.detach()).to(device)
+                recycle_z += self.recycle_disto(recycle_bins.detach()).to(device)
+
+                s_s, s_z = trunk_iter(s_s_0 + recycle_s, s_z_0 + recycle_z, residx, mask)
+
+                # === Structure module ===
+                structure = self.structure_module(
+                    {"single": self.trunk2sm_s(s_s), "pair": self.trunk2sm_z(s_z)},
+                    true_aa,
+                    mask.float(),
+                )
+
+                recycle_s = s_s
+                recycle_z = s_z
+                # Distogram needs the N, CA, C coordinates, and bin constants same as alphafold.
+                recycle_bins = EsmFoldingTrunk.distogram(
+                    structure["positions"][-1][:, :, :3],
+                    3.375,
+                    21.375,
+                    self.recycle_bins,
+                )
+
+        structure["s_s"] = s_s
+        structure["s_z"] = s_z
+
+        return structure
+
+    @staticmethod
+    def distogram(coords, min_bin, max_bin, num_bins):
+        # Coords are [... L x 3 x 3], where it's [N, CA, C] x 3 coordinates.
+        boundaries = torch.linspace(
+            min_bin,
+            max_bin,
+            num_bins - 1,
+            device=coords.device,
+        )
+        boundaries = boundaries**2
+        N, CA, C = [x.squeeze(-2) for x in coords.chunk(3, dim=-2)]
+        # Infer CB coordinates.
+        b = CA - N
+        c = C - CA
+        a = b.cross(c, dim=-1)
+        CB = -0.58273431 * a + 0.56802827 * b - 0.54067466 * c + CA
+        dists = (CB[..., None, :, :] - CB[..., :, None, :]).pow(2).sum(dim=-1, keepdims=True)
+        bins = torch.sum(dists > boundaries, dim=-1)  # [..., L, L]
+        return bins
+
+
+# TODO Add information to the docstring about any methods that convert to PDB format, or otherwise prepare
+#      the outputs for downstream use.
+
+
+@add_start_docstrings(
+    """
+    ESMForProteinFolding is the HuggingFace port of the original ESMFold model. It consists of an ESM-2 "stem" followed
+    by a protein folding "head", although unlike most other output heads, this "head" is similar in size and runtime to
+    the rest of the model combined! It outputs a dictionary containing predicted structural information about the input
+    protein(s).
+    """,
+    ESM_START_DOCSTRING,
+)
+class EsmForProteinFolding(EsmPreTrainedModel):
+    _no_split_modules = ["EsmFoldStructureModule", "EsmFoldTriangularSelfAttentionBlock"]
+
+    def __init__(self, config):
+        super().__init__(config)
+
+        self.config = config
+
+        self.distogram_bins = 64
+
+        self.esm = EsmModel(config, add_pooling_layer=False)
+
+        self.esm.requires_grad_(False)
+        if self.config.esmfold_config.fp16_esm:
+            self.esm.half()
+
+        self.esm_feats = self.config.hidden_size
+        self.esm_attns = self.config.num_hidden_layers * self.config.num_attention_heads
+        self.esm_layers = self.config.num_hidden_layers
+        self.register_buffer("af2_to_esm", self._af2_to_esm_from_vocab_list(config.vocab_list))
+        self.esm_s_combine = nn.Parameter(torch.zeros(self.esm_layers + 1))
+
+        trunk_config = self.config.esmfold_config.trunk
+        c_s = trunk_config.sequence_state_dim
+        c_z = trunk_config.pairwise_state_dim
+        self.esm_s_mlp = nn.Sequential(
+            LayerNorm(self.esm_feats),
+            nn.Linear(self.esm_feats, c_s),
+            nn.ReLU(),
+            nn.Linear(c_s, c_s),
+        )
+
+        # 0 is padding, N is unknown residues, N + 1 is mask.
+        self.n_tokens_embed = residue_constants.restype_num + 3
+        self.pad_idx = 0
+        self.unk_idx = self.n_tokens_embed - 2
+        self.mask_idx = self.n_tokens_embed - 1
+        self.esm_dict_cls_idx = self.config.vocab_list.index("<cls>")
+        self.esm_dict_mask_idx = self.config.vocab_list.index("<mask>")
+        self.esm_dict_eos_idx = self.config.vocab_list.index("<eos>")
+        self.esm_dict_padding_idx = self.config.vocab_list.index("<pad>")
+        if self.config.esmfold_config.embed_aa:
+            self.embedding = nn.Embedding(self.n_tokens_embed, c_s, padding_idx=0)
+
+        self.trunk = EsmFoldingTrunk(trunk_config)
+
+        self.distogram_head = nn.Linear(c_z, self.distogram_bins)
+        self.ptm_head = nn.Linear(c_z, self.distogram_bins)
+        self.lm_head = nn.Linear(c_s, self.n_tokens_embed)
+        self.lddt_bins = 50
+        structure_module_config = trunk_config.structure_module
+        self.lddt_head = nn.Sequential(
+            nn.LayerNorm(structure_module_config.sequence_dim),
+            nn.Linear(structure_module_config.sequence_dim, self.config.esmfold_config.lddt_head_hid_dim),
+            nn.Linear(self.config.esmfold_config.lddt_head_hid_dim, self.config.esmfold_config.lddt_head_hid_dim),
+            nn.Linear(self.config.esmfold_config.lddt_head_hid_dim, 37 * self.lddt_bins),
+        )
+
+    @staticmethod
+    def _af2_to_esm_from_vocab_list(vocab_list: List[str]) -> torch.Tensor:
+        # Remember that t is shifted from residue_constants by 1 (0 is padding).
+        esm_reorder = [vocab_list.index("<pad>")] + [vocab_list.index(v) for v in residue_constants.restypes_with_x]
+        return torch.tensor(esm_reorder)
+
+    @add_start_docstrings_to_model_forward(ESMFOLD_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @replace_return_docstrings(output_type=EsmForProteinFoldingOutput, config_class=EsmConfig)
+    def forward(
+        self,
+        input_ids: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        masking_pattern: Optional[torch.Tensor] = None,
+        num_recycles: Optional[int] = None,
+    ) -> EsmForProteinFoldingOutput:
+        r"""
+        Returns:
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, EsmForProteinFolding
+
+        >>> model = EsmForProteinFolding.from_pretrained("facebook/esmfold_v1")
+        >>> tokenizer = AutoTokenizer.from_pretrained("facebook/esmfold_v1")
+        >>> inputs = tokenizer(["MLKNVQVQLV"], return_tensors="pt", add_special_tokens=False)  # A tiny random peptide
+        >>> outputs = model(**inputs)
+        >>> folded_positions = outputs.positions
+        ```
+
+        """
+        cfg = self.config.esmfold_config
+
+        aa = input_ids  # B x L
+        B = aa.shape[0]
+        L = aa.shape[1]
+        device = input_ids.device
+        if attention_mask is None:
+            attention_mask = torch.ones_like(aa, device=device)
+        if position_ids is None:
+            position_ids = torch.arange(L, device=device).expand_as(input_ids)
+
+        # === ESM ===
+        esmaa = self.af2_idx_to_esm_idx(aa, attention_mask)
+
+        if masking_pattern is not None:
+            masked_aa, esmaa, mlm_targets = self.bert_mask(aa, esmaa, attention_mask, masking_pattern)
+        else:
+            masked_aa = aa
+            mlm_targets = None
+
+        # We get sequence and pair representations from whatever version of ESM /
+        # configuration we are using. The sequence representation esm_s is always
+        # present. The pair embedding esm_z may be present depending on the
+        # configuration of the model. If esm_z is not used by the model then it
+        # is returned as None here.
+        esm_s = self.compute_language_model_representations(esmaa)
+
+        # Convert esm_s and esm_z, if present, to the precision used by the trunk and
+        # the structure module. These tensors may be a lower precision if, for example,
+        # we're running the language model in fp16 precision.
+        esm_s = esm_s.to(self.esm_s_combine.dtype)
+
+        if cfg.esm_ablate_sequence:
+            esm_s = esm_s * 0
+
+        esm_s = esm_s.detach()
+
+        # === preprocessing ===
+        esm_s = (self.esm_s_combine.softmax(0).unsqueeze(0) @ esm_s).squeeze(2)
+        s_s_0 = self.esm_s_mlp(esm_s)
+
+        s_z_0 = s_s_0.new_zeros(B, L, L, cfg.trunk.pairwise_state_dim)
+
+        if self.config.esmfold_config.embed_aa:
+            s_s_0 += self.embedding(masked_aa)
+
+        structure: dict = self.trunk(s_s_0, s_z_0, aa, position_ids, attention_mask, no_recycles=num_recycles)
+        # Documenting what we expect:
+        structure = {
+            k: v
+            for k, v in structure.items()
+            if k
+            in [
+                "s_z",
+                "s_s",
+                "frames",
+                "sidechain_frames",
+                "unnormalized_angles",
+                "angles",
+                "positions",
+                "states",
+            ]
+        }
+
+        # Add BERT mask for the loss to use, if available.
+        if mlm_targets:
+            structure["mlm_targets"] = mlm_targets
+
+        disto_logits = self.distogram_head(structure["s_z"])
+        disto_logits = (disto_logits + disto_logits.transpose(1, 2)) / 2
+        structure["distogram_logits"] = disto_logits
+
+        lm_logits = self.lm_head(structure["s_s"])
+        structure["lm_logits"] = lm_logits
+
+        structure["aatype"] = aa
+        make_atom14_masks(structure)
+        # Of course, this doesn't respect the true mask because it doesn't know about it...
+        # We're not going to properly mask change of index tensors:
+        #    "residx_atom14_to_atom37",
+        #    "residx_atom37_to_atom14",
+        for k in [
+            "atom14_atom_exists",
+            "atom37_atom_exists",
+        ]:
+            structure[k] *= attention_mask.unsqueeze(-1)
+        structure["residue_index"] = position_ids
+
+        lddt_head = self.lddt_head(structure["states"]).reshape(structure["states"].shape[0], B, L, -1, self.lddt_bins)
+        structure["lddt_head"] = lddt_head
+        plddt = categorical_lddt(lddt_head[-1], bins=self.lddt_bins)
+        structure["plddt"] = plddt
+
+        ptm_logits = self.ptm_head(structure["s_z"])
+        structure["ptm_logits"] = ptm_logits
+        structure["ptm"] = compute_tm(ptm_logits, max_bin=31, no_bins=self.distogram_bins)
+        structure.update(compute_predicted_aligned_error(ptm_logits, max_bin=31, no_bins=self.distogram_bins))
+
+        return EsmForProteinFoldingOutput(**structure)
+
+    def af2_idx_to_esm_idx(self, aa, mask):
+        # avoid indexing on different devices
+        if self.af2_to_esm.device != aa.device:
+            self.af2_to_esm = self.af2_to_esm.to(aa.device)
+        aa = (aa + 1).masked_fill(mask != 1, 0)
+        return self.af2_to_esm[aa]
+
+    def compute_language_model_representations(self, esmaa: torch.Tensor) -> torch.Tensor:
+        device = next(self.parameters()).device
+        B, L = esmaa.shape  # B = batch size, L = sequence length.
+
+        if self.config.esmfold_config.bypass_lm:
+            esm_s = torch.zeros(B, L, self.esm_s_combine.size[0], -1, self.esm_feats, device=device)
+            return esm_s
+
+        bosi, eosi = self.esm_dict_cls_idx, self.esm_dict_eos_idx
+        bos = esmaa.new_full((B, 1), bosi)
+        eos = esmaa.new_full((B, 1), self.esm_dict_padding_idx)
+        esmaa = torch.cat([bos, esmaa, eos], dim=1)
+        # Use the first padding index as eos during inference.
+        esmaa[range(B), (esmaa != 1).sum(1)] = eosi
+
+        # _, esm_z, esm_s = self.esm(esmaa, return_pairs=self.config.esmfold_config.use_esm_attn_map)
+        # Because we do not support use_esm_attn_map in the HF port as it is not used in any public models,
+        # esm_z is always None
+        esm_hidden_states = self.esm(esmaa, attention_mask=esmaa != 1, output_hidden_states=True)["hidden_states"]
+        esm_s = torch.stack(esm_hidden_states, dim=2)
+
+        esm_s = esm_s[:, 1:-1]  # B, L, nLayers, C
+
+        return esm_s
+
+    def bert_mask(self, aa, esmaa, mask, pattern):
+        new_aa = aa.clone()
+        target = aa.clone()
+        new_esmaa = esmaa.clone()
+        new_aa[pattern == 1] = self.mask_idx
+        target[pattern != 1] = 0
+        new_esmaa[pattern == 1] = self.esm_dict_mask_idx
+        return new_aa, new_esmaa, target
+
+    @torch.no_grad()
+    def infer(
+        self,
+        seqs: Union[str, List[str]],
+        position_ids=None,
+    ):
+        if isinstance(seqs, str):
+            lst = [seqs]
+        else:
+            lst = seqs
+        # Returns the raw outputs of the model given an input sequence.
+        device = next(self.parameters()).device
+        aatype = collate_dense_tensors(
+            [
+                torch.from_numpy(
+                    residue_constants.sequence_to_onehot(
+                        sequence=seq,
+                        mapping=residue_constants.restype_order_with_x,
+                        map_unknown_to_x=True,
+                    )
+                )
+                .to(device)
+                .argmax(dim=1)
+                for seq in lst
+            ]
+        )  # B=1 x L
+        mask = collate_dense_tensors([aatype.new_ones(len(seq)) for seq in lst])
+        position_ids = (
+            torch.arange(aatype.shape[1], device=device).expand(len(lst), -1)
+            if position_ids is None
+            else position_ids.to(device)
+        )
+        if position_ids.ndim == 1:
+            position_ids = position_ids.unsqueeze(0)
+        return self.forward(
+            aatype,
+            mask,
+            position_ids=position_ids,
+        )
+
+    @staticmethod
+    def output_to_pdb(output: Dict) -> List[str]:
+        """Returns the pbd (file) string from the model given the model output."""
+        output = {k: v.to("cpu").numpy() for k, v in output.items()}
+        pdbs = []
+        final_atom_positions = atom14_to_atom37(output["positions"][-1], output)
+        final_atom_mask = output["atom37_atom_exists"]
+        for i in range(output["aatype"].shape[0]):
+            aa = output["aatype"][i]
+            pred_pos = final_atom_positions[i]
+            mask = final_atom_mask[i]
+            resid = output["residue_index"][i] + 1
+            pred = OFProtein(
+                aatype=aa,
+                atom_positions=pred_pos,
+                atom_mask=mask,
+                residue_index=resid,
+                b_factors=output["plddt"][i],
+            )
+            pdbs.append(to_pdb(pred))
+        return pdbs
+
+    def infer_pdb(self, seqs, *args, **kwargs) -> str:
+        """Returns the pdb (file) string from the model given an input sequence."""
+        assert isinstance(seqs, str)
+        output = self.infer(seqs, *args, **kwargs)
+        return self.output_to_pdb(output)[0]
+
+    def infer_pdbs(self, seqs: List[str], *args, **kwargs) -> List[str]:
+        """Returns the pdb (file) string from the model given an input sequence."""
+        output = self.infer(seqs, *args, **kwargs)
+        return self.output_to_pdb(output)

venv/lib/python3.10/site-packages/transformers/models/esm/modeling_tf_esm.py

@@ -0,0 +1,1567 @@
+# coding=utf-8
+# Copyright 2022 Meta and The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" PyTorch ESM model."""
+
+
+from __future__ import annotations
+
+import os
+from typing import Optional, Tuple, Union
+
+import numpy as np
+import tensorflow as tf
+
+from ...file_utils import add_code_sample_docstrings, add_start_docstrings, add_start_docstrings_to_model_forward
+from ...modeling_tf_outputs import (
+    TFBaseModelOutputWithPastAndCrossAttentions,
+    TFBaseModelOutputWithPoolingAndCrossAttentions,
+    TFMaskedLMOutput,
+    TFSequenceClassifierOutput,
+    TFTokenClassifierOutput,
+)
+from ...modeling_tf_utils import (
+    TFMaskedLanguageModelingLoss,
+    TFModelInputType,
+    TFPreTrainedModel,
+    TFSequenceClassificationLoss,
+    TFTokenClassificationLoss,
+    get_initializer,
+    keras,
+    shape_list,
+    unpack_inputs,
+)
+from ...tf_utils import check_embeddings_within_bounds, stable_softmax
+from ...utils import logging
+from .configuration_esm import EsmConfig
+
+
+logger = logging.get_logger(__name__)
+
+_CHECKPOINT_FOR_DOC = "facebook/esm2_t6_8M_UR50D"
+_CONFIG_FOR_DOC = "EsmConfig"
+
+
+def rotate_half(x):
+    x1, x2 = tf.split(x, 2, axis=-1)
+    return tf.concat((-x2, x1), axis=-1)
+
+
+def apply_rotary_pos_emb(x, cos, sin):
+    cos = cos[:, :, : tf.shape(x)[-2], :]
+    sin = sin[:, :, : tf.shape(x)[-2], :]
+
+    return (x * cos) + (rotate_half(x) * sin)
+
+
+def symmetrize(x):
+    "Make layer symmetric in final two dimensions, used for contact prediction."
+    return x + tf.linalg.matrix_transpose(x)  # Transposes last two dimensions only
+
+
+def average_product_correct(x):
+    "Perform average product correct, used for contact prediction."
+    a1 = tf.reduce_sum(x, -1, keepdims=True)
+    a2 = tf.reduce_sum(x, -2, keepdims=True)
+    a12 = tf.reduce_sum(x, (-1, -2), keepdims=True)
+
+    avg = a1 * a2
+    avg = avg / a12
+    normalized = x - avg
+    return normalized
+
+
+class TFRotaryEmbedding(keras.layers.Layer):
+    """
+    Rotary position embeddings based on those in
+    [RoFormer](https://huggingface.co/docs/transformers/model_doc/roformer). Query and keys are transformed by rotation
+    matrices which depend on their relative positions.
+    """
+
+    def __init__(self, dim: int, name=None):
+        super().__init__(name=name)
+        # Matt: The PyTorch version of this layer does a lot of work to cache values, but we just rely on TF compilation
+        # and/or XLA to sort out constants like that. It actually may not seem like this layer needs to be stateful at
+        # all when we benefit from TF compilation, but it does. The reason is that self.inv_freq is a buffer in the
+        # original implementation, but all the shared ESM checkpoints were trained with fp16 params. This means that
+        # the inv_freq tensor was stored as a float16, and we need to replicate those lower-precision values or our
+        # models give different outputs from the original.
+        self.dim = dim
+
+    def build(self, input_shape):
+        super().build(input_shape)
+        self.inv_freq = self.add_weight(
+            "inv_freq", shape=(self.dim // 2,), dtype=tf.float32, initializer=get_initializer(1.0), trainable=False
+        )
+        self.inv_freq.assign(
+            1.0 / (10000 ** (tf.range(start=0, limit=self.dim, delta=2, dtype=tf.float32) / self.dim))
+        )
+
+    def _compute_cos_sin(self, x, seq_dimension=2):
+        seq_len = tf.shape(x)[seq_dimension]
+
+        t = tf.range(seq_len, dtype=self.inv_freq.dtype)
+        freqs = tf.einsum("i, j -> ij", t, self.inv_freq)  # Outer multiplication
+        emb = tf.concat((freqs, freqs), axis=-1)[None, None, :, :]
+
+        return tf.cos(emb), tf.sin(emb)
+
+    def call(self, q: tf.Tensor, k: tf.Tensor) -> Tuple[tf.Tensor, tf.Tensor]:
+        cos_emb, sin_emb = self._compute_cos_sin(k, seq_dimension=-2)
+
+        return (
+            apply_rotary_pos_emb(q, cos_emb, sin_emb),
+            apply_rotary_pos_emb(k, cos_emb, sin_emb),
+        )
+
+
+class TFEsmContactPredictionHead(keras.layers.Layer):
+    """Performs symmetrization, apc, and computes a logistic regression on the output features"""
+
+    def __init__(
+        self,
+        in_features: int,
+        bias=True,
+        eos_idx: int = 2,
+        name=None,
+    ):
+        super().__init__(name=name)
+        self.eos_idx = eos_idx
+        self.in_features = in_features
+        self.regression = keras.layers.Dense(1, use_bias=bias, activation="sigmoid", name="regression")
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "regression", None) is not None:
+            with tf.name_scope(self.regression.name):
+                self.regression.build((None, self.in_features))
+
+    def call(self, tokens, attentions):
+        # remove eos token attentions
+        eos_mask = tf.cast(tokens != self.eos_idx, attentions.dtype)
+        eos_mask = tf.expand_dims(eos_mask, 1) * tf.expand_dims(eos_mask, 2)
+        attentions = attentions * eos_mask[:, None, None, :, :]
+        attentions = attentions[..., :-1, :-1]
+        # remove cls token attentions
+        attentions = attentions[..., 1:, 1:]
+        batch_size, layers, heads, seqlen, _ = shape_list(attentions)
+        attentions = tf.reshape(attentions, (batch_size, layers * heads, seqlen, seqlen))
+
+        # features: batch x channels x tokens x tokens (symmetric)
+        attentions = average_product_correct(symmetrize(attentions))
+        attentions = tf.transpose(attentions, perm=(0, 2, 3, 1))
+        return tf.squeeze(self.regression(attentions), 3)
+
+
+class TFEsmEmbeddings(keras.layers.Layer):
+    """
+    Same as BertEmbeddings with a tiny tweak for positional embeddings indexing.
+    """
+
+    def __init__(self, config, name=None):
+        super().__init__(name=name)
+        self.word_embeddings = keras.layers.Embedding(
+            config.vocab_size,
+            config.hidden_size,
+            embeddings_initializer=get_initializer(config.initializer_range),
+            name="word_embeddings",
+        )
+        self.position_embeddings = keras.layers.Embedding(
+            config.max_position_embeddings,
+            config.hidden_size,
+            embeddings_initializer=get_initializer(config.initializer_range),
+            name="position_embeddings",
+        )
+
+        if config.emb_layer_norm_before:
+            self.layer_norm = keras.layers.LayerNormalization(epsilon=config.layer_norm_eps, name="layer_norm")
+        else:
+            self.layer_norm = None
+        # Matt: I think this line was copied incorrectly from BERT, disabling for now
+        # self.dropout = Dropout(config.hidden_dropout_prob)
+        self.position_embedding_type = getattr(config, "position_embedding_type", "absolute")
+
+        self.position_ids = tf.range(config.max_position_embeddings)[None, :]
+
+        self.padding_idx = config.pad_token_id
+        self.token_dropout = config.token_dropout
+        self.mask_token_id = config.mask_token_id
+        self.config = config
+
+    def call(
+        self, input_ids=None, attention_mask=None, position_ids=None, inputs_embeds=None, past_key_values_length=0
+    ):
+        if position_ids is None:
+            if input_ids is not None:
+                # Create the position ids from the input token ids. Any padded tokens remain padded.
+                position_ids = create_position_ids_from_input_ids(input_ids, self.padding_idx, past_key_values_length)
+            else:
+                position_ids = self.create_position_ids_from_inputs_embeds(inputs_embeds)
+
+        if inputs_embeds is None:
+            check_embeddings_within_bounds(input_ids, self.config.vocab_size)
+            inputs_embeds = self.word_embeddings(input_ids)
+
+        # Note that if we want to support ESM-1 (not 1b!) in future then we need to support an
+        # embedding_scale factor here.
+        embeddings = inputs_embeds
+
+        # Matt: ESM has the option to handle masking in MLM in a slightly unusual way. If the token_dropout
+        # flag is False then it is handled in the same was as BERT/RoBERTa. If it is set to True, however,
+        # masked tokens are treated as if they were selected for input dropout and zeroed out.
+        # This "mask-dropout" is compensated for when masked tokens are not present, by scaling embeddings by
+        # a factor of (fraction of unmasked tokens during training) / (fraction of unmasked tokens in sample).
+        # This is analogous to the way that dropout layers scale down outputs during evaluation when not
+        # actually dropping out values (or, equivalently, scale up their un-dropped outputs in training).
+        if self.token_dropout:
+            embeddings = tf.where((input_ids == self.mask_token_id)[:, :, None], 0.0, embeddings)
+            mask_ratio_train = 0.15 * 0.8  # Hardcoded as the ratio used in all ESM model training runs
+            src_lengths = tf.cast(tf.reduce_sum(attention_mask, axis=-1), tf.float32)
+            masked_tokens = input_ids == self.mask_token_id
+            mask_ratio_observed = tf.math.count_nonzero(masked_tokens, dtype=tf.float32, axis=-1) / src_lengths
+            embeddings = embeddings * (1 - mask_ratio_train) / (1 - mask_ratio_observed)[:, None, None]
+
+        if self.position_embedding_type == "absolute":
+            position_embeddings = self.position_embeddings(position_ids)
+            embeddings += position_embeddings
+
+        if self.layer_norm is not None:
+            embeddings = self.layer_norm(embeddings)
+        if attention_mask is not None:
+            embeddings = embeddings * tf.cast(tf.expand_dims(attention_mask, -1), embeddings.dtype)
+        # Matt: I think this line was copied incorrectly from BERT, disabling it for now.
+        # embeddings = self.dropout(embeddings)
+        return embeddings
+
+    def create_position_ids_from_inputs_embeds(self, inputs_embeds):
+        """
+        We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids.
+
+        Args:
+            inputs_embeds: tf.Tensor
+
+        Returns: tf.Tensor
+        """
+        input_shape = shape_list(inputs_embeds)[:-1]
+        sequence_length = input_shape[1]
+
+        position_ids = tf.range(
+            start=self.padding_idx + 1, limit=sequence_length + self.padding_idx + 1, dtype=tf.int64
+        )
+        return tf.broadcast_to(tf.expand_dims(position_ids, 0), input_shape)
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "word_embeddings", None) is not None:
+            with tf.name_scope(self.word_embeddings.name):
+                self.word_embeddings.build(None)
+        if getattr(self, "position_embeddings", None) is not None:
+            with tf.name_scope(self.position_embeddings.name):
+                self.position_embeddings.build(None)
+        if getattr(self, "layer_norm", None) is not None:
+            with tf.name_scope(self.layer_norm.name):
+                self.layer_norm.build([None, None, self.config.hidden_size])
+
+
+class TFEsmSelfAttention(keras.layers.Layer):
+    def __init__(self, config, position_embedding_type=None, name=None):
+        super().__init__(name=name)
+        if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"):
+            raise ValueError(
+                f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
+                f"heads ({config.num_attention_heads})"
+            )
+
+        self.num_attention_heads = config.num_attention_heads
+        self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
+        self.all_head_size = self.num_attention_heads * self.attention_head_size
+
+        self.query = keras.layers.Dense(
+            self.all_head_size, kernel_initializer=get_initializer(config.initializer_range), name="query"
+        )
+        self.key = keras.layers.Dense(
+            self.all_head_size, kernel_initializer=get_initializer(config.initializer_range), name="key"
+        )
+        self.value = keras.layers.Dense(
+            self.all_head_size, kernel_initializer=get_initializer(config.initializer_range), name="value"
+        )
+
+        self.dropout = keras.layers.Dropout(config.attention_probs_dropout_prob)
+        self.position_embedding_type = position_embedding_type or getattr(
+            config, "position_embedding_type", "absolute"
+        )
+        self.rotary_embeddings = None
+        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
+            self.max_position_embeddings = config.max_position_embeddings
+            self.distance_embedding = keras.layers.Embedding(
+                2 * config.max_position_embeddings - 1,
+                self.attention_head_size,
+                embeddings_initializer=get_initializer(config.initializer_range),
+            )
+        elif self.position_embedding_type == "rotary":
+            self.rotary_embeddings = TFRotaryEmbedding(dim=self.attention_head_size, name="rotary_embeddings")
+
+        self.is_decoder = config.is_decoder
+        self.config = config
+
+    def transpose_for_scores(self, x: tf.Tensor) -> tf.Tensor:
+        new_x_shape = shape_list(x)[:-1] + [self.num_attention_heads, self.attention_head_size]
+        x = tf.reshape(x, new_x_shape)
+        return tf.transpose(x, perm=(0, 2, 1, 3))
+
+    def call(
+        self,
+        hidden_states: tf.Tensor,
+        attention_mask: tf.Tensor | None = None,
+        head_mask: tf.Tensor | None = None,
+        encoder_hidden_states: tf.Tensor | None = None,
+        encoder_attention_mask: tf.Tensor | None = None,
+        past_key_value: Tuple[Tuple[tf.Tensor]] | None = None,
+        output_attentions: Optional[bool] = False,
+        training: bool = False,
+    ) -> Tuple[tf.Tensor]:
+        mixed_query_layer = self.query(hidden_states)
+
+        # If this is instantiated as a cross-attention module, the keys
+        # and values come from an encoder; the attention mask needs to be
+        # such that the encoder's padding tokens are not attended to.
+        is_cross_attention = encoder_hidden_states is not None
+
+        if is_cross_attention and past_key_value is not None:
+            # reuse k,v, cross_attentions
+            key_layer = past_key_value[0]
+            value_layer = past_key_value[1]
+            attention_mask = encoder_attention_mask
+        elif is_cross_attention:
+            key_layer = self.transpose_for_scores(self.key(encoder_hidden_states))
+            value_layer = self.transpose_for_scores(self.value(encoder_hidden_states))
+            attention_mask = encoder_attention_mask
+        elif past_key_value is not None:
+            key_layer = self.transpose_for_scores(self.key(hidden_states))
+            value_layer = self.transpose_for_scores(self.value(hidden_states))
+            key_layer = tf.concat([past_key_value[0], key_layer], axis=2)
+            value_layer = tf.concat([past_key_value[1], value_layer], axis=2)
+        else:
+            key_layer = self.transpose_for_scores(self.key(hidden_states))
+            value_layer = self.transpose_for_scores(self.value(hidden_states))
+
+        query_layer = self.transpose_for_scores(mixed_query_layer)
+
+        # Matt: Our BERT model (which this code was derived from) scales attention logits down by sqrt(head_dim).
+        # ESM scales the query down by the same factor instead. Modulo numerical stability these are equivalent,
+        # but not when rotary embeddings get involved. Therefore, we scale the query here to match the original
+        # ESM code and fix rotary embeddings.
+        query_layer = query_layer * self.attention_head_size**-0.5
+
+        if self.is_decoder:
+            # if cross_attention save Tuple(tf.Tensor, tf.Tensor) of all cross attention key/value_states.
+            # Further calls to cross_attention layer can then reuse all cross-attention
+            # key/value_states (first "if" case)
+            # if uni-directional self-attention (decoder) save Tuple(tf.Tensor, tf.Tensor) of
+            # all previous decoder key/value_states. Further calls to uni-directional self-attention
+            # can concat previous decoder key/value_states to current projected key/value_states (third "elif" case)
+            # if encoder bi-directional self-attention `past_key_value` is always `None`
+            past_key_value = (key_layer, value_layer)
+
+        if self.position_embedding_type == "rotary":
+            query_layer, key_layer = self.rotary_embeddings(query_layer, key_layer)
+
+        # Take the dot product between "query" and "key" to get the raw attention scores.
+        attention_scores = tf.matmul(query_layer, key_layer, transpose_b=True)
+
+        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
+            seq_length = shape_list(hidden_states)[1]
+            position_ids_l = tf.expand_dims(tf.range(seq_length, dtype=tf.int64), -1)
+            position_ids_r = tf.expand_dims(tf.range(seq_length, dtype=tf.int64), 0)
+            distance = position_ids_l - position_ids_r
+            positional_embedding = self.distance_embedding(distance + self.max_position_embeddings - 1)
+            positional_embedding = tf.cast(positional_embedding, query_layer.dtype)  # fp16 compatibility
+
+            if self.position_embedding_type == "relative_key":
+                relative_position_scores = tf.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
+                attention_scores = attention_scores + relative_position_scores
+            elif self.position_embedding_type == "relative_key_query":
+                relative_position_scores_query = tf.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
+                relative_position_scores_key = tf.einsum("bhrd,lrd->bhlr", key_layer, positional_embedding)
+                attention_scores = attention_scores + relative_position_scores_query + relative_position_scores_key
+
+        if attention_mask is not None:
+            # Apply the attention mask is (precomputed for all layers in EsmModel forward() function)
+            attention_scores = attention_scores + attention_mask
+
+        # Normalize the attention scores to probabilities.
+        attention_probs = stable_softmax(attention_scores, axis=-1)
+
+        # This is actually dropping out entire tokens to attend to, which might
+        # seem a bit unusual, but is taken from the original Transformer paper.
+        attention_probs = self.dropout(attention_probs, training=training)
+
+        # Mask heads if we want to
+        if head_mask is not None:
+            attention_probs = attention_probs * head_mask
+
+        context_layer = attention_probs @ value_layer
+
+        context_layer = tf.transpose(context_layer, perm=(0, 2, 1, 3))
+        new_context_layer_shape = shape_list(context_layer)[:-2] + [self.all_head_size]
+        context_layer = tf.reshape(context_layer, new_context_layer_shape)
+
+        outputs = (context_layer, attention_probs) if output_attentions else (context_layer,)
+
+        if self.is_decoder:
+            outputs = outputs + (past_key_value,)
+        return outputs
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "query", None) is not None:
+            with tf.name_scope(self.query.name):
+                self.query.build([None, None, self.config.hidden_size])
+        if getattr(self, "key", None) is not None:
+            with tf.name_scope(self.key.name):
+                self.key.build([None, None, self.config.hidden_size])
+        if getattr(self, "value", None) is not None:
+            with tf.name_scope(self.value.name):
+                self.value.build([None, None, self.config.hidden_size])
+        if getattr(self, "rotary_embeddings", None) is not None:
+            with tf.name_scope(self.rotary_embeddings.name):
+                self.rotary_embeddings.build(None)
+
+
+class TFEsmSelfOutput(keras.layers.Layer):
+    def __init__(self, config, name=None):
+        super().__init__(name=name)
+        self.dense = keras.layers.Dense(
+            config.hidden_size, kernel_initializer=get_initializer(config.initializer_range), name="dense"
+        )
+        self.dropout = keras.layers.Dropout(config.hidden_dropout_prob)
+        self.config = config
+
+    def call(self, hidden_states, input_tensor, training=False):
+        hidden_states = self.dense(hidden_states)
+        hidden_states = self.dropout(hidden_states, training=training)
+        hidden_states += input_tensor
+        return hidden_states
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "dense", None) is not None:
+            with tf.name_scope(self.dense.name):
+                self.dense.build([None, None, self.config.hidden_size])
+
+
+class TFEsmAttention(keras.layers.Layer):
+    def __init__(self, config, name=None):
+        super().__init__(name=name)
+        self.self = TFEsmSelfAttention(config, name="self")
+        self.output_layer = TFEsmSelfOutput(config, name="output")
+        self.pruned_heads = set()
+        self.LayerNorm = keras.layers.LayerNormalization(epsilon=config.layer_norm_eps, name="LayerNorm")
+        self.config = config
+
+    def prune_heads(self, heads):
+        raise NotImplementedError
+
+    def call(
+        self,
+        hidden_states,
+        attention_mask=None,
+        head_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_value=None,
+        output_attentions=False,
+        training=False,
+    ):
+        hidden_states_ln = self.LayerNorm(hidden_states)
+        self_outputs = self.self(
+            hidden_states_ln,
+            attention_mask,
+            head_mask,
+            encoder_hidden_states,
+            encoder_attention_mask,
+            past_key_value,
+            output_attentions,
+            training,
+        )
+        attention_output = self.output_layer(self_outputs[0], hidden_states)
+        outputs = (attention_output,) + self_outputs[1:]  # add attentions if we output them
+        return outputs
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "self", None) is not None:
+            with tf.name_scope(self.self.name):
+                self.self.build(None)
+        if getattr(self, "output_layer", None) is not None:
+            with tf.name_scope(self.output_layer.name):
+                self.output_layer.build(None)
+        if getattr(self, "LayerNorm", None) is not None:
+            with tf.name_scope(self.LayerNorm.name):
+                self.LayerNorm.build([None, None, self.config.hidden_size])
+
+
+class TFEsmIntermediate(keras.layers.Layer):
+    def __init__(self, config: EsmConfig, **kwargs):
+        super().__init__(**kwargs)
+
+        self.dense = keras.layers.Dense(
+            units=config.intermediate_size,
+            kernel_initializer=get_initializer(config.initializer_range),
+            name="dense",
+        )
+        self.config = config
+
+    def call(self, hidden_states: tf.Tensor) -> tf.Tensor:
+        hidden_states = self.dense(inputs=hidden_states)
+        hidden_states = tf.nn.gelu(hidden_states)
+        return hidden_states
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "dense", None) is not None:
+            with tf.name_scope(self.dense.name):
+                self.dense.build([None, None, self.config.hidden_size])
+
+
+class TFEsmOutput(keras.layers.Layer):
+    def __init__(self, config, name=None):
+        super().__init__(name=name)
+        self.dense = keras.layers.Dense(
+            config.hidden_size, kernel_initializer=get_initializer(config.initializer_range), name="dense"
+        )
+        self.dropout = keras.layers.Dropout(config.hidden_dropout_prob)
+        self.config = config
+
+    def call(self, hidden_states, input_tensor, training=False):
+        hidden_states = self.dense(hidden_states)
+        hidden_states = self.dropout(hidden_states, training=training)
+        hidden_states += input_tensor
+        return hidden_states
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "dense", None) is not None:
+            with tf.name_scope(self.dense.name):
+                self.dense.build([None, None, self.config.intermediate_size])
+
+
+class TFEsmLayer(keras.layers.Layer):
+    def __init__(self, config, name=None):
+        super().__init__(name=name)
+        self.chunk_size_feed_forward = config.chunk_size_feed_forward
+        self.seq_len_dim = 1
+        self.attention = TFEsmAttention(config, name="attention")
+        self.is_decoder = config.is_decoder
+        self.add_cross_attention = config.add_cross_attention
+        if self.add_cross_attention:
+            if not self.is_decoder:
+                raise RuntimeError(f"{self} should be used as a decoder model if cross attention is added")
+            self.crossattention = TFEsmAttention(config)
+        self.intermediate = TFEsmIntermediate(config, name="intermediate")
+        self.output_layer = TFEsmOutput(config, name="output")
+        self.LayerNorm = keras.layers.LayerNormalization(epsilon=config.layer_norm_eps, name="LayerNorm")
+        self.config = config
+
+    def call(
+        self,
+        hidden_states,
+        attention_mask=None,
+        head_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_value=None,
+        output_attentions=False,
+        training=False,
+    ):
+        # decoder uni-directional self-attention cached key/values tuple is at positions 1,2
+        self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None
+        self_attention_outputs = self.attention(
+            hidden_states,
+            attention_mask,
+            head_mask,
+            output_attentions=output_attentions,
+            past_key_value=self_attn_past_key_value,
+            training=training,
+        )
+        attention_output = self_attention_outputs[0]
+
+        # if decoder, the last output is tuple of self-attn cache
+        if self.is_decoder:
+            outputs = self_attention_outputs[1:-1]
+            present_key_value = self_attention_outputs[-1]
+        else:
+            outputs = self_attention_outputs[1:]  # add self attentions if we output attention weights
+
+        cross_attn_present_key_value = None
+        if self.is_decoder and encoder_hidden_states is not None:
+            if not hasattr(self, "crossattention"):
+                raise AttributeError(
+                    f"If `encoder_hidden_states` are passed, {self} has to be instantiated"
+                    " with cross-attention layers by setting `config.add_cross_attention=True`"
+                )
+
+            # cross_attn cached key/values tuple is at positions 3,4 of past_key_value tuple
+            cross_attn_past_key_value = past_key_value[-2:] if past_key_value is not None else None
+            cross_attention_outputs = self.crossattention(
+                attention_output,
+                attention_mask,
+                head_mask,
+                encoder_hidden_states,
+                encoder_attention_mask,
+                cross_attn_past_key_value,
+                output_attentions,
+                training=training,
+            )
+            attention_output = cross_attention_outputs[0]
+            outputs = outputs + cross_attention_outputs[1:-1]  # add cross attentions if we output attention weights
+
+            # add cross-attn cache to positions 3,4 of present_key_value tuple
+            cross_attn_present_key_value = cross_attention_outputs[-1]
+            present_key_value = present_key_value + cross_attn_present_key_value
+
+        layernorm_output = self.LayerNorm(attention_output)
+        intermediate_output = self.intermediate(hidden_states=layernorm_output)
+        layer_output = self.output_layer(
+            hidden_states=intermediate_output, input_tensor=attention_output, training=training
+        )
+        outputs = (layer_output,) + outputs  # add attentions if we output them
+
+        # if decoder, return the attn key/values as the last output
+        if self.is_decoder:
+            outputs = outputs + (present_key_value,)
+
+        return outputs
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "attention", None) is not None:
+            with tf.name_scope(self.attention.name):
+                self.attention.build(None)
+        if getattr(self, "intermediate", None) is not None:
+            with tf.name_scope(self.intermediate.name):
+                self.intermediate.build(None)
+        if getattr(self, "output_layer", None) is not None:
+            with tf.name_scope(self.output_layer.name):
+                self.output_layer.build(None)
+        if getattr(self, "LayerNorm", None) is not None:
+            with tf.name_scope(self.LayerNorm.name):
+                self.LayerNorm.build([None, None, self.config.hidden_size])
+
+
+class TFEsmEncoder(keras.layers.Layer):
+    def __init__(self, config, name=None):
+        super().__init__(name=name)
+        self.config = config
+        self.layer = [TFEsmLayer(config, name=f"layer_._{i}") for i in range(config.num_hidden_layers)]
+        self.emb_layer_norm_after = keras.layers.LayerNormalization(
+            epsilon=config.layer_norm_eps, name="emb_layer_norm_after"
+        )
+
+    def call(
+        self,
+        hidden_states,
+        attention_mask=None,
+        head_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_values=None,
+        use_cache=None,
+        output_attentions=False,
+        output_hidden_states=False,
+        return_dict=True,
+        training=False,
+    ):
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attentions = () if output_attentions else None
+        all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None
+
+        next_decoder_cache = () if use_cache else None
+        for i, layer_module in enumerate(self.layer):
+            if output_hidden_states:
+                all_hidden_states = all_hidden_states + (hidden_states,)
+
+            layer_head_mask = head_mask[i] if head_mask is not None else None
+            past_key_value = past_key_values[i] if past_key_values is not None else None
+
+            layer_outputs = layer_module(
+                hidden_states,
+                attention_mask,
+                layer_head_mask,
+                encoder_hidden_states,
+                encoder_attention_mask,
+                past_key_value,
+                output_attentions,
+                training,
+            )
+
+            hidden_states = layer_outputs[0]
+            if use_cache:
+                next_decoder_cache += (layer_outputs[-1],)
+            if output_attentions:
+                all_self_attentions = all_self_attentions + (layer_outputs[1],)
+                if self.config.add_cross_attention:
+                    all_cross_attentions = all_cross_attentions + (layer_outputs[2],)
+
+        if self.emb_layer_norm_after:
+            hidden_states = self.emb_layer_norm_after(hidden_states)
+
+        if output_hidden_states:
+            all_hidden_states = all_hidden_states + (hidden_states,)
+
+        if not return_dict:
+            return tuple(
+                v
+                for v in [
+                    hidden_states,
+                    next_decoder_cache,
+                    all_hidden_states,
+                    all_self_attentions,
+                    all_cross_attentions,
+                ]
+                if v is not None
+            )
+        return TFBaseModelOutputWithPastAndCrossAttentions(
+            last_hidden_state=hidden_states,
+            past_key_values=next_decoder_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attentions,
+            cross_attentions=all_cross_attentions,
+        )
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "emb_layer_norm_after", None) is not None:
+            with tf.name_scope(self.emb_layer_norm_after.name):
+                self.emb_layer_norm_after.build([None, None, self.config.hidden_size])
+        if getattr(self, "layer", None) is not None:
+            for layer in self.layer:
+                with tf.name_scope(layer.name):
+                    layer.build(None)
+
+
+# Copied from transformers.models.bert.modeling_tf_bert.TFBertPooler with Bert->Esm
+class TFEsmPooler(keras.layers.Layer):
+    def __init__(self, config: EsmConfig, **kwargs):
+        super().__init__(**kwargs)
+
+        self.dense = keras.layers.Dense(
+            units=config.hidden_size,
+            kernel_initializer=get_initializer(config.initializer_range),
+            activation="tanh",
+            name="dense",
+        )
+        self.config = config
+
+    def call(self, hidden_states: tf.Tensor) -> tf.Tensor:
+        # We "pool" the model by simply taking the hidden state corresponding
+        # to the first token.
+        first_token_tensor = hidden_states[:, 0]
+        pooled_output = self.dense(inputs=first_token_tensor)
+
+        return pooled_output
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "dense", None) is not None:
+            with tf.name_scope(self.dense.name):
+                self.dense.build([None, None, self.config.hidden_size])
+
+
+class TFEsmPreTrainedModel(TFPreTrainedModel):
+    """
+    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
+    models.
+    """
+
+    config_class = EsmConfig
+    base_model_prefix = "esm"
+
+
+ESM_START_DOCSTRING = r"""
+
+    This model inherits from [`TFPreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+
+    This model is also a Keras [Model](https://www.tensorflow.org/api_docs/python/tf/keras/Model) subclass. Use it as a
+    regular Keras model and refer to the TF/Keras documentation for all matters related to general usage and behavior.
+
+    Parameters:
+        config ([`EsmConfig`]): Model configuration class with all the parameters of the
+            model. Initializing with a config file does not load the weights associated with the model, only the
+            configuration. Check out the [`~TFPreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+
+ESM_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`tf.Tensor` of shape `({0})`):
+            Indices of input sequence tokens in the vocabulary.
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`tf.Tensor` of shape `({0})`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            [What are attention masks?](../glossary#attention-mask)
+        position_ids (`tf.Tensor` of shape `({0})`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.max_position_embeddings - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        head_mask (`tf.Tensor` of shape `(num_heads,)` or `(num_layers, num_heads)`, *optional*):
+            Mask to nullify selected heads of the self-attention modules. Mask values selected in `[0, 1]`:
+
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+
+        inputs_embeds (`tf.Tensor` of shape `({0}, hidden_size)`, *optional*):
+            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
+            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
+            model's internal embedding lookup matrix.
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~file_utils.ModelOutput`] instead of a plain tuple.
+"""
+
+
+@add_start_docstrings(
+    "The bare ESM Model transformer outputting raw hidden-states without any specific head on top.",
+    ESM_START_DOCSTRING,
+)
+class TFEsmMainLayer(keras.layers.Layer):
+    """
+
+    The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of
+    cross-attention is added between the self-attention layers, following the architecture described in [Attention is
+    all you need](https://arxiv.org/abs/1706.03762) by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit,
+    Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin.
+
+    To behave as an decoder the model needs to be initialized with the `is_decoder` argument of the configuration set
+    to `True`. To be used in a Seq2Seq model, the model needs to initialized with both `is_decoder` argument and
+    `add_cross_attention` set to `True`; an `encoder_hidden_states` is then expected as an input to the forward pass.
+    """
+
+    _keys_to_ignore_on_load_missing = [r"position_ids"]
+
+    def __init__(self, config, add_pooling_layer=True, name=None, **kwargs):
+        super().__init__(name=name, **kwargs)
+
+        self.config = config
+        self.is_decoder = config.is_decoder
+
+        self.embeddings = TFEsmEmbeddings(config, name="embeddings")
+        self.encoder = TFEsmEncoder(config, name="encoder")
+        self.pooler = TFEsmPooler(config, name="pooler") if add_pooling_layer else None
+
+        self.contact_head = TFEsmContactPredictionHead(
+            in_features=self.config.num_hidden_layers * self.config.num_attention_heads, bias=True, name="contact_head"
+        )
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "embeddings", None) is not None:
+            with tf.name_scope(self.embeddings.name):
+                self.embeddings.build(None)
+        if getattr(self, "encoder", None) is not None:
+            with tf.name_scope(self.encoder.name):
+                self.encoder.build(None)
+        if getattr(self, "pooler", None) is not None:
+            with tf.name_scope(self.pooler.name):
+                self.pooler.build(None)
+        if getattr(self, "contact_head", None) is not None:
+            with tf.name_scope(self.contact_head.name):
+                self.contact_head.build(None)
+
+    def get_input_embeddings(self):
+        return self.embeddings.word_embeddings
+
+    def set_input_embeddings(self, value: tf.Variable):
+        self.embeddings.word_embeddings.weight = value
+        self.embeddings.vocab_size = shape_list(value)[0]
+
+    def _prune_heads(self, heads_to_prune):
+        raise NotImplementedError
+
+    def call(
+        self,
+        input_ids: TFModelInputType | None = None,
+        attention_mask: np.ndarray | tf.Tensor | None = None,
+        position_ids: np.ndarray | tf.Tensor | None = None,
+        head_mask: np.ndarray | tf.Tensor | None = None,
+        inputs_embeds: np.ndarray | tf.Tensor | None = None,
+        encoder_hidden_states: np.ndarray | tf.Tensor | None = None,
+        encoder_attention_mask: np.ndarray | tf.Tensor | None = None,
+        past_key_values: Optional[Tuple[Tuple[Union[np.ndarray, tf.Tensor]]]] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        training: bool = False,
+    ) -> Union[TFBaseModelOutputWithPoolingAndCrossAttentions, Tuple[tf.Tensor]]:
+        if not self.config.is_decoder:
+            use_cache = False
+
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
+        elif input_ids is not None:
+            input_shape = shape_list(input_ids)
+        elif inputs_embeds is not None:
+            input_shape = shape_list(inputs_embeds)[:-1]
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        batch_size, seq_length = input_shape
+
+        if past_key_values is None:
+            past_key_values_length = 0
+            past_key_values = [None] * len(self.encoder.layer)
+        else:
+            past_key_values_length = shape_list(past_key_values[0][0])[-2]
+
+        if attention_mask is None:
+            attention_mask = tf.fill(dims=(batch_size, seq_length + past_key_values_length), value=1)
+
+        embedding_output = self.embeddings(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            inputs_embeds=inputs_embeds,
+            past_key_values_length=past_key_values_length,
+            training=training,
+        )
+
+        # We create a 3D attention mask from a 2D tensor mask.
+        # Sizes are [batch_size, 1, 1, to_seq_length]
+        # So we can broadcast to [batch_size, num_heads, from_seq_length, to_seq_length]
+        # this attention mask is more simple than the triangular masking of causal attention
+        # used in OpenAI GPT, we just need to prepare the broadcast dimension here.
+        attention_mask_shape = shape_list(attention_mask)
+
+        mask_seq_length = seq_length + past_key_values_length
+        # Copied from `modeling_tf_t5.py`
+        # Provided a padding mask of dimensions [batch_size, mask_seq_length]
+        # - if the model is a decoder, apply a causal mask in addition to the padding mask
+        # - if the model is an encoder, make the mask broadcastable to [batch_size, num_heads, mask_seq_length, mask_seq_length]
+        if self.is_decoder:
+            seq_ids = tf.range(mask_seq_length)
+            causal_mask = tf.less_equal(
+                tf.tile(seq_ids[None, None, :], (batch_size, mask_seq_length, 1)),
+                seq_ids[None, :, None],
+            )
+            causal_mask = tf.cast(causal_mask, dtype=attention_mask.dtype)
+            extended_attention_mask = causal_mask * attention_mask[:, None, :]
+            attention_mask_shape = shape_list(extended_attention_mask)
+            extended_attention_mask = tf.reshape(
+                extended_attention_mask, (attention_mask_shape[0], 1, attention_mask_shape[1], attention_mask_shape[2])
+            )
+            if past_key_values[0] is not None:
+                # attention_mask needs to be sliced to the shape `[batch_size, 1, from_seq_length - cached_seq_length, to_seq_length]
+                extended_attention_mask = extended_attention_mask[:, :, -seq_length:, :]
+        else:
+            extended_attention_mask = tf.reshape(
+                attention_mask, (attention_mask_shape[0], 1, 1, attention_mask_shape[1])
+            )
+
+        # Since attention_mask is 1.0 for positions we want to attend and 0.0 for
+        # masked positions, this operation will create a tensor which is 0.0 for
+        # positions we want to attend and -10000.0 for masked positions.
+        # Since we are adding it to the raw scores before the softmax, this is
+        # effectively the same as removing these entirely.
+        extended_attention_mask = tf.cast(extended_attention_mask, dtype=embedding_output.dtype)
+        one_cst = tf.constant(1.0, dtype=embedding_output.dtype)
+        ten_thousand_cst = tf.constant(-10000.0, dtype=embedding_output.dtype)
+        extended_attention_mask = tf.multiply(tf.subtract(one_cst, extended_attention_mask), ten_thousand_cst)
+
+        # Copied from `modeling_tf_t5.py` with -1e9 -> -10000
+        if self.is_decoder and encoder_attention_mask is not None:
+            # If a 2D ou 3D attention mask is provided for the cross-attention
+            # we need to make broadcastable to [batch_size, num_heads, mask_seq_length, mask_seq_length]
+            # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
+            encoder_attention_mask = tf.cast(encoder_attention_mask, dtype=extended_attention_mask.dtype)
+            num_dims_encoder_attention_mask = len(shape_list(encoder_attention_mask))
+            if num_dims_encoder_attention_mask == 3:
+                encoder_extended_attention_mask = encoder_attention_mask[:, None, :, :]
+            if num_dims_encoder_attention_mask == 2:
+                encoder_extended_attention_mask = encoder_attention_mask[:, None, None, :]
+
+            # T5 has a mask that can compare sequence ids, we can simulate this here with this transposition
+            # Cf. https://github.com/tensorflow/mesh/blob/8d2465e9bc93129b913b5ccc6a59aa97abd96ec6/mesh_tensorflow/transformer/transformer_layers.py#L270
+            # encoder_extended_attention_mask = tf.math.equal(encoder_extended_attention_mask,
+            #                                         tf.transpose(encoder_extended_attention_mask, perm=(-1, -2)))
+
+            encoder_extended_attention_mask = (1.0 - encoder_extended_attention_mask) * -10000.0
+        else:
+            encoder_extended_attention_mask = None
+
+        # Prepare head mask if needed
+        # 1.0 in head_mask indicate we keep the head
+        # attention_probs has shape bsz x n_heads x N x N
+        # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
+        # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
+        if head_mask is not None:
+            raise NotImplementedError
+        else:
+            head_mask = [None] * self.config.num_hidden_layers
+
+        encoder_outputs = self.encoder(
+            hidden_states=embedding_output,
+            attention_mask=extended_attention_mask,
+            head_mask=head_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_extended_attention_mask,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            training=training,
+        )
+
+        sequence_output = encoder_outputs[0]
+        pooled_output = self.pooler(hidden_states=sequence_output) if self.pooler is not None else None
+
+        if not return_dict:
+            return (
+                sequence_output,
+                pooled_output,
+            ) + encoder_outputs[1:]
+
+        return TFBaseModelOutputWithPoolingAndCrossAttentions(
+            last_hidden_state=sequence_output,
+            pooler_output=pooled_output,
+            past_key_values=encoder_outputs.past_key_values,
+            hidden_states=encoder_outputs.hidden_states,
+            attentions=encoder_outputs.attentions,
+            cross_attentions=encoder_outputs.cross_attentions,
+        )
+
+    def predict_contacts(self, tokens, attention_mask):
+        attns = self(tokens, attention_mask=attention_mask, return_dict=True, output_attentions=True).attentions
+        attns = tf.stack(attns, axis=1)  # Matches the original model layout
+        # In the original model, attentions for padding tokens are completely zeroed out.
+        # This makes no difference most of the time because the other tokens won't attend to them,
+        # but it does for the contact prediction task, which takes attentions as input,
+        # so we have to mimic that here.
+        attention_mask = tf.cast(attention_mask, attns.dtype)
+        attns *= attention_mask[:, None, None, None]
+        attns *= attention_mask[:, None, None, :, None]
+        return self.contact_head(tokens, attns)
+
+
+@add_start_docstrings(
+    "The bare ESM Model transformer outputting raw hidden-states without any specific head on top.",
+    ESM_START_DOCSTRING,
+)
+class TFEsmModel(TFEsmPreTrainedModel):
+    def __init__(self, config: EsmConfig, add_pooling_layer=True, *inputs, **kwargs):
+        super().__init__(config, *inputs, **kwargs)
+
+        self.esm = TFEsmMainLayer(config, add_pooling_layer=add_pooling_layer, name="esm")
+
+    @unpack_inputs
+    @add_start_docstrings_to_model_forward(ESM_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=TFBaseModelOutputWithPoolingAndCrossAttentions,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def call(
+        self,
+        input_ids: TFModelInputType | None = None,
+        attention_mask: np.ndarray | tf.Tensor | None = None,
+        position_ids: np.ndarray | tf.Tensor | None = None,
+        head_mask: np.ndarray | tf.Tensor | None = None,
+        inputs_embeds: np.ndarray | tf.Tensor | None = None,
+        encoder_hidden_states: np.ndarray | tf.Tensor | None = None,
+        encoder_attention_mask: np.ndarray | tf.Tensor | None = None,
+        past_key_values: Optional[Tuple[Tuple[Union[np.ndarray, tf.Tensor]]]] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        training: Optional[bool] = False,
+    ) -> Union[TFBaseModelOutputWithPoolingAndCrossAttentions, Tuple[tf.Tensor]]:
+        r"""
+        encoder_hidden_states  (`tf.Tensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if
+            the model is configured as a decoder.
+        encoder_attention_mask (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in
+            the cross-attention if the model is configured as a decoder. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+        past_key_values (`Tuple[Tuple[tf.Tensor]]` of length `config.n_layers`)
+            contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding.
+            If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
+            don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
+            `decoder_input_ids` of shape `(batch_size, sequence_length)`.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`). Set to `False` during training, `True` during generation
+        """
+        outputs = self.esm(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            training=training,
+        )
+        return outputs
+
+    def predict_contacts(self, tokens, attention_mask):
+        return self.esm.predict_contacts(tokens, attention_mask)
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "esm", None) is not None:
+            with tf.name_scope(self.esm.name):
+                self.esm.build(None)
+
+
+@add_start_docstrings("""ESM Model with a `language modeling` head on top.""", ESM_START_DOCSTRING)
+class TFEsmForMaskedLM(TFEsmPreTrainedModel, TFMaskedLanguageModelingLoss):
+    _keys_to_ignore_on_load_missing = [r"position_ids"]
+    _keys_to_ignore_on_load_unexpected = [r"pooler"]
+
+    def __init__(self, config):
+        super().__init__(config)
+
+        if config.is_decoder:
+            logger.warning(
+                "If you want to use `EsmForMaskedLM` make sure `config.is_decoder=False` for "
+                "bi-directional self-attention."
+            )
+
+        self.esm = TFEsmMainLayer(config, add_pooling_layer=False, name="esm")
+        self.lm_head = TFEsmLMHead(config, name="lm_head")
+        if config.tie_word_embeddings:
+            # Ensure word embeddings are built so that we actually have something to tie
+            with tf.name_scope(os.path.join(self._name_scope(), "esm", "embeddings", "word_embeddings")):
+                self.esm.embeddings.word_embeddings.build((None, None))
+            self.lm_head.decoder = self.esm.embeddings.word_embeddings.weights[0]
+
+    def get_output_embeddings(self):
+        return self.lm_head.decoder
+
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head.decoder = new_embeddings
+
+    def get_lm_head(self):
+        return self.lm_head
+
+    @unpack_inputs
+    @add_start_docstrings_to_model_forward(ESM_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=TFMaskedLMOutput,
+        config_class=_CONFIG_FOR_DOC,
+        mask="<mask>",
+    )
+    def call(
+        self,
+        input_ids: TFModelInputType | None = None,
+        attention_mask: np.ndarray | tf.Tensor | None = None,
+        position_ids: np.ndarray | tf.Tensor | None = None,
+        head_mask: np.ndarray | tf.Tensor | None = None,
+        inputs_embeds: np.ndarray | tf.Tensor | None = None,
+        encoder_hidden_states: np.ndarray | tf.Tensor | None = None,
+        encoder_attention_mask: np.ndarray | tf.Tensor | None = None,
+        labels: np.ndarray | tf.Tensor | None = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        training: bool = False,
+    ) -> Union[TFMaskedLMOutput, Tuple[tf.Tensor]]:
+        r"""
+        labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
+            config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
+            loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
+        kwargs (`Dict[str, any]`, optional, defaults to *{}*):
+            Used to hide legacy arguments that have been deprecated.
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.esm(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            training=training,
+        )
+        sequence_output = outputs[0]
+        prediction_scores = self.lm_head(sequence_output)
+
+        masked_lm_loss = None
+        if labels is not None:
+            masked_lm_loss = self.hf_compute_loss(labels=labels, logits=prediction_scores)
+
+        if not return_dict:
+            output = (prediction_scores,) + outputs[2:]
+            return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output
+
+        return TFMaskedLMOutput(
+            loss=masked_lm_loss,
+            logits=prediction_scores,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    def predict_contacts(self, tokens, attention_mask):
+        return self.esm.predict_contacts(tokens, attention_mask)
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "esm", None) is not None:
+            with tf.name_scope(self.esm.name):
+                self.esm.build(None)
+        if getattr(self, "lm_head", None) is not None:
+            with tf.name_scope(self.lm_head.name):
+                self.lm_head.build(None)
+
+
+class TFEsmLMHead(keras.layers.Layer):
+    """ESM Head for masked language modeling."""
+
+    def __init__(self, config, name=None):
+        super().__init__(name=name)
+        self.dense = keras.layers.Dense(
+            config.hidden_size, kernel_initializer=get_initializer(config.initializer_range), name="dense"
+        )
+
+        self.layer_norm = keras.layers.LayerNormalization(epsilon=config.layer_norm_eps, name="layer_norm")
+        if config.tie_word_embeddings:
+            self.decoder = None
+        else:
+            self.decoder = keras.layers.Dense(
+                config.vocab_size,
+                kernel_initializer=get_initializer(config.initializer_range),
+                name="decoder",
+                use_bias=False,
+            )
+        self.config = config
+
+    def build(self, input_shape=None):
+        # Separate bias to match the PT model and allow weight cross-loading to work
+        # Put it in the build so it gets the right name when adding it as a weight
+        if self.built:
+            return
+        self.built = True
+        self.bias = self.add_weight("bias", shape=(self.config.vocab_size,), initializer="zeros", trainable=True)
+        if getattr(self, "dense", None) is not None:
+            with tf.name_scope(self.dense.name):
+                self.dense.build([None, None, self.config.hidden_size])
+        if getattr(self, "layer_norm", None) is not None:
+            with tf.name_scope(self.layer_norm.name):
+                self.layer_norm.build([None, None, self.config.hidden_size])
+        if getattr(self, "decoder", None) is not None and not self.config.tie_word_embeddings:
+            with tf.name_scope(self.decoder.name):
+                self.decoder.build([None, None, self.config.hidden_size])
+
+    def get_bias(self):
+        return {"bias": self.bias}
+
+    def call(self, features):
+        x = self.dense(features)
+        x = tf.nn.gelu(x)
+        x = self.layer_norm(x)
+
+        # project back to size of vocabulary with bias
+        if self.config.tie_word_embeddings:
+            x = tf.matmul(x, self.decoder, transpose_b=True) + self.bias
+        else:
+            x = self.decoder(x) + self.bias
+        return x
+
+
+@add_start_docstrings(
+    """
+    ESM Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled
+    output) e.g. for GLUE tasks.
+    """,
+    ESM_START_DOCSTRING,
+)
+class TFEsmForSequenceClassification(TFEsmPreTrainedModel, TFSequenceClassificationLoss):
+    _keys_to_ignore_on_load_missing = [r"position_ids"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.config = config
+
+        self.esm = TFEsmMainLayer(config, add_pooling_layer=False, name="esm")
+        self.classifier = TFEsmClassificationHead(config, name="classifier")
+
+    @unpack_inputs
+    @add_start_docstrings_to_model_forward(ESM_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=TFSequenceClassifierOutput,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def call(
+        self,
+        input_ids: TFModelInputType | None = None,
+        attention_mask: np.ndarray | tf.Tensor | None = None,
+        position_ids: np.ndarray | tf.Tensor | None = None,
+        head_mask: np.ndarray | tf.Tensor | None = None,
+        inputs_embeds: np.ndarray | tf.Tensor | None = None,
+        labels: np.ndarray | tf.Tensor | None = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        training: bool = False,
+    ) -> Union[TFSequenceClassifierOutput, Tuple[tf.Tensor]]:
+        r"""
+        labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.esm(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            training=training,
+        )
+        sequence_output = outputs[0]
+        logits = self.classifier(sequence_output)
+
+        loss = None if labels is None else self.hf_compute_loss(labels, logits)
+
+        if not return_dict:
+            output = (logits,) + outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return TFSequenceClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "esm", None) is not None:
+            with tf.name_scope(self.esm.name):
+                self.esm.build(None)
+        if getattr(self, "classifier", None) is not None:
+            with tf.name_scope(self.classifier.name):
+                self.classifier.build(None)
+
+
+@add_start_docstrings(
+    """
+    ESM Model with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for
+    Named-Entity-Recognition (NER) tasks.
+    """,
+    ESM_START_DOCSTRING,
+)
+class TFEsmForTokenClassification(TFEsmPreTrainedModel, TFTokenClassificationLoss):
+    _keys_to_ignore_on_load_unexpected = [r"pooler"]
+    _keys_to_ignore_on_load_missing = [r"position_ids"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.esm = TFEsmMainLayer(config, add_pooling_layer=False, name="esm")
+        self.dropout = keras.layers.Dropout(config.hidden_dropout_prob)
+        self.classifier = keras.layers.Dense(config.num_labels, name="classifier")
+        self.config = config
+
+    @unpack_inputs
+    @add_start_docstrings_to_model_forward(ESM_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=TFTokenClassifierOutput,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def call(
+        self,
+        input_ids: TFModelInputType | None = None,
+        attention_mask: np.ndarray | tf.Tensor | None = None,
+        position_ids: np.ndarray | tf.Tensor | None = None,
+        head_mask: np.ndarray | tf.Tensor | None = None,
+        inputs_embeds: np.ndarray | tf.Tensor | None = None,
+        labels: np.ndarray | tf.Tensor | None = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        training: bool = False,
+    ) -> Union[TFTokenClassifierOutput, Tuple[tf.Tensor]]:
+        r"""
+        labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.esm(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            training=training,
+        )
+
+        sequence_output = outputs[0]
+
+        sequence_output = self.dropout(sequence_output, training=training)
+        logits = self.classifier(sequence_output)
+
+        loss = None if labels is None else self.hf_compute_loss(labels, logits)
+
+        if not return_dict:
+            output = (logits,) + outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return TFTokenClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "esm", None) is not None:
+            with tf.name_scope(self.esm.name):
+                self.esm.build(None)
+        if getattr(self, "classifier", None) is not None:
+            with tf.name_scope(self.classifier.name):
+                self.classifier.build([None, None, self.config.hidden_size])
+
+
+class TFEsmClassificationHead(keras.layers.Layer):
+    """Head for sentence-level classification tasks."""
+
+    def __init__(self, config, name=None):
+        super().__init__(name=name)
+        self.dense = keras.layers.Dense(
+            config.hidden_size,
+            kernel_initializer=get_initializer(config.initializer_range),
+            activation="tanh",
+            name="dense",
+        )
+        self.dropout = keras.layers.Dropout(config.hidden_dropout_prob)
+        self.out_proj = keras.layers.Dense(
+            config.num_labels,
+            kernel_initializer=get_initializer(config.initializer_range),
+            activation="linear",
+            name="out_proj",
+        )
+        self.config = config
+
+    def call(self, features, training=False):
+        x = features[:, 0, :]  # take <s> token (equiv. to [CLS])
+        x = self.dropout(x, training=training)
+        x = self.dense(x)
+        x = self.dropout(x, training=training)
+        x = self.out_proj(x)
+        return x
+
+    def build(self, input_shape=None):
+        if self.built:
+            return
+        self.built = True
+        if getattr(self, "dense", None) is not None:
+            with tf.name_scope(self.dense.name):
+                self.dense.build([None, None, self.config.hidden_size])
+        if getattr(self, "out_proj", None) is not None:
+            with tf.name_scope(self.out_proj.name):
+                self.out_proj.build([None, None, self.config.hidden_size])
+
+
+def create_position_ids_from_input_ids(input_ids, padding_idx, past_key_values_length=0):
+    """
+    Replace non-padding symbols with their position numbers. Position numbers begin at padding_idx+1. Padding symbols
+    are ignored. This is modified from fairseq's `utils.make_positions`.
+
+    Args:
+        x: tf.Tensor x:
+
+    Returns: tf.Tensor
+    """
+    # The series of casts and type-conversions here are carefully balanced to both work with ONNX export and XLA.
+    mask = tf.cast(input_ids != padding_idx, tf.int64)
+    incremental_indices = (tf.cumsum(mask, axis=1) + past_key_values_length) * mask
+    return incremental_indices + padding_idx

venv/lib/python3.10/site-packages/transformers/models/esm/openfold_utils/__init__.py

@@ -0,0 +1,8 @@
+from .chunk_utils import chunk_layer
+from .data_transforms import make_atom14_masks
+from .feats import atom14_to_atom37, frames_and_literature_positions_to_atom14_pos, torsion_angles_to_frames
+from .loss import compute_predicted_aligned_error, compute_tm
+from .protein import Protein as OFProtein
+from .protein import to_pdb
+from .rigid_utils import Rigid, Rotation
+from .tensor_utils import dict_multimap, flatten_final_dims, permute_final_dims

venv/lib/python3.10/site-packages/transformers/models/esm/openfold_utils/chunk_utils.py

@@ -0,0 +1,397 @@
+# Copyright 2021 AlQuraishi Laboratory
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import logging
+import math
+from functools import partial
+from typing import Any, Callable, Dict, Iterable, List, Optional, Sequence, Tuple, Union
+
+import torch
+
+from .tensor_utils import tensor_tree_map, tree_map
+
+
+def _fetch_dims(tree: Union[dict, list, tuple, torch.Tensor]) -> List[Tuple[int, ...]]:
+    shapes = []
+    if isinstance(tree, dict):
+        for v in tree.values():
+            shapes.extend(_fetch_dims(v))
+    elif isinstance(tree, (list, tuple)):
+        for t in tree:
+            shapes.extend(_fetch_dims(t))
+    elif isinstance(tree, torch.Tensor):
+        shapes.append(tree.shape)
+    else:
+        raise ValueError("Not supported")
+
+    return shapes
+
+
+@torch.jit.ignore
+def _flat_idx_to_idx(flat_idx: int, dims: Tuple[int, ...]) -> Tuple[int, ...]:
+    idx = []
+    for d in reversed(dims):
+        idx.append(flat_idx % d)
+        flat_idx = flat_idx // d
+
+    return tuple(reversed(idx))
+
+
+@torch.jit.ignore
+def _get_minimal_slice_set(
+    start: Sequence[int],
+    end: Sequence[int],
+    dims: Sequence[int],
+    start_edges: Optional[Sequence[bool]] = None,
+    end_edges: Optional[Sequence[bool]] = None,
+) -> List[Tuple[slice, ...]]:
+    """
+    Produces an ordered sequence of tensor slices that, when used in sequence on a tensor with shape dims, yields
+    tensors that contain every leaf in the contiguous range [start, end]. Care is taken to yield a short sequence of
+    slices, and perhaps even the shortest possible (I'm pretty sure it's the latter).
+
+    end is INCLUSIVE.
+    """
+
+    # start_edges and end_edges both indicate whether, starting from any given
+    # dimension, the start/end index is at the top/bottom edge of the
+    # corresponding tensor, modeled as a tree
+    def reduce_edge_list(l: List[bool]) -> None:
+        tally = True
+        for i in range(len(l)):
+            reversed_idx = -1 * (i + 1)
+            l[reversed_idx] &= tally
+            tally = l[reversed_idx]
+
+    if start_edges is None:
+        start_edges = [s == 0 for s in start]
+        reduce_edge_list(start_edges)
+    if end_edges is None:
+        end_edges = [e == (d - 1) for e, d in zip(end, dims)]
+        reduce_edge_list(end_edges)
+
+    # Base cases. Either start/end are empty and we're done, or the final,
+    # one-dimensional tensor can be simply sliced
+    if len(start) == 0:
+        return [()]
+    elif len(start) == 1:
+        return [(slice(start[0], end[0] + 1),)]
+
+    slices: List[Tuple[slice, ...]] = []
+    path_list: List[slice] = []
+
+    # Dimensions common to start and end can be selected directly
+    for s, e in zip(start, end):
+        if s == e:
+            path_list.append(slice(s, s + 1))
+        else:
+            break
+
+    path: Tuple[slice, ...] = tuple(path_list)
+    divergence_idx = len(path)
+
+    # start == end, and we're done
+    if divergence_idx == len(dims):
+        return [path]
+
+    def upper() -> Tuple[Tuple[slice, ...], ...]:
+        assert start_edges is not None
+        assert end_edges is not None
+
+        sdi = start[divergence_idx]
+        return tuple(
+            path + (slice(sdi, sdi + 1),) + s
+            for s in _get_minimal_slice_set(
+                start[divergence_idx + 1 :],
+                [d - 1 for d in dims[divergence_idx + 1 :]],
+                dims[divergence_idx + 1 :],
+                start_edges=start_edges[divergence_idx + 1 :],
+                end_edges=[True for _ in end_edges[divergence_idx + 1 :]],
+            )
+        )
+
+    def lower() -> Tuple[Tuple[slice, ...], ...]:
+        assert start_edges is not None
+        assert end_edges is not None
+
+        edi = end[divergence_idx]
+        return tuple(
+            path + (slice(edi, edi + 1),) + s
+            for s in _get_minimal_slice_set(
+                [0 for _ in start[divergence_idx + 1 :]],
+                end[divergence_idx + 1 :],
+                dims[divergence_idx + 1 :],
+                start_edges=[True for _ in start_edges[divergence_idx + 1 :]],
+                end_edges=end_edges[divergence_idx + 1 :],
+            )
+        )
+
+    # If both start and end are at the edges of the subtree rooted at
+    # divergence_idx, we can just select the whole subtree at once
+    if start_edges[divergence_idx] and end_edges[divergence_idx]:
+        slices.append(path + (slice(start[divergence_idx], end[divergence_idx] + 1),))
+    # If just start is at the edge, we can grab almost all of the subtree,
+    # treating only the ragged bottom edge as an edge case
+    elif start_edges[divergence_idx]:
+        slices.append(path + (slice(start[divergence_idx], end[divergence_idx]),))
+        slices.extend(lower())
+    # Analogous to the previous case, but the top is ragged this time
+    elif end_edges[divergence_idx]:
+        slices.extend(upper())
+        slices.append(path + (slice(start[divergence_idx] + 1, end[divergence_idx] + 1),))
+    # If both sides of the range are ragged, we need to handle both sides
+    # separately. If there's contiguous meat in between them, we can index it
+    # in one big chunk
+    else:
+        slices.extend(upper())
+        middle_ground = end[divergence_idx] - start[divergence_idx]
+        if middle_ground > 1:
+            slices.append(path + (slice(start[divergence_idx] + 1, end[divergence_idx]),))
+        slices.extend(lower())
+
+    return slices
+
+
+@torch.jit.ignore
+def _chunk_slice(t: torch.Tensor, flat_start: int, flat_end: int, no_batch_dims: int) -> torch.Tensor:
+    """
+    Equivalent to
+
+        t.reshape((-1,) + t.shape[no_batch_dims:])[flat_start:flat_end]
+
+    but without the need for the initial reshape call, which can be memory-intensive in certain situations. The only
+    reshape operations in this function are performed on sub-tensors that scale with (flat_end - flat_start), the chunk
+    size.
+    """
+
+    batch_dims = t.shape[:no_batch_dims]
+    start_idx = list(_flat_idx_to_idx(flat_start, batch_dims))
+    # _get_minimal_slice_set is inclusive
+    end_idx = list(_flat_idx_to_idx(flat_end - 1, batch_dims))
+
+    # Get an ordered list of slices to perform
+    slices = _get_minimal_slice_set(
+        start_idx,
+        end_idx,
+        batch_dims,
+    )
+
+    sliced_tensors = [t[s] for s in slices]
+
+    return torch.cat([s.view((-1,) + t.shape[no_batch_dims:]) for s in sliced_tensors])
+
+
+def chunk_layer(
+    layer: Callable,
+    inputs: Dict[str, Any],
+    chunk_size: int,
+    no_batch_dims: int,
+    low_mem: bool = False,
+    _out: Any = None,
+    _add_into_out: bool = False,
+) -> Any:
+    """
+    Implements the "chunking" procedure described in section 1.11.8.
+
+    Layer outputs and inputs are assumed to be simple "pytrees," consisting only of (arbitrarily nested) lists, tuples,
+    and dicts with torch.Tensor leaves.
+
+    Args:
+        layer:
+            The layer to be applied chunk-wise
+        inputs:
+            A (non-nested) dictionary of keyworded inputs. All leaves must be tensors and must share the same batch
+            dimensions.
+        chunk_size:
+            The number of sub-batches per chunk. If multiple batch dimensions are specified, a "sub-batch" is defined
+            as a single indexing of all batch dimensions simultaneously (s.t. the number of sub-batches is the product
+            of the batch dimensions).
+        no_batch_dims:
+            How many of the initial dimensions of each input tensor can be considered batch dimensions.
+        low_mem:
+            Avoids flattening potentially large input tensors. Unnecessary in most cases, and is ever so slightly
+            slower than the default setting.
+    Returns:
+        The reassembled output of the layer on the inputs.
+    """
+    if not (len(inputs) > 0):
+        raise ValueError("Must provide at least one input")
+
+    initial_dims = [shape[:no_batch_dims] for shape in _fetch_dims(inputs)]
+    orig_batch_dims = tuple([max(s) for s in zip(*initial_dims)])
+
+    def _prep_inputs(t: torch.Tensor) -> torch.Tensor:
+        if not low_mem:
+            if not sum(t.shape[:no_batch_dims]) == no_batch_dims:
+                t = t.expand(orig_batch_dims + t.shape[no_batch_dims:])
+            t = t.reshape(-1, *t.shape[no_batch_dims:])
+        else:
+            t = t.expand(orig_batch_dims + t.shape[no_batch_dims:])
+        return t
+
+    prepped_inputs: Dict[str, Any] = tensor_tree_map(_prep_inputs, inputs)
+    prepped_outputs = None
+    if _out is not None:
+        prepped_outputs = tensor_tree_map(lambda t: t.view([-1] + list(t.shape[no_batch_dims:])), _out)
+
+    flat_batch_dim = 1
+    for d in orig_batch_dims:
+        flat_batch_dim *= d
+
+    no_chunks = flat_batch_dim // chunk_size + (flat_batch_dim % chunk_size != 0)
+
+    def _select_chunk(t: torch.Tensor) -> torch.Tensor:
+        return t[i : i + chunk_size] if t.shape[0] != 1 else t
+
+    i = 0
+    out = prepped_outputs
+    for _ in range(no_chunks):
+        # Chunk the input
+        if not low_mem:
+            select_chunk = _select_chunk
+        else:
+            select_chunk = partial(
+                _chunk_slice,
+                flat_start=i,
+                flat_end=min(flat_batch_dim, i + chunk_size),
+                no_batch_dims=len(orig_batch_dims),
+            )
+
+        chunks: Dict[str, Any] = tensor_tree_map(select_chunk, prepped_inputs)
+
+        # Run the layer on the chunk
+        output_chunk = layer(**chunks)
+
+        # Allocate space for the output
+        if out is None:
+            out = tensor_tree_map(lambda t: t.new_zeros((flat_batch_dim,) + t.shape[1:]), output_chunk)
+
+        # Put the chunk in its pre-allocated space
+        if isinstance(output_chunk, dict):
+
+            def assign(d1: dict, d2: dict) -> None:
+                for k, v in d1.items():
+                    if isinstance(v, dict):
+                        assign(v, d2[k])
+                    else:
+                        if _add_into_out:
+                            v[i : i + chunk_size] += d2[k]
+                        else:
+                            v[i : i + chunk_size] = d2[k]
+
+            assign(out, output_chunk)
+        elif isinstance(output_chunk, tuple):
+            for x1, x2 in zip(out, output_chunk):
+                if _add_into_out:
+                    x1[i : i + chunk_size] += x2
+                else:
+                    x1[i : i + chunk_size] = x2
+        elif isinstance(output_chunk, torch.Tensor):
+            if _add_into_out:
+                out[i : i + chunk_size] += output_chunk
+            else:
+                out[i : i + chunk_size] = output_chunk
+        else:
+            raise ValueError("Not supported")
+
+        i += chunk_size
+
+    out = tensor_tree_map(lambda t: t.view(orig_batch_dims + t.shape[1:]), out)
+
+    return out
+
+
+class ChunkSizeTuner:
+    def __init__(
+        self,
+        # Heuristically, runtimes for most of the modules in the network
+        # plateau earlier than this on all GPUs I've run the model on.
+        max_chunk_size: int = 512,
+    ):
+        self.max_chunk_size = max_chunk_size
+        self.cached_chunk_size: Optional[int] = None
+        self.cached_arg_data: Optional[tuple] = None
+
+    def _determine_favorable_chunk_size(self, fn: Callable, args: tuple, min_chunk_size: int) -> int:
+        logging.info("Tuning chunk size...")
+
+        if min_chunk_size >= self.max_chunk_size:
+            return min_chunk_size
+
+        candidates: List[int] = [2**l for l in range(int(math.log(self.max_chunk_size, 2)) + 1)]
+        candidates = [c for c in candidates if c > min_chunk_size]
+        candidates = [min_chunk_size] + candidates
+        candidates[-1] += 4
+
+        def test_chunk_size(chunk_size: int) -> bool:
+            try:
+                with torch.no_grad():
+                    fn(*args, chunk_size=chunk_size)
+                return True
+            except RuntimeError:
+                return False
+
+        min_viable_chunk_size_index = 0
+        i = len(candidates) - 1
+        while i > min_viable_chunk_size_index:
+            viable = test_chunk_size(candidates[i])
+            if not viable:
+                i = (min_viable_chunk_size_index + i) // 2
+            else:
+                min_viable_chunk_size_index = i
+                i = (i + len(candidates) - 1) // 2
+
+        return candidates[min_viable_chunk_size_index]
+
+    def _compare_arg_caches(self, ac1: Iterable, ac2: Iterable) -> bool:
+        consistent = True
+        for a1, a2 in zip(ac1, ac2):
+            assert type(ac1) == type(ac2)
+            if isinstance(ac1, (list, tuple)):
+                consistent &= self._compare_arg_caches(a1, a2)
+            elif isinstance(ac1, dict):
+                a1_items = [v for _, v in sorted(a1.items(), key=lambda x: x[0])]
+                a2_items = [v for _, v in sorted(a2.items(), key=lambda x: x[0])]
+                consistent &= self._compare_arg_caches(a1_items, a2_items)
+            else:
+                consistent &= a1 == a2
+
+        return consistent
+
+    def tune_chunk_size(
+        self,
+        representative_fn: Callable,
+        args: tuple,
+        min_chunk_size: int,
+    ) -> int:
+        consistent = True
+        arg_data: tuple = tree_map(lambda a: a.shape if isinstance(a, torch.Tensor) else a, args, object)
+        if self.cached_arg_data is not None:
+            # If args have changed shape/value, we need to re-tune
+            assert len(self.cached_arg_data) == len(arg_data)
+            consistent = self._compare_arg_caches(self.cached_arg_data, arg_data)
+        else:
+            # Otherwise, we can reuse the precomputed value
+            consistent = False
+
+        if not consistent:
+            self.cached_chunk_size = self._determine_favorable_chunk_size(
+                representative_fn,
+                args,
+                min_chunk_size,
+            )
+            self.cached_arg_data = arg_data
+
+        assert self.cached_chunk_size is not None
+
+        return self.cached_chunk_size

venv/lib/python3.10/site-packages/transformers/models/esm/openfold_utils/data_transforms.py

@@ -0,0 +1,93 @@
+# Copyright 2021 AlQuraishi Laboratory
+# Copyright 2021 DeepMind Technologies Limited
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Dict
+
+import numpy as np
+import torch
+
+from . import residue_constants as rc
+from .tensor_utils import tensor_tree_map, tree_map
+
+
+def make_atom14_masks(protein: Dict[str, torch.Tensor]) -> Dict[str, torch.Tensor]:
+    """Construct denser atom positions (14 dimensions instead of 37)."""
+    restype_atom14_to_atom37_list = []
+    restype_atom37_to_atom14_list = []
+    restype_atom14_mask_list = []
+
+    for rt in rc.restypes:
+        atom_names = rc.restype_name_to_atom14_names[rc.restype_1to3[rt]]
+        restype_atom14_to_atom37_list.append([(rc.atom_order[name] if name else 0) for name in atom_names])
+        atom_name_to_idx14 = {name: i for i, name in enumerate(atom_names)}
+        restype_atom37_to_atom14_list.append(
+            [(atom_name_to_idx14[name] if name in atom_name_to_idx14 else 0) for name in rc.atom_types]
+        )
+
+        restype_atom14_mask_list.append([(1.0 if name else 0.0) for name in atom_names])
+
+    # Add dummy mapping for restype 'UNK'
+    restype_atom14_to_atom37_list.append([0] * 14)
+    restype_atom37_to_atom14_list.append([0] * 37)
+    restype_atom14_mask_list.append([0.0] * 14)
+
+    restype_atom14_to_atom37 = torch.tensor(
+        restype_atom14_to_atom37_list,
+        dtype=torch.int32,
+        device=protein["aatype"].device,
+    )
+    restype_atom37_to_atom14 = torch.tensor(
+        restype_atom37_to_atom14_list,
+        dtype=torch.int32,
+        device=protein["aatype"].device,
+    )
+    restype_atom14_mask = torch.tensor(
+        restype_atom14_mask_list,
+        dtype=torch.float32,
+        device=protein["aatype"].device,
+    )
+    protein_aatype = protein["aatype"].to(torch.long)
+
+    # create the mapping for (residx, atom14) --> atom37, i.e. an array
+    # with shape (num_res, 14) containing the atom37 indices for this protein
+    residx_atom14_to_atom37 = restype_atom14_to_atom37[protein_aatype]
+    residx_atom14_mask = restype_atom14_mask[protein_aatype]
+
+    protein["atom14_atom_exists"] = residx_atom14_mask
+    protein["residx_atom14_to_atom37"] = residx_atom14_to_atom37.long()
+
+    # create the gather indices for mapping back
+    residx_atom37_to_atom14 = restype_atom37_to_atom14[protein_aatype]
+    protein["residx_atom37_to_atom14"] = residx_atom37_to_atom14.long()
+
+    # create the corresponding mask
+    restype_atom37_mask = torch.zeros([21, 37], dtype=torch.float32, device=protein["aatype"].device)
+    for restype, restype_letter in enumerate(rc.restypes):
+        restype_name = rc.restype_1to3[restype_letter]
+        atom_names = rc.residue_atoms[restype_name]
+        for atom_name in atom_names:
+            atom_type = rc.atom_order[atom_name]
+            restype_atom37_mask[restype, atom_type] = 1
+
+    residx_atom37_mask = restype_atom37_mask[protein_aatype]
+    protein["atom37_atom_exists"] = residx_atom37_mask
+
+    return protein
+
+
+def make_atom14_masks_np(batch: Dict[str, torch.Tensor]) -> Dict[str, np.ndarray]:
+    batch = tree_map(lambda n: torch.tensor(n, device=batch["aatype"].device), batch, np.ndarray)
+    out = tensor_tree_map(lambda t: np.array(t), make_atom14_masks(batch))
+    return out

venv/lib/python3.10/site-packages/transformers/models/esm/openfold_utils/feats.py

@@ -0,0 +1,255 @@
+# Copyright 2021 AlQuraishi Laboratory
+# Copyright 2021 DeepMind Technologies Limited
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Dict, Tuple, overload
+
+import torch
+import torch.types
+from torch import nn
+
+from . import residue_constants as rc
+from .rigid_utils import Rigid, Rotation
+from .tensor_utils import batched_gather
+
+
+@overload
+def pseudo_beta_fn(aatype: torch.Tensor, all_atom_positions: torch.Tensor, all_atom_masks: None) -> torch.Tensor:
+    ...
+
+
+@overload
+def pseudo_beta_fn(
+    aatype: torch.Tensor, all_atom_positions: torch.Tensor, all_atom_masks: torch.Tensor
+) -> Tuple[torch.Tensor, torch.Tensor]:
+    ...
+
+
+def pseudo_beta_fn(aatype, all_atom_positions, all_atom_masks):
+    is_gly = aatype == rc.restype_order["G"]
+    ca_idx = rc.atom_order["CA"]
+    cb_idx = rc.atom_order["CB"]
+    pseudo_beta = torch.where(
+        is_gly[..., None].expand(*((-1,) * len(is_gly.shape)), 3),
+        all_atom_positions[..., ca_idx, :],
+        all_atom_positions[..., cb_idx, :],
+    )
+
+    if all_atom_masks is not None:
+        pseudo_beta_mask = torch.where(
+            is_gly,
+            all_atom_masks[..., ca_idx],
+            all_atom_masks[..., cb_idx],
+        )
+        return pseudo_beta, pseudo_beta_mask
+    else:
+        return pseudo_beta
+
+
+def atom14_to_atom37(atom14: torch.Tensor, batch: Dict[str, torch.Tensor]) -> torch.Tensor:
+    atom37_data = batched_gather(
+        atom14,
+        batch["residx_atom37_to_atom14"],
+        dim=-2,
+        no_batch_dims=len(atom14.shape[:-2]),
+    )
+
+    atom37_data = atom37_data * batch["atom37_atom_exists"][..., None]
+
+    return atom37_data
+
+
+def build_template_angle_feat(template_feats: Dict[str, torch.Tensor]) -> torch.Tensor:
+    template_aatype = template_feats["template_aatype"]
+    torsion_angles_sin_cos = template_feats["template_torsion_angles_sin_cos"]
+    alt_torsion_angles_sin_cos = template_feats["template_alt_torsion_angles_sin_cos"]
+    torsion_angles_mask = template_feats["template_torsion_angles_mask"]
+    template_angle_feat = torch.cat(
+        [
+            nn.functional.one_hot(template_aatype, 22),
+            torsion_angles_sin_cos.reshape(*torsion_angles_sin_cos.shape[:-2], 14),
+            alt_torsion_angles_sin_cos.reshape(*alt_torsion_angles_sin_cos.shape[:-2], 14),
+            torsion_angles_mask,
+        ],
+        dim=-1,
+    )
+
+    return template_angle_feat
+
+
+def build_template_pair_feat(
+    batch: Dict[str, torch.Tensor],
+    min_bin: torch.types.Number,
+    max_bin: torch.types.Number,
+    no_bins: int,
+    use_unit_vector: bool = False,
+    eps: float = 1e-20,
+    inf: float = 1e8,
+) -> torch.Tensor:
+    template_mask = batch["template_pseudo_beta_mask"]
+    template_mask_2d = template_mask[..., None] * template_mask[..., None, :]
+
+    # Compute distogram (this seems to differ slightly from Alg. 5)
+    tpb = batch["template_pseudo_beta"]
+    dgram = torch.sum((tpb[..., None, :] - tpb[..., None, :, :]) ** 2, dim=-1, keepdim=True)
+    lower = torch.linspace(min_bin, max_bin, no_bins, device=tpb.device) ** 2
+    upper = torch.cat([lower[1:], lower.new_tensor([inf])], dim=-1)
+    dgram = ((dgram > lower) * (dgram < upper)).type(dgram.dtype)
+
+    to_concat = [dgram, template_mask_2d[..., None]]
+
+    aatype_one_hot: torch.LongTensor = nn.functional.one_hot(
+        batch["template_aatype"],
+        rc.restype_num + 2,
+    )
+
+    n_res = batch["template_aatype"].shape[-1]
+    to_concat.append(aatype_one_hot[..., None, :, :].expand(*aatype_one_hot.shape[:-2], n_res, -1, -1))
+    to_concat.append(aatype_one_hot[..., None, :].expand(*aatype_one_hot.shape[:-2], -1, n_res, -1))
+
+    n, ca, c = [rc.atom_order[a] for a in ["N", "CA", "C"]]
+    rigids = Rigid.make_transform_from_reference(
+        n_xyz=batch["template_all_atom_positions"][..., n, :],
+        ca_xyz=batch["template_all_atom_positions"][..., ca, :],
+        c_xyz=batch["template_all_atom_positions"][..., c, :],
+        eps=eps,
+    )
+    points = rigids.get_trans()[..., None, :, :]
+    rigid_vec = rigids[..., None].invert_apply(points)
+
+    inv_distance_scalar = torch.rsqrt(eps + torch.sum(rigid_vec**2, dim=-1))
+
+    t_aa_masks = batch["template_all_atom_mask"]
+    template_mask = t_aa_masks[..., n] * t_aa_masks[..., ca] * t_aa_masks[..., c]
+    template_mask_2d = template_mask[..., None] * template_mask[..., None, :]
+
+    inv_distance_scalar = inv_distance_scalar * template_mask_2d
+    unit_vector = rigid_vec * inv_distance_scalar[..., None]
+
+    if not use_unit_vector:
+        unit_vector = unit_vector * 0.0
+
+    to_concat.extend(torch.unbind(unit_vector[..., None, :], dim=-1))
+    to_concat.append(template_mask_2d[..., None])
+
+    act = torch.cat(to_concat, dim=-1)
+    act = act * template_mask_2d[..., None]
+
+    return act
+
+
+def build_extra_msa_feat(batch: Dict[str, torch.Tensor]) -> torch.Tensor:
+    msa_1hot: torch.LongTensor = nn.functional.one_hot(batch["extra_msa"], 23)
+    msa_feat = [
+        msa_1hot,
+        batch["extra_has_deletion"].unsqueeze(-1),
+        batch["extra_deletion_value"].unsqueeze(-1),
+    ]
+    return torch.cat(msa_feat, dim=-1)
+
+
+def torsion_angles_to_frames(
+    r: Rigid,
+    alpha: torch.Tensor,
+    aatype: torch.Tensor,
+    rrgdf: torch.Tensor,
+) -> Rigid:
+    # [*, N, 8, 4, 4]
+    default_4x4 = rrgdf[aatype, ...]
+
+    # [*, N, 8] transformations, i.e.
+    #   One [*, N, 8, 3, 3] rotation matrix and
+    #   One [*, N, 8, 3]    translation matrix
+    default_r = r.from_tensor_4x4(default_4x4)
+
+    bb_rot = alpha.new_zeros((*((1,) * len(alpha.shape[:-1])), 2))
+    bb_rot[..., 1] = 1
+
+    # [*, N, 8, 2]
+    alpha = torch.cat([bb_rot.expand(*alpha.shape[:-2], -1, -1), alpha], dim=-2)
+
+    # [*, N, 8, 3, 3]
+    # Produces rotation matrices of the form:
+    # [
+    #   [1, 0  , 0  ],
+    #   [0, a_2,-a_1],
+    #   [0, a_1, a_2]
+    # ]
+    # This follows the original code rather than the supplement, which uses
+    # different indices.
+
+    all_rots = alpha.new_zeros(default_r.get_rots().get_rot_mats().shape)
+    all_rots[..., 0, 0] = 1
+    all_rots[..., 1, 1] = alpha[..., 1]
+    all_rots[..., 1, 2] = -alpha[..., 0]
+    all_rots[..., 2, 1:] = alpha
+
+    all_frames = default_r.compose(Rigid(Rotation(rot_mats=all_rots), None))
+
+    chi2_frame_to_frame = all_frames[..., 5]
+    chi3_frame_to_frame = all_frames[..., 6]
+    chi4_frame_to_frame = all_frames[..., 7]
+
+    chi1_frame_to_bb = all_frames[..., 4]
+    chi2_frame_to_bb = chi1_frame_to_bb.compose(chi2_frame_to_frame)
+    chi3_frame_to_bb = chi2_frame_to_bb.compose(chi3_frame_to_frame)
+    chi4_frame_to_bb = chi3_frame_to_bb.compose(chi4_frame_to_frame)
+
+    all_frames_to_bb = Rigid.cat(
+        [
+            all_frames[..., :5],
+            chi2_frame_to_bb.unsqueeze(-1),
+            chi3_frame_to_bb.unsqueeze(-1),
+            chi4_frame_to_bb.unsqueeze(-1),
+        ],
+        dim=-1,
+    )
+
+    all_frames_to_global = r[..., None].compose(all_frames_to_bb)
+
+    return all_frames_to_global
+
+
+def frames_and_literature_positions_to_atom14_pos(
+    r: Rigid,
+    aatype: torch.Tensor,
+    default_frames: torch.Tensor,
+    group_idx: torch.Tensor,
+    atom_mask: torch.Tensor,
+    lit_positions: torch.Tensor,
+) -> torch.Tensor:
+    # [*, N, 14]
+    group_mask = group_idx[aatype, ...]
+
+    # [*, N, 14, 8]
+    group_mask_one_hot: torch.LongTensor = nn.functional.one_hot(
+        group_mask,
+        num_classes=default_frames.shape[-3],
+    )
+
+    # [*, N, 14, 8]
+    t_atoms_to_global = r[..., None, :] * group_mask_one_hot
+
+    # [*, N, 14]
+    t_atoms_to_global = t_atoms_to_global.map_tensor_fn(lambda x: torch.sum(x, dim=-1))
+
+    # [*, N, 14, 1]
+    atom_mask = atom_mask[aatype, ...].unsqueeze(-1)
+
+    # [*, N, 14, 3]
+    lit_positions = lit_positions[aatype, ...]
+    pred_positions = t_atoms_to_global.apply(lit_positions)
+    pred_positions = pred_positions * atom_mask
+
+    return pred_positions

venv/lib/python3.10/site-packages/transformers/models/esm/openfold_utils/loss.py

@@ -0,0 +1,105 @@
+# Copyright 2021 AlQuraishi Laboratory
+# Copyright 2021 DeepMind Technologies Limited
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Dict, Optional, Tuple
+
+import torch
+
+
+def _calculate_bin_centers(boundaries: torch.Tensor) -> torch.Tensor:
+    step = boundaries[1] - boundaries[0]
+    bin_centers = boundaries + step / 2
+    bin_centers = torch.cat([bin_centers, (bin_centers[-1] + step).unsqueeze(-1)], dim=0)
+    return bin_centers
+
+
+def _calculate_expected_aligned_error(
+    alignment_confidence_breaks: torch.Tensor,
+    aligned_distance_error_probs: torch.Tensor,
+) -> Tuple[torch.Tensor, torch.Tensor]:
+    bin_centers = _calculate_bin_centers(alignment_confidence_breaks)
+    return (
+        torch.sum(aligned_distance_error_probs * bin_centers, dim=-1),
+        bin_centers[-1],
+    )
+
+
+def compute_predicted_aligned_error(
+    logits: torch.Tensor,
+    max_bin: int = 31,
+    no_bins: int = 64,
+    **kwargs,
+) -> Dict[str, torch.Tensor]:
+    """Computes aligned confidence metrics from logits.
+
+    Args:
+      logits: [*, num_res, num_res, num_bins] the logits output from
+        PredictedAlignedErrorHead.
+      max_bin: Maximum bin value
+      no_bins: Number of bins
+    Returns:
+      aligned_confidence_probs: [*, num_res, num_res, num_bins] the predicted
+        aligned error probabilities over bins for each residue pair.
+      predicted_aligned_error: [*, num_res, num_res] the expected aligned distance
+        error for each pair of residues.
+      max_predicted_aligned_error: [*] the maximum predicted error possible.
+    """
+    boundaries = torch.linspace(0, max_bin, steps=(no_bins - 1), device=logits.device)
+
+    aligned_confidence_probs = torch.nn.functional.softmax(logits, dim=-1)
+    predicted_aligned_error, max_predicted_aligned_error = _calculate_expected_aligned_error(
+        alignment_confidence_breaks=boundaries,
+        aligned_distance_error_probs=aligned_confidence_probs,
+    )
+
+    return {
+        "aligned_confidence_probs": aligned_confidence_probs,
+        "predicted_aligned_error": predicted_aligned_error,
+        "max_predicted_aligned_error": max_predicted_aligned_error,
+    }
+
+
+def compute_tm(
+    logits: torch.Tensor,
+    residue_weights: Optional[torch.Tensor] = None,
+    max_bin: int = 31,
+    no_bins: int = 64,
+    eps: float = 1e-8,
+    **kwargs,
+) -> torch.Tensor:
+    if residue_weights is None:
+        residue_weights = logits.new_ones(logits.shape[-2])
+
+    boundaries = torch.linspace(0, max_bin, steps=(no_bins - 1), device=logits.device)
+
+    bin_centers = _calculate_bin_centers(boundaries)
+    torch.sum(residue_weights)
+    n = logits.shape[-2]
+    clipped_n = max(n, 19)
+
+    d0 = 1.24 * (clipped_n - 15) ** (1.0 / 3) - 1.8
+
+    probs = torch.nn.functional.softmax(logits, dim=-1)
+
+    tm_per_bin = 1.0 / (1 + (bin_centers**2) / (d0**2))
+    predicted_tm_term = torch.sum(probs * tm_per_bin, dim=-1)
+
+    normed_residue_mask = residue_weights / (eps + residue_weights.sum())
+    per_alignment = torch.sum(predicted_tm_term * normed_residue_mask, dim=-1)
+
+    weighted = per_alignment * residue_weights
+
+    argmax = (weighted == torch.max(weighted)).nonzero()[0]
+    return per_alignment[tuple(argmax)]

venv/lib/python3.10/site-packages/transformers/models/esm/openfold_utils/protein.py

@@ -0,0 +1,329 @@
+# Copyright 2021 AlQuraishi Laboratory
+# Copyright 2021 DeepMind Technologies Limited
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Protein data type."""
+import dataclasses
+import re
+import string
+from typing import Any, Dict, Iterator, List, Mapping, Optional, Sequence, Tuple
+
+import numpy as np
+
+from . import residue_constants
+
+
+FeatureDict = Mapping[str, np.ndarray]
+ModelOutput = Mapping[str, Any]  # Is a nested dict.
+PICO_TO_ANGSTROM = 0.01
+
+
+@dataclasses.dataclass(frozen=True)
+class Protein:
+    """Protein structure representation."""
+
+    # Cartesian coordinates of atoms in angstroms. The atom types correspond to
+    # residue_constants.atom_types, i.e. the first three are N, CA, CB.
+    atom_positions: np.ndarray  # [num_res, num_atom_type, 3]
+
+    # Amino-acid type for each residue represented as an integer between 0 and
+    # 20, where 20 is 'X'.
+    aatype: np.ndarray  # [num_res]
+
+    # Binary float mask to indicate presence of a particular atom. 1.0 if an atom
+    # is present and 0.0 if not. This should be used for loss masking.
+    atom_mask: np.ndarray  # [num_res, num_atom_type]
+
+    # Residue index as used in PDB. It is not necessarily continuous or 0-indexed.
+    residue_index: np.ndarray  # [num_res]
+
+    # B-factors, or temperature factors, of each residue (in sq. angstroms units),
+    # representing the displacement of the residue from its ground truth mean
+    # value.
+    b_factors: np.ndarray  # [num_res, num_atom_type]
+
+    # Chain indices for multi-chain predictions
+    chain_index: Optional[np.ndarray] = None
+
+    # Optional remark about the protein. Included as a comment in output PDB
+    # files
+    remark: Optional[str] = None
+
+    # Templates used to generate this protein (prediction-only)
+    parents: Optional[Sequence[str]] = None
+
+    # Chain corresponding to each parent
+    parents_chain_index: Optional[Sequence[int]] = None
+
+
+def from_proteinnet_string(proteinnet_str: str) -> Protein:
+    tag_re = r"(\[[A-Z]+\]\n)"
+    tags: List[str] = [tag.strip() for tag in re.split(tag_re, proteinnet_str) if len(tag) > 0]
+    groups: Iterator[Tuple[str, List[str]]] = zip(tags[0::2], [l.split("\n") for l in tags[1::2]])
+
+    atoms: List[str] = ["N", "CA", "C"]
+    aatype = None
+    atom_positions = None
+    atom_mask = None
+    for g in groups:
+        if "[PRIMARY]" == g[0]:
+            seq = g[1][0].strip()
+            for i in range(len(seq)):
+                if seq[i] not in residue_constants.restypes:
+                    seq[i] = "X"  # FIXME: strings are immutable
+            aatype = np.array(
+                [residue_constants.restype_order.get(res_symbol, residue_constants.restype_num) for res_symbol in seq]
+            )
+        elif "[TERTIARY]" == g[0]:
+            tertiary: List[List[float]] = []
+            for axis in range(3):
+                tertiary.append(list(map(float, g[1][axis].split())))
+            tertiary_np = np.array(tertiary)
+            atom_positions = np.zeros((len(tertiary[0]) // 3, residue_constants.atom_type_num, 3)).astype(np.float32)
+            for i, atom in enumerate(atoms):
+                atom_positions[:, residue_constants.atom_order[atom], :] = np.transpose(tertiary_np[:, i::3])
+            atom_positions *= PICO_TO_ANGSTROM
+        elif "[MASK]" == g[0]:
+            mask = np.array(list(map({"-": 0, "+": 1}.get, g[1][0].strip())))
+            atom_mask = np.zeros(
+                (
+                    len(mask),
+                    residue_constants.atom_type_num,
+                )
+            ).astype(np.float32)
+            for i, atom in enumerate(atoms):
+                atom_mask[:, residue_constants.atom_order[atom]] = 1
+            atom_mask *= mask[..., None]
+
+    assert aatype is not None
+
+    return Protein(
+        atom_positions=atom_positions,
+        atom_mask=atom_mask,
+        aatype=aatype,
+        residue_index=np.arange(len(aatype)),
+        b_factors=None,
+    )
+
+
+def get_pdb_headers(prot: Protein, chain_id: int = 0) -> List[str]:
+    pdb_headers: List[str] = []
+
+    remark = prot.remark
+    if remark is not None:
+        pdb_headers.append(f"REMARK {remark}")
+
+    parents = prot.parents
+    parents_chain_index = prot.parents_chain_index
+    if parents is not None and parents_chain_index is not None:
+        parents = [p for i, p in zip(parents_chain_index, parents) if i == chain_id]
+
+    if parents is None or len(parents) == 0:
+        parents = ["N/A"]
+
+    pdb_headers.append(f"PARENT {' '.join(parents)}")
+
+    return pdb_headers
+
+
+def add_pdb_headers(prot: Protein, pdb_str: str) -> str:
+    """Add pdb headers to an existing PDB string. Useful during multi-chain
+    recycling
+    """
+    out_pdb_lines: List[str] = []
+    lines = pdb_str.split("\n")
+
+    remark = prot.remark
+    if remark is not None:
+        out_pdb_lines.append(f"REMARK {remark}")
+
+    parents_per_chain: List[List[str]]
+    if prot.parents is not None and len(prot.parents) > 0:
+        parents_per_chain = []
+        if prot.parents_chain_index is not None:
+            parent_dict: Dict[str, List[str]] = {}
+            for p, i in zip(prot.parents, prot.parents_chain_index):
+                parent_dict.setdefault(str(i), [])
+                parent_dict[str(i)].append(p)
+
+            max_idx = max([int(chain_idx) for chain_idx in parent_dict])
+            for i in range(max_idx + 1):
+                chain_parents = parent_dict.get(str(i), ["N/A"])
+                parents_per_chain.append(chain_parents)
+        else:
+            parents_per_chain.append(list(prot.parents))
+    else:
+        parents_per_chain = [["N/A"]]
+
+    def make_parent_line(p: Sequence[str]) -> str:
+        return f"PARENT {' '.join(p)}"
+
+    out_pdb_lines.append(make_parent_line(parents_per_chain[0]))
+
+    chain_counter = 0
+    for i, l in enumerate(lines):
+        if "PARENT" not in l and "REMARK" not in l:
+            out_pdb_lines.append(l)
+        if "TER" in l and "END" not in lines[i + 1]:
+            chain_counter += 1
+            if not chain_counter >= len(parents_per_chain):
+                chain_parents = parents_per_chain[chain_counter]
+            else:
+                chain_parents = ["N/A"]
+
+            out_pdb_lines.append(make_parent_line(chain_parents))
+
+    return "\n".join(out_pdb_lines)
+
+
+def to_pdb(prot: Protein) -> str:
+    """Converts a `Protein` instance to a PDB string.
+
+    Args:
+      prot: The protein to convert to PDB.
+
+    Returns:
+      PDB string.
+    """
+    restypes = residue_constants.restypes + ["X"]
+
+    def res_1to3(r: int) -> str:
+        return residue_constants.restype_1to3.get(restypes[r], "UNK")
+
+    atom_types = residue_constants.atom_types
+
+    pdb_lines: List[str] = []
+
+    atom_mask = prot.atom_mask
+    aatype = prot.aatype
+    atom_positions = prot.atom_positions
+    residue_index = prot.residue_index.astype(np.int32)
+    b_factors = prot.b_factors
+    chain_index = prot.chain_index
+
+    if np.any(aatype > residue_constants.restype_num):
+        raise ValueError("Invalid aatypes.")
+
+    headers = get_pdb_headers(prot)
+    if len(headers) > 0:
+        pdb_lines.extend(headers)
+
+    n = aatype.shape[0]
+    atom_index = 1
+    prev_chain_index = 0
+    chain_tags = string.ascii_uppercase
+    chain_tag = None
+    # Add all atom sites.
+    for i in range(n):
+        res_name_3 = res_1to3(aatype[i])
+        for atom_name, pos, mask, b_factor in zip(atom_types, atom_positions[i], atom_mask[i], b_factors[i]):
+            if mask < 0.5:
+                continue
+
+            record_type = "ATOM"
+            name = atom_name if len(atom_name) == 4 else f" {atom_name}"
+            alt_loc = ""
+            insertion_code = ""
+            occupancy = 1.00
+            element = atom_name[0]  # Protein supports only C, N, O, S, this works.
+            charge = ""
+
+            chain_tag = "A"
+            if chain_index is not None:
+                chain_tag = chain_tags[chain_index[i]]
+
+            # PDB is a columnar format, every space matters here!
+            atom_line = (
+                f"{record_type:<6}{atom_index:>5} {name:<4}{alt_loc:>1}"
+                f"{res_name_3:>3} {chain_tag:>1}"
+                f"{residue_index[i]:>4}{insertion_code:>1}   "
+                f"{pos[0]:>8.3f}{pos[1]:>8.3f}{pos[2]:>8.3f}"
+                f"{occupancy:>6.2f}{b_factor:>6.2f}          "
+                f"{element:>2}{charge:>2}"
+            )
+            pdb_lines.append(atom_line)
+            atom_index += 1
+
+        should_terminate = i == n - 1
+        if chain_index is not None:
+            if i != n - 1 and chain_index[i + 1] != prev_chain_index:
+                should_terminate = True
+                prev_chain_index = chain_index[i + 1]
+
+        if should_terminate:
+            # Close the chain.
+            chain_end = "TER"
+            chain_termination_line = (
+                f"{chain_end:<6}{atom_index:>5}      {res_1to3(aatype[i]):>3} {chain_tag:>1}{residue_index[i]:>4}"
+            )
+            pdb_lines.append(chain_termination_line)
+            atom_index += 1
+
+            if i != n - 1:
+                # "prev" is a misnomer here. This happens at the beginning of
+                # each new chain.
+                pdb_lines.extend(get_pdb_headers(prot, prev_chain_index))
+
+    pdb_lines.append("END")
+    pdb_lines.append("")
+    return "\n".join(pdb_lines)
+
+
+def ideal_atom_mask(prot: Protein) -> np.ndarray:
+    """Computes an ideal atom mask.
+
+    `Protein.atom_mask` typically is defined according to the atoms that are reported in the PDB. This function
+    computes a mask according to heavy atoms that should be present in the given sequence of amino acids.
+
+    Args:
+      prot: `Protein` whose fields are `numpy.ndarray` objects.
+
+    Returns:
+      An ideal atom mask.
+    """
+    return residue_constants.STANDARD_ATOM_MASK[prot.aatype]
+
+
+def from_prediction(
+    features: FeatureDict,
+    result: ModelOutput,
+    b_factors: Optional[np.ndarray] = None,
+    chain_index: Optional[np.ndarray] = None,
+    remark: Optional[str] = None,
+    parents: Optional[Sequence[str]] = None,
+    parents_chain_index: Optional[Sequence[int]] = None,
+) -> Protein:
+    """Assembles a protein from a prediction.
+
+    Args:
+      features: Dictionary holding model inputs.
+      result: Dictionary holding model outputs.
+      b_factors: (Optional) B-factors to use for the protein.
+      chain_index: (Optional) Chain indices for multi-chain predictions
+      remark: (Optional) Remark about the prediction
+      parents: (Optional) List of template names
+    Returns:
+      A protein instance.
+    """
+    return Protein(
+        aatype=features["aatype"],
+        atom_positions=result["final_atom_positions"],
+        atom_mask=result["final_atom_mask"],
+        residue_index=features["residue_index"] + 1,
+        b_factors=b_factors if b_factors is not None else np.zeros_like(result["final_atom_mask"]),
+        chain_index=chain_index,
+        remark=remark,
+        parents=parents,
+        parents_chain_index=parents_chain_index,
+    )

venv/lib/python3.10/site-packages/transformers/models/esm/openfold_utils/residue_constants.py

@@ -0,0 +1,983 @@
+# Copyright 2021 AlQuraishi Laboratory
+# Copyright 2021 DeepMind Technologies Limited
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Constants used in AlphaFold."""
+
+import collections
+import copy
+import functools
+from importlib import resources
+from typing import Dict, List, Mapping, Sequence, Tuple
+
+import numpy as np
+
+
+# Internal import (35fd).
+
+
+# Distance from one CA to next CA [trans configuration: omega = 180].
+ca_ca = 3.80209737096
+
+# Format: The list for each AA type contains chi1, chi2, chi3, chi4 in
+# this order (or a relevant subset from chi1 onwards). ALA and GLY don't have
+# chi angles so their chi angle lists are empty.
+chi_angles_atoms: Dict[str, List[List[str]]] = {
+    "ALA": [],
+    # Chi5 in arginine is always 0 +- 5 degrees, so ignore it.
+    "ARG": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "CD"], ["CB", "CG", "CD", "NE"], ["CG", "CD", "NE", "CZ"]],
+    "ASN": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "OD1"]],
+    "ASP": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "OD1"]],
+    "CYS": [["N", "CA", "CB", "SG"]],
+    "GLN": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "CD"], ["CB", "CG", "CD", "OE1"]],
+    "GLU": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "CD"], ["CB", "CG", "CD", "OE1"]],
+    "GLY": [],
+    "HIS": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "ND1"]],
+    "ILE": [["N", "CA", "CB", "CG1"], ["CA", "CB", "CG1", "CD1"]],
+    "LEU": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "CD1"]],
+    "LYS": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "CD"], ["CB", "CG", "CD", "CE"], ["CG", "CD", "CE", "NZ"]],
+    "MET": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "SD"], ["CB", "CG", "SD", "CE"]],
+    "PHE": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "CD1"]],
+    "PRO": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "CD"]],
+    "SER": [["N", "CA", "CB", "OG"]],
+    "THR": [["N", "CA", "CB", "OG1"]],
+    "TRP": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "CD1"]],
+    "TYR": [["N", "CA", "CB", "CG"], ["CA", "CB", "CG", "CD1"]],
+    "VAL": [["N", "CA", "CB", "CG1"]],
+}
+
+# If chi angles given in fixed-length array, this matrix determines how to mask
+# them for each AA type. The order is as per restype_order (see below).
+chi_angles_mask: List[List[float]] = [
+    [0.0, 0.0, 0.0, 0.0],  # ALA
+    [1.0, 1.0, 1.0, 1.0],  # ARG
+    [1.0, 1.0, 0.0, 0.0],  # ASN
+    [1.0, 1.0, 0.0, 0.0],  # ASP
+    [1.0, 0.0, 0.0, 0.0],  # CYS
+    [1.0, 1.0, 1.0, 0.0],  # GLN
+    [1.0, 1.0, 1.0, 0.0],  # GLU
+    [0.0, 0.0, 0.0, 0.0],  # GLY
+    [1.0, 1.0, 0.0, 0.0],  # HIS
+    [1.0, 1.0, 0.0, 0.0],  # ILE
+    [1.0, 1.0, 0.0, 0.0],  # LEU
+    [1.0, 1.0, 1.0, 1.0],  # LYS
+    [1.0, 1.0, 1.0, 0.0],  # MET
+    [1.0, 1.0, 0.0, 0.0],  # PHE
+    [1.0, 1.0, 0.0, 0.0],  # PRO
+    [1.0, 0.0, 0.0, 0.0],  # SER
+    [1.0, 0.0, 0.0, 0.0],  # THR
+    [1.0, 1.0, 0.0, 0.0],  # TRP
+    [1.0, 1.0, 0.0, 0.0],  # TYR
+    [1.0, 0.0, 0.0, 0.0],  # VAL
+]
+
+# The following chi angles are pi periodic: they can be rotated by a multiple
+# of pi without affecting the structure.
+chi_pi_periodic: List[List[float]] = [
+    [0.0, 0.0, 0.0, 0.0],  # ALA
+    [0.0, 0.0, 0.0, 0.0],  # ARG
+    [0.0, 0.0, 0.0, 0.0],  # ASN
+    [0.0, 1.0, 0.0, 0.0],  # ASP
+    [0.0, 0.0, 0.0, 0.0],  # CYS
+    [0.0, 0.0, 0.0, 0.0],  # GLN
+    [0.0, 0.0, 1.0, 0.0],  # GLU
+    [0.0, 0.0, 0.0, 0.0],  # GLY
+    [0.0, 0.0, 0.0, 0.0],  # HIS
+    [0.0, 0.0, 0.0, 0.0],  # ILE
+    [0.0, 0.0, 0.0, 0.0],  # LEU
+    [0.0, 0.0, 0.0, 0.0],  # LYS
+    [0.0, 0.0, 0.0, 0.0],  # MET
+    [0.0, 1.0, 0.0, 0.0],  # PHE
+    [0.0, 0.0, 0.0, 0.0],  # PRO
+    [0.0, 0.0, 0.0, 0.0],  # SER
+    [0.0, 0.0, 0.0, 0.0],  # THR
+    [0.0, 0.0, 0.0, 0.0],  # TRP
+    [0.0, 1.0, 0.0, 0.0],  # TYR
+    [0.0, 0.0, 0.0, 0.0],  # VAL
+    [0.0, 0.0, 0.0, 0.0],  # UNK
+]
+
+# Atoms positions relative to the 8 rigid groups, defined by the pre-omega, phi,
+# psi and chi angles:
+# 0: 'backbone group',
+# 1: 'pre-omega-group', (empty)
+# 2: 'phi-group', (currently empty, because it defines only hydrogens)
+# 3: 'psi-group',
+# 4,5,6,7: 'chi1,2,3,4-group'
+# The atom positions are relative to the axis-end-atom of the corresponding
+# rotation axis. The x-axis is in direction of the rotation axis, and the y-axis
+# is defined such that the dihedral-angle-definiting atom (the last entry in
+# chi_angles_atoms above) is in the xy-plane (with a positive y-coordinate).
+# format: [atomname, group_idx, rel_position]
+rigid_group_atom_positions: Dict[str, List[Tuple[str, int, Tuple[float, float, float]]]] = {
+    "ALA": [
+        ("N", 0, (-0.525, 1.363, 0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.526, -0.000, -0.000)),
+        ("CB", 0, (-0.529, -0.774, -1.205)),
+        ("O", 3, (0.627, 1.062, 0.000)),
+    ],
+    "ARG": [
+        ("N", 0, (-0.524, 1.362, -0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.525, -0.000, -0.000)),
+        ("CB", 0, (-0.524, -0.778, -1.209)),
+        ("O", 3, (0.626, 1.062, 0.000)),
+        ("CG", 4, (0.616, 1.390, -0.000)),
+        ("CD", 5, (0.564, 1.414, 0.000)),
+        ("NE", 6, (0.539, 1.357, -0.000)),
+        ("NH1", 7, (0.206, 2.301, 0.000)),
+        ("NH2", 7, (2.078, 0.978, -0.000)),
+        ("CZ", 7, (0.758, 1.093, -0.000)),
+    ],
+    "ASN": [
+        ("N", 0, (-0.536, 1.357, 0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.526, -0.000, -0.000)),
+        ("CB", 0, (-0.531, -0.787, -1.200)),
+        ("O", 3, (0.625, 1.062, 0.000)),
+        ("CG", 4, (0.584, 1.399, 0.000)),
+        ("ND2", 5, (0.593, -1.188, 0.001)),
+        ("OD1", 5, (0.633, 1.059, 0.000)),
+    ],
+    "ASP": [
+        ("N", 0, (-0.525, 1.362, -0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.527, 0.000, -0.000)),
+        ("CB", 0, (-0.526, -0.778, -1.208)),
+        ("O", 3, (0.626, 1.062, -0.000)),
+        ("CG", 4, (0.593, 1.398, -0.000)),
+        ("OD1", 5, (0.610, 1.091, 0.000)),
+        ("OD2", 5, (0.592, -1.101, -0.003)),
+    ],
+    "CYS": [
+        ("N", 0, (-0.522, 1.362, -0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.524, 0.000, 0.000)),
+        ("CB", 0, (-0.519, -0.773, -1.212)),
+        ("O", 3, (0.625, 1.062, -0.000)),
+        ("SG", 4, (0.728, 1.653, 0.000)),
+    ],
+    "GLN": [
+        ("N", 0, (-0.526, 1.361, -0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.526, 0.000, 0.000)),
+        ("CB", 0, (-0.525, -0.779, -1.207)),
+        ("O", 3, (0.626, 1.062, -0.000)),
+        ("CG", 4, (0.615, 1.393, 0.000)),
+        ("CD", 5, (0.587, 1.399, -0.000)),
+        ("NE2", 6, (0.593, -1.189, -0.001)),
+        ("OE1", 6, (0.634, 1.060, 0.000)),
+    ],
+    "GLU": [
+        ("N", 0, (-0.528, 1.361, 0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.526, -0.000, -0.000)),
+        ("CB", 0, (-0.526, -0.781, -1.207)),
+        ("O", 3, (0.626, 1.062, 0.000)),
+        ("CG", 4, (0.615, 1.392, 0.000)),
+        ("CD", 5, (0.600, 1.397, 0.000)),
+        ("OE1", 6, (0.607, 1.095, -0.000)),
+        ("OE2", 6, (0.589, -1.104, -0.001)),
+    ],
+    "GLY": [
+        ("N", 0, (-0.572, 1.337, 0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.517, -0.000, -0.000)),
+        ("O", 3, (0.626, 1.062, -0.000)),
+    ],
+    "HIS": [
+        ("N", 0, (-0.527, 1.360, 0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.525, 0.000, 0.000)),
+        ("CB", 0, (-0.525, -0.778, -1.208)),
+        ("O", 3, (0.625, 1.063, 0.000)),
+        ("CG", 4, (0.600, 1.370, -0.000)),
+        ("CD2", 5, (0.889, -1.021, 0.003)),
+        ("ND1", 5, (0.744, 1.160, -0.000)),
+        ("CE1", 5, (2.030, 0.851, 0.002)),
+        ("NE2", 5, (2.145, -0.466, 0.004)),
+    ],
+    "ILE": [
+        ("N", 0, (-0.493, 1.373, -0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.527, -0.000, -0.000)),
+        ("CB", 0, (-0.536, -0.793, -1.213)),
+        ("O", 3, (0.627, 1.062, -0.000)),
+        ("CG1", 4, (0.534, 1.437, -0.000)),
+        ("CG2", 4, (0.540, -0.785, -1.199)),
+        ("CD1", 5, (0.619, 1.391, 0.000)),
+    ],
+    "LEU": [
+        ("N", 0, (-0.520, 1.363, 0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.525, -0.000, -0.000)),
+        ("CB", 0, (-0.522, -0.773, -1.214)),
+        ("O", 3, (0.625, 1.063, -0.000)),
+        ("CG", 4, (0.678, 1.371, 0.000)),
+        ("CD1", 5, (0.530, 1.430, -0.000)),
+        ("CD2", 5, (0.535, -0.774, 1.200)),
+    ],
+    "LYS": [
+        ("N", 0, (-0.526, 1.362, -0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.526, 0.000, 0.000)),
+        ("CB", 0, (-0.524, -0.778, -1.208)),
+        ("O", 3, (0.626, 1.062, -0.000)),
+        ("CG", 4, (0.619, 1.390, 0.000)),
+        ("CD", 5, (0.559, 1.417, 0.000)),
+        ("CE", 6, (0.560, 1.416, 0.000)),
+        ("NZ", 7, (0.554, 1.387, 0.000)),
+    ],
+    "MET": [
+        ("N", 0, (-0.521, 1.364, -0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.525, 0.000, 0.000)),
+        ("CB", 0, (-0.523, -0.776, -1.210)),
+        ("O", 3, (0.625, 1.062, -0.000)),
+        ("CG", 4, (0.613, 1.391, -0.000)),
+        ("SD", 5, (0.703, 1.695, 0.000)),
+        ("CE", 6, (0.320, 1.786, -0.000)),
+    ],
+    "PHE": [
+        ("N", 0, (-0.518, 1.363, 0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.524, 0.000, -0.000)),
+        ("CB", 0, (-0.525, -0.776, -1.212)),
+        ("O", 3, (0.626, 1.062, -0.000)),
+        ("CG", 4, (0.607, 1.377, 0.000)),
+        ("CD1", 5, (0.709, 1.195, -0.000)),
+        ("CD2", 5, (0.706, -1.196, 0.000)),
+        ("CE1", 5, (2.102, 1.198, -0.000)),
+        ("CE2", 5, (2.098, -1.201, -0.000)),
+        ("CZ", 5, (2.794, -0.003, -0.001)),
+    ],
+    "PRO": [
+        ("N", 0, (-0.566, 1.351, -0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.527, -0.000, 0.000)),
+        ("CB", 0, (-0.546, -0.611, -1.293)),
+        ("O", 3, (0.621, 1.066, 0.000)),
+        ("CG", 4, (0.382, 1.445, 0.0)),
+        # ('CD', 5, (0.427, 1.440, 0.0)),
+        ("CD", 5, (0.477, 1.424, 0.0)),  # manually made angle 2 degrees larger
+    ],
+    "SER": [
+        ("N", 0, (-0.529, 1.360, -0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.525, -0.000, -0.000)),
+        ("CB", 0, (-0.518, -0.777, -1.211)),
+        ("O", 3, (0.626, 1.062, -0.000)),
+        ("OG", 4, (0.503, 1.325, 0.000)),
+    ],
+    "THR": [
+        ("N", 0, (-0.517, 1.364, 0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.526, 0.000, -0.000)),
+        ("CB", 0, (-0.516, -0.793, -1.215)),
+        ("O", 3, (0.626, 1.062, 0.000)),
+        ("CG2", 4, (0.550, -0.718, -1.228)),
+        ("OG1", 4, (0.472, 1.353, 0.000)),
+    ],
+    "TRP": [
+        ("N", 0, (-0.521, 1.363, 0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.525, -0.000, 0.000)),
+        ("CB", 0, (-0.523, -0.776, -1.212)),
+        ("O", 3, (0.627, 1.062, 0.000)),
+        ("CG", 4, (0.609, 1.370, -0.000)),
+        ("CD1", 5, (0.824, 1.091, 0.000)),
+        ("CD2", 5, (0.854, -1.148, -0.005)),
+        ("CE2", 5, (2.186, -0.678, -0.007)),
+        ("CE3", 5, (0.622, -2.530, -0.007)),
+        ("NE1", 5, (2.140, 0.690, -0.004)),
+        ("CH2", 5, (3.028, -2.890, -0.013)),
+        ("CZ2", 5, (3.283, -1.543, -0.011)),
+        ("CZ3", 5, (1.715, -3.389, -0.011)),
+    ],
+    "TYR": [
+        ("N", 0, (-0.522, 1.362, 0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.524, -0.000, -0.000)),
+        ("CB", 0, (-0.522, -0.776, -1.213)),
+        ("O", 3, (0.627, 1.062, -0.000)),
+        ("CG", 4, (0.607, 1.382, -0.000)),
+        ("CD1", 5, (0.716, 1.195, -0.000)),
+        ("CD2", 5, (0.713, -1.194, -0.001)),
+        ("CE1", 5, (2.107, 1.200, -0.002)),
+        ("CE2", 5, (2.104, -1.201, -0.003)),
+        ("OH", 5, (4.168, -0.002, -0.005)),
+        ("CZ", 5, (2.791, -0.001, -0.003)),
+    ],
+    "VAL": [
+        ("N", 0, (-0.494, 1.373, -0.000)),
+        ("CA", 0, (0.000, 0.000, 0.000)),
+        ("C", 0, (1.527, -0.000, -0.000)),
+        ("CB", 0, (-0.533, -0.795, -1.213)),
+        ("O", 3, (0.627, 1.062, -0.000)),
+        ("CG1", 4, (0.540, 1.429, -0.000)),
+        ("CG2", 4, (0.533, -0.776, 1.203)),
+    ],
+}
+
+# A list of atoms (excluding hydrogen) for each AA type. PDB naming convention.
+residue_atoms: Dict[str, List[str]] = {
+    "ALA": ["C", "CA", "CB", "N", "O"],
+    "ARG": ["C", "CA", "CB", "CG", "CD", "CZ", "N", "NE", "O", "NH1", "NH2"],
+    "ASP": ["C", "CA", "CB", "CG", "N", "O", "OD1", "OD2"],
+    "ASN": ["C", "CA", "CB", "CG", "N", "ND2", "O", "OD1"],
+    "CYS": ["C", "CA", "CB", "N", "O", "SG"],
+    "GLU": ["C", "CA", "CB", "CG", "CD", "N", "O", "OE1", "OE2"],
+    "GLN": ["C", "CA", "CB", "CG", "CD", "N", "NE2", "O", "OE1"],
+    "GLY": ["C", "CA", "N", "O"],
+    "HIS": ["C", "CA", "CB", "CG", "CD2", "CE1", "N", "ND1", "NE2", "O"],
+    "ILE": ["C", "CA", "CB", "CG1", "CG2", "CD1", "N", "O"],
+    "LEU": ["C", "CA", "CB", "CG", "CD1", "CD2", "N", "O"],
+    "LYS": ["C", "CA", "CB", "CG", "CD", "CE", "N", "NZ", "O"],
+    "MET": ["C", "CA", "CB", "CG", "CE", "N", "O", "SD"],
+    "PHE": ["C", "CA", "CB", "CG", "CD1", "CD2", "CE1", "CE2", "CZ", "N", "O"],
+    "PRO": ["C", "CA", "CB", "CG", "CD", "N", "O"],
+    "SER": ["C", "CA", "CB", "N", "O", "OG"],
+    "THR": ["C", "CA", "CB", "CG2", "N", "O", "OG1"],
+    "TRP": ["C", "CA", "CB", "CG", "CD1", "CD2", "CE2", "CE3", "CZ2", "CZ3", "CH2", "N", "NE1", "O"],
+    "TYR": ["C", "CA", "CB", "CG", "CD1", "CD2", "CE1", "CE2", "CZ", "N", "O", "OH"],
+    "VAL": ["C", "CA", "CB", "CG1", "CG2", "N", "O"],
+}
+
+# Naming swaps for ambiguous atom names.
+# Due to symmetries in the amino acids the naming of atoms is ambiguous in
+# 4 of the 20 amino acids.
+# (The LDDT paper lists 7 amino acids as ambiguous, but the naming ambiguities
+# in LEU, VAL and ARG can be resolved by using the 3d constellations of
+# the 'ambiguous' atoms and their neighbours)
+# TODO: ^ interpret this
+residue_atom_renaming_swaps: Dict[str, Dict[str, str]] = {
+    "ASP": {"OD1": "OD2"},
+    "GLU": {"OE1": "OE2"},
+    "PHE": {"CD1": "CD2", "CE1": "CE2"},
+    "TYR": {"CD1": "CD2", "CE1": "CE2"},
+}
+
+# Van der Waals radii [Angstroem] of the atoms (from Wikipedia)
+van_der_waals_radius: Dict[str, float] = {
+    "C": 1.7,
+    "N": 1.55,
+    "O": 1.52,
+    "S": 1.8,
+}
+
+Bond = collections.namedtuple("Bond", ["atom1_name", "atom2_name", "length", "stddev"])
+BondAngle = collections.namedtuple(
+    "BondAngle",
+    ["atom1_name", "atom2_name", "atom3name", "angle_rad", "stddev"],
+)
+
+
+def map_structure_with_atom_order(in_list: list, first_call: bool = True) -> list:
+    # Maps strings in a nested list structure to their corresponding index in atom_order
+    if first_call:
+        in_list = copy.deepcopy(in_list)
+    for i in range(len(in_list)):
+        if isinstance(in_list[i], list):
+            in_list[i] = map_structure_with_atom_order(in_list[i], first_call=False)
+        elif isinstance(in_list[i], str):
+            in_list[i] = atom_order[in_list[i]]
+        else:
+            raise ValueError("Unexpected type when mapping nested lists!")
+    return in_list
+
+
+@functools.lru_cache(maxsize=None)
+def load_stereo_chemical_props() -> (
+    Tuple[
+        Mapping[str, List[Bond]],
+        Mapping[str, List[Bond]],
+        Mapping[str, List[BondAngle]],
+    ]
+):
+    """Load stereo_chemical_props.txt into a nice structure.
+
+    Load literature values for bond lengths and bond angles and translate bond angles into the length of the opposite
+    edge of the triangle ("residue_virtual_bonds").
+
+    Returns:
+      residue_bonds: dict that maps resname --> list of Bond tuples residue_virtual_bonds: dict that maps resname -->
+      list of Bond tuples residue_bond_angles: dict that maps resname --> list of BondAngle tuples
+    """
+    # TODO: this file should be downloaded in a setup script
+    stereo_chemical_props = resources.read_text("openfold.resources", "stereo_chemical_props.txt")
+
+    lines_iter = iter(stereo_chemical_props.splitlines())
+    # Load bond lengths.
+    residue_bonds: Dict[str, List[Bond]] = {}
+    next(lines_iter)  # Skip header line.
+    for line in lines_iter:
+        if line.strip() == "-":
+            break
+        bond, resname, bond_length, stddev = line.split()
+        atom1, atom2 = bond.split("-")
+        if resname not in residue_bonds:
+            residue_bonds[resname] = []
+        residue_bonds[resname].append(Bond(atom1, atom2, float(bond_length), float(stddev)))
+    residue_bonds["UNK"] = []
+
+    # Load bond angles.
+    residue_bond_angles: Dict[str, List[BondAngle]] = {}
+    next(lines_iter)  # Skip empty line.
+    next(lines_iter)  # Skip header line.
+    for line in lines_iter:
+        if line.strip() == "-":
+            break
+        bond, resname, angle_degree, stddev_degree = line.split()
+        atom1, atom2, atom3 = bond.split("-")
+        if resname not in residue_bond_angles:
+            residue_bond_angles[resname] = []
+        residue_bond_angles[resname].append(
+            BondAngle(
+                atom1,
+                atom2,
+                atom3,
+                float(angle_degree) / 180.0 * np.pi,
+                float(stddev_degree) / 180.0 * np.pi,
+            )
+        )
+    residue_bond_angles["UNK"] = []
+
+    def make_bond_key(atom1_name: str, atom2_name: str) -> str:
+        """Unique key to lookup bonds."""
+        return "-".join(sorted([atom1_name, atom2_name]))
+
+    # Translate bond angles into distances ("virtual bonds").
+    residue_virtual_bonds: Dict[str, List[Bond]] = {}
+    for resname, bond_angles in residue_bond_angles.items():
+        # Create a fast lookup dict for bond lengths.
+        bond_cache: Dict[str, Bond] = {}
+        for b in residue_bonds[resname]:
+            bond_cache[make_bond_key(b.atom1_name, b.atom2_name)] = b
+        residue_virtual_bonds[resname] = []
+        for ba in bond_angles:
+            bond1 = bond_cache[make_bond_key(ba.atom1_name, ba.atom2_name)]
+            bond2 = bond_cache[make_bond_key(ba.atom2_name, ba.atom3name)]
+
+            # Compute distance between atom1 and atom3 using the law of cosines
+            # c^2 = a^2 + b^2 - 2ab*cos(gamma).
+            gamma = ba.angle_rad
+            length = np.sqrt(bond1.length**2 + bond2.length**2 - 2 * bond1.length * bond2.length * np.cos(gamma))
+
+            # Propagation of uncertainty assuming uncorrelated errors.
+            dl_outer = 0.5 / length
+            dl_dgamma = (2 * bond1.length * bond2.length * np.sin(gamma)) * dl_outer
+            dl_db1 = (2 * bond1.length - 2 * bond2.length * np.cos(gamma)) * dl_outer
+            dl_db2 = (2 * bond2.length - 2 * bond1.length * np.cos(gamma)) * dl_outer
+            stddev = np.sqrt(
+                (dl_dgamma * ba.stddev) ** 2 + (dl_db1 * bond1.stddev) ** 2 + (dl_db2 * bond2.stddev) ** 2
+            )
+            residue_virtual_bonds[resname].append(Bond(ba.atom1_name, ba.atom3name, length, stddev))
+
+    return (residue_bonds, residue_virtual_bonds, residue_bond_angles)
+
+
+# Between-residue bond lengths for general bonds (first element) and for Proline
+# (second element).
+between_res_bond_length_c_n: Tuple[float, float] = (1.329, 1.341)
+between_res_bond_length_stddev_c_n: Tuple[float, float] = (0.014, 0.016)
+
+# Between-residue cos_angles.
+between_res_cos_angles_c_n_ca: Tuple[float, float] = (-0.5203, 0.0353)  # degrees: 121.352 +- 2.315
+between_res_cos_angles_ca_c_n: Tuple[float, float] = (-0.4473, 0.0311)  # degrees: 116.568 +- 1.995
+
+# This mapping is used when we need to store atom data in a format that requires
+# fixed atom data size for every residue (e.g. a numpy array).
+atom_types: List[str] = [
+    "N",
+    "CA",
+    "C",
+    "CB",
+    "O",
+    "CG",
+    "CG1",
+    "CG2",
+    "OG",
+    "OG1",
+    "SG",
+    "CD",
+    "CD1",
+    "CD2",
+    "ND1",
+    "ND2",
+    "OD1",
+    "OD2",
+    "SD",
+    "CE",
+    "CE1",
+    "CE2",
+    "CE3",
+    "NE",
+    "NE1",
+    "NE2",
+    "OE1",
+    "OE2",
+    "CH2",
+    "NH1",
+    "NH2",
+    "OH",
+    "CZ",
+    "CZ2",
+    "CZ3",
+    "NZ",
+    "OXT",
+]
+atom_order: Dict[str, int] = {atom_type: i for i, atom_type in enumerate(atom_types)}
+atom_type_num = len(atom_types)  # := 37.
+
+# A compact atom encoding with 14 columns
+# pylint: disable=line-too-long
+# pylint: disable=bad-whitespace
+restype_name_to_atom14_names: Dict[str, List[str]] = {
+    "ALA": ["N", "CA", "C", "O", "CB", "", "", "", "", "", "", "", "", ""],
+    "ARG": ["N", "CA", "C", "O", "CB", "CG", "CD", "NE", "CZ", "NH1", "NH2", "", "", ""],
+    "ASN": ["N", "CA", "C", "O", "CB", "CG", "OD1", "ND2", "", "", "", "", "", ""],
+    "ASP": ["N", "CA", "C", "O", "CB", "CG", "OD1", "OD2", "", "", "", "", "", ""],
+    "CYS": ["N", "CA", "C", "O", "CB", "SG", "", "", "", "", "", "", "", ""],
+    "GLN": ["N", "CA", "C", "O", "CB", "CG", "CD", "OE1", "NE2", "", "", "", "", ""],
+    "GLU": ["N", "CA", "C", "O", "CB", "CG", "CD", "OE1", "OE2", "", "", "", "", ""],
+    "GLY": ["N", "CA", "C", "O", "", "", "", "", "", "", "", "", "", ""],
+    "HIS": ["N", "CA", "C", "O", "CB", "CG", "ND1", "CD2", "CE1", "NE2", "", "", "", ""],
+    "ILE": ["N", "CA", "C", "O", "CB", "CG1", "CG2", "CD1", "", "", "", "", "", ""],
+    "LEU": ["N", "CA", "C", "O", "CB", "CG", "CD1", "CD2", "", "", "", "", "", ""],
+    "LYS": ["N", "CA", "C", "O", "CB", "CG", "CD", "CE", "NZ", "", "", "", "", ""],
+    "MET": ["N", "CA", "C", "O", "CB", "CG", "SD", "CE", "", "", "", "", "", ""],
+    "PHE": ["N", "CA", "C", "O", "CB", "CG", "CD1", "CD2", "CE1", "CE2", "CZ", "", "", ""],
+    "PRO": ["N", "CA", "C", "O", "CB", "CG", "CD", "", "", "", "", "", "", ""],
+    "SER": ["N", "CA", "C", "O", "CB", "OG", "", "", "", "", "", "", "", ""],
+    "THR": ["N", "CA", "C", "O", "CB", "OG1", "CG2", "", "", "", "", "", "", ""],
+    "TRP": ["N", "CA", "C", "O", "CB", "CG", "CD1", "CD2", "NE1", "CE2", "CE3", "CZ2", "CZ3", "CH2"],
+    "TYR": ["N", "CA", "C", "O", "CB", "CG", "CD1", "CD2", "CE1", "CE2", "CZ", "OH", "", ""],
+    "VAL": ["N", "CA", "C", "O", "CB", "CG1", "CG2", "", "", "", "", "", "", ""],
+    "UNK": ["", "", "", "", "", "", "", "", "", "", "", "", "", ""],
+}
+# pylint: enable=line-too-long
+# pylint: enable=bad-whitespace
+
+
+# This is the standard residue order when coding AA type as a number.
+# Reproduce it by taking 3-letter AA codes and sorting them alphabetically.
+restypes: List[str] = [
+    "A",
+    "R",
+    "N",
+    "D",
+    "C",
+    "Q",
+    "E",
+    "G",
+    "H",
+    "I",
+    "L",
+    "K",
+    "M",
+    "F",
+    "P",
+    "S",
+    "T",
+    "W",
+    "Y",
+    "V",
+]
+restype_order: Dict[str, int] = {restype: i for i, restype in enumerate(restypes)}
+restype_num = len(restypes)  # := 20.
+unk_restype_index = restype_num  # Catch-all index for unknown restypes.
+
+restypes_with_x: List[str] = restypes + ["X"]
+restype_order_with_x: Dict[str, int] = {restype: i for i, restype in enumerate(restypes_with_x)}
+
+
+def sequence_to_onehot(sequence: str, mapping: Mapping[str, int], map_unknown_to_x: bool = False) -> np.ndarray:
+    """Maps the given sequence into a one-hot encoded matrix.
+
+    Args:
+      sequence: An amino acid sequence.
+      mapping: A dictionary mapping amino acids to integers.
+      map_unknown_to_x: If True, any amino acid that is not in the mapping will be
+        mapped to the unknown amino acid 'X'. If the mapping doesn't contain amino acid 'X', an error will be thrown.
+        If False, any amino acid not in the mapping will throw an error.
+
+    Returns:
+      A numpy array of shape (seq_len, num_unique_aas) with one-hot encoding of the sequence.
+
+    Raises:
+      ValueError: If the mapping doesn't contain values from 0 to
+        num_unique_aas - 1 without any gaps.
+    """
+    num_entries = max(mapping.values()) + 1
+
+    if sorted(set(mapping.values())) != list(range(num_entries)):
+        raise ValueError(
+            "The mapping must have values from 0 to num_unique_aas-1 without any gaps. Got: %s"
+            % sorted(mapping.values())
+        )
+
+    one_hot_arr = np.zeros((len(sequence), num_entries), dtype=np.int32)
+
+    for aa_index, aa_type in enumerate(sequence):
+        if map_unknown_to_x:
+            if aa_type.isalpha() and aa_type.isupper():
+                aa_id = mapping.get(aa_type, mapping["X"])
+            else:
+                raise ValueError(f"Invalid character in the sequence: {aa_type}")
+        else:
+            aa_id = mapping[aa_type]
+        one_hot_arr[aa_index, aa_id] = 1
+
+    return one_hot_arr
+
+
+restype_1to3: Dict[str, str] = {
+    "A": "ALA",
+    "R": "ARG",
+    "N": "ASN",
+    "D": "ASP",
+    "C": "CYS",
+    "Q": "GLN",
+    "E": "GLU",
+    "G": "GLY",
+    "H": "HIS",
+    "I": "ILE",
+    "L": "LEU",
+    "K": "LYS",
+    "M": "MET",
+    "F": "PHE",
+    "P": "PRO",
+    "S": "SER",
+    "T": "THR",
+    "W": "TRP",
+    "Y": "TYR",
+    "V": "VAL",
+}
+
+
+# NB: restype_3to1 differs from Bio.PDB.protein_letters_3to1 by being a simple
+# 1-to-1 mapping of 3 letter names to one letter names. The latter contains
+# many more, and less common, three letter names as keys and maps many of these
+# to the same one letter name (including 'X' and 'U' which we don't use here).
+restype_3to1: Dict[str, str] = {v: k for k, v in restype_1to3.items()}
+
+# Define a restype name for all unknown residues.
+unk_restype = "UNK"
+
+resnames: List[str] = [restype_1to3[r] for r in restypes] + [unk_restype]
+resname_to_idx: Dict[str, int] = {resname: i for i, resname in enumerate(resnames)}
+
+
+# The mapping here uses hhblits convention, so that B is mapped to D, J and O
+# are mapped to X, U is mapped to C, and Z is mapped to E. Other than that the
+# remaining 20 amino acids are kept in alphabetical order.
+# There are 2 non-amino acid codes, X (representing any amino acid) and
+# "-" representing a missing amino acid in an alignment.  The id for these
+# codes is put at the end (20 and 21) so that they can easily be ignored if
+# desired.
+HHBLITS_AA_TO_ID: Dict[str, int] = {
+    "A": 0,
+    "B": 2,
+    "C": 1,
+    "D": 2,
+    "E": 3,
+    "F": 4,
+    "G": 5,
+    "H": 6,
+    "I": 7,
+    "J": 20,
+    "K": 8,
+    "L": 9,
+    "M": 10,
+    "N": 11,
+    "O": 20,
+    "P": 12,
+    "Q": 13,
+    "R": 14,
+    "S": 15,
+    "T": 16,
+    "U": 1,
+    "V": 17,
+    "W": 18,
+    "X": 20,
+    "Y": 19,
+    "Z": 3,
+    "-": 21,
+}
+
+# Partial inversion of HHBLITS_AA_TO_ID.
+ID_TO_HHBLITS_AA: Dict[int, str] = {
+    0: "A",
+    1: "C",  # Also U.
+    2: "D",  # Also B.
+    3: "E",  # Also Z.
+    4: "F",
+    5: "G",
+    6: "H",
+    7: "I",
+    8: "K",
+    9: "L",
+    10: "M",
+    11: "N",
+    12: "P",
+    13: "Q",
+    14: "R",
+    15: "S",
+    16: "T",
+    17: "V",
+    18: "W",
+    19: "Y",
+    20: "X",  # Includes J and O.
+    21: "-",
+}
+
+restypes_with_x_and_gap: List[str] = restypes + ["X", "-"]
+MAP_HHBLITS_AATYPE_TO_OUR_AATYPE: Tuple[int, ...] = tuple(
+    restypes_with_x_and_gap.index(ID_TO_HHBLITS_AA[i]) for i in range(len(restypes_with_x_and_gap))
+)
+
+
+def _make_standard_atom_mask() -> np.ndarray:
+    """Returns [num_res_types, num_atom_types] mask array."""
+    # +1 to account for unknown (all 0s).
+    mask = np.zeros([restype_num + 1, atom_type_num], dtype=np.int32)
+    for restype, restype_letter in enumerate(restypes):
+        restype_name = restype_1to3[restype_letter]
+        atom_names = residue_atoms[restype_name]
+        for atom_name in atom_names:
+            atom_type = atom_order[atom_name]
+            mask[restype, atom_type] = 1
+    return mask
+
+
+STANDARD_ATOM_MASK = _make_standard_atom_mask()
+
+
+# A one hot representation for the first and second atoms defining the axis
+# of rotation for each chi-angle in each residue.
+def chi_angle_atom(atom_index: int) -> np.ndarray:
+    """Define chi-angle rigid groups via one-hot representations."""
+    chi_angles_index = {}
+    one_hots = []
+
+    for k, v in chi_angles_atoms.items():
+        indices = [atom_types.index(s[atom_index]) for s in v]
+        indices.extend([-1] * (4 - len(indices)))
+        chi_angles_index[k] = indices
+
+    for r in restypes:
+        res3 = restype_1to3[r]
+        one_hot = np.eye(atom_type_num)[chi_angles_index[res3]]
+        one_hots.append(one_hot)
+
+    one_hots.append(np.zeros([4, atom_type_num]))  # Add zeros for residue `X`.
+    one_hot = np.stack(one_hots, axis=0)
+    one_hot = np.transpose(one_hot, [0, 2, 1])
+
+    return one_hot
+
+
+chi_atom_1_one_hot = chi_angle_atom(1)
+chi_atom_2_one_hot = chi_angle_atom(2)
+
+# An array like chi_angles_atoms but using indices rather than names.
+chi_angles_atom_indices_list: List[List[List[str]]] = [chi_angles_atoms[restype_1to3[r]] for r in restypes]
+chi_angles_atom_indices_ours: list = map_structure_with_atom_order(chi_angles_atom_indices_list)
+chi_angles_atom_indices = np.array(
+    [chi_atoms + ([[0, 0, 0, 0]] * (4 - len(chi_atoms))) for chi_atoms in chi_angles_atom_indices_list]
+)
+
+# Mapping from (res_name, atom_name) pairs to the atom's chi group index
+# and atom index within that group.
+chi_groups_for_atom: Dict[Tuple[str, str], List[Tuple[int, int]]] = collections.defaultdict(list)
+for res_name, chi_angle_atoms_for_res in chi_angles_atoms.items():
+    for chi_group_i, chi_group in enumerate(chi_angle_atoms_for_res):
+        for atom_i, atom in enumerate(chi_group):
+            chi_groups_for_atom[(res_name, atom)].append((chi_group_i, atom_i))
+chi_groups_for_atom = dict(chi_groups_for_atom)
+
+
+def _make_rigid_transformation_4x4(ex: np.ndarray, ey: np.ndarray, translation: np.ndarray) -> np.ndarray:
+    """Create a rigid 4x4 transformation matrix from two axes and transl."""
+    # Normalize ex.
+    ex_normalized = ex / np.linalg.norm(ex)
+
+    # make ey perpendicular to ex
+    ey_normalized = ey - np.dot(ey, ex_normalized) * ex_normalized
+    ey_normalized /= np.linalg.norm(ey_normalized)
+
+    # compute ez as cross product
+    eznorm = np.cross(ex_normalized, ey_normalized)
+    m = np.stack([ex_normalized, ey_normalized, eznorm, translation]).transpose()
+    m = np.concatenate([m, [[0.0, 0.0, 0.0, 1.0]]], axis=0)
+    return m
+
+
+# create an array with (restype, atomtype) --> rigid_group_idx
+# and an array with (restype, atomtype, coord) for the atom positions
+# and compute affine transformation matrices (4,4) from one rigid group to the
+# previous group
+restype_atom37_to_rigid_group = np.zeros([21, 37], dtype=int)
+restype_atom37_mask = np.zeros([21, 37], dtype=np.float32)
+restype_atom37_rigid_group_positions = np.zeros([21, 37, 3], dtype=np.float32)
+restype_atom14_to_rigid_group = np.zeros([21, 14], dtype=int)
+restype_atom14_mask = np.zeros([21, 14], dtype=np.float32)
+restype_atom14_rigid_group_positions = np.zeros([21, 14, 3], dtype=np.float32)
+restype_rigid_group_default_frame = np.zeros([21, 8, 4, 4], dtype=np.float32)
+
+
+def _make_rigid_group_constants() -> None:
+    """Fill the arrays above."""
+    for restype, restype_letter in enumerate(restypes):
+        resname = restype_1to3[restype_letter]
+        for atomname, group_idx, atom_position in rigid_group_atom_positions[resname]:
+            atomtype = atom_order[atomname]
+            restype_atom37_to_rigid_group[restype, atomtype] = group_idx
+            restype_atom37_mask[restype, atomtype] = 1
+            restype_atom37_rigid_group_positions[restype, atomtype, :] = atom_position
+
+            atom14idx = restype_name_to_atom14_names[resname].index(atomname)
+            restype_atom14_to_rigid_group[restype, atom14idx] = group_idx
+            restype_atom14_mask[restype, atom14idx] = 1
+            restype_atom14_rigid_group_positions[restype, atom14idx, :] = atom_position
+
+    for restype, restype_letter in enumerate(restypes):
+        resname = restype_1to3[restype_letter]
+        atom_positions: Dict[str, np.ndarray] = {
+            name: np.array(pos) for name, _, pos in rigid_group_atom_positions[resname]
+        }
+
+        # backbone to backbone is the identity transform
+        restype_rigid_group_default_frame[restype, 0, :, :] = np.eye(4)
+
+        # pre-omega-frame to backbone (currently dummy identity matrix)
+        restype_rigid_group_default_frame[restype, 1, :, :] = np.eye(4)
+
+        # phi-frame to backbone
+        mat = _make_rigid_transformation_4x4(
+            ex=atom_positions["N"] - atom_positions["CA"],
+            ey=np.array([1.0, 0.0, 0.0]),
+            translation=atom_positions["N"],
+        )
+        restype_rigid_group_default_frame[restype, 2, :, :] = mat
+
+        # psi-frame to backbone
+        mat = _make_rigid_transformation_4x4(
+            ex=atom_positions["C"] - atom_positions["CA"],
+            ey=atom_positions["CA"] - atom_positions["N"],
+            translation=atom_positions["C"],
+        )
+        restype_rigid_group_default_frame[restype, 3, :, :] = mat
+
+        # chi1-frame to backbone
+        if chi_angles_mask[restype][0]:
+            base_atom_names = chi_angles_atoms[resname][0]
+            base_atom_positions = [atom_positions[name] for name in base_atom_names]
+            mat = _make_rigid_transformation_4x4(
+                ex=base_atom_positions[2] - base_atom_positions[1],
+                ey=base_atom_positions[0] - base_atom_positions[1],
+                translation=base_atom_positions[2],
+            )
+            restype_rigid_group_default_frame[restype, 4, :, :] = mat
+
+        # chi2-frame to chi1-frame
+        # chi3-frame to chi2-frame
+        # chi4-frame to chi3-frame
+        # luckily all rotation axes for the next frame start at (0,0,0) of the
+        # previous frame
+        for chi_idx in range(1, 4):
+            if chi_angles_mask[restype][chi_idx]:
+                axis_end_atom_name = chi_angles_atoms[resname][chi_idx][2]
+                axis_end_atom_position = atom_positions[axis_end_atom_name]
+                mat = _make_rigid_transformation_4x4(
+                    ex=axis_end_atom_position,
+                    ey=np.array([-1.0, 0.0, 0.0]),
+                    translation=axis_end_atom_position,
+                )
+                restype_rigid_group_default_frame[restype, 4 + chi_idx, :, :] = mat
+
+
+_make_rigid_group_constants()
+
+
+def make_atom14_dists_bounds(
+    overlap_tolerance: float = 1.5,
+    bond_length_tolerance_factor: int = 15,
+) -> Dict[str, np.ndarray]:
+    """compute upper and lower bounds for bonds to assess violations."""
+    restype_atom14_bond_lower_bound = np.zeros([21, 14, 14], np.float32)
+    restype_atom14_bond_upper_bound = np.zeros([21, 14, 14], np.float32)
+    restype_atom14_bond_stddev = np.zeros([21, 14, 14], np.float32)
+    residue_bonds, residue_virtual_bonds, _ = load_stereo_chemical_props()
+    for restype, restype_letter in enumerate(restypes):
+        resname = restype_1to3[restype_letter]
+        atom_list = restype_name_to_atom14_names[resname]
+
+        # create lower and upper bounds for clashes
+        for atom1_idx, atom1_name in enumerate(atom_list):
+            if not atom1_name:
+                continue
+            atom1_radius = van_der_waals_radius[atom1_name[0]]
+            for atom2_idx, atom2_name in enumerate(atom_list):
+                if (not atom2_name) or atom1_idx == atom2_idx:
+                    continue
+                atom2_radius = van_der_waals_radius[atom2_name[0]]
+                lower = atom1_radius + atom2_radius - overlap_tolerance
+                upper = 1e10
+                restype_atom14_bond_lower_bound[restype, atom1_idx, atom2_idx] = lower
+                restype_atom14_bond_lower_bound[restype, atom2_idx, atom1_idx] = lower
+                restype_atom14_bond_upper_bound[restype, atom1_idx, atom2_idx] = upper
+                restype_atom14_bond_upper_bound[restype, atom2_idx, atom1_idx] = upper
+
+        # overwrite lower and upper bounds for bonds and angles
+        for b in residue_bonds[resname] + residue_virtual_bonds[resname]:
+            atom1_idx = atom_list.index(b.atom1_name)
+            atom2_idx = atom_list.index(b.atom2_name)
+            lower = b.length - bond_length_tolerance_factor * b.stddev
+            upper = b.length + bond_length_tolerance_factor * b.stddev
+            restype_atom14_bond_lower_bound[restype, atom1_idx, atom2_idx] = lower
+            restype_atom14_bond_lower_bound[restype, atom2_idx, atom1_idx] = lower
+            restype_atom14_bond_upper_bound[restype, atom1_idx, atom2_idx] = upper
+            restype_atom14_bond_upper_bound[restype, atom2_idx, atom1_idx] = upper
+            restype_atom14_bond_stddev[restype, atom1_idx, atom2_idx] = b.stddev
+            restype_atom14_bond_stddev[restype, atom2_idx, atom1_idx] = b.stddev
+    return {
+        "lower_bound": restype_atom14_bond_lower_bound,  # shape (21,14,14)
+        "upper_bound": restype_atom14_bond_upper_bound,  # shape (21,14,14)
+        "stddev": restype_atom14_bond_stddev,  # shape (21,14,14)
+    }
+
+
+restype_atom14_ambiguous_atoms = np.zeros((21, 14), dtype=np.float32)
+restype_atom14_ambiguous_atoms_swap_idx: np.ndarray = np.tile(np.arange(14, dtype=int), (21, 1))
+
+
+def _make_atom14_ambiguity_feats() -> None:
+    for res, pairs in residue_atom_renaming_swaps.items():
+        res_idx = restype_order[restype_3to1[res]]
+        for atom1, atom2 in pairs.items():
+            atom1_idx = restype_name_to_atom14_names[res].index(atom1)
+            atom2_idx = restype_name_to_atom14_names[res].index(atom2)
+            restype_atom14_ambiguous_atoms[res_idx, atom1_idx] = 1
+            restype_atom14_ambiguous_atoms[res_idx, atom2_idx] = 1
+            restype_atom14_ambiguous_atoms_swap_idx[res_idx, atom1_idx] = atom2_idx
+            restype_atom14_ambiguous_atoms_swap_idx[res_idx, atom2_idx] = atom1_idx
+
+
+_make_atom14_ambiguity_feats()
+
+
+def aatype_to_str_sequence(aatype: Sequence[int]) -> str:
+    return "".join([restypes_with_x[aatype[i]] for i in range(len(aatype))])

venv/lib/python3.10/site-packages/transformers/models/esm/openfold_utils/rigid_utils.py

@@ -0,0 +1,1242 @@
+# Copyright 2021 AlQuraishi Laboratory
+# Copyright 2021 DeepMind Technologies Limited
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import annotations
+
+from functools import lru_cache
+from typing import Any, Callable, Dict, List, Optional, Sequence, Tuple
+
+import numpy as np
+import torch
+
+
+def rot_matmul(a: torch.Tensor, b: torch.Tensor) -> torch.Tensor:
+    """
+    Performs matrix multiplication of two rotation matrix tensors. Written out by hand to avoid AMP downcasting.
+
+    Args:
+        a: [*, 3, 3] left multiplicand
+        b: [*, 3, 3] right multiplicand
+    Returns:
+        The product ab
+    """
+
+    def row_mul(i: int) -> torch.Tensor:
+        return torch.stack(
+            [
+                a[..., i, 0] * b[..., 0, 0] + a[..., i, 1] * b[..., 1, 0] + a[..., i, 2] * b[..., 2, 0],
+                a[..., i, 0] * b[..., 0, 1] + a[..., i, 1] * b[..., 1, 1] + a[..., i, 2] * b[..., 2, 1],
+                a[..., i, 0] * b[..., 0, 2] + a[..., i, 1] * b[..., 1, 2] + a[..., i, 2] * b[..., 2, 2],
+            ],
+            dim=-1,
+        )
+
+    return torch.stack(
+        [
+            row_mul(0),
+            row_mul(1),
+            row_mul(2),
+        ],
+        dim=-2,
+    )
+
+
+def rot_vec_mul(r: torch.Tensor, t: torch.Tensor) -> torch.Tensor:
+    """
+    Applies a rotation to a vector. Written out by hand to avoid transfer to avoid AMP downcasting.
+
+    Args:
+        r: [*, 3, 3] rotation matrices
+        t: [*, 3] coordinate tensors
+    Returns:
+        [*, 3] rotated coordinates
+    """
+    x, y, z = torch.unbind(t, dim=-1)
+    return torch.stack(
+        [
+            r[..., 0, 0] * x + r[..., 0, 1] * y + r[..., 0, 2] * z,
+            r[..., 1, 0] * x + r[..., 1, 1] * y + r[..., 1, 2] * z,
+            r[..., 2, 0] * x + r[..., 2, 1] * y + r[..., 2, 2] * z,
+        ],
+        dim=-1,
+    )
+
+
+@lru_cache(maxsize=None)
+def identity_rot_mats(
+    batch_dims: Tuple[int, ...],
+    dtype: Optional[torch.dtype] = None,
+    device: Optional[torch.device] = None,
+    requires_grad: bool = True,
+) -> torch.Tensor:
+    rots = torch.eye(3, dtype=dtype, device=device, requires_grad=requires_grad)
+    rots = rots.view(*((1,) * len(batch_dims)), 3, 3)
+    rots = rots.expand(*batch_dims, -1, -1)
+    rots = rots.contiguous()
+
+    return rots
+
+
+@lru_cache(maxsize=None)
+def identity_trans(
+    batch_dims: Tuple[int, ...],
+    dtype: Optional[torch.dtype] = None,
+    device: Optional[torch.device] = None,
+    requires_grad: bool = True,
+) -> torch.Tensor:
+    trans = torch.zeros((*batch_dims, 3), dtype=dtype, device=device, requires_grad=requires_grad)
+    return trans
+
+
+@lru_cache(maxsize=None)
+def identity_quats(
+    batch_dims: Tuple[int, ...],
+    dtype: Optional[torch.dtype] = None,
+    device: Optional[torch.device] = None,
+    requires_grad: bool = True,
+) -> torch.Tensor:
+    quat = torch.zeros((*batch_dims, 4), dtype=dtype, device=device, requires_grad=requires_grad)
+
+    with torch.no_grad():
+        quat[..., 0] = 1
+
+    return quat
+
+
+_quat_elements: List[str] = ["a", "b", "c", "d"]
+_qtr_keys: List[str] = [l1 + l2 for l1 in _quat_elements for l2 in _quat_elements]
+_qtr_ind_dict: Dict[str, int] = {key: ind for ind, key in enumerate(_qtr_keys)}
+
+
+def _to_mat(pairs: List[Tuple[str, int]]) -> np.ndarray:
+    mat = np.zeros((4, 4))
+    for key, value in pairs:
+        ind = _qtr_ind_dict[key]
+        mat[ind // 4][ind % 4] = value
+
+    return mat
+
+
+_QTR_MAT = np.zeros((4, 4, 3, 3))
+_QTR_MAT[..., 0, 0] = _to_mat([("aa", 1), ("bb", 1), ("cc", -1), ("dd", -1)])
+_QTR_MAT[..., 0, 1] = _to_mat([("bc", 2), ("ad", -2)])
+_QTR_MAT[..., 0, 2] = _to_mat([("bd", 2), ("ac", 2)])
+_QTR_MAT[..., 1, 0] = _to_mat([("bc", 2), ("ad", 2)])
+_QTR_MAT[..., 1, 1] = _to_mat([("aa", 1), ("bb", -1), ("cc", 1), ("dd", -1)])
+_QTR_MAT[..., 1, 2] = _to_mat([("cd", 2), ("ab", -2)])
+_QTR_MAT[..., 2, 0] = _to_mat([("bd", 2), ("ac", -2)])
+_QTR_MAT[..., 2, 1] = _to_mat([("cd", 2), ("ab", 2)])
+_QTR_MAT[..., 2, 2] = _to_mat([("aa", 1), ("bb", -1), ("cc", -1), ("dd", 1)])
+
+
+def quat_to_rot(quat: torch.Tensor) -> torch.Tensor:
+    """
+    Converts a quaternion to a rotation matrix.
+
+    Args:
+        quat: [*, 4] quaternions
+    Returns:
+        [*, 3, 3] rotation matrices
+    """
+    # [*, 4, 4]
+    quat = quat[..., None] * quat[..., None, :]
+
+    # [4, 4, 3, 3]
+    mat = _get_quat("_QTR_MAT", dtype=quat.dtype, device=quat.device)
+
+    # [*, 4, 4, 3, 3]
+    shaped_qtr_mat = mat.view((1,) * len(quat.shape[:-2]) + mat.shape)
+    quat = quat[..., None, None] * shaped_qtr_mat
+
+    # [*, 3, 3]
+    return torch.sum(quat, dim=(-3, -4))
+
+
+def rot_to_quat(rot: torch.Tensor) -> torch.Tensor:
+    if rot.shape[-2:] != (3, 3):
+        raise ValueError("Input rotation is incorrectly shaped")
+
+    [[xx, xy, xz], [yx, yy, yz], [zx, zy, zz]] = [[rot[..., i, j] for j in range(3)] for i in range(3)]
+
+    k = [
+        [
+            xx + yy + zz,
+            zy - yz,
+            xz - zx,
+            yx - xy,
+        ],
+        [
+            zy - yz,
+            xx - yy - zz,
+            xy + yx,
+            xz + zx,
+        ],
+        [
+            xz - zx,
+            xy + yx,
+            yy - xx - zz,
+            yz + zy,
+        ],
+        [
+            yx - xy,
+            xz + zx,
+            yz + zy,
+            zz - xx - yy,
+        ],
+    ]
+
+    _, vectors = torch.linalg.eigh((1.0 / 3.0) * torch.stack([torch.stack(t, dim=-1) for t in k], dim=-2))
+    return vectors[..., -1]
+
+
+_QUAT_MULTIPLY = np.zeros((4, 4, 4))
+_QUAT_MULTIPLY[:, :, 0] = [[1, 0, 0, 0], [0, -1, 0, 0], [0, 0, -1, 0], [0, 0, 0, -1]]
+
+_QUAT_MULTIPLY[:, :, 1] = [[0, 1, 0, 0], [1, 0, 0, 0], [0, 0, 0, 1], [0, 0, -1, 0]]
+
+_QUAT_MULTIPLY[:, :, 2] = [[0, 0, 1, 0], [0, 0, 0, -1], [1, 0, 0, 0], [0, 1, 0, 0]]
+
+_QUAT_MULTIPLY[:, :, 3] = [[0, 0, 0, 1], [0, 0, 1, 0], [0, -1, 0, 0], [1, 0, 0, 0]]
+
+_QUAT_MULTIPLY_BY_VEC = _QUAT_MULTIPLY[:, 1:, :]
+
+_CACHED_QUATS: Dict[str, np.ndarray] = {
+    "_QTR_MAT": _QTR_MAT,
+    "_QUAT_MULTIPLY": _QUAT_MULTIPLY,
+    "_QUAT_MULTIPLY_BY_VEC": _QUAT_MULTIPLY_BY_VEC,
+}
+
+
+@lru_cache(maxsize=None)
+def _get_quat(quat_key: str, dtype: torch.dtype, device: torch.device) -> torch.Tensor:
+    return torch.tensor(_CACHED_QUATS[quat_key], dtype=dtype, device=device)
+
+
+def quat_multiply(quat1: torch.Tensor, quat2: torch.Tensor) -> torch.Tensor:
+    """Multiply a quaternion by another quaternion."""
+    mat = _get_quat("_QUAT_MULTIPLY", dtype=quat1.dtype, device=quat1.device)
+    reshaped_mat = mat.view((1,) * len(quat1.shape[:-1]) + mat.shape)
+    return torch.sum(reshaped_mat * quat1[..., :, None, None] * quat2[..., None, :, None], dim=(-3, -2))
+
+
+def quat_multiply_by_vec(quat: torch.Tensor, vec: torch.Tensor) -> torch.Tensor:
+    """Multiply a quaternion by a pure-vector quaternion."""
+    mat = _get_quat("_QUAT_MULTIPLY_BY_VEC", dtype=quat.dtype, device=quat.device)
+    reshaped_mat = mat.view((1,) * len(quat.shape[:-1]) + mat.shape)
+    return torch.sum(reshaped_mat * quat[..., :, None, None] * vec[..., None, :, None], dim=(-3, -2))
+
+
+def invert_rot_mat(rot_mat: torch.Tensor) -> torch.Tensor:
+    return rot_mat.transpose(-1, -2)
+
+
+def invert_quat(quat: torch.Tensor) -> torch.Tensor:
+    quat_prime = quat.clone()
+    quat_prime[..., 1:] *= -1
+    inv = quat_prime / torch.sum(quat**2, dim=-1, keepdim=True)
+    return inv
+
+
+class Rotation:
+    """
+    A 3D rotation. Depending on how the object is initialized, the rotation is represented by either a rotation matrix
+    or a quaternion, though both formats are made available by helper functions. To simplify gradient computation, the
+    underlying format of the rotation cannot be changed in-place. Like Rigid, the class is designed to mimic the
+    behavior of a torch Tensor, almost as if each Rotation object were a tensor of rotations, in one format or another.
+    """
+
+    def __init__(
+        self,
+        rot_mats: Optional[torch.Tensor] = None,
+        quats: Optional[torch.Tensor] = None,
+        normalize_quats: bool = True,
+    ):
+        """
+        Args:
+            rot_mats:
+                A [*, 3, 3] rotation matrix tensor. Mutually exclusive with quats
+            quats:
+                A [*, 4] quaternion. Mutually exclusive with rot_mats. If normalize_quats is not True, must be a unit
+                quaternion
+            normalize_quats:
+                If quats is specified, whether to normalize quats
+        """
+        if (rot_mats is None and quats is None) or (rot_mats is not None and quats is not None):
+            raise ValueError("Exactly one input argument must be specified")
+
+        if (rot_mats is not None and rot_mats.shape[-2:] != (3, 3)) or (quats is not None and quats.shape[-1] != 4):
+            raise ValueError("Incorrectly shaped rotation matrix or quaternion")
+
+        # Force full-precision
+        if quats is not None:
+            quats = quats.to(dtype=torch.float32)
+        if rot_mats is not None:
+            rot_mats = rot_mats.to(dtype=torch.float32)
+
+        if quats is not None and normalize_quats:
+            quats = quats / torch.linalg.norm(quats, dim=-1, keepdim=True)
+
+        self._rot_mats = rot_mats
+        self._quats = quats
+
+    @staticmethod
+    def identity(
+        shape,
+        dtype: Optional[torch.dtype] = None,
+        device: Optional[torch.device] = None,
+        requires_grad: bool = True,
+        fmt: str = "quat",
+    ) -> Rotation:
+        """
+        Returns an identity Rotation.
+
+        Args:
+            shape:
+                The "shape" of the resulting Rotation object. See documentation for the shape property
+            dtype:
+                The torch dtype for the rotation
+            device:
+                The torch device for the new rotation
+            requires_grad:
+                Whether the underlying tensors in the new rotation object should require gradient computation
+            fmt:
+                One of "quat" or "rot_mat". Determines the underlying format of the new object's rotation
+        Returns:
+            A new identity rotation
+        """
+        if fmt == "rot_mat":
+            rot_mats = identity_rot_mats(
+                shape,
+                dtype,
+                device,
+                requires_grad,
+            )
+            return Rotation(rot_mats=rot_mats, quats=None)
+        elif fmt == "quat":
+            quats = identity_quats(shape, dtype, device, requires_grad)
+            return Rotation(rot_mats=None, quats=quats, normalize_quats=False)
+        else:
+            raise ValueError(f"Invalid format: f{fmt}")
+
+    # Magic methods
+
+    def __getitem__(self, index: Any) -> Rotation:
+        """
+        Allows torch-style indexing over the virtual shape of the rotation object. See documentation for the shape
+        property.
+
+        Args:
+            index:
+                A torch index. E.g. (1, 3, 2), or (slice(None,))
+        Returns:
+            The indexed rotation
+        """
+        if type(index) != tuple:
+            index = (index,)
+
+        if self._rot_mats is not None:
+            rot_mats = self._rot_mats[index + (slice(None), slice(None))]
+            return Rotation(rot_mats=rot_mats)
+        elif self._quats is not None:
+            quats = self._quats[index + (slice(None),)]
+            return Rotation(quats=quats, normalize_quats=False)
+        else:
+            raise ValueError("Both rotations are None")
+
+    def __mul__(self, right: torch.Tensor) -> Rotation:
+        """
+        Pointwise left multiplication of the rotation with a tensor. Can be used to e.g. mask the Rotation.
+
+        Args:
+            right:
+                The tensor multiplicand
+        Returns:
+            The product
+        """
+        if not (isinstance(right, torch.Tensor)):
+            raise TypeError("The other multiplicand must be a Tensor")
+
+        if self._rot_mats is not None:
+            rot_mats = self._rot_mats * right[..., None, None]
+            return Rotation(rot_mats=rot_mats, quats=None)
+        elif self._quats is not None:
+            quats = self._quats * right[..., None]
+            return Rotation(rot_mats=None, quats=quats, normalize_quats=False)
+        else:
+            raise ValueError("Both rotations are None")
+
+    def __rmul__(self, left: torch.Tensor) -> Rotation:
+        """
+        Reverse pointwise multiplication of the rotation with a tensor.
+
+        Args:
+            left:
+                The left multiplicand
+        Returns:
+            The product
+        """
+        return self.__mul__(left)
+
+    # Properties
+
+    @property
+    def shape(self) -> torch.Size:
+        """
+        Returns the virtual shape of the rotation object. This shape is defined as the batch dimensions of the
+        underlying rotation matrix or quaternion. If the Rotation was initialized with a [10, 3, 3] rotation matrix
+        tensor, for example, the resulting shape would be [10].
+
+        Returns:
+            The virtual shape of the rotation object
+        """
+        if self._rot_mats is not None:
+            return self._rot_mats.shape[:-2]
+        elif self._quats is not None:
+            return self._quats.shape[:-1]
+        else:
+            raise ValueError("Both rotations are None")
+
+    @property
+    def dtype(self) -> torch.dtype:
+        """
+        Returns the dtype of the underlying rotation.
+
+        Returns:
+            The dtype of the underlying rotation
+        """
+        if self._rot_mats is not None:
+            return self._rot_mats.dtype
+        elif self._quats is not None:
+            return self._quats.dtype
+        else:
+            raise ValueError("Both rotations are None")
+
+    @property
+    def device(self) -> torch.device:
+        """
+        The device of the underlying rotation
+
+        Returns:
+            The device of the underlying rotation
+        """
+        if self._rot_mats is not None:
+            return self._rot_mats.device
+        elif self._quats is not None:
+            return self._quats.device
+        else:
+            raise ValueError("Both rotations are None")
+
+    @property
+    def requires_grad(self) -> bool:
+        """
+        Returns the requires_grad property of the underlying rotation
+
+        Returns:
+            The requires_grad property of the underlying tensor
+        """
+        if self._rot_mats is not None:
+            return self._rot_mats.requires_grad
+        elif self._quats is not None:
+            return self._quats.requires_grad
+        else:
+            raise ValueError("Both rotations are None")
+
+    def get_rot_mats(self) -> torch.Tensor:
+        """
+        Returns the underlying rotation as a rotation matrix tensor.
+
+        Returns:
+            The rotation as a rotation matrix tensor
+        """
+        if self._rot_mats is not None:
+            return self._rot_mats
+        elif self._quats is not None:
+            return quat_to_rot(self._quats)
+        else:
+            raise ValueError("Both rotations are None")
+
+    def get_quats(self) -> torch.Tensor:
+        """
+        Returns the underlying rotation as a quaternion tensor.
+
+        Depending on whether the Rotation was initialized with a quaternion, this function may call torch.linalg.eigh.
+
+        Returns:
+            The rotation as a quaternion tensor.
+        """
+        if self._rot_mats is not None:
+            return rot_to_quat(self._rot_mats)
+        elif self._quats is not None:
+            return self._quats
+        else:
+            raise ValueError("Both rotations are None")
+
+    def get_cur_rot(self) -> torch.Tensor:
+        """
+        Return the underlying rotation in its current form
+
+        Returns:
+            The stored rotation
+        """
+        if self._rot_mats is not None:
+            return self._rot_mats
+        elif self._quats is not None:
+            return self._quats
+        else:
+            raise ValueError("Both rotations are None")
+
+    # Rotation functions
+
+    def compose_q_update_vec(self, q_update_vec: torch.Tensor, normalize_quats: bool = True) -> Rotation:
+        """
+        Returns a new quaternion Rotation after updating the current object's underlying rotation with a quaternion
+        update, formatted as a [*, 3] tensor whose final three columns represent x, y, z such that (1, x, y, z) is the
+        desired (not necessarily unit) quaternion update.
+
+        Args:
+            q_update_vec:
+                A [*, 3] quaternion update tensor
+            normalize_quats:
+                Whether to normalize the output quaternion
+        Returns:
+            An updated Rotation
+        """
+        quats = self.get_quats()
+        new_quats = quats + quat_multiply_by_vec(quats, q_update_vec)
+        return Rotation(
+            rot_mats=None,
+            quats=new_quats,
+            normalize_quats=normalize_quats,
+        )
+
+    def compose_r(self, r: Rotation) -> Rotation:
+        """
+        Compose the rotation matrices of the current Rotation object with those of another.
+
+        Args:
+            r:
+                An update rotation object
+        Returns:
+            An updated rotation object
+        """
+        r1 = self.get_rot_mats()
+        r2 = r.get_rot_mats()
+        new_rot_mats = rot_matmul(r1, r2)
+        return Rotation(rot_mats=new_rot_mats, quats=None)
+
+    def compose_q(self, r: Rotation, normalize_quats: bool = True) -> Rotation:
+        """
+        Compose the quaternions of the current Rotation object with those of another.
+
+        Depending on whether either Rotation was initialized with quaternions, this function may call
+        torch.linalg.eigh.
+
+        Args:
+            r:
+                An update rotation object
+        Returns:
+            An updated rotation object
+        """
+        q1 = self.get_quats()
+        q2 = r.get_quats()
+        new_quats = quat_multiply(q1, q2)
+        return Rotation(rot_mats=None, quats=new_quats, normalize_quats=normalize_quats)
+
+    def apply(self, pts: torch.Tensor) -> torch.Tensor:
+        """
+        Apply the current Rotation as a rotation matrix to a set of 3D coordinates.
+
+        Args:
+            pts:
+                A [*, 3] set of points
+        Returns:
+            [*, 3] rotated points
+        """
+        rot_mats = self.get_rot_mats()
+        return rot_vec_mul(rot_mats, pts)
+
+    def invert_apply(self, pts: torch.Tensor) -> torch.Tensor:
+        """
+        The inverse of the apply() method.
+
+        Args:
+            pts:
+                A [*, 3] set of points
+        Returns:
+            [*, 3] inverse-rotated points
+        """
+        rot_mats = self.get_rot_mats()
+        inv_rot_mats = invert_rot_mat(rot_mats)
+        return rot_vec_mul(inv_rot_mats, pts)
+
+    def invert(self) -> Rotation:
+        """
+        Returns the inverse of the current Rotation.
+
+        Returns:
+            The inverse of the current Rotation
+        """
+        if self._rot_mats is not None:
+            return Rotation(rot_mats=invert_rot_mat(self._rot_mats), quats=None)
+        elif self._quats is not None:
+            return Rotation(
+                rot_mats=None,
+                quats=invert_quat(self._quats),
+                normalize_quats=False,
+            )
+        else:
+            raise ValueError("Both rotations are None")
+
+    # "Tensor" stuff
+
+    def unsqueeze(self, dim: int) -> Rotation:
+        """
+        Analogous to torch.unsqueeze. The dimension is relative to the shape of the Rotation object.
+
+        Args:
+            dim: A positive or negative dimension index.
+        Returns:
+            The unsqueezed Rotation.
+        """
+        if dim >= len(self.shape):
+            raise ValueError("Invalid dimension")
+
+        if self._rot_mats is not None:
+            rot_mats = self._rot_mats.unsqueeze(dim if dim >= 0 else dim - 2)
+            return Rotation(rot_mats=rot_mats, quats=None)
+        elif self._quats is not None:
+            quats = self._quats.unsqueeze(dim if dim >= 0 else dim - 1)
+            return Rotation(rot_mats=None, quats=quats, normalize_quats=False)
+        else:
+            raise ValueError("Both rotations are None")
+
+    @staticmethod
+    def cat(rs: Sequence[Rotation], dim: int) -> Rotation:
+        """
+        Concatenates rotations along one of the batch dimensions. Analogous to torch.cat().
+
+        Note that the output of this operation is always a rotation matrix, regardless of the format of input
+        rotations.
+
+        Args:
+            rs:
+                A list of rotation objects
+            dim:
+                The dimension along which the rotations should be concatenated
+        Returns:
+            A concatenated Rotation object in rotation matrix format
+        """
+        rot_mats = torch.cat(
+            [r.get_rot_mats() for r in rs],
+            dim=dim if dim >= 0 else dim - 2,
+        )
+
+        return Rotation(rot_mats=rot_mats, quats=None)
+
+    def map_tensor_fn(self, fn: Callable[[torch.Tensor], torch.Tensor]) -> Rotation:
+        """
+        Apply a Tensor -> Tensor function to underlying rotation tensors, mapping over the rotation dimension(s). Can
+        be used e.g. to sum out a one-hot batch dimension.
+
+        Args:
+            fn:
+                A Tensor -> Tensor function to be mapped over the Rotation
+        Returns:
+            The transformed Rotation object
+        """
+        if self._rot_mats is not None:
+            rot_mats = self._rot_mats.view(self._rot_mats.shape[:-2] + (9,))
+            rot_mats = torch.stack(list(map(fn, torch.unbind(rot_mats, dim=-1))), dim=-1)
+            rot_mats = rot_mats.view(rot_mats.shape[:-1] + (3, 3))
+            return Rotation(rot_mats=rot_mats, quats=None)
+        elif self._quats is not None:
+            quats = torch.stack(list(map(fn, torch.unbind(self._quats, dim=-1))), dim=-1)
+            return Rotation(rot_mats=None, quats=quats, normalize_quats=False)
+        else:
+            raise ValueError("Both rotations are None")
+
+    def cuda(self) -> Rotation:
+        """
+        Analogous to the cuda() method of torch Tensors
+
+        Returns:
+            A copy of the Rotation in CUDA memory
+        """
+        if self._rot_mats is not None:
+            return Rotation(rot_mats=self._rot_mats.cuda(), quats=None)
+        elif self._quats is not None:
+            return Rotation(rot_mats=None, quats=self._quats.cuda(), normalize_quats=False)
+        else:
+            raise ValueError("Both rotations are None")
+
+    def to(self, device: Optional[torch.device], dtype: Optional[torch.dtype]) -> Rotation:
+        """
+        Analogous to the to() method of torch Tensors
+
+        Args:
+            device:
+                A torch device
+            dtype:
+                A torch dtype
+        Returns:
+            A copy of the Rotation using the new device and dtype
+        """
+        if self._rot_mats is not None:
+            return Rotation(
+                rot_mats=self._rot_mats.to(device=device, dtype=dtype),
+                quats=None,
+            )
+        elif self._quats is not None:
+            return Rotation(
+                rot_mats=None,
+                quats=self._quats.to(device=device, dtype=dtype),
+                normalize_quats=False,
+            )
+        else:
+            raise ValueError("Both rotations are None")
+
+    def detach(self) -> Rotation:
+        """
+        Returns a copy of the Rotation whose underlying Tensor has been detached from its torch graph.
+
+        Returns:
+            A copy of the Rotation whose underlying Tensor has been detached from its torch graph
+        """
+        if self._rot_mats is not None:
+            return Rotation(rot_mats=self._rot_mats.detach(), quats=None)
+        elif self._quats is not None:
+            return Rotation(
+                rot_mats=None,
+                quats=self._quats.detach(),
+                normalize_quats=False,
+            )
+        else:
+            raise ValueError("Both rotations are None")
+
+
+class Rigid:
+    """
+    A class representing a rigid transformation. Little more than a wrapper around two objects: a Rotation object and a
+    [*, 3] translation Designed to behave approximately like a single torch tensor with the shape of the shared batch
+    dimensions of its component parts.
+    """
+
+    def __init__(self, rots: Optional[Rotation], trans: Optional[torch.Tensor]):
+        """
+        Args:
+            rots: A [*, 3, 3] rotation tensor
+            trans: A corresponding [*, 3] translation tensor
+        """
+        # (we need device, dtype, etc. from at least one input)
+
+        batch_dims, dtype, device, requires_grad = None, None, None, None
+        if trans is not None:
+            batch_dims = trans.shape[:-1]
+            dtype = trans.dtype
+            device = trans.device
+            requires_grad = trans.requires_grad
+        elif rots is not None:
+            batch_dims = rots.shape
+            dtype = rots.dtype
+            device = rots.device
+            requires_grad = rots.requires_grad
+        else:
+            raise ValueError("At least one input argument must be specified")
+
+        if rots is None:
+            rots = Rotation.identity(
+                batch_dims,
+                dtype,
+                device,
+                requires_grad,
+            )
+        elif trans is None:
+            trans = identity_trans(
+                batch_dims,
+                dtype,
+                device,
+                requires_grad,
+            )
+
+        assert rots is not None
+        assert trans is not None
+
+        if (rots.shape != trans.shape[:-1]) or (rots.device != trans.device):
+            raise ValueError("Rots and trans incompatible")
+
+        # Force full precision. Happens to the rotations automatically.
+        trans = trans.to(dtype=torch.float32)
+
+        self._rots = rots
+        self._trans = trans
+
+    @staticmethod
+    def identity(
+        shape: Tuple[int, ...],
+        dtype: Optional[torch.dtype] = None,
+        device: Optional[torch.device] = None,
+        requires_grad: bool = True,
+        fmt: str = "quat",
+    ) -> Rigid:
+        """
+        Constructs an identity transformation.
+
+        Args:
+            shape:
+                The desired shape
+            dtype:
+                The dtype of both internal tensors
+            device:
+                The device of both internal tensors
+            requires_grad:
+                Whether grad should be enabled for the internal tensors
+        Returns:
+            The identity transformation
+        """
+        return Rigid(
+            Rotation.identity(shape, dtype, device, requires_grad, fmt=fmt),
+            identity_trans(shape, dtype, device, requires_grad),
+        )
+
+    def __getitem__(self, index: Any) -> Rigid:
+        """
+        Indexes the affine transformation with PyTorch-style indices. The index is applied to the shared dimensions of
+        both the rotation and the translation.
+
+        E.g.::
+
+            r = Rotation(rot_mats=torch.rand(10, 10, 3, 3), quats=None) t = Rigid(r, torch.rand(10, 10, 3)) indexed =
+            t[3, 4:6] assert(indexed.shape == (2,)) assert(indexed.get_rots().shape == (2,))
+            assert(indexed.get_trans().shape == (2, 3))
+
+        Args:
+            index: A standard torch tensor index. E.g. 8, (10, None, 3),
+            or (3, slice(0, 1, None))
+        Returns:
+            The indexed tensor
+        """
+        if type(index) != tuple:
+            index = (index,)
+
+        return Rigid(
+            self._rots[index],
+            self._trans[index + (slice(None),)],
+        )
+
+    def __mul__(self, right: torch.Tensor) -> Rigid:
+        """
+        Pointwise left multiplication of the transformation with a tensor. Can be used to e.g. mask the Rigid.
+
+        Args:
+            right:
+                The tensor multiplicand
+        Returns:
+            The product
+        """
+        if not (isinstance(right, torch.Tensor)):
+            raise TypeError("The other multiplicand must be a Tensor")
+
+        new_rots = self._rots * right
+        new_trans = self._trans * right[..., None]
+
+        return Rigid(new_rots, new_trans)
+
+    def __rmul__(self, left: torch.Tensor) -> Rigid:
+        """
+        Reverse pointwise multiplication of the transformation with a tensor.
+
+        Args:
+            left:
+                The left multiplicand
+        Returns:
+            The product
+        """
+        return self.__mul__(left)
+
+    @property
+    def shape(self) -> torch.Size:
+        """
+        Returns the shape of the shared dimensions of the rotation and the translation.
+
+        Returns:
+            The shape of the transformation
+        """
+        return self._trans.shape[:-1]
+
+    @property
+    def device(self) -> torch.device:
+        """
+        Returns the device on which the Rigid's tensors are located.
+
+        Returns:
+            The device on which the Rigid's tensors are located
+        """
+        return self._trans.device
+
+    def get_rots(self) -> Rotation:
+        """
+        Getter for the rotation.
+
+        Returns:
+            The rotation object
+        """
+        return self._rots
+
+    def get_trans(self) -> torch.Tensor:
+        """
+        Getter for the translation.
+
+        Returns:
+            The stored translation
+        """
+        return self._trans
+
+    def compose_q_update_vec(self, q_update_vec: torch.Tensor) -> Rigid:
+        """
+        Composes the transformation with a quaternion update vector of shape [*, 6], where the final 6 columns
+        represent the x, y, and z values of a quaternion of form (1, x, y, z) followed by a 3D translation.
+
+        Args:
+            q_vec: The quaternion update vector.
+        Returns:
+            The composed transformation.
+        """
+        q_vec, t_vec = q_update_vec[..., :3], q_update_vec[..., 3:]
+        new_rots = self._rots.compose_q_update_vec(q_vec)
+
+        trans_update = self._rots.apply(t_vec)
+        new_translation = self._trans + trans_update
+
+        return Rigid(new_rots, new_translation)
+
+    def compose(self, r: Rigid) -> Rigid:
+        """
+        Composes the current rigid object with another.
+
+        Args:
+            r:
+                Another Rigid object
+        Returns:
+            The composition of the two transformations
+        """
+        new_rot = self._rots.compose_r(r._rots)
+        new_trans = self._rots.apply(r._trans) + self._trans
+        return Rigid(new_rot, new_trans)
+
+    def apply(self, pts: torch.Tensor) -> torch.Tensor:
+        """
+        Applies the transformation to a coordinate tensor.
+
+        Args:
+            pts: A [*, 3] coordinate tensor.
+        Returns:
+            The transformed points.
+        """
+        rotated = self._rots.apply(pts)
+        return rotated + self._trans
+
+    def invert_apply(self, pts: torch.Tensor) -> torch.Tensor:
+        """
+        Applies the inverse of the transformation to a coordinate tensor.
+
+        Args:
+            pts: A [*, 3] coordinate tensor
+        Returns:
+            The transformed points.
+        """
+        pts = pts - self._trans
+        return self._rots.invert_apply(pts)
+
+    def invert(self) -> Rigid:
+        """
+        Inverts the transformation.
+
+        Returns:
+            The inverse transformation.
+        """
+        rot_inv = self._rots.invert()
+        trn_inv = rot_inv.apply(self._trans)
+
+        return Rigid(rot_inv, -1 * trn_inv)
+
+    def map_tensor_fn(self, fn: Callable[[torch.Tensor], torch.Tensor]) -> Rigid:
+        """
+        Apply a Tensor -> Tensor function to underlying translation and rotation tensors, mapping over the
+        translation/rotation dimensions respectively.
+
+        Args:
+            fn:
+                A Tensor -> Tensor function to be mapped over the Rigid
+        Returns:
+            The transformed Rigid object
+        """
+        new_rots = self._rots.map_tensor_fn(fn)
+        new_trans = torch.stack(list(map(fn, torch.unbind(self._trans, dim=-1))), dim=-1)
+
+        return Rigid(new_rots, new_trans)
+
+    def to_tensor_4x4(self) -> torch.Tensor:
+        """
+        Converts a transformation to a homogenous transformation tensor.
+
+        Returns:
+            A [*, 4, 4] homogenous transformation tensor
+        """
+        tensor = self._trans.new_zeros((*self.shape, 4, 4))
+        tensor[..., :3, :3] = self._rots.get_rot_mats()
+        tensor[..., :3, 3] = self._trans
+        tensor[..., 3, 3] = 1
+        return tensor
+
+    @staticmethod
+    def from_tensor_4x4(t: torch.Tensor) -> Rigid:
+        """
+        Constructs a transformation from a homogenous transformation tensor.
+
+        Args:
+            t: [*, 4, 4] homogenous transformation tensor
+        Returns:
+            T object with shape [*]
+        """
+        if t.shape[-2:] != (4, 4):
+            raise ValueError("Incorrectly shaped input tensor")
+
+        rots = Rotation(rot_mats=t[..., :3, :3], quats=None)
+        trans = t[..., :3, 3]
+
+        return Rigid(rots, trans)
+
+    def to_tensor_7(self) -> torch.Tensor:
+        """
+        Converts a transformation to a tensor with 7 final columns, four for the quaternion followed by three for the
+        translation.
+
+        Returns:
+            A [*, 7] tensor representation of the transformation
+        """
+        tensor = self._trans.new_zeros((*self.shape, 7))
+        tensor[..., :4] = self._rots.get_quats()
+        tensor[..., 4:] = self._trans
+
+        return tensor
+
+    @staticmethod
+    def from_tensor_7(t: torch.Tensor, normalize_quats: bool = False) -> Rigid:
+        if t.shape[-1] != 7:
+            raise ValueError("Incorrectly shaped input tensor")
+
+        quats, trans = t[..., :4], t[..., 4:]
+
+        rots = Rotation(rot_mats=None, quats=quats, normalize_quats=normalize_quats)
+
+        return Rigid(rots, trans)
+
+    @staticmethod
+    def from_3_points(
+        p_neg_x_axis: torch.Tensor, origin: torch.Tensor, p_xy_plane: torch.Tensor, eps: float = 1e-8
+    ) -> Rigid:
+        """
+        Implements algorithm 21. Constructs transformations from sets of 3 points using the Gram-Schmidt algorithm.
+
+        Args:
+            p_neg_x_axis: [*, 3] coordinates
+            origin: [*, 3] coordinates used as frame origins
+            p_xy_plane: [*, 3] coordinates
+            eps: Small epsilon value
+        Returns:
+            A transformation object of shape [*]
+        """
+        p_neg_x_axis_unbound = torch.unbind(p_neg_x_axis, dim=-1)
+        origin_unbound = torch.unbind(origin, dim=-1)
+        p_xy_plane_unbound = torch.unbind(p_xy_plane, dim=-1)
+
+        e0 = [c1 - c2 for c1, c2 in zip(origin_unbound, p_neg_x_axis_unbound)]
+        e1 = [c1 - c2 for c1, c2 in zip(p_xy_plane_unbound, origin_unbound)]
+
+        denom = torch.sqrt(sum(c * c for c in e0) + eps * torch.ones_like(e0[0]))
+        e0 = [c / denom for c in e0]
+        dot = sum((c1 * c2 for c1, c2 in zip(e0, e1)))
+        e1 = [c2 - c1 * dot for c1, c2 in zip(e0, e1)]
+        denom = torch.sqrt(sum((c * c for c in e1)) + eps * torch.ones_like(e1[0]))
+        e1 = [c / denom for c in e1]
+        e2 = [
+            e0[1] * e1[2] - e0[2] * e1[1],
+            e0[2] * e1[0] - e0[0] * e1[2],
+            e0[0] * e1[1] - e0[1] * e1[0],
+        ]
+
+        rots = torch.stack([c for tup in zip(e0, e1, e2) for c in tup], dim=-1)
+        rots = rots.reshape(rots.shape[:-1] + (3, 3))
+
+        rot_obj = Rotation(rot_mats=rots, quats=None)
+
+        return Rigid(rot_obj, torch.stack(origin_unbound, dim=-1))
+
+    def unsqueeze(self, dim: int) -> Rigid:
+        """
+        Analogous to torch.unsqueeze. The dimension is relative to the shared dimensions of the rotation/translation.
+
+        Args:
+            dim: A positive or negative dimension index.
+        Returns:
+            The unsqueezed transformation.
+        """
+        if dim >= len(self.shape):
+            raise ValueError("Invalid dimension")
+        rots = self._rots.unsqueeze(dim)
+        trans = self._trans.unsqueeze(dim if dim >= 0 else dim - 1)
+
+        return Rigid(rots, trans)
+
+    @staticmethod
+    def cat(ts: Sequence[Rigid], dim: int) -> Rigid:
+        """
+        Concatenates transformations along a new dimension.
+
+        Args:
+            ts:
+                A list of T objects
+            dim:
+                The dimension along which the transformations should be concatenated
+        Returns:
+            A concatenated transformation object
+        """
+        rots = Rotation.cat([t._rots for t in ts], dim)
+        trans = torch.cat([t._trans for t in ts], dim=dim if dim >= 0 else dim - 1)
+
+        return Rigid(rots, trans)
+
+    def apply_rot_fn(self, fn: Callable[[Rotation], Rotation]) -> Rigid:
+        """
+        Applies a Rotation -> Rotation function to the stored rotation object.
+
+        Args:
+            fn: A function of type Rotation -> Rotation
+        Returns:
+            A transformation object with a transformed rotation.
+        """
+        return Rigid(fn(self._rots), self._trans)
+
+    def apply_trans_fn(self, fn: Callable[[torch.Tensor], torch.Tensor]) -> Rigid:
+        """
+        Applies a Tensor -> Tensor function to the stored translation.
+
+        Args:
+            fn:
+                A function of type Tensor -> Tensor to be applied to the translation
+        Returns:
+            A transformation object with a transformed translation.
+        """
+        return Rigid(self._rots, fn(self._trans))
+
+    def scale_translation(self, trans_scale_factor: float) -> Rigid:
+        """
+        Scales the translation by a constant factor.
+
+        Args:
+            trans_scale_factor:
+                The constant factor
+        Returns:
+            A transformation object with a scaled translation.
+        """
+        return self.apply_trans_fn(lambda t: t * trans_scale_factor)
+
+    def stop_rot_gradient(self) -> Rigid:
+        """
+        Detaches the underlying rotation object
+
+        Returns:
+            A transformation object with detached rotations
+        """
+        return self.apply_rot_fn(lambda r: r.detach())
+
+    @staticmethod
+    def make_transform_from_reference(
+        n_xyz: torch.Tensor, ca_xyz: torch.Tensor, c_xyz: torch.Tensor, eps: float = 1e-20
+    ) -> Rigid:
+        """
+        Returns a transformation object from reference coordinates.
+
+        Note that this method does not take care of symmetries. If you provide the atom positions in the non-standard
+        way, the N atom will end up not at [-0.527250, 1.359329, 0.0] but instead at [-0.527250, -1.359329, 0.0]. You
+        need to take care of such cases in your code.
+
+        Args:
+            n_xyz: A [*, 3] tensor of nitrogen xyz coordinates.
+            ca_xyz: A [*, 3] tensor of carbon alpha xyz coordinates.
+            c_xyz: A [*, 3] tensor of carbon xyz coordinates.
+        Returns:
+            A transformation object. After applying the translation and rotation to the reference backbone, the
+            coordinates will approximately equal to the input coordinates.
+        """
+        translation = -1 * ca_xyz
+        n_xyz = n_xyz + translation
+        c_xyz = c_xyz + translation
+
+        c_x, c_y, c_z = [c_xyz[..., i] for i in range(3)]
+        norm = torch.sqrt(eps + c_x**2 + c_y**2)
+        sin_c1 = -c_y / norm
+        cos_c1 = c_x / norm
+
+        c1_rots = sin_c1.new_zeros((*sin_c1.shape, 3, 3))
+        c1_rots[..., 0, 0] = cos_c1
+        c1_rots[..., 0, 1] = -1 * sin_c1
+        c1_rots[..., 1, 0] = sin_c1
+        c1_rots[..., 1, 1] = cos_c1
+        c1_rots[..., 2, 2] = 1
+
+        norm = torch.sqrt(eps + c_x**2 + c_y**2 + c_z**2)
+        sin_c2 = c_z / norm
+        cos_c2 = torch.sqrt(c_x**2 + c_y**2) / norm
+
+        c2_rots = sin_c2.new_zeros((*sin_c2.shape, 3, 3))
+        c2_rots[..., 0, 0] = cos_c2
+        c2_rots[..., 0, 2] = sin_c2
+        c2_rots[..., 1, 1] = 1
+        c2_rots[..., 2, 0] = -1 * sin_c2
+        c2_rots[..., 2, 2] = cos_c2
+
+        c_rots = rot_matmul(c2_rots, c1_rots)
+        n_xyz = rot_vec_mul(c_rots, n_xyz)
+
+        _, n_y, n_z = [n_xyz[..., i] for i in range(3)]
+        norm = torch.sqrt(eps + n_y**2 + n_z**2)
+        sin_n = -n_z / norm
+        cos_n = n_y / norm
+
+        n_rots = sin_c2.new_zeros((*sin_c2.shape, 3, 3))
+        n_rots[..., 0, 0] = 1
+        n_rots[..., 1, 1] = cos_n
+        n_rots[..., 1, 2] = -1 * sin_n
+        n_rots[..., 2, 1] = sin_n
+        n_rots[..., 2, 2] = cos_n
+
+        rots = rot_matmul(n_rots, c_rots)
+
+        rots = rots.transpose(-1, -2)
+        translation = -1 * translation
+
+        rot_obj = Rotation(rot_mats=rots, quats=None)
+
+        return Rigid(rot_obj, translation)
+
+    def cuda(self) -> Rigid:
+        """
+        Moves the transformation object to GPU memory
+
+        Returns:
+            A version of the transformation on GPU
+        """
+        return Rigid(self._rots.cuda(), self._trans.cuda())