Spaces:
Running
on
A100
Running
on
A100
File size: 2,902 Bytes
174ae06 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 |
# Copyright (c) 2025 NVIDIA CORPORATION.
# Licensed under the MIT license.
# Adapted from https://github.com/NVlabs/VILA/tree/main under the Apache 2.0 license.
# LICENSE is in incl_licenses directory.
# Copyright 2024 NVIDIA CORPORATION & AFFILIATES
#
# 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.
#
# SPDX-License-Identifier: Apache-2.0
# This file is modified from https://github.com/haotian-liu/LLaVA/
import os
import torch
from transformers import PretrainedConfig, PreTrainedModel
from .base_projector import MultimodalProjector, MultimodalProjectorConfig
from .speech_base_projector import SpeechMultimodalProjector, SpeechMultimodalProjectorConfig
from .sound_base_projector import SoundMultimodalProjector, SoundMultimodalProjectorConfig
def build_speech_mm_projector(model_type_or_path: str, config: PretrainedConfig) -> PreTrainedModel:
if model_type_or_path is None:
return None
## load from pretrained model
if config.resume_path:
assert os.path.exists(model_type_or_path), f"Resume speech mm projector path {model_type_or_path} does not exist!"
return SpeechMultimodalProjector.from_pretrained(model_type_or_path, config, torch_dtype=eval(config.model_dtype))
## build from scratch
else:
print("WARNING: Building speech multimodal projector from scratch!")
speech_mm_projector_cfg = SpeechMultimodalProjectorConfig(model_type_or_path)
speech_mm_projector = SpeechMultimodalProjector(speech_mm_projector_cfg, config).to(eval(config.model_dtype))
return speech_mm_projector
def build_sound_mm_projector(model_type_or_path: str, config: PretrainedConfig) -> PreTrainedModel:
if model_type_or_path is None:
return None
## load from pretrained model
if config.resume_path:
print(config.resume_path)
assert os.path.exists(model_type_or_path), f"Resume sound mm projector path {model_type_or_path} does not exist!"
return SoundMultimodalProjector.from_pretrained(model_type_or_path, config, torch_dtype=eval(config.model_dtype))
# build from scratch
else:
print("WARNING: Building sound multimodal projector from scratch!")
sound_mm_projector_cfg = SoundMultimodalProjectorConfig(model_type_or_path)
sound_mm_projector = SoundMultimodalProjector(sound_mm_projector_cfg, config).to(eval(config.model_dtype))
return sound_mm_projector
|