Upload folder using huggingface_hub

checkpoint-284/README.md

@@ -0,0 +1,220 @@
+---
+library_name: peft
+base_model: mistralai/Mixtral-8x7B-v0.1
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure 
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+
+
+## Training procedure
+
+
+The following `bitsandbytes` quantization config was used during training:
+- quant_method: bitsandbytes
+- load_in_8bit: False
+- load_in_4bit: True
+- llm_int8_threshold: 6.0
+- llm_int8_skip_modules: None
+- llm_int8_enable_fp32_cpu_offload: False
+- llm_int8_has_fp16_weight: False
+- bnb_4bit_quant_type: nf4
+- bnb_4bit_use_double_quant: True
+- bnb_4bit_compute_dtype: bfloat16
+
+### Framework versions
+
+
+- PEFT 0.6.0

checkpoint-284/adapter_config.json

@@ -0,0 +1,29 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "mistralai/Mixtral-8x7B-v0.1",
+  "bias": "none",
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "lora_alpha": 64,
+  "lora_dropout": 0,
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate",
+    "w1",
+    "v_proj",
+    "k_proj",
+    "w3",
+    "o_proj",
+    "q_proj",
+    "w2"
+  ],
+  "task_type": "CAUSAL_LM"
+}

checkpoint-284/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7de8dbe58c960a8a3e90f5ebe902b4830193773274caf5e1c44cf80fdf5a71f7
+size 969176736

checkpoint-284/latest

@@ -0,0 +1 @@
+global_step284

checkpoint-284/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70c0b62bd953493d0675c34feecb6cb4564541e1efc60bb1704ce38255044908
+size 15607

checkpoint-284/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:241736953f399f6929c6ad898eaf28227b020331cba8da2c020aaaff41594b10
+size 15607

checkpoint-284/trainer_state.json

@@ -0,0 +1,1725 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.5026548672566372,
+  "eval_steps": 500,
+  "global_step": 284,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.807692307692308e-06,
+      "loss": 2.0125,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.615384615384616e-06,
+      "loss": 1.9307,
+      "step": 2
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4423076923076924e-05,
+      "loss": 1.6104,
+      "step": 3
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.923076923076923e-05,
+      "loss": 1.9208,
+      "step": 4
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4038461538461542e-05,
+      "loss": 1.7502,
+      "step": 5
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.884615384615385e-05,
+      "loss": 2.2112,
+      "step": 6
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.365384615384615e-05,
+      "loss": 1.5895,
+      "step": 7
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.846153846153846e-05,
+      "loss": 1.9856,
+      "step": 8
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.3269230769230766e-05,
+      "loss": 2.1565,
+      "step": 9
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.8076923076923084e-05,
+      "loss": 1.6874,
+      "step": 10
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.288461538461539e-05,
+      "loss": 1.614,
+      "step": 11
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.76923076923077e-05,
+      "loss": 1.8226,
+      "step": 12
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.25e-05,
+      "loss": 1.4058,
+      "step": 13
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.73076923076923e-05,
+      "loss": 1.4717,
+      "step": 14
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 7.211538461538461e-05,
+      "loss": 1.5335,
+      "step": 15
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 7.692307692307693e-05,
+      "loss": 2.1125,
+      "step": 16
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.173076923076923e-05,
+      "loss": 1.9451,
+      "step": 17
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.653846153846153e-05,
+      "loss": 1.7484,
+      "step": 18
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.134615384615384e-05,
+      "loss": 1.7573,
+      "step": 19
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.615384615384617e-05,
+      "loss": 2.1968,
+      "step": 20
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00010096153846153847,
+      "loss": 1.7941,
+      "step": 21
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00010576923076923077,
+      "loss": 1.8685,
+      "step": 22
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00011057692307692308,
+      "loss": 2.0065,
+      "step": 23
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001153846153846154,
+      "loss": 1.9018,
+      "step": 24
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001201923076923077,
+      "loss": 2.0752,
+      "step": 25
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000125,
+      "loss": 1.716,
+      "step": 26
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00012980769230769233,
+      "loss": 1.6542,
+      "step": 27
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001346153846153846,
+      "loss": 1.7198,
+      "step": 28
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013942307692307694,
+      "loss": 1.8383,
+      "step": 29
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014423076923076922,
+      "loss": 1.6938,
+      "step": 30
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014903846153846155,
+      "loss": 1.9142,
+      "step": 31
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015384615384615385,
+      "loss": 1.7715,
+      "step": 32
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015865384615384616,
+      "loss": 1.467,
+      "step": 33
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016346153846153846,
+      "loss": 1.7608,
+      "step": 34
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001682692307692308,
+      "loss": 1.5371,
+      "step": 35
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017307692307692307,
+      "loss": 1.6211,
+      "step": 36
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001778846153846154,
+      "loss": 1.7275,
+      "step": 37
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00018269230769230767,
+      "loss": 1.7063,
+      "step": 38
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001875,
+      "loss": 1.9367,
+      "step": 39
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019230769230769233,
+      "loss": 1.6608,
+      "step": 40
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001971153846153846,
+      "loss": 1.9212,
+      "step": 41
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00020192307692307694,
+      "loss": 1.611,
+      "step": 42
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00020673076923076922,
+      "loss": 1.8582,
+      "step": 43
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00021153846153846155,
+      "loss": 1.5658,
+      "step": 44
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00021634615384615385,
+      "loss": 1.8543,
+      "step": 45
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00022115384615384616,
+      "loss": 1.7291,
+      "step": 46
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00022596153846153846,
+      "loss": 1.9411,
+      "step": 47
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002307692307692308,
+      "loss": 1.8138,
+      "step": 48
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00023557692307692307,
+      "loss": 1.708,
+      "step": 49
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002403846153846154,
+      "loss": 1.675,
+      "step": 50
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002451923076923077,
+      "loss": 1.658,
+      "step": 51
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00025,
+      "loss": 1.9797,
+      "step": 52
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000249997656075194,
+      "loss": 1.5857,
+      "step": 53
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002499906243886798,
+      "loss": 1.8837,
+      "step": 54
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00024997890520416535,
+      "loss": 1.8022,
+      "step": 55
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002499624989611527,
+      "loss": 1.9101,
+      "step": 56
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00024994140627492207,
+      "loss": 1.5614,
+      "step": 57
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00024991562793650793,
+      "loss": 1.8988,
+      "step": 58
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002498851649126703,
+      "loss": 1.7089,
+      "step": 59
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00024985001834585763,
+      "loss": 1.7782,
+      "step": 60
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002498101895541645,
+      "loss": 1.5338,
+      "step": 61
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002497656800312821,
+      "loss": 1.6484,
+      "step": 62
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002497164914464419,
+      "loss": 1.733,
+      "step": 63
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00024966262564435343,
+      "loss": 1.8893,
+      "step": 64
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000249604084645135,
+      "loss": 1.6361,
+      "step": 65
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002495408706442377,
+      "loss": 1.9827,
+      "step": 66
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00024947298601236343,
+      "loss": 1.6801,
+      "step": 67
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002494004332953758,
+      "loss": 1.9678,
+      "step": 68
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00024932321521420456,
+      "loss": 1.9245,
+      "step": 69
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002492413346647437,
+      "loss": 1.5254,
+      "step": 70
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00024915479471774286,
+      "loss": 1.7283,
+      "step": 71
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00024906359861869216,
+      "loss": 1.9968,
+      "step": 72
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002489677497877003,
+      "loss": 1.6559,
+      "step": 73
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002488672518193665,
+      "loss": 1.5518,
+      "step": 74
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002487621084826458,
+      "loss": 1.6201,
+      "step": 75
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002486523237207072,
+      "loss": 1.7592,
+      "step": 76
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00024853790165078654,
+      "loss": 1.5929,
+      "step": 77
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002484188465640313,
+      "loss": 1.4261,
+      "step": 78
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002482951629253403,
+      "loss": 1.8929,
+      "step": 79
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002481668553731959,
+      "loss": 1.5752,
+      "step": 80
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00024803392871949013,
+      "loss": 1.8596,
+      "step": 81
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00024789638794934436,
+      "loss": 1.838,
+      "step": 82
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00024775423822092214,
+      "loss": 1.7938,
+      "step": 83
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002476074848652358,
+      "loss": 1.6448,
+      "step": 84
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002474561333859467,
+      "loss": 1.7674,
+      "step": 85
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00024730018945915864,
+      "loss": 1.6526,
+      "step": 86
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000247139658933205,
+      "loss": 1.6921,
+      "step": 87
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00024697454782842944,
+      "loss": 1.8208,
+      "step": 88
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002468048623369603,
+      "loss": 1.9528,
+      "step": 89
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00024663060882247796,
+      "loss": 1.8188,
+      "step": 90
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00024645179381997673,
+      "loss": 1.921,
+      "step": 91
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00024626842403551927,
+      "loss": 1.9332,
+      "step": 92
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002460805063459853,
+      "loss": 1.967,
+      "step": 93
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00024588804779881383,
+      "loss": 1.4963,
+      "step": 94
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00024569105561173866,
+      "loss": 1.6944,
+      "step": 95
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00024548953717251783,
+      "loss": 1.9083,
+      "step": 96
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002452835000386563,
+      "loss": 1.7368,
+      "step": 97
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000245072951937123,
+      "loss": 1.6183,
+      "step": 98
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00024485790076406047,
+      "loss": 1.7917,
+      "step": 99
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00024463835458448925,
+      "loss": 2.1032,
+      "step": 100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002444143216320052,
+      "loss": 1.8451,
+      "step": 101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002441858103084705,
+      "loss": 1.7479,
+      "step": 102
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000243952829183699,
+      "loss": 1.8158,
+      "step": 103
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00024371538699513443,
+      "loss": 1.9275,
+      "step": 104
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00024347349264752303,
+      "loss": 1.7759,
+      "step": 105
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00024322715521257933,
+      "loss": 2.0328,
+      "step": 106
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00024297638392864617,
+      "loss": 1.8839,
+      "step": 107
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00024272118820034804,
+      "loss": 1.8443,
+      "step": 108
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00024246157759823855,
+      "loss": 1.9779,
+      "step": 109
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00024219756185844132,
+      "loss": 1.6495,
+      "step": 110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00024192915088228512,
+      "loss": 1.6645,
+      "step": 111
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00024165635473593215,
+      "loss": 1.4214,
+      "step": 112
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00024137918365000095,
+      "loss": 1.6603,
+      "step": 113
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00024109764801918244,
+      "loss": 1.4227,
+      "step": 114
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002408117584018502,
+      "loss": 1.3033,
+      "step": 115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024052152551966457,
+      "loss": 2.0312,
+      "step": 116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024022696025717023,
+      "loss": 1.8867,
+      "step": 117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023992807366138847,
+      "loss": 1.6906,
+      "step": 118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023962487694140263,
+      "loss": 1.595,
+      "step": 119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023931738146793763,
+      "loss": 1.5583,
+      "step": 120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023900559877293383,
+      "loss": 1.9068,
+      "step": 121
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00023868954054911428,
+      "loss": 1.4536,
+      "step": 122
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00023836921864954635,
+      "loss": 1.7105,
+      "step": 123
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002380446450871972,
+      "loss": 1.4355,
+      "step": 124
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00023771583203448322,
+      "loss": 1.7782,
+      "step": 125
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00023738279182281352,
+      "loss": 1.9977,
+      "step": 126
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00023704553694212752,
+      "loss": 1.5658,
+      "step": 127
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00023670408004042653,
+      "loss": 1.5389,
+      "step": 128
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00023635843392329938,
+      "loss": 1.7435,
+      "step": 129
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00023600861155344223,
+      "loss": 1.5706,
+      "step": 130
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00023565462605017228,
+      "loss": 1.8396,
+      "step": 131
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00023529649068893598,
+      "loss": 1.8073,
+      "step": 132
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00023493421890081112,
+      "loss": 1.9954,
+      "step": 133
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00023456782427200295,
+      "loss": 1.9553,
+      "step": 134
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002341973205433348,
+      "loss": 1.2249,
+      "step": 135
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002338227216097328,
+      "loss": 1.8021,
+      "step": 136
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00023344404151970464,
+      "loss": 1.8086,
+      "step": 137
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00023306129447481282,
+      "loss": 1.4283,
+      "step": 138
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00023267449482914203,
+      "loss": 1.8477,
+      "step": 139
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002322836570887608,
+      "loss": 1.3233,
+      "step": 140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002318887959111776,
+      "loss": 1.7511,
+      "step": 141
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002314899261047909,
+      "loss": 1.9737,
+      "step": 142
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00023108706262833407,
+      "loss": 1.6347,
+      "step": 143
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00023068022059031425,
+      "loss": 1.9142,
+      "step": 144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00023026941524844572,
+      "loss": 1.7408,
+      "step": 145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00022985466200907783,
+      "loss": 1.3576,
+      "step": 146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00022943597642661705,
+      "loss": 1.6058,
+      "step": 147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00022901337420294378,
+      "loss": 1.6575,
+      "step": 148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002285868711868233,
+      "loss": 1.3605,
+      "step": 149
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00022815648337331168,
+      "loss": 1.4417,
+      "step": 150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00022772222690315563,
+      "loss": 1.684,
+      "step": 151
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002272841180621874,
+      "loss": 1.8368,
+      "step": 152
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00022684217328071383,
+      "loss": 1.6624,
+      "step": 153
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00022639640913290027,
+      "loss": 1.8384,
+      "step": 154
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00022594684233614908,
+      "loss": 1.4881,
+      "step": 155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00022549348975047252,
+      "loss": 1.7802,
+      "step": 156
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00022503636837786052,
+      "loss": 1.6981,
+      "step": 157
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00022457549536164306,
+      "loss": 1.6845,
+      "step": 158
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00022411088798584728,
+      "loss": 1.8524,
+      "step": 159
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00022364256367454923,
+      "loss": 1.7893,
+      "step": 160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00022317053999122038,
+      "loss": 1.5371,
+      "step": 161
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00022269483463806917,
+      "loss": 1.7511,
+      "step": 162
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00022221546545537674,
+      "loss": 1.4536,
+      "step": 163
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00022173245042082822,
+      "loss": 1.7809,
+      "step": 164
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002212458076488384,
+      "loss": 1.7447,
+      "step": 165
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00022075555538987224,
+      "loss": 2.1818,
+      "step": 166
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002202617120297607,
+      "loss": 1.8901,
+      "step": 167
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00021976429608901093,
+      "loss": 1.6585,
+      "step": 168
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00021926332622211205,
+      "loss": 1.9877,
+      "step": 169
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002187588212168351,
+      "loss": 1.4926,
+      "step": 170
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00021825079999352893,
+      "loss": 1.7696,
+      "step": 171
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002177392816044102,
+      "loss": 1.9484,
+      "step": 172
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00021722428523284927,
+      "loss": 1.6379,
+      "step": 173
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00021670583019265034,
+      "loss": 1.4783,
+      "step": 174
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002161839359273276,
+      "loss": 1.6047,
+      "step": 175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00021565862200937565,
+      "loss": 1.6504,
+      "step": 176
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00021512990813953562,
+      "loss": 1.6248,
+      "step": 177
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00021459781414605642,
+      "loss": 2.0182,
+      "step": 178
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000214062359983951,
+      "loss": 1.7979,
+      "step": 179
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00021352356573424807,
+      "loss": 1.7023,
+      "step": 180
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00021298145160323896,
+      "loss": 1.6911,
+      "step": 181
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00021243603792171976,
+      "loss": 1.8985,
+      "step": 182
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00021188734514422902,
+      "loss": 2.1085,
+      "step": 183
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00021133539384828054,
+      "loss": 1.8299,
+      "step": 184
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00021078020473359172,
+      "loss": 1.8975,
+      "step": 185
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00021022179862130704,
+      "loss": 1.5791,
+      "step": 186
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00020966019645321765,
+      "loss": 1.7475,
+      "step": 187
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002090954192909755,
+      "loss": 1.6978,
+      "step": 188
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00020852748831530382,
+      "loss": 1.7039,
+      "step": 189
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00020795642482520266,
+      "loss": 1.5206,
+      "step": 190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00020738225023715013,
+      "loss": 1.6684,
+      "step": 191
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00020680498608429914,
+      "loss": 1.7339,
+      "step": 192
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00020622465401566999,
+      "loss": 1.6174,
+      "step": 193
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00020564127579533831,
+      "loss": 1.6451,
+      "step": 194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00020505487330161915,
+      "loss": 1.6214,
+      "step": 195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00020446546852624604,
+      "loss": 1.7493,
+      "step": 196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00020387308357354655,
+      "loss": 1.7067,
+      "step": 197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002032777406596133,
+      "loss": 1.6918,
+      "step": 198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00020267946211147058,
+      "loss": 1.9448,
+      "step": 199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00020207827036623744,
+      "loss": 1.8517,
+      "step": 200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002014741879702856,
+      "loss": 2.0169,
+      "step": 201
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002008672375783946,
+      "loss": 2.1721,
+      "step": 202
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00020025744195290167,
+      "loss": 1.5175,
+      "step": 203
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019964482396284827,
+      "loss": 1.6105,
+      "step": 204
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001990294065831225,
+      "loss": 1.9778,
+      "step": 205
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019841121289359737,
+      "loss": 1.6379,
+      "step": 206
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001977902660782652,
+      "loss": 1.7973,
+      "step": 207
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019716658942436834,
+      "loss": 1.6022,
+      "step": 208
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019654020632152563,
+      "loss": 1.4371,
+      "step": 209
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019591114026085537,
+      "loss": 1.8489,
+      "step": 210
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001952794148340943,
+      "loss": 1.6484,
+      "step": 211
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019464505373271274,
+      "loss": 1.841,
+      "step": 212
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019400808074702624,
+      "loss": 1.4165,
+      "step": 213
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019336851976530338,
+      "loss": 1.8259,
+      "step": 214
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001927263947728697,
+      "loss": 1.8661,
+      "step": 215
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019208172985120837,
+      "loss": 1.4957,
+      "step": 216
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000191434549177057,
+      "loss": 1.3717,
+      "step": 217
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019078487702150102,
+      "loss": 1.5568,
+      "step": 218
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001901327377490633,
+      "loss": 1.512,
+      "step": 219
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018947815581679052,
+      "loss": 1.7289,
+      "step": 220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018882115577333592,
+      "loss": 1.6628,
+      "step": 221
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00018816176225803876,
+      "loss": 1.8897,
+      "step": 222
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001875,
+      "loss": 1.2269,
+      "step": 223
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018683589381715532,
+      "loss": 1.9278,
+      "step": 224
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018616946861534396,
+      "loss": 1.5643,
+      "step": 225
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001855007493873749,
+      "loss": 1.8058,
+      "step": 226
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001848297612120895,
+      "loss": 1.6883,
+      "step": 227
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00018415652925342105,
+      "loss": 1.4985,
+      "step": 228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000183481078759451,
+      "loss": 1.9467,
+      "step": 229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018280343506146197,
+      "loss": 1.6599,
+      "step": 230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018212362357298797,
+      "loss": 1.6791,
+      "step": 231
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001814416697888612,
+      "loss": 2.1714,
+      "step": 232
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018075759928425582,
+      "loss": 1.6677,
+      "step": 233
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018007143771372916,
+      "loss": 1.9834,
+      "step": 234
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00017938321081025917,
+      "loss": 1.8803,
+      "step": 235
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00017869294438427964,
+      "loss": 1.8268,
+      "step": 236
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001780006643227121,
+      "loss": 1.9126,
+      "step": 237
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00017730639658799512,
+      "loss": 1.3397,
+      "step": 238
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00017661016721711063,
+      "loss": 1.5539,
+      "step": 239
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00017591200232060719,
+      "loss": 1.8692,
+      "step": 240
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001752119280816212,
+      "loss": 1.5348,
+      "step": 241
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00017450997075489462,
+      "loss": 1.9957,
+      "step": 242
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00017380615666579054,
+      "loss": 1.7454,
+      "step": 243
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00017310051220930574,
+      "loss": 1.8623,
+      "step": 244
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00017239306384908096,
+      "loss": 1.9291,
+      "step": 245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017168383811640842,
+      "loss": 1.6059,
+      "step": 246
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017097286160923668,
+      "loss": 1.8271,
+      "step": 247
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001702601609911733,
+      "loss": 1.8032,
+      "step": 248
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001695457629904848,
+      "loss": 1.6174,
+      "step": 249
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016882969439909433,
+      "loss": 1.4768,
+      "step": 250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001681119820715768,
+      "loss": 1.9159,
+      "step": 251
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016739265292415185,
+      "loss": 1.5289,
+      "step": 252
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016667173393367446,
+      "loss": 1.8026,
+      "step": 253
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016594925213662303,
+      "loss": 1.6027,
+      "step": 254
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016522523462808572,
+      "loss": 1.8634,
+      "step": 255
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001644997085607441,
+      "loss": 1.7965,
+      "step": 256
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016377270114385485,
+      "loss": 1.8084,
+      "step": 257
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001630442396422295,
+      "loss": 1.5158,
+      "step": 258
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016231435137521182,
+      "loss": 1.8523,
+      "step": 259
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016158306371565322,
+      "loss": 1.2613,
+      "step": 260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001608504040888863,
+      "loss": 1.6135,
+      "step": 261
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001601163999716962,
+      "loss": 1.4411,
+      "step": 262
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015938107889129023,
+      "loss": 1.7133,
+      "step": 263
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015864446842426554,
+      "loss": 1.5927,
+      "step": 264
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001579065961955749,
+      "loss": 1.5994,
+      "step": 265
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015716748987749065,
+      "loss": 1.6315,
+      "step": 266
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001564271771885668,
+      "loss": 2.0437,
+      "step": 267
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001556856858925999,
+      "loss": 1.9095,
+      "step": 268
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015494304379758735,
+      "loss": 1.5502,
+      "step": 269
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015419927875468485,
+      "loss": 1.4151,
+      "step": 270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001534544186571617,
+      "loss": 1.6274,
+      "step": 271
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015270849143935483,
+      "loss": 1.6849,
+      "step": 272
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015196152507562127,
+      "loss": 1.9244,
+      "step": 273
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001512135475792888,
+      "loss": 2.0862,
+      "step": 274
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015046458700160553,
+      "loss": 1.7078,
+      "step": 275
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00014971467143068791,
+      "loss": 1.4565,
+      "step": 276
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001489638289904673,
+      "loss": 1.8075,
+      "step": 277
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00014821208783963522,
+      "loss": 2.2061,
+      "step": 278
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00014745947617058735,
+      "loss": 1.6676,
+      "step": 279
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014670602220836633,
+      "loss": 2.0041,
+      "step": 280
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014595175420960293,
+      "loss": 1.5538,
+      "step": 281
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014519670046145685,
+      "loss": 1.5781,
+      "step": 282
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001444408892805554,
+      "loss": 1.703,
+      "step": 283
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001436843490119318,
+      "loss": 1.8105,
+      "step": 284
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 565,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 142,
+  "total_flos": 2.627459909913936e+18,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-284/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88c1454fe3e0c5f38812eb2f4667c494fbefda796a453029074b9ce12e2fe147
+size 6011

checkpoint-284/zero_to_fp32.py

@@ -0,0 +1,587 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel()
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file, tag=args.tag)