Upload folder using huggingface_hub

14_128_e5_3e-5/checkpoint-1040/README.md

@@ -0,0 +1,202 @@
+---
+base_model: ibm-granite/granite-3.3-8b-base
+library_name: peft
+---
+
+# 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]
+### Framework versions
+
+- PEFT 0.15.2

14_128_e5_3e-5/checkpoint-1040/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "v_proj",
+    "q_proj",
+    "k_proj",
+    "up_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

14_128_e5_3e-5/checkpoint-1040/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b58e96a981dca7803306739b8e397eb2ac77dcea6a920cf94eec5e43a1d2b90a
+size 791751704

14_128_e5_3e-5/checkpoint-1040/latest

@@ -0,0 +1 @@
+global_step1040

14_128_e5_3e-5/checkpoint-1040/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cdb6c1fa3153f84d11400d54d8c83791b839380e5fa07c909e8fa791c354fff
+size 15920

14_128_e5_3e-5/checkpoint-1040/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:635b856239aa7caba9b2ff261397499eac18b0b81f906b4b404dea44ce32e121
+size 15920

14_128_e5_3e-5/checkpoint-1040/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:246419b60d791be0cbae79666766bdff3e4ab307cf547d5f0765ad4a04995026
+size 15920

14_128_e5_3e-5/checkpoint-1040/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7723b567f37e70d861aec180454c69e2ab73bf94db3b783c27788278d0c37801
+size 15920

14_128_e5_3e-5/checkpoint-1040/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6a6b0f52ed9395fdb055bd5006a746320a1567c4b2c6e4b7d31ab124221a0e5
+size 15920

14_128_e5_3e-5/checkpoint-1040/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:17d336d993f01d77978d3ed89dd7c322eed577988976ca2be7f40fe81f53866e
+size 15920

14_128_e5_3e-5/checkpoint-1040/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97834dacd5f821a27b8e16668cb2ed61dde2853877f5b7e2e2f036ad33f19d6f
+size 15920

14_128_e5_3e-5/checkpoint-1040/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92a0bbe0c316785984917e8819b070139b94633585427c2fce41439a86dde988
+size 15920

14_128_e5_3e-5/checkpoint-1040/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab47d59cc0e7966553505f40b8970d638d2f1a59504bc59effcd578237756eb8
+size 1064

14_128_e5_3e-5/checkpoint-1040/special_tokens_map.json

@@ -0,0 +1,45 @@
+{
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<reponame>",
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

14_128_e5_3e-5/checkpoint-1040/tokenizer_config.json

@@ -0,0 +1,188 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<fim_prefix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<fim_middle>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<fim_suffix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<fim_pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<filename>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<gh_stars>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<issue_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "8": {
+      "content": "<issue_comment>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "9": {
+      "content": "<issue_closed>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "10": {
+      "content": "<jupyter_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "11": {
+      "content": "<jupyter_text>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "12": {
+      "content": "<jupyter_code>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "13": {
+      "content": "<jupyter_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "14": {
+      "content": "<empty_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "15": {
+      "content": "<commit_before>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "16": {
+      "content": "<commit_msg>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "17": {
+      "content": "<commit_after>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "18": {
+      "content": "<reponame>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<reponame>",
+  "padding_side": "left",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

14_128_e5_3e-5/checkpoint-1040/trainer_state.json

@@ -0,0 +1,1490 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 1040,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.019230769230769232,
+      "grad_norm": 1.4232831001281738,
+      "learning_rate": 1.8461538461538462e-06,
+      "loss": 1.3571,
+      "step": 5
+    },
+    {
+      "epoch": 0.038461538461538464,
+      "grad_norm": 1.0114805698394775,
+      "learning_rate": 4.1538461538461545e-06,
+      "loss": 1.3519,
+      "step": 10
+    },
+    {
+      "epoch": 0.057692307692307696,
+      "grad_norm": 0.7543752789497375,
+      "learning_rate": 6.461538461538462e-06,
+      "loss": 1.2802,
+      "step": 15
+    },
+    {
+      "epoch": 0.07692307692307693,
+      "grad_norm": 0.5682826042175293,
+      "learning_rate": 8.76923076923077e-06,
+      "loss": 1.2099,
+      "step": 20
+    },
+    {
+      "epoch": 0.09615384615384616,
+      "grad_norm": 0.5222262144088745,
+      "learning_rate": 1.1076923076923077e-05,
+      "loss": 1.2572,
+      "step": 25
+    },
+    {
+      "epoch": 0.11538461538461539,
+      "grad_norm": 0.547173261642456,
+      "learning_rate": 1.3384615384615386e-05,
+      "loss": 1.2272,
+      "step": 30
+    },
+    {
+      "epoch": 0.1346153846153846,
+      "grad_norm": 0.4619555175304413,
+      "learning_rate": 1.5692307692307693e-05,
+      "loss": 1.1852,
+      "step": 35
+    },
+    {
+      "epoch": 0.15384615384615385,
+      "grad_norm": 0.5007979273796082,
+      "learning_rate": 1.8e-05,
+      "loss": 1.2387,
+      "step": 40
+    },
+    {
+      "epoch": 0.17307692307692307,
+      "grad_norm": 0.4946490526199341,
+      "learning_rate": 2.0307692307692308e-05,
+      "loss": 1.1931,
+      "step": 45
+    },
+    {
+      "epoch": 0.19230769230769232,
+      "grad_norm": 0.5487769842147827,
+      "learning_rate": 2.2615384615384615e-05,
+      "loss": 1.2114,
+      "step": 50
+    },
+    {
+      "epoch": 0.21153846153846154,
+      "grad_norm": 0.46380364894866943,
+      "learning_rate": 2.4923076923076926e-05,
+      "loss": 1.1068,
+      "step": 55
+    },
+    {
+      "epoch": 0.23076923076923078,
+      "grad_norm": 0.5553277134895325,
+      "learning_rate": 2.7230769230769233e-05,
+      "loss": 1.1258,
+      "step": 60
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.5975745320320129,
+      "learning_rate": 2.953846153846154e-05,
+      "loss": 1.1133,
+      "step": 65
+    },
+    {
+      "epoch": 0.2692307692307692,
+      "grad_norm": 0.4654400646686554,
+      "learning_rate": 2.9999223496658065e-05,
+      "loss": 1.0798,
+      "step": 70
+    },
+    {
+      "epoch": 0.28846153846153844,
+      "grad_norm": 0.5261916518211365,
+      "learning_rate": 2.9996069089616144e-05,
+      "loss": 1.105,
+      "step": 75
+    },
+    {
+      "epoch": 0.3076923076923077,
+      "grad_norm": 0.5102323293685913,
+      "learning_rate": 2.9990488757313582e-05,
+      "loss": 1.1385,
+      "step": 80
+    },
+    {
+      "epoch": 0.3269230769230769,
+      "grad_norm": 0.49091285467147827,
+      "learning_rate": 2.99824834024849e-05,
+      "loss": 1.0717,
+      "step": 85
+    },
+    {
+      "epoch": 0.34615384615384615,
+      "grad_norm": 0.5449517369270325,
+      "learning_rate": 2.9972054320162297e-05,
+      "loss": 1.0408,
+      "step": 90
+    },
+    {
+      "epoch": 0.36538461538461536,
+      "grad_norm": 0.5686984062194824,
+      "learning_rate": 2.9959203197466193e-05,
+      "loss": 1.0707,
+      "step": 95
+    },
+    {
+      "epoch": 0.38461538461538464,
+      "grad_norm": 0.6011113524436951,
+      "learning_rate": 2.9943932113332267e-05,
+      "loss": 1.0017,
+      "step": 100
+    },
+    {
+      "epoch": 0.40384615384615385,
+      "grad_norm": 0.5413767695426941,
+      "learning_rate": 2.9926243538175172e-05,
+      "loss": 1.017,
+      "step": 105
+    },
+    {
+      "epoch": 0.4230769230769231,
+      "grad_norm": 0.5385787487030029,
+      "learning_rate": 2.990614033348888e-05,
+      "loss": 1.0517,
+      "step": 110
+    },
+    {
+      "epoch": 0.4423076923076923,
+      "grad_norm": 0.7817659974098206,
+      "learning_rate": 2.9883625751383776e-05,
+      "loss": 0.9498,
+      "step": 115
+    },
+    {
+      "epoch": 0.46153846153846156,
+      "grad_norm": 0.636008620262146,
+      "learning_rate": 2.985870343406057e-05,
+      "loss": 0.9977,
+      "step": 120
+    },
+    {
+      "epoch": 0.4807692307692308,
+      "grad_norm": 0.6735332608222961,
+      "learning_rate": 2.983137741322109e-05,
+      "loss": 0.951,
+      "step": 125
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.7509468197822571,
+      "learning_rate": 2.980165210941607e-05,
+      "loss": 0.9125,
+      "step": 130
+    },
+    {
+      "epoch": 0.5192307692307693,
+      "grad_norm": 0.7218055725097656,
+      "learning_rate": 2.9769532331330037e-05,
+      "loss": 0.9313,
+      "step": 135
+    },
+    {
+      "epoch": 0.5384615384615384,
+      "grad_norm": 0.7202168703079224,
+      "learning_rate": 2.97350232750034e-05,
+      "loss": 0.8895,
+      "step": 140
+    },
+    {
+      "epoch": 0.5576923076923077,
+      "grad_norm": 0.7541701197624207,
+      "learning_rate": 2.96981305229919e-05,
+      "loss": 0.9392,
+      "step": 145
+    },
+    {
+      "epoch": 0.5769230769230769,
+      "grad_norm": 0.7201166152954102,
+      "learning_rate": 2.9658860043463485e-05,
+      "loss": 0.8911,
+      "step": 150
+    },
+    {
+      "epoch": 0.5961538461538461,
+      "grad_norm": 0.8302095532417297,
+      "learning_rate": 2.9617218189232867e-05,
+      "loss": 0.9174,
+      "step": 155
+    },
+    {
+      "epoch": 0.6153846153846154,
+      "grad_norm": 0.762272298336029,
+      "learning_rate": 2.9573211696733812e-05,
+      "loss": 0.8463,
+      "step": 160
+    },
+    {
+      "epoch": 0.6346153846153846,
+      "grad_norm": 0.8819282054901123,
+      "learning_rate": 2.9526847684929363e-05,
+      "loss": 0.8511,
+      "step": 165
+    },
+    {
+      "epoch": 0.6538461538461539,
+      "grad_norm": 0.8169696927070618,
+      "learning_rate": 2.947813365416023e-05,
+      "loss": 0.8837,
+      "step": 170
+    },
+    {
+      "epoch": 0.6730769230769231,
+      "grad_norm": 0.7911653518676758,
+      "learning_rate": 2.9427077484931443e-05,
+      "loss": 0.8293,
+      "step": 175
+    },
+    {
+      "epoch": 0.6923076923076923,
+      "grad_norm": 0.7542786002159119,
+      "learning_rate": 2.9373687436637492e-05,
+      "loss": 0.8309,
+      "step": 180
+    },
+    {
+      "epoch": 0.7115384615384616,
+      "grad_norm": 0.8718588948249817,
+      "learning_rate": 2.9317972146226255e-05,
+      "loss": 0.8319,
+      "step": 185
+    },
+    {
+      "epoch": 0.7307692307692307,
+      "grad_norm": 0.8762234449386597,
+      "learning_rate": 2.9259940626801725e-05,
+      "loss": 0.8576,
+      "step": 190
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.7861353158950806,
+      "learning_rate": 2.919960226616601e-05,
+      "loss": 0.8158,
+      "step": 195
+    },
+    {
+      "epoch": 0.7692307692307693,
+      "grad_norm": 1.0044094324111938,
+      "learning_rate": 2.9136966825300622e-05,
+      "loss": 0.8114,
+      "step": 200
+    },
+    {
+      "epoch": 0.7884615384615384,
+      "grad_norm": 0.8322351574897766,
+      "learning_rate": 2.907204443678747e-05,
+      "loss": 0.7426,
+      "step": 205
+    },
+    {
+      "epoch": 0.8076923076923077,
+      "grad_norm": 0.9439869523048401,
+      "learning_rate": 2.900484560316967e-05,
+      "loss": 0.7483,
+      "step": 210
+    },
+    {
+      "epoch": 0.8269230769230769,
+      "grad_norm": 1.0131150484085083,
+      "learning_rate": 2.8935381195252562e-05,
+      "loss": 0.7888,
+      "step": 215
+    },
+    {
+      "epoch": 0.8461538461538461,
+      "grad_norm": 0.8887054324150085,
+      "learning_rate": 2.8863662450345126e-05,
+      "loss": 0.7283,
+      "step": 220
+    },
+    {
+      "epoch": 0.8653846153846154,
+      "grad_norm": 0.987267255783081,
+      "learning_rate": 2.8789700970442124e-05,
+      "loss": 0.7665,
+      "step": 225
+    },
+    {
+      "epoch": 0.8846153846153846,
+      "grad_norm": 0.8717510104179382,
+      "learning_rate": 2.8713508720347224e-05,
+      "loss": 0.7308,
+      "step": 230
+    },
+    {
+      "epoch": 0.9038461538461539,
+      "grad_norm": 0.881127655506134,
+      "learning_rate": 2.863509802573744e-05,
+      "loss": 0.7231,
+      "step": 235
+    },
+    {
+      "epoch": 0.9230769230769231,
+      "grad_norm": 0.9518759250640869,
+      "learning_rate": 2.8554481571169208e-05,
+      "loss": 0.6872,
+      "step": 240
+    },
+    {
+      "epoch": 0.9423076923076923,
+      "grad_norm": 0.8832864761352539,
+      "learning_rate": 2.8471672398026414e-05,
+      "loss": 0.7028,
+      "step": 245
+    },
+    {
+      "epoch": 0.9615384615384616,
+      "grad_norm": 0.9555291533470154,
+      "learning_rate": 2.838668390241063e-05,
+      "loss": 0.6354,
+      "step": 250
+    },
+    {
+      "epoch": 0.9807692307692307,
+      "grad_norm": 1.1762830018997192,
+      "learning_rate": 2.8299529832974076e-05,
+      "loss": 0.6844,
+      "step": 255
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.9526051878929138,
+      "learning_rate": 2.8210224288695437e-05,
+      "loss": 0.6363,
+      "step": 260
+    },
+    {
+      "epoch": 1.0192307692307692,
+      "grad_norm": 1.0443257093429565,
+      "learning_rate": 2.81187817165991e-05,
+      "loss": 0.6202,
+      "step": 265
+    },
+    {
+      "epoch": 1.0384615384615385,
+      "grad_norm": 1.000659465789795,
+      "learning_rate": 2.8025216909418033e-05,
+      "loss": 0.5572,
+      "step": 270
+    },
+    {
+      "epoch": 1.0576923076923077,
+      "grad_norm": 0.9184067249298096,
+      "learning_rate": 2.7929545003200738e-05,
+      "loss": 0.5833,
+      "step": 275
+    },
+    {
+      "epoch": 1.0769230769230769,
+      "grad_norm": 0.9467416405677795,
+      "learning_rate": 2.783178147486272e-05,
+      "loss": 0.5685,
+      "step": 280
+    },
+    {
+      "epoch": 1.0961538461538463,
+      "grad_norm": 1.153429627418518,
+      "learning_rate": 2.7731942139682732e-05,
+      "loss": 0.5922,
+      "step": 285
+    },
+    {
+      "epoch": 1.1153846153846154,
+      "grad_norm": 1.072616457939148,
+      "learning_rate": 2.763004314874435e-05,
+      "loss": 0.5749,
+      "step": 290
+    },
+    {
+      "epoch": 1.1346153846153846,
+      "grad_norm": 1.0735993385314941,
+      "learning_rate": 2.752610098632321e-05,
+      "loss": 0.5514,
+      "step": 295
+    },
+    {
+      "epoch": 1.1538461538461537,
+      "grad_norm": 1.027702808380127,
+      "learning_rate": 2.74201324672203e-05,
+      "loss": 0.5768,
+      "step": 300
+    },
+    {
+      "epoch": 1.1730769230769231,
+      "grad_norm": 1.050849199295044,
+      "learning_rate": 2.7312154734041845e-05,
+      "loss": 0.5781,
+      "step": 305
+    },
+    {
+      "epoch": 1.1923076923076923,
+      "grad_norm": 0.9099370837211609,
+      "learning_rate": 2.720218525442613e-05,
+      "loss": 0.5156,
+      "step": 310
+    },
+    {
+      "epoch": 1.2115384615384615,
+      "grad_norm": 1.1377861499786377,
+      "learning_rate": 2.7090241818217732e-05,
+      "loss": 0.4876,
+      "step": 315
+    },
+    {
+      "epoch": 1.2307692307692308,
+      "grad_norm": 0.9942741394042969,
+      "learning_rate": 2.6976342534589672e-05,
+      "loss": 0.5006,
+      "step": 320
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 1.0567532777786255,
+      "learning_rate": 2.6860505829113855e-05,
+      "loss": 0.5222,
+      "step": 325
+    },
+    {
+      "epoch": 1.2692307692307692,
+      "grad_norm": 1.1886547803878784,
+      "learning_rate": 2.6742750440780364e-05,
+      "loss": 0.5178,
+      "step": 330
+    },
+    {
+      "epoch": 1.2884615384615383,
+      "grad_norm": 1.0619785785675049,
+      "learning_rate": 2.662309541896603e-05,
+      "loss": 0.5096,
+      "step": 335
+    },
+    {
+      "epoch": 1.3076923076923077,
+      "grad_norm": 1.1066687107086182,
+      "learning_rate": 2.6501560120352805e-05,
+      "loss": 0.536,
+      "step": 340
+    },
+    {
+      "epoch": 1.3269230769230769,
+      "grad_norm": 1.0737420320510864,
+      "learning_rate": 2.637816420579644e-05,
+      "loss": 0.488,
+      "step": 345
+    },
+    {
+      "epoch": 1.3461538461538463,
+      "grad_norm": 1.301080584526062,
+      "learning_rate": 2.6252927637145885e-05,
+      "loss": 0.4923,
+      "step": 350
+    },
+    {
+      "epoch": 1.3653846153846154,
+      "grad_norm": 1.1474460363388062,
+      "learning_rate": 2.6125870674014082e-05,
+      "loss": 0.518,
+      "step": 355
+    },
+    {
+      "epoch": 1.3846153846153846,
+      "grad_norm": 1.2663733959197998,
+      "learning_rate": 2.599701387050056e-05,
+      "loss": 0.5176,
+      "step": 360
+    },
+    {
+      "epoch": 1.4038461538461537,
+      "grad_norm": 1.1481361389160156,
+      "learning_rate": 2.5866378071866338e-05,
+      "loss": 0.4349,
+      "step": 365
+    },
+    {
+      "epoch": 1.4230769230769231,
+      "grad_norm": 1.082078456878662,
+      "learning_rate": 2.5733984411161818e-05,
+      "loss": 0.4671,
+      "step": 370
+    },
+    {
+      "epoch": 1.4423076923076923,
+      "grad_norm": 1.4580280780792236,
+      "learning_rate": 2.559985430580806e-05,
+      "loss": 0.4701,
+      "step": 375
+    },
+    {
+      "epoch": 1.4615384615384617,
+      "grad_norm": 1.1313014030456543,
+      "learning_rate": 2.5464009454132056e-05,
+      "loss": 0.4608,
+      "step": 380
+    },
+    {
+      "epoch": 1.4807692307692308,
+      "grad_norm": 1.0687332153320312,
+      "learning_rate": 2.5326471831856593e-05,
+      "loss": 0.4786,
+      "step": 385
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 1.138247013092041,
+      "learning_rate": 2.5187263688545223e-05,
+      "loss": 0.4662,
+      "step": 390
+    },
+    {
+      "epoch": 1.5192307692307692,
+      "grad_norm": 1.1429054737091064,
+      "learning_rate": 2.5046407544002923e-05,
+      "loss": 0.4892,
+      "step": 395
+    },
+    {
+      "epoch": 1.5384615384615383,
+      "grad_norm": 1.2546718120574951,
+      "learning_rate": 2.490392618463305e-05,
+      "loss": 0.4601,
+      "step": 400
+    },
+    {
+      "epoch": 1.5576923076923077,
+      "grad_norm": 1.0453137159347534,
+      "learning_rate": 2.4759842659751196e-05,
+      "loss": 0.406,
+      "step": 405
+    },
+    {
+      "epoch": 1.5769230769230769,
+      "grad_norm": 1.2192074060440063,
+      "learning_rate": 2.4614180277856437e-05,
+      "loss": 0.4452,
+      "step": 410
+    },
+    {
+      "epoch": 1.5961538461538463,
+      "grad_norm": 1.0450655221939087,
+      "learning_rate": 2.446696260286073e-05,
+      "loss": 0.4179,
+      "step": 415
+    },
+    {
+      "epoch": 1.6153846153846154,
+      "grad_norm": 1.2793571949005127,
+      "learning_rate": 2.431821345027696e-05,
+      "loss": 0.4236,
+      "step": 420
+    },
+    {
+      "epoch": 1.6346153846153846,
+      "grad_norm": 1.083042860031128,
+      "learning_rate": 2.416795688336628e-05,
+      "loss": 0.4281,
+      "step": 425
+    },
+    {
+      "epoch": 1.6538461538461537,
+      "grad_norm": 1.165944218635559,
+      "learning_rate": 2.4016217209245377e-05,
+      "loss": 0.3969,
+      "step": 430
+    },
+    {
+      "epoch": 1.6730769230769231,
+      "grad_norm": 1.1456364393234253,
+      "learning_rate": 2.38630189749543e-05,
+      "loss": 0.4283,
+      "step": 435
+    },
+    {
+      "epoch": 1.6923076923076923,
+      "grad_norm": 1.3166793584823608,
+      "learning_rate": 2.3708386963485463e-05,
+      "loss": 0.4097,
+      "step": 440
+    },
+    {
+      "epoch": 1.7115384615384617,
+      "grad_norm": 1.1229557991027832,
+      "learning_rate": 2.3552346189774473e-05,
+      "loss": 0.3977,
+      "step": 445
+    },
+    {
+      "epoch": 1.7307692307692308,
+      "grad_norm": 1.085475206375122,
+      "learning_rate": 2.3394921896653458e-05,
+      "loss": 0.4038,
+      "step": 450
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 1.0684691667556763,
+      "learning_rate": 2.3236139550767527e-05,
+      "loss": 0.4019,
+      "step": 455
+    },
+    {
+      "epoch": 1.7692307692307692,
+      "grad_norm": 1.2656275033950806,
+      "learning_rate": 2.3076024838454966e-05,
+      "loss": 0.4045,
+      "step": 460
+    },
+    {
+      "epoch": 1.7884615384615383,
+      "grad_norm": 1.1343746185302734,
+      "learning_rate": 2.291460366159199e-05,
+      "loss": 0.4252,
+      "step": 465
+    },
+    {
+      "epoch": 1.8076923076923077,
+      "grad_norm": 1.22384774684906,
+      "learning_rate": 2.2751902133402556e-05,
+      "loss": 0.4102,
+      "step": 470
+    },
+    {
+      "epoch": 1.8269230769230769,
+      "grad_norm": 1.090930461883545,
+      "learning_rate": 2.2587946574234e-05,
+      "loss": 0.3758,
+      "step": 475
+    },
+    {
+      "epoch": 1.8461538461538463,
+      "grad_norm": 1.1665537357330322,
+      "learning_rate": 2.242276350729919e-05,
+      "loss": 0.4026,
+      "step": 480
+    },
+    {
+      "epoch": 1.8653846153846154,
+      "grad_norm": 0.9968311190605164,
+      "learning_rate": 2.2256379654385836e-05,
+      "loss": 0.3408,
+      "step": 485
+    },
+    {
+      "epoch": 1.8846153846153846,
+      "grad_norm": 1.222080111503601,
+      "learning_rate": 2.208882193153369e-05,
+      "loss": 0.3371,
+      "step": 490
+    },
+    {
+      "epoch": 1.9038461538461537,
+      "grad_norm": 1.0169144868850708,
+      "learning_rate": 2.1920117444680317e-05,
+      "loss": 0.3572,
+      "step": 495
+    },
+    {
+      "epoch": 1.9230769230769231,
+      "grad_norm": 1.1043157577514648,
+      "learning_rate": 2.175029348527615e-05,
+      "loss": 0.3479,
+      "step": 500
+    },
+    {
+      "epoch": 1.9423076923076923,
+      "grad_norm": 1.3724185228347778,
+      "learning_rate": 2.1579377525869532e-05,
+      "loss": 0.3546,
+      "step": 505
+    },
+    {
+      "epoch": 1.9615384615384617,
+      "grad_norm": 1.1697155237197876,
+      "learning_rate": 2.1407397215662455e-05,
+      "loss": 0.338,
+      "step": 510
+    },
+    {
+      "epoch": 1.9807692307692308,
+      "grad_norm": 1.1002439260482788,
+      "learning_rate": 2.1234380376037732e-05,
+      "loss": 0.3481,
+      "step": 515
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.3132089376449585,
+      "learning_rate": 2.1060354996058315e-05,
+      "loss": 0.3685,
+      "step": 520
+    },
+    {
+      "epoch": 2.019230769230769,
+      "grad_norm": 1.1034177541732788,
+      "learning_rate": 2.088534922793948e-05,
+      "loss": 0.2488,
+      "step": 525
+    },
+    {
+      "epoch": 2.0384615384615383,
+      "grad_norm": 1.1746792793273926,
+      "learning_rate": 2.0709391382494618e-05,
+      "loss": 0.2968,
+      "step": 530
+    },
+    {
+      "epoch": 2.0576923076923075,
+      "grad_norm": 1.2488741874694824,
+      "learning_rate": 2.0532509924555386e-05,
+      "loss": 0.2731,
+      "step": 535
+    },
+    {
+      "epoch": 2.076923076923077,
+      "grad_norm": 1.1873801946640015,
+      "learning_rate": 2.0354733468366913e-05,
+      "loss": 0.2538,
+      "step": 540
+    },
+    {
+      "epoch": 2.0961538461538463,
+      "grad_norm": 1.2156885862350464,
+      "learning_rate": 2.017609077295886e-05,
+      "loss": 0.2895,
+      "step": 545
+    },
+    {
+      "epoch": 2.1153846153846154,
+      "grad_norm": 1.0242483615875244,
+      "learning_rate": 1.9996610737493063e-05,
+      "loss": 0.2416,
+      "step": 550
+    },
+    {
+      "epoch": 2.1346153846153846,
+      "grad_norm": 1.2181206941604614,
+      "learning_rate": 1.9816322396588472e-05,
+      "loss": 0.2367,
+      "step": 555
+    },
+    {
+      "epoch": 2.1538461538461537,
+      "grad_norm": 1.0379295349121094,
+      "learning_rate": 1.963525491562421e-05,
+      "loss": 0.2528,
+      "step": 560
+    },
+    {
+      "epoch": 2.173076923076923,
+      "grad_norm": 1.2297924757003784,
+      "learning_rate": 1.945343758602148e-05,
+      "loss": 0.2294,
+      "step": 565
+    },
+    {
+      "epoch": 2.1923076923076925,
+      "grad_norm": 1.1392128467559814,
+      "learning_rate": 1.927089982050504e-05,
+      "loss": 0.2773,
+      "step": 570
+    },
+    {
+      "epoch": 2.2115384615384617,
+      "grad_norm": 1.131394386291504,
+      "learning_rate": 1.9087671148345113e-05,
+      "loss": 0.2586,
+      "step": 575
+    },
+    {
+      "epoch": 2.230769230769231,
+      "grad_norm": 1.1836904287338257,
+      "learning_rate": 1.8903781210580402e-05,
+      "loss": 0.2437,
+      "step": 580
+    },
+    {
+      "epoch": 2.25,
+      "grad_norm": 1.1704728603363037,
+      "learning_rate": 1.8719259755223036e-05,
+      "loss": 0.2709,
+      "step": 585
+    },
+    {
+      "epoch": 2.269230769230769,
+      "grad_norm": 1.042198896408081,
+      "learning_rate": 1.8534136632446218e-05,
+      "loss": 0.2602,
+      "step": 590
+    },
+    {
+      "epoch": 2.2884615384615383,
+      "grad_norm": 1.3327559232711792,
+      "learning_rate": 1.8348441789755337e-05,
+      "loss": 0.2754,
+      "step": 595
+    },
+    {
+      "epoch": 2.3076923076923075,
+      "grad_norm": 1.2408418655395508,
+      "learning_rate": 1.8162205267143333e-05,
+      "loss": 0.2541,
+      "step": 600
+    },
+    {
+      "epoch": 2.326923076923077,
+      "grad_norm": 1.1976670026779175,
+      "learning_rate": 1.7975457192231132e-05,
+      "loss": 0.244,
+      "step": 605
+    },
+    {
+      "epoch": 2.3461538461538463,
+      "grad_norm": 1.4825388193130493,
+      "learning_rate": 1.7788227775393815e-05,
+      "loss": 0.2734,
+      "step": 610
+    },
+    {
+      "epoch": 2.3653846153846154,
+      "grad_norm": 1.1539435386657715,
+      "learning_rate": 1.7600547304873552e-05,
+      "loss": 0.289,
+      "step": 615
+    },
+    {
+      "epoch": 2.3846153846153846,
+      "grad_norm": 1.5810664892196655,
+      "learning_rate": 1.7412446141879766e-05,
+      "loss": 0.236,
+      "step": 620
+    },
+    {
+      "epoch": 2.4038461538461537,
+      "grad_norm": 1.108161449432373,
+      "learning_rate": 1.722395471567763e-05,
+      "loss": 0.205,
+      "step": 625
+    },
+    {
+      "epoch": 2.423076923076923,
+      "grad_norm": 1.1394013166427612,
+      "learning_rate": 1.7035103518665488e-05,
+      "loss": 0.2297,
+      "step": 630
+    },
+    {
+      "epoch": 2.4423076923076925,
+      "grad_norm": 1.1658471822738647,
+      "learning_rate": 1.684592310144208e-05,
+      "loss": 0.2278,
+      "step": 635
+    },
+    {
+      "epoch": 2.4615384615384617,
+      "grad_norm": 1.1106714010238647,
+      "learning_rate": 1.6656444067864323e-05,
+      "loss": 0.2206,
+      "step": 640
+    },
+    {
+      "epoch": 2.480769230769231,
+      "grad_norm": 1.2031956911087036,
+      "learning_rate": 1.6466697070096555e-05,
+      "loss": 0.214,
+      "step": 645
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 1.0847914218902588,
+      "learning_rate": 1.627671280365185e-05,
+      "loss": 0.2248,
+      "step": 650
+    },
+    {
+      "epoch": 2.519230769230769,
+      "grad_norm": 1.2272170782089233,
+      "learning_rate": 1.6086522002426437e-05,
+      "loss": 0.2334,
+      "step": 655
+    },
+    {
+      "epoch": 2.5384615384615383,
+      "grad_norm": 1.1179710626602173,
+      "learning_rate": 1.5896155433727814e-05,
+      "loss": 0.241,
+      "step": 660
+    },
+    {
+      "epoch": 2.5576923076923075,
+      "grad_norm": 1.2092406749725342,
+      "learning_rate": 1.5705643893297526e-05,
+      "loss": 0.2261,
+      "step": 665
+    },
+    {
+      "epoch": 2.5769230769230766,
+      "grad_norm": 1.1971381902694702,
+      "learning_rate": 1.5515018200329295e-05,
+      "loss": 0.2216,
+      "step": 670
+    },
+    {
+      "epoch": 2.5961538461538463,
+      "grad_norm": 1.197493553161621,
+      "learning_rate": 1.532430919248338e-05,
+      "loss": 0.2313,
+      "step": 675
+    },
+    {
+      "epoch": 2.6153846153846154,
+      "grad_norm": 1.0707098245620728,
+      "learning_rate": 1.5133547720897963e-05,
+      "loss": 0.1937,
+      "step": 680
+    },
+    {
+      "epoch": 2.6346153846153846,
+      "grad_norm": 1.232611060142517,
+      "learning_rate": 1.4942764645198317e-05,
+      "loss": 0.2248,
+      "step": 685
+    },
+    {
+      "epoch": 2.6538461538461537,
+      "grad_norm": 1.374160885810852,
+      "learning_rate": 1.4751990828504623e-05,
+      "loss": 0.2354,
+      "step": 690
+    },
+    {
+      "epoch": 2.6730769230769234,
+      "grad_norm": 1.3755944967269897,
+      "learning_rate": 1.4561257132439237e-05,
+      "loss": 0.2018,
+      "step": 695
+    },
+    {
+      "epoch": 2.6923076923076925,
+      "grad_norm": 1.0376676321029663,
+      "learning_rate": 1.4370594412134168e-05,
+      "loss": 0.2061,
+      "step": 700
+    },
+    {
+      "epoch": 2.7115384615384617,
+      "grad_norm": 1.2844074964523315,
+      "learning_rate": 1.4180033511239609e-05,
+      "loss": 0.2224,
+      "step": 705
+    },
+    {
+      "epoch": 2.730769230769231,
+      "grad_norm": 1.2904430627822876,
+      "learning_rate": 1.3989605256934383e-05,
+      "loss": 0.2185,
+      "step": 710
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 1.1227009296417236,
+      "learning_rate": 1.3799340454938932e-05,
+      "loss": 0.2187,
+      "step": 715
+    },
+    {
+      "epoch": 2.769230769230769,
+      "grad_norm": 1.2107775211334229,
+      "learning_rate": 1.3609269884531935e-05,
+      "loss": 0.2177,
+      "step": 720
+    },
+    {
+      "epoch": 2.7884615384615383,
+      "grad_norm": 1.090583086013794,
+      "learning_rate": 1.3419424293571067e-05,
+      "loss": 0.1973,
+      "step": 725
+    },
+    {
+      "epoch": 2.8076923076923075,
+      "grad_norm": 1.045013189315796,
+      "learning_rate": 1.3229834393518903e-05,
+      "loss": 0.2127,
+      "step": 730
+    },
+    {
+      "epoch": 2.8269230769230766,
+      "grad_norm": 1.2943766117095947,
+      "learning_rate": 1.304053085447472e-05,
+      "loss": 0.2101,
+      "step": 735
+    },
+    {
+      "epoch": 2.8461538461538463,
+      "grad_norm": 1.050495982170105,
+      "learning_rate": 1.2851544300212965e-05,
+      "loss": 0.219,
+      "step": 740
+    },
+    {
+      "epoch": 2.8653846153846154,
+      "grad_norm": 1.23055100440979,
+      "learning_rate": 1.2662905303229207e-05,
+      "loss": 0.1913,
+      "step": 745
+    },
+    {
+      "epoch": 2.8846153846153846,
+      "grad_norm": 1.290574550628662,
+      "learning_rate": 1.2474644379794421e-05,
+      "loss": 0.1936,
+      "step": 750
+    },
+    {
+      "epoch": 2.9038461538461537,
+      "grad_norm": 0.9990605115890503,
+      "learning_rate": 1.2286791985018356e-05,
+      "loss": 0.2027,
+      "step": 755
+    },
+    {
+      "epoch": 2.9230769230769234,
+      "grad_norm": 1.294940710067749,
+      "learning_rate": 1.209937850792275e-05,
+      "loss": 0.2048,
+      "step": 760
+    },
+    {
+      "epoch": 2.9423076923076925,
+      "grad_norm": 1.292446255683899,
+      "learning_rate": 1.191243426652532e-05,
+      "loss": 0.1737,
+      "step": 765
+    },
+    {
+      "epoch": 2.9615384615384617,
+      "grad_norm": 1.1711947917938232,
+      "learning_rate": 1.1725989502935175e-05,
+      "loss": 0.1758,
+      "step": 770
+    },
+    {
+      "epoch": 2.980769230769231,
+      "grad_norm": 1.2699744701385498,
+      "learning_rate": 1.1540074378460516e-05,
+      "loss": 0.1859,
+      "step": 775
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.1081920862197876,
+      "learning_rate": 1.135471896872944e-05,
+      "loss": 0.1901,
+      "step": 780
+    },
+    {
+      "epoch": 3.019230769230769,
+      "grad_norm": 1.0495061874389648,
+      "learning_rate": 1.1169953258824567e-05,
+      "loss": 0.1403,
+      "step": 785
+    },
+    {
+      "epoch": 3.0384615384615383,
+      "grad_norm": 1.2737641334533691,
+      "learning_rate": 1.0985807138432336e-05,
+      "loss": 0.1351,
+      "step": 790
+    },
+    {
+      "epoch": 3.0576923076923075,
+      "grad_norm": 1.1091535091400146,
+      "learning_rate": 1.080231039700774e-05,
+      "loss": 0.136,
+      "step": 795
+    },
+    {
+      "epoch": 3.076923076923077,
+      "grad_norm": 1.0284589529037476,
+      "learning_rate": 1.0619492718955242e-05,
+      "loss": 0.1283,
+      "step": 800
+    },
+    {
+      "epoch": 3.0961538461538463,
+      "grad_norm": 1.054133653640747,
+      "learning_rate": 1.0437383678826723e-05,
+      "loss": 0.1324,
+      "step": 805
+    },
+    {
+      "epoch": 3.1153846153846154,
+      "grad_norm": 1.0665591955184937,
+      "learning_rate": 1.0256012736537172e-05,
+      "loss": 0.1448,
+      "step": 810
+    },
+    {
+      "epoch": 3.1346153846153846,
+      "grad_norm": 1.1725926399230957,
+      "learning_rate": 1.007540923259896e-05,
+      "loss": 0.1448,
+      "step": 815
+    },
+    {
+      "epoch": 3.1538461538461537,
+      "grad_norm": 0.9875220060348511,
+      "learning_rate": 9.895602383375355e-06,
+      "loss": 0.1361,
+      "step": 820
+    },
+    {
+      "epoch": 3.173076923076923,
+      "grad_norm": 1.078674077987671,
+      "learning_rate": 9.716621276354225e-06,
+      "loss": 0.1348,
+      "step": 825
+    },
+    {
+      "epoch": 3.1923076923076925,
+      "grad_norm": 1.1486966609954834,
+      "learning_rate": 9.5384948654425e-06,
+      "loss": 0.1522,
+      "step": 830
+    },
+    {
+      "epoch": 3.2115384615384617,
+      "grad_norm": 1.0026553869247437,
+      "learning_rate": 9.361251966282276e-06,
+      "loss": 0.1311,
+      "step": 835
+    },
+    {
+      "epoch": 3.230769230769231,
+      "grad_norm": 1.2294085025787354,
+      "learning_rate": 9.184921251589299e-06,
+      "loss": 0.1277,
+      "step": 840
+    },
+    {
+      "epoch": 3.25,
+      "grad_norm": 0.8330014944076538,
+      "learning_rate": 9.009531246514569e-06,
+      "loss": 0.1327,
+      "step": 845
+    },
+    {
+      "epoch": 3.269230769230769,
+      "grad_norm": 0.8544556498527527,
+      "learning_rate": 8.835110324029762e-06,
+      "loss": 0.1288,
+      "step": 850
+    },
+    {
+      "epoch": 3.2884615384615383,
+      "grad_norm": 1.1136083602905273,
+      "learning_rate": 8.661686700337368e-06,
+      "loss": 0.1297,
+      "step": 855
+    },
+    {
+      "epoch": 3.3076923076923075,
+      "grad_norm": 1.009947657585144,
+      "learning_rate": 8.489288430306097e-06,
+      "loss": 0.1289,
+      "step": 860
+    },
+    {
+      "epoch": 3.326923076923077,
+      "grad_norm": 1.0761442184448242,
+      "learning_rate": 8.317943402932434e-06,
+      "loss": 0.1287,
+      "step": 865
+    },
+    {
+      "epoch": 3.3461538461538463,
+      "grad_norm": 1.067358374595642,
+      "learning_rate": 8.147679336829019e-06,
+      "loss": 0.1185,
+      "step": 870
+    },
+    {
+      "epoch": 3.3653846153846154,
+      "grad_norm": 0.9998955726623535,
+      "learning_rate": 7.97852377574058e-06,
+      "loss": 0.1223,
+      "step": 875
+    },
+    {
+      "epoch": 3.3846153846153846,
+      "grad_norm": 1.0905925035476685,
+      "learning_rate": 7.810504084088168e-06,
+      "loss": 0.122,
+      "step": 880
+    },
+    {
+      "epoch": 3.4038461538461537,
+      "grad_norm": 0.9051513671875,
+      "learning_rate": 7.643647442542383e-06,
+      "loss": 0.1272,
+      "step": 885
+    },
+    {
+      "epoch": 3.423076923076923,
+      "grad_norm": 1.091073751449585,
+      "learning_rate": 7.477980843626337e-06,
+      "loss": 0.1242,
+      "step": 890
+    },
+    {
+      "epoch": 3.4423076923076925,
+      "grad_norm": 1.00954270362854,
+      "learning_rate": 7.313531087349085e-06,
+      "loss": 0.118,
+      "step": 895
+    },
+    {
+      "epoch": 3.4615384615384617,
+      "grad_norm": 1.1877714395523071,
+      "learning_rate": 7.150324776870125e-06,
+      "loss": 0.1127,
+      "step": 900
+    },
+    {
+      "epoch": 3.480769230769231,
+      "grad_norm": 1.0862516164779663,
+      "learning_rate": 6.988388314195793e-06,
+      "loss": 0.1221,
+      "step": 905
+    },
+    {
+      "epoch": 3.5,
+      "grad_norm": 1.2270557880401611,
+      "learning_rate": 6.827747895908212e-06,
+      "loss": 0.1194,
+      "step": 910
+    },
+    {
+      "epoch": 3.519230769230769,
+      "grad_norm": 1.0056644678115845,
+      "learning_rate": 6.6684295089274565e-06,
+      "loss": 0.1217,
+      "step": 915
+    },
+    {
+      "epoch": 3.5384615384615383,
+      "grad_norm": 1.038124918937683,
+      "learning_rate": 6.510458926307577e-06,
+      "loss": 0.1194,
+      "step": 920
+    },
+    {
+      "epoch": 3.5576923076923075,
+      "grad_norm": 0.970857560634613,
+      "learning_rate": 6.353861703067332e-06,
+      "loss": 0.1127,
+      "step": 925
+    },
+    {
+      "epoch": 3.5769230769230766,
+      "grad_norm": 1.0015406608581543,
+      "learning_rate": 6.198663172056109e-06,
+      "loss": 0.1263,
+      "step": 930
+    },
+    {
+      "epoch": 3.5961538461538463,
+      "grad_norm": 0.9921681880950928,
+      "learning_rate": 6.044888439855806e-06,
+      "loss": 0.1341,
+      "step": 935
+    },
+    {
+      "epoch": 3.6153846153846154,
+      "grad_norm": 1.0572803020477295,
+      "learning_rate": 5.892562382719333e-06,
+      "loss": 0.1299,
+      "step": 940
+    },
+    {
+      "epoch": 3.6346153846153846,
+      "grad_norm": 1.028430700302124,
+      "learning_rate": 5.741709642546352e-06,
+      "loss": 0.1112,
+      "step": 945
+    },
+    {
+      "epoch": 3.6538461538461537,
+      "grad_norm": 1.1680800914764404,
+      "learning_rate": 5.592354622896944e-06,
+      "loss": 0.1118,
+      "step": 950
+    },
+    {
+      "epoch": 3.6730769230769234,
+      "grad_norm": 0.914523184299469,
+      "learning_rate": 5.444521485043831e-06,
+      "loss": 0.1073,
+      "step": 955
+    },
+    {
+      "epoch": 3.6923076923076925,
+      "grad_norm": 0.8823187351226807,
+      "learning_rate": 5.2982341440637804e-06,
+      "loss": 0.1217,
+      "step": 960
+    },
+    {
+      "epoch": 3.7115384615384617,
+      "grad_norm": 1.0132150650024414,
+      "learning_rate": 5.153516264968841e-06,
+      "loss": 0.1198,
+      "step": 965
+    },
+    {
+      "epoch": 3.730769230769231,
+      "grad_norm": 1.0348647832870483,
+      "learning_rate": 5.010391258878049e-06,
+      "loss": 0.1189,
+      "step": 970
+    },
+    {
+      "epoch": 3.75,
+      "grad_norm": 1.1547722816467285,
+      "learning_rate": 4.868882279230208e-06,
+      "loss": 0.1255,
+      "step": 975
+    },
+    {
+      "epoch": 3.769230769230769,
+      "grad_norm": 0.9065052270889282,
+      "learning_rate": 4.729012218038287e-06,
+      "loss": 0.1161,
+      "step": 980
+    },
+    {
+      "epoch": 3.7884615384615383,
+      "grad_norm": 0.8526612520217896,
+      "learning_rate": 4.590803702186219e-06,
+      "loss": 0.0981,
+      "step": 985
+    },
+    {
+      "epoch": 3.8076923076923075,
+      "grad_norm": 0.9093031883239746,
+      "learning_rate": 4.454279089768503e-06,
+      "loss": 0.1082,
+      "step": 990
+    },
+    {
+      "epoch": 3.8269230769230766,
+      "grad_norm": 0.9617008566856384,
+      "learning_rate": 4.319460466473328e-06,
+      "loss": 0.1013,
+      "step": 995
+    },
+    {
+      "epoch": 3.8461538461538463,
+      "grad_norm": 1.0160218477249146,
+      "learning_rate": 4.186369642009753e-06,
+      "loss": 0.116,
+      "step": 1000
+    },
+    {
+      "epoch": 3.8653846153846154,
+      "grad_norm": 0.9817861318588257,
+      "learning_rate": 4.055028146579532e-06,
+      "loss": 0.1139,
+      "step": 1005
+    },
+    {
+      "epoch": 3.8846153846153846,
+      "grad_norm": 0.9847229719161987,
+      "learning_rate": 3.925457227394158e-06,
+      "loss": 0.1148,
+      "step": 1010
+    },
+    {
+      "epoch": 3.9038461538461537,
+      "grad_norm": 1.0290164947509766,
+      "learning_rate": 3.7976778452376966e-06,
+      "loss": 0.1196,
+      "step": 1015
+    },
+    {
+      "epoch": 3.9230769230769234,
+      "grad_norm": 0.7677153944969177,
+      "learning_rate": 3.671710671075913e-06,
+      "loss": 0.1116,
+      "step": 1020
+    },
+    {
+      "epoch": 3.9423076923076925,
+      "grad_norm": 0.9820197820663452,
+      "learning_rate": 3.547576082712351e-06,
+      "loss": 0.0969,
+      "step": 1025
+    },
+    {
+      "epoch": 3.9615384615384617,
+      "grad_norm": 0.8637183904647827,
+      "learning_rate": 3.425294161491796e-06,
+      "loss": 0.1065,
+      "step": 1030
+    },
+    {
+      "epoch": 3.980769230769231,
+      "grad_norm": 0.9800280928611755,
+      "learning_rate": 3.3048846890516658e-06,
+      "loss": 0.1113,
+      "step": 1035
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 0.785662055015564,
+      "learning_rate": 3.1863671441219634e-06,
+      "loss": 0.1047,
+      "step": 1040
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1300,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5446139105926185e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

14_128_e5_3e-5/checkpoint-1040/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dbfb4ffc8d0db1e939d7417e2f349ee624f0febeeb1e1b8c00f77c5c3770afcd
+size 7736

14_128_e5_3e-5/checkpoint-1040/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/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, exclude_frozen_parameters):
+    """
+    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,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+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 _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+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() if _has_callable(shape, 'numel') else math.prod(shape)
+            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,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _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,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _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, exclude_frozen_parameters=False):
+    """
+    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``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    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, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    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``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    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("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    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,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

14_128_e5_3e-5/checkpoint-1300/README.md

@@ -0,0 +1,202 @@
+---
+base_model: ibm-granite/granite-3.3-8b-base
+library_name: peft
+---
+
+# 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]
+### Framework versions
+
+- PEFT 0.15.2

14_128_e5_3e-5/checkpoint-1300/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "v_proj",
+    "q_proj",
+    "k_proj",
+    "up_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

14_128_e5_3e-5/checkpoint-1300/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30d4e6336ecc4d2f4803cfd7063bba778ad8fb3c39ec9cc9c7dfb1c8fa89df75
+size 791751704

14_128_e5_3e-5/checkpoint-1300/latest

@@ -0,0 +1 @@
+global_step1300

14_128_e5_3e-5/checkpoint-1300/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9ba8067c69d1e3725c1df5ff37f7e09053d1abf670d40b273a42405a8c5b136
+size 15920

14_128_e5_3e-5/checkpoint-1300/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7bab0f4e69afde86f1bd68159afe956d0c9be7638eed01cc289088a439db1c3
+size 15920

14_128_e5_3e-5/checkpoint-1300/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ceb060b8b3bbf10b2b7e678eb3c9d5b1f50f64decdd2d47fbf5ce8fb12257909
+size 15920

14_128_e5_3e-5/checkpoint-1300/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0de07b98eba75a4d9256a767e7b3a4546bcab465c0a673e0c66b5f2fb79e6511
+size 15920

14_128_e5_3e-5/checkpoint-1300/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a43bdfc307b2248d7f40087b7e9a208410640e3f64b4a23433607e23a55b5dd9
+size 15920

14_128_e5_3e-5/checkpoint-1300/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcbe5ea887513bd18875286d24294774d1403020d6d996050886b082d9f2dec2
+size 15920

14_128_e5_3e-5/checkpoint-1300/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcdb1f8666c840daadd9553e2aec1b82e49b2ef1a289f2a5af73ae020a21692b
+size 15920

14_128_e5_3e-5/checkpoint-1300/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80c9d52cd878711e831caa7afd8652a03b5ef8f4507d0b6a8116b0c5c0c9f233
+size 15920

14_128_e5_3e-5/checkpoint-1300/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9625068f8cef5fb1ebcdb56b8b470048aa364ea1d0262d6af12d81070293e84
+size 1064

14_128_e5_3e-5/checkpoint-1300/special_tokens_map.json

@@ -0,0 +1,45 @@
+{
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<reponame>",
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

14_128_e5_3e-5/checkpoint-1300/tokenizer_config.json

@@ -0,0 +1,188 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<fim_prefix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<fim_middle>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<fim_suffix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<fim_pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<filename>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<gh_stars>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<issue_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "8": {
+      "content": "<issue_comment>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "9": {
+      "content": "<issue_closed>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "10": {
+      "content": "<jupyter_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "11": {
+      "content": "<jupyter_text>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "12": {
+      "content": "<jupyter_code>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "13": {
+      "content": "<jupyter_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "14": {
+      "content": "<empty_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "15": {
+      "content": "<commit_before>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "16": {
+      "content": "<commit_msg>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "17": {
+      "content": "<commit_after>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "18": {
+      "content": "<reponame>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<reponame>",
+  "padding_side": "left",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

14_128_e5_3e-5/checkpoint-1300/trainer_state.json

@@ -0,0 +1,1854 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.0,
+  "eval_steps": 500,
+  "global_step": 1300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.019230769230769232,
+      "grad_norm": 1.4232831001281738,
+      "learning_rate": 1.8461538461538462e-06,
+      "loss": 1.3571,
+      "step": 5
+    },
+    {
+      "epoch": 0.038461538461538464,
+      "grad_norm": 1.0114805698394775,
+      "learning_rate": 4.1538461538461545e-06,
+      "loss": 1.3519,
+      "step": 10
+    },
+    {
+      "epoch": 0.057692307692307696,
+      "grad_norm": 0.7543752789497375,
+      "learning_rate": 6.461538461538462e-06,
+      "loss": 1.2802,
+      "step": 15
+    },
+    {
+      "epoch": 0.07692307692307693,
+      "grad_norm": 0.5682826042175293,
+      "learning_rate": 8.76923076923077e-06,
+      "loss": 1.2099,
+      "step": 20
+    },
+    {
+      "epoch": 0.09615384615384616,
+      "grad_norm": 0.5222262144088745,
+      "learning_rate": 1.1076923076923077e-05,
+      "loss": 1.2572,
+      "step": 25
+    },
+    {
+      "epoch": 0.11538461538461539,
+      "grad_norm": 0.547173261642456,
+      "learning_rate": 1.3384615384615386e-05,
+      "loss": 1.2272,
+      "step": 30
+    },
+    {
+      "epoch": 0.1346153846153846,
+      "grad_norm": 0.4619555175304413,
+      "learning_rate": 1.5692307692307693e-05,
+      "loss": 1.1852,
+      "step": 35
+    },
+    {
+      "epoch": 0.15384615384615385,
+      "grad_norm": 0.5007979273796082,
+      "learning_rate": 1.8e-05,
+      "loss": 1.2387,
+      "step": 40
+    },
+    {
+      "epoch": 0.17307692307692307,
+      "grad_norm": 0.4946490526199341,
+      "learning_rate": 2.0307692307692308e-05,
+      "loss": 1.1931,
+      "step": 45
+    },
+    {
+      "epoch": 0.19230769230769232,
+      "grad_norm": 0.5487769842147827,
+      "learning_rate": 2.2615384615384615e-05,
+      "loss": 1.2114,
+      "step": 50
+    },
+    {
+      "epoch": 0.21153846153846154,
+      "grad_norm": 0.46380364894866943,
+      "learning_rate": 2.4923076923076926e-05,
+      "loss": 1.1068,
+      "step": 55
+    },
+    {
+      "epoch": 0.23076923076923078,
+      "grad_norm": 0.5553277134895325,
+      "learning_rate": 2.7230769230769233e-05,
+      "loss": 1.1258,
+      "step": 60
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.5975745320320129,
+      "learning_rate": 2.953846153846154e-05,
+      "loss": 1.1133,
+      "step": 65
+    },
+    {
+      "epoch": 0.2692307692307692,
+      "grad_norm": 0.4654400646686554,
+      "learning_rate": 2.9999223496658065e-05,
+      "loss": 1.0798,
+      "step": 70
+    },
+    {
+      "epoch": 0.28846153846153844,
+      "grad_norm": 0.5261916518211365,
+      "learning_rate": 2.9996069089616144e-05,
+      "loss": 1.105,
+      "step": 75
+    },
+    {
+      "epoch": 0.3076923076923077,
+      "grad_norm": 0.5102323293685913,
+      "learning_rate": 2.9990488757313582e-05,
+      "loss": 1.1385,
+      "step": 80
+    },
+    {
+      "epoch": 0.3269230769230769,
+      "grad_norm": 0.49091285467147827,
+      "learning_rate": 2.99824834024849e-05,
+      "loss": 1.0717,
+      "step": 85
+    },
+    {
+      "epoch": 0.34615384615384615,
+      "grad_norm": 0.5449517369270325,
+      "learning_rate": 2.9972054320162297e-05,
+      "loss": 1.0408,
+      "step": 90
+    },
+    {
+      "epoch": 0.36538461538461536,
+      "grad_norm": 0.5686984062194824,
+      "learning_rate": 2.9959203197466193e-05,
+      "loss": 1.0707,
+      "step": 95
+    },
+    {
+      "epoch": 0.38461538461538464,
+      "grad_norm": 0.6011113524436951,
+      "learning_rate": 2.9943932113332267e-05,
+      "loss": 1.0017,
+      "step": 100
+    },
+    {
+      "epoch": 0.40384615384615385,
+      "grad_norm": 0.5413767695426941,
+      "learning_rate": 2.9926243538175172e-05,
+      "loss": 1.017,
+      "step": 105
+    },
+    {
+      "epoch": 0.4230769230769231,
+      "grad_norm": 0.5385787487030029,
+      "learning_rate": 2.990614033348888e-05,
+      "loss": 1.0517,
+      "step": 110
+    },
+    {
+      "epoch": 0.4423076923076923,
+      "grad_norm": 0.7817659974098206,
+      "learning_rate": 2.9883625751383776e-05,
+      "loss": 0.9498,
+      "step": 115
+    },
+    {
+      "epoch": 0.46153846153846156,
+      "grad_norm": 0.636008620262146,
+      "learning_rate": 2.985870343406057e-05,
+      "loss": 0.9977,
+      "step": 120
+    },
+    {
+      "epoch": 0.4807692307692308,
+      "grad_norm": 0.6735332608222961,
+      "learning_rate": 2.983137741322109e-05,
+      "loss": 0.951,
+      "step": 125
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.7509468197822571,
+      "learning_rate": 2.980165210941607e-05,
+      "loss": 0.9125,
+      "step": 130
+    },
+    {
+      "epoch": 0.5192307692307693,
+      "grad_norm": 0.7218055725097656,
+      "learning_rate": 2.9769532331330037e-05,
+      "loss": 0.9313,
+      "step": 135
+    },
+    {
+      "epoch": 0.5384615384615384,
+      "grad_norm": 0.7202168703079224,
+      "learning_rate": 2.97350232750034e-05,
+      "loss": 0.8895,
+      "step": 140
+    },
+    {
+      "epoch": 0.5576923076923077,
+      "grad_norm": 0.7541701197624207,
+      "learning_rate": 2.96981305229919e-05,
+      "loss": 0.9392,
+      "step": 145
+    },
+    {
+      "epoch": 0.5769230769230769,
+      "grad_norm": 0.7201166152954102,
+      "learning_rate": 2.9658860043463485e-05,
+      "loss": 0.8911,
+      "step": 150
+    },
+    {
+      "epoch": 0.5961538461538461,
+      "grad_norm": 0.8302095532417297,
+      "learning_rate": 2.9617218189232867e-05,
+      "loss": 0.9174,
+      "step": 155
+    },
+    {
+      "epoch": 0.6153846153846154,
+      "grad_norm": 0.762272298336029,
+      "learning_rate": 2.9573211696733812e-05,
+      "loss": 0.8463,
+      "step": 160
+    },
+    {
+      "epoch": 0.6346153846153846,
+      "grad_norm": 0.8819282054901123,
+      "learning_rate": 2.9526847684929363e-05,
+      "loss": 0.8511,
+      "step": 165
+    },
+    {
+      "epoch": 0.6538461538461539,
+      "grad_norm": 0.8169696927070618,
+      "learning_rate": 2.947813365416023e-05,
+      "loss": 0.8837,
+      "step": 170
+    },
+    {
+      "epoch": 0.6730769230769231,
+      "grad_norm": 0.7911653518676758,
+      "learning_rate": 2.9427077484931443e-05,
+      "loss": 0.8293,
+      "step": 175
+    },
+    {
+      "epoch": 0.6923076923076923,
+      "grad_norm": 0.7542786002159119,
+      "learning_rate": 2.9373687436637492e-05,
+      "loss": 0.8309,
+      "step": 180
+    },
+    {
+      "epoch": 0.7115384615384616,
+      "grad_norm": 0.8718588948249817,
+      "learning_rate": 2.9317972146226255e-05,
+      "loss": 0.8319,
+      "step": 185
+    },
+    {
+      "epoch": 0.7307692307692307,
+      "grad_norm": 0.8762234449386597,
+      "learning_rate": 2.9259940626801725e-05,
+      "loss": 0.8576,
+      "step": 190
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.7861353158950806,
+      "learning_rate": 2.919960226616601e-05,
+      "loss": 0.8158,
+      "step": 195
+    },
+    {
+      "epoch": 0.7692307692307693,
+      "grad_norm": 1.0044094324111938,
+      "learning_rate": 2.9136966825300622e-05,
+      "loss": 0.8114,
+      "step": 200
+    },
+    {
+      "epoch": 0.7884615384615384,
+      "grad_norm": 0.8322351574897766,
+      "learning_rate": 2.907204443678747e-05,
+      "loss": 0.7426,
+      "step": 205
+    },
+    {
+      "epoch": 0.8076923076923077,
+      "grad_norm": 0.9439869523048401,
+      "learning_rate": 2.900484560316967e-05,
+      "loss": 0.7483,
+      "step": 210
+    },
+    {
+      "epoch": 0.8269230769230769,
+      "grad_norm": 1.0131150484085083,
+      "learning_rate": 2.8935381195252562e-05,
+      "loss": 0.7888,
+      "step": 215
+    },
+    {
+      "epoch": 0.8461538461538461,
+      "grad_norm": 0.8887054324150085,
+      "learning_rate": 2.8863662450345126e-05,
+      "loss": 0.7283,
+      "step": 220
+    },
+    {
+      "epoch": 0.8653846153846154,
+      "grad_norm": 0.987267255783081,
+      "learning_rate": 2.8789700970442124e-05,
+      "loss": 0.7665,
+      "step": 225
+    },
+    {
+      "epoch": 0.8846153846153846,
+      "grad_norm": 0.8717510104179382,
+      "learning_rate": 2.8713508720347224e-05,
+      "loss": 0.7308,
+      "step": 230
+    },
+    {
+      "epoch": 0.9038461538461539,
+      "grad_norm": 0.881127655506134,
+      "learning_rate": 2.863509802573744e-05,
+      "loss": 0.7231,
+      "step": 235
+    },
+    {
+      "epoch": 0.9230769230769231,
+      "grad_norm": 0.9518759250640869,
+      "learning_rate": 2.8554481571169208e-05,
+      "loss": 0.6872,
+      "step": 240
+    },
+    {
+      "epoch": 0.9423076923076923,
+      "grad_norm": 0.8832864761352539,
+      "learning_rate": 2.8471672398026414e-05,
+      "loss": 0.7028,
+      "step": 245
+    },
+    {
+      "epoch": 0.9615384615384616,
+      "grad_norm": 0.9555291533470154,
+      "learning_rate": 2.838668390241063e-05,
+      "loss": 0.6354,
+      "step": 250
+    },
+    {
+      "epoch": 0.9807692307692307,
+      "grad_norm": 1.1762830018997192,
+      "learning_rate": 2.8299529832974076e-05,
+      "loss": 0.6844,
+      "step": 255
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.9526051878929138,
+      "learning_rate": 2.8210224288695437e-05,
+      "loss": 0.6363,
+      "step": 260
+    },
+    {
+      "epoch": 1.0192307692307692,
+      "grad_norm": 1.0443257093429565,
+      "learning_rate": 2.81187817165991e-05,
+      "loss": 0.6202,
+      "step": 265
+    },
+    {
+      "epoch": 1.0384615384615385,
+      "grad_norm": 1.000659465789795,
+      "learning_rate": 2.8025216909418033e-05,
+      "loss": 0.5572,
+      "step": 270
+    },
+    {
+      "epoch": 1.0576923076923077,
+      "grad_norm": 0.9184067249298096,
+      "learning_rate": 2.7929545003200738e-05,
+      "loss": 0.5833,
+      "step": 275
+    },
+    {
+      "epoch": 1.0769230769230769,
+      "grad_norm": 0.9467416405677795,
+      "learning_rate": 2.783178147486272e-05,
+      "loss": 0.5685,
+      "step": 280
+    },
+    {
+      "epoch": 1.0961538461538463,
+      "grad_norm": 1.153429627418518,
+      "learning_rate": 2.7731942139682732e-05,
+      "loss": 0.5922,
+      "step": 285
+    },
+    {
+      "epoch": 1.1153846153846154,
+      "grad_norm": 1.072616457939148,
+      "learning_rate": 2.763004314874435e-05,
+      "loss": 0.5749,
+      "step": 290
+    },
+    {
+      "epoch": 1.1346153846153846,
+      "grad_norm": 1.0735993385314941,
+      "learning_rate": 2.752610098632321e-05,
+      "loss": 0.5514,
+      "step": 295
+    },
+    {
+      "epoch": 1.1538461538461537,
+      "grad_norm": 1.027702808380127,
+      "learning_rate": 2.74201324672203e-05,
+      "loss": 0.5768,
+      "step": 300
+    },
+    {
+      "epoch": 1.1730769230769231,
+      "grad_norm": 1.050849199295044,
+      "learning_rate": 2.7312154734041845e-05,
+      "loss": 0.5781,
+      "step": 305
+    },
+    {
+      "epoch": 1.1923076923076923,
+      "grad_norm": 0.9099370837211609,
+      "learning_rate": 2.720218525442613e-05,
+      "loss": 0.5156,
+      "step": 310
+    },
+    {
+      "epoch": 1.2115384615384615,
+      "grad_norm": 1.1377861499786377,
+      "learning_rate": 2.7090241818217732e-05,
+      "loss": 0.4876,
+      "step": 315
+    },
+    {
+      "epoch": 1.2307692307692308,
+      "grad_norm": 0.9942741394042969,
+      "learning_rate": 2.6976342534589672e-05,
+      "loss": 0.5006,
+      "step": 320
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 1.0567532777786255,
+      "learning_rate": 2.6860505829113855e-05,
+      "loss": 0.5222,
+      "step": 325
+    },
+    {
+      "epoch": 1.2692307692307692,
+      "grad_norm": 1.1886547803878784,
+      "learning_rate": 2.6742750440780364e-05,
+      "loss": 0.5178,
+      "step": 330
+    },
+    {
+      "epoch": 1.2884615384615383,
+      "grad_norm": 1.0619785785675049,
+      "learning_rate": 2.662309541896603e-05,
+      "loss": 0.5096,
+      "step": 335
+    },
+    {
+      "epoch": 1.3076923076923077,
+      "grad_norm": 1.1066687107086182,
+      "learning_rate": 2.6501560120352805e-05,
+      "loss": 0.536,
+      "step": 340
+    },
+    {
+      "epoch": 1.3269230769230769,
+      "grad_norm": 1.0737420320510864,
+      "learning_rate": 2.637816420579644e-05,
+      "loss": 0.488,
+      "step": 345
+    },
+    {
+      "epoch": 1.3461538461538463,
+      "grad_norm": 1.301080584526062,
+      "learning_rate": 2.6252927637145885e-05,
+      "loss": 0.4923,
+      "step": 350
+    },
+    {
+      "epoch": 1.3653846153846154,
+      "grad_norm": 1.1474460363388062,
+      "learning_rate": 2.6125870674014082e-05,
+      "loss": 0.518,
+      "step": 355
+    },
+    {
+      "epoch": 1.3846153846153846,
+      "grad_norm": 1.2663733959197998,
+      "learning_rate": 2.599701387050056e-05,
+      "loss": 0.5176,
+      "step": 360
+    },
+    {
+      "epoch": 1.4038461538461537,
+      "grad_norm": 1.1481361389160156,
+      "learning_rate": 2.5866378071866338e-05,
+      "loss": 0.4349,
+      "step": 365
+    },
+    {
+      "epoch": 1.4230769230769231,
+      "grad_norm": 1.082078456878662,
+      "learning_rate": 2.5733984411161818e-05,
+      "loss": 0.4671,
+      "step": 370
+    },
+    {
+      "epoch": 1.4423076923076923,
+      "grad_norm": 1.4580280780792236,
+      "learning_rate": 2.559985430580806e-05,
+      "loss": 0.4701,
+      "step": 375
+    },
+    {
+      "epoch": 1.4615384615384617,
+      "grad_norm": 1.1313014030456543,
+      "learning_rate": 2.5464009454132056e-05,
+      "loss": 0.4608,
+      "step": 380
+    },
+    {
+      "epoch": 1.4807692307692308,
+      "grad_norm": 1.0687332153320312,
+      "learning_rate": 2.5326471831856593e-05,
+      "loss": 0.4786,
+      "step": 385
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 1.138247013092041,
+      "learning_rate": 2.5187263688545223e-05,
+      "loss": 0.4662,
+      "step": 390
+    },
+    {
+      "epoch": 1.5192307692307692,
+      "grad_norm": 1.1429054737091064,
+      "learning_rate": 2.5046407544002923e-05,
+      "loss": 0.4892,
+      "step": 395
+    },
+    {
+      "epoch": 1.5384615384615383,
+      "grad_norm": 1.2546718120574951,
+      "learning_rate": 2.490392618463305e-05,
+      "loss": 0.4601,
+      "step": 400
+    },
+    {
+      "epoch": 1.5576923076923077,
+      "grad_norm": 1.0453137159347534,
+      "learning_rate": 2.4759842659751196e-05,
+      "loss": 0.406,
+      "step": 405
+    },
+    {
+      "epoch": 1.5769230769230769,
+      "grad_norm": 1.2192074060440063,
+      "learning_rate": 2.4614180277856437e-05,
+      "loss": 0.4452,
+      "step": 410
+    },
+    {
+      "epoch": 1.5961538461538463,
+      "grad_norm": 1.0450655221939087,
+      "learning_rate": 2.446696260286073e-05,
+      "loss": 0.4179,
+      "step": 415
+    },
+    {
+      "epoch": 1.6153846153846154,
+      "grad_norm": 1.2793571949005127,
+      "learning_rate": 2.431821345027696e-05,
+      "loss": 0.4236,
+      "step": 420
+    },
+    {
+      "epoch": 1.6346153846153846,
+      "grad_norm": 1.083042860031128,
+      "learning_rate": 2.416795688336628e-05,
+      "loss": 0.4281,
+      "step": 425
+    },
+    {
+      "epoch": 1.6538461538461537,
+      "grad_norm": 1.165944218635559,
+      "learning_rate": 2.4016217209245377e-05,
+      "loss": 0.3969,
+      "step": 430
+    },
+    {
+      "epoch": 1.6730769230769231,
+      "grad_norm": 1.1456364393234253,
+      "learning_rate": 2.38630189749543e-05,
+      "loss": 0.4283,
+      "step": 435
+    },
+    {
+      "epoch": 1.6923076923076923,
+      "grad_norm": 1.3166793584823608,
+      "learning_rate": 2.3708386963485463e-05,
+      "loss": 0.4097,
+      "step": 440
+    },
+    {
+      "epoch": 1.7115384615384617,
+      "grad_norm": 1.1229557991027832,
+      "learning_rate": 2.3552346189774473e-05,
+      "loss": 0.3977,
+      "step": 445
+    },
+    {
+      "epoch": 1.7307692307692308,
+      "grad_norm": 1.085475206375122,
+      "learning_rate": 2.3394921896653458e-05,
+      "loss": 0.4038,
+      "step": 450
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 1.0684691667556763,
+      "learning_rate": 2.3236139550767527e-05,
+      "loss": 0.4019,
+      "step": 455
+    },
+    {
+      "epoch": 1.7692307692307692,
+      "grad_norm": 1.2656275033950806,
+      "learning_rate": 2.3076024838454966e-05,
+      "loss": 0.4045,
+      "step": 460
+    },
+    {
+      "epoch": 1.7884615384615383,
+      "grad_norm": 1.1343746185302734,
+      "learning_rate": 2.291460366159199e-05,
+      "loss": 0.4252,
+      "step": 465
+    },
+    {
+      "epoch": 1.8076923076923077,
+      "grad_norm": 1.22384774684906,
+      "learning_rate": 2.2751902133402556e-05,
+      "loss": 0.4102,
+      "step": 470
+    },
+    {
+      "epoch": 1.8269230769230769,
+      "grad_norm": 1.090930461883545,
+      "learning_rate": 2.2587946574234e-05,
+      "loss": 0.3758,
+      "step": 475
+    },
+    {
+      "epoch": 1.8461538461538463,
+      "grad_norm": 1.1665537357330322,
+      "learning_rate": 2.242276350729919e-05,
+      "loss": 0.4026,
+      "step": 480
+    },
+    {
+      "epoch": 1.8653846153846154,
+      "grad_norm": 0.9968311190605164,
+      "learning_rate": 2.2256379654385836e-05,
+      "loss": 0.3408,
+      "step": 485
+    },
+    {
+      "epoch": 1.8846153846153846,
+      "grad_norm": 1.222080111503601,
+      "learning_rate": 2.208882193153369e-05,
+      "loss": 0.3371,
+      "step": 490
+    },
+    {
+      "epoch": 1.9038461538461537,
+      "grad_norm": 1.0169144868850708,
+      "learning_rate": 2.1920117444680317e-05,
+      "loss": 0.3572,
+      "step": 495
+    },
+    {
+      "epoch": 1.9230769230769231,
+      "grad_norm": 1.1043157577514648,
+      "learning_rate": 2.175029348527615e-05,
+      "loss": 0.3479,
+      "step": 500
+    },
+    {
+      "epoch": 1.9423076923076923,
+      "grad_norm": 1.3724185228347778,
+      "learning_rate": 2.1579377525869532e-05,
+      "loss": 0.3546,
+      "step": 505
+    },
+    {
+      "epoch": 1.9615384615384617,
+      "grad_norm": 1.1697155237197876,
+      "learning_rate": 2.1407397215662455e-05,
+      "loss": 0.338,
+      "step": 510
+    },
+    {
+      "epoch": 1.9807692307692308,
+      "grad_norm": 1.1002439260482788,
+      "learning_rate": 2.1234380376037732e-05,
+      "loss": 0.3481,
+      "step": 515
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.3132089376449585,
+      "learning_rate": 2.1060354996058315e-05,
+      "loss": 0.3685,
+      "step": 520
+    },
+    {
+      "epoch": 2.019230769230769,
+      "grad_norm": 1.1034177541732788,
+      "learning_rate": 2.088534922793948e-05,
+      "loss": 0.2488,
+      "step": 525
+    },
+    {
+      "epoch": 2.0384615384615383,
+      "grad_norm": 1.1746792793273926,
+      "learning_rate": 2.0709391382494618e-05,
+      "loss": 0.2968,
+      "step": 530
+    },
+    {
+      "epoch": 2.0576923076923075,
+      "grad_norm": 1.2488741874694824,
+      "learning_rate": 2.0532509924555386e-05,
+      "loss": 0.2731,
+      "step": 535
+    },
+    {
+      "epoch": 2.076923076923077,
+      "grad_norm": 1.1873801946640015,
+      "learning_rate": 2.0354733468366913e-05,
+      "loss": 0.2538,
+      "step": 540
+    },
+    {
+      "epoch": 2.0961538461538463,
+      "grad_norm": 1.2156885862350464,
+      "learning_rate": 2.017609077295886e-05,
+      "loss": 0.2895,
+      "step": 545
+    },
+    {
+      "epoch": 2.1153846153846154,
+      "grad_norm": 1.0242483615875244,
+      "learning_rate": 1.9996610737493063e-05,
+      "loss": 0.2416,
+      "step": 550
+    },
+    {
+      "epoch": 2.1346153846153846,
+      "grad_norm": 1.2181206941604614,
+      "learning_rate": 1.9816322396588472e-05,
+      "loss": 0.2367,
+      "step": 555
+    },
+    {
+      "epoch": 2.1538461538461537,
+      "grad_norm": 1.0379295349121094,
+      "learning_rate": 1.963525491562421e-05,
+      "loss": 0.2528,
+      "step": 560
+    },
+    {
+      "epoch": 2.173076923076923,
+      "grad_norm": 1.2297924757003784,
+      "learning_rate": 1.945343758602148e-05,
+      "loss": 0.2294,
+      "step": 565
+    },
+    {
+      "epoch": 2.1923076923076925,
+      "grad_norm": 1.1392128467559814,
+      "learning_rate": 1.927089982050504e-05,
+      "loss": 0.2773,
+      "step": 570
+    },
+    {
+      "epoch": 2.2115384615384617,
+      "grad_norm": 1.131394386291504,
+      "learning_rate": 1.9087671148345113e-05,
+      "loss": 0.2586,
+      "step": 575
+    },
+    {
+      "epoch": 2.230769230769231,
+      "grad_norm": 1.1836904287338257,
+      "learning_rate": 1.8903781210580402e-05,
+      "loss": 0.2437,
+      "step": 580
+    },
+    {
+      "epoch": 2.25,
+      "grad_norm": 1.1704728603363037,
+      "learning_rate": 1.8719259755223036e-05,
+      "loss": 0.2709,
+      "step": 585
+    },
+    {
+      "epoch": 2.269230769230769,
+      "grad_norm": 1.042198896408081,
+      "learning_rate": 1.8534136632446218e-05,
+      "loss": 0.2602,
+      "step": 590
+    },
+    {
+      "epoch": 2.2884615384615383,
+      "grad_norm": 1.3327559232711792,
+      "learning_rate": 1.8348441789755337e-05,
+      "loss": 0.2754,
+      "step": 595
+    },
+    {
+      "epoch": 2.3076923076923075,
+      "grad_norm": 1.2408418655395508,
+      "learning_rate": 1.8162205267143333e-05,
+      "loss": 0.2541,
+      "step": 600
+    },
+    {
+      "epoch": 2.326923076923077,
+      "grad_norm": 1.1976670026779175,
+      "learning_rate": 1.7975457192231132e-05,
+      "loss": 0.244,
+      "step": 605
+    },
+    {
+      "epoch": 2.3461538461538463,
+      "grad_norm": 1.4825388193130493,
+      "learning_rate": 1.7788227775393815e-05,
+      "loss": 0.2734,
+      "step": 610
+    },
+    {
+      "epoch": 2.3653846153846154,
+      "grad_norm": 1.1539435386657715,
+      "learning_rate": 1.7600547304873552e-05,
+      "loss": 0.289,
+      "step": 615
+    },
+    {
+      "epoch": 2.3846153846153846,
+      "grad_norm": 1.5810664892196655,
+      "learning_rate": 1.7412446141879766e-05,
+      "loss": 0.236,
+      "step": 620
+    },
+    {
+      "epoch": 2.4038461538461537,
+      "grad_norm": 1.108161449432373,
+      "learning_rate": 1.722395471567763e-05,
+      "loss": 0.205,
+      "step": 625
+    },
+    {
+      "epoch": 2.423076923076923,
+      "grad_norm": 1.1394013166427612,
+      "learning_rate": 1.7035103518665488e-05,
+      "loss": 0.2297,
+      "step": 630
+    },
+    {
+      "epoch": 2.4423076923076925,
+      "grad_norm": 1.1658471822738647,
+      "learning_rate": 1.684592310144208e-05,
+      "loss": 0.2278,
+      "step": 635
+    },
+    {
+      "epoch": 2.4615384615384617,
+      "grad_norm": 1.1106714010238647,
+      "learning_rate": 1.6656444067864323e-05,
+      "loss": 0.2206,
+      "step": 640
+    },
+    {
+      "epoch": 2.480769230769231,
+      "grad_norm": 1.2031956911087036,
+      "learning_rate": 1.6466697070096555e-05,
+      "loss": 0.214,
+      "step": 645
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 1.0847914218902588,
+      "learning_rate": 1.627671280365185e-05,
+      "loss": 0.2248,
+      "step": 650
+    },
+    {
+      "epoch": 2.519230769230769,
+      "grad_norm": 1.2272170782089233,
+      "learning_rate": 1.6086522002426437e-05,
+      "loss": 0.2334,
+      "step": 655
+    },
+    {
+      "epoch": 2.5384615384615383,
+      "grad_norm": 1.1179710626602173,
+      "learning_rate": 1.5896155433727814e-05,
+      "loss": 0.241,
+      "step": 660
+    },
+    {
+      "epoch": 2.5576923076923075,
+      "grad_norm": 1.2092406749725342,
+      "learning_rate": 1.5705643893297526e-05,
+      "loss": 0.2261,
+      "step": 665
+    },
+    {
+      "epoch": 2.5769230769230766,
+      "grad_norm": 1.1971381902694702,
+      "learning_rate": 1.5515018200329295e-05,
+      "loss": 0.2216,
+      "step": 670
+    },
+    {
+      "epoch": 2.5961538461538463,
+      "grad_norm": 1.197493553161621,
+      "learning_rate": 1.532430919248338e-05,
+      "loss": 0.2313,
+      "step": 675
+    },
+    {
+      "epoch": 2.6153846153846154,
+      "grad_norm": 1.0707098245620728,
+      "learning_rate": 1.5133547720897963e-05,
+      "loss": 0.1937,
+      "step": 680
+    },
+    {
+      "epoch": 2.6346153846153846,
+      "grad_norm": 1.232611060142517,
+      "learning_rate": 1.4942764645198317e-05,
+      "loss": 0.2248,
+      "step": 685
+    },
+    {
+      "epoch": 2.6538461538461537,
+      "grad_norm": 1.374160885810852,
+      "learning_rate": 1.4751990828504623e-05,
+      "loss": 0.2354,
+      "step": 690
+    },
+    {
+      "epoch": 2.6730769230769234,
+      "grad_norm": 1.3755944967269897,
+      "learning_rate": 1.4561257132439237e-05,
+      "loss": 0.2018,
+      "step": 695
+    },
+    {
+      "epoch": 2.6923076923076925,
+      "grad_norm": 1.0376676321029663,
+      "learning_rate": 1.4370594412134168e-05,
+      "loss": 0.2061,
+      "step": 700
+    },
+    {
+      "epoch": 2.7115384615384617,
+      "grad_norm": 1.2844074964523315,
+      "learning_rate": 1.4180033511239609e-05,
+      "loss": 0.2224,
+      "step": 705
+    },
+    {
+      "epoch": 2.730769230769231,
+      "grad_norm": 1.2904430627822876,
+      "learning_rate": 1.3989605256934383e-05,
+      "loss": 0.2185,
+      "step": 710
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 1.1227009296417236,
+      "learning_rate": 1.3799340454938932e-05,
+      "loss": 0.2187,
+      "step": 715
+    },
+    {
+      "epoch": 2.769230769230769,
+      "grad_norm": 1.2107775211334229,
+      "learning_rate": 1.3609269884531935e-05,
+      "loss": 0.2177,
+      "step": 720
+    },
+    {
+      "epoch": 2.7884615384615383,
+      "grad_norm": 1.090583086013794,
+      "learning_rate": 1.3419424293571067e-05,
+      "loss": 0.1973,
+      "step": 725
+    },
+    {
+      "epoch": 2.8076923076923075,
+      "grad_norm": 1.045013189315796,
+      "learning_rate": 1.3229834393518903e-05,
+      "loss": 0.2127,
+      "step": 730
+    },
+    {
+      "epoch": 2.8269230769230766,
+      "grad_norm": 1.2943766117095947,
+      "learning_rate": 1.304053085447472e-05,
+      "loss": 0.2101,
+      "step": 735
+    },
+    {
+      "epoch": 2.8461538461538463,
+      "grad_norm": 1.050495982170105,
+      "learning_rate": 1.2851544300212965e-05,
+      "loss": 0.219,
+      "step": 740
+    },
+    {
+      "epoch": 2.8653846153846154,
+      "grad_norm": 1.23055100440979,
+      "learning_rate": 1.2662905303229207e-05,
+      "loss": 0.1913,
+      "step": 745
+    },
+    {
+      "epoch": 2.8846153846153846,
+      "grad_norm": 1.290574550628662,
+      "learning_rate": 1.2474644379794421e-05,
+      "loss": 0.1936,
+      "step": 750
+    },
+    {
+      "epoch": 2.9038461538461537,
+      "grad_norm": 0.9990605115890503,
+      "learning_rate": 1.2286791985018356e-05,
+      "loss": 0.2027,
+      "step": 755
+    },
+    {
+      "epoch": 2.9230769230769234,
+      "grad_norm": 1.294940710067749,
+      "learning_rate": 1.209937850792275e-05,
+      "loss": 0.2048,
+      "step": 760
+    },
+    {
+      "epoch": 2.9423076923076925,
+      "grad_norm": 1.292446255683899,
+      "learning_rate": 1.191243426652532e-05,
+      "loss": 0.1737,
+      "step": 765
+    },
+    {
+      "epoch": 2.9615384615384617,
+      "grad_norm": 1.1711947917938232,
+      "learning_rate": 1.1725989502935175e-05,
+      "loss": 0.1758,
+      "step": 770
+    },
+    {
+      "epoch": 2.980769230769231,
+      "grad_norm": 1.2699744701385498,
+      "learning_rate": 1.1540074378460516e-05,
+      "loss": 0.1859,
+      "step": 775
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.1081920862197876,
+      "learning_rate": 1.135471896872944e-05,
+      "loss": 0.1901,
+      "step": 780
+    },
+    {
+      "epoch": 3.019230769230769,
+      "grad_norm": 1.0495061874389648,
+      "learning_rate": 1.1169953258824567e-05,
+      "loss": 0.1403,
+      "step": 785
+    },
+    {
+      "epoch": 3.0384615384615383,
+      "grad_norm": 1.2737641334533691,
+      "learning_rate": 1.0985807138432336e-05,
+      "loss": 0.1351,
+      "step": 790
+    },
+    {
+      "epoch": 3.0576923076923075,
+      "grad_norm": 1.1091535091400146,
+      "learning_rate": 1.080231039700774e-05,
+      "loss": 0.136,
+      "step": 795
+    },
+    {
+      "epoch": 3.076923076923077,
+      "grad_norm": 1.0284589529037476,
+      "learning_rate": 1.0619492718955242e-05,
+      "loss": 0.1283,
+      "step": 800
+    },
+    {
+      "epoch": 3.0961538461538463,
+      "grad_norm": 1.054133653640747,
+      "learning_rate": 1.0437383678826723e-05,
+      "loss": 0.1324,
+      "step": 805
+    },
+    {
+      "epoch": 3.1153846153846154,
+      "grad_norm": 1.0665591955184937,
+      "learning_rate": 1.0256012736537172e-05,
+      "loss": 0.1448,
+      "step": 810
+    },
+    {
+      "epoch": 3.1346153846153846,
+      "grad_norm": 1.1725926399230957,
+      "learning_rate": 1.007540923259896e-05,
+      "loss": 0.1448,
+      "step": 815
+    },
+    {
+      "epoch": 3.1538461538461537,
+      "grad_norm": 0.9875220060348511,
+      "learning_rate": 9.895602383375355e-06,
+      "loss": 0.1361,
+      "step": 820
+    },
+    {
+      "epoch": 3.173076923076923,
+      "grad_norm": 1.078674077987671,
+      "learning_rate": 9.716621276354225e-06,
+      "loss": 0.1348,
+      "step": 825
+    },
+    {
+      "epoch": 3.1923076923076925,
+      "grad_norm": 1.1486966609954834,
+      "learning_rate": 9.5384948654425e-06,
+      "loss": 0.1522,
+      "step": 830
+    },
+    {
+      "epoch": 3.2115384615384617,
+      "grad_norm": 1.0026553869247437,
+      "learning_rate": 9.361251966282276e-06,
+      "loss": 0.1311,
+      "step": 835
+    },
+    {
+      "epoch": 3.230769230769231,
+      "grad_norm": 1.2294085025787354,
+      "learning_rate": 9.184921251589299e-06,
+      "loss": 0.1277,
+      "step": 840
+    },
+    {
+      "epoch": 3.25,
+      "grad_norm": 0.8330014944076538,
+      "learning_rate": 9.009531246514569e-06,
+      "loss": 0.1327,
+      "step": 845
+    },
+    {
+      "epoch": 3.269230769230769,
+      "grad_norm": 0.8544556498527527,
+      "learning_rate": 8.835110324029762e-06,
+      "loss": 0.1288,
+      "step": 850
+    },
+    {
+      "epoch": 3.2884615384615383,
+      "grad_norm": 1.1136083602905273,
+      "learning_rate": 8.661686700337368e-06,
+      "loss": 0.1297,
+      "step": 855
+    },
+    {
+      "epoch": 3.3076923076923075,
+      "grad_norm": 1.009947657585144,
+      "learning_rate": 8.489288430306097e-06,
+      "loss": 0.1289,
+      "step": 860
+    },
+    {
+      "epoch": 3.326923076923077,
+      "grad_norm": 1.0761442184448242,
+      "learning_rate": 8.317943402932434e-06,
+      "loss": 0.1287,
+      "step": 865
+    },
+    {
+      "epoch": 3.3461538461538463,
+      "grad_norm": 1.067358374595642,
+      "learning_rate": 8.147679336829019e-06,
+      "loss": 0.1185,
+      "step": 870
+    },
+    {
+      "epoch": 3.3653846153846154,
+      "grad_norm": 0.9998955726623535,
+      "learning_rate": 7.97852377574058e-06,
+      "loss": 0.1223,
+      "step": 875
+    },
+    {
+      "epoch": 3.3846153846153846,
+      "grad_norm": 1.0905925035476685,
+      "learning_rate": 7.810504084088168e-06,
+      "loss": 0.122,
+      "step": 880
+    },
+    {
+      "epoch": 3.4038461538461537,
+      "grad_norm": 0.9051513671875,
+      "learning_rate": 7.643647442542383e-06,
+      "loss": 0.1272,
+      "step": 885
+    },
+    {
+      "epoch": 3.423076923076923,
+      "grad_norm": 1.091073751449585,
+      "learning_rate": 7.477980843626337e-06,
+      "loss": 0.1242,
+      "step": 890
+    },
+    {
+      "epoch": 3.4423076923076925,
+      "grad_norm": 1.00954270362854,
+      "learning_rate": 7.313531087349085e-06,
+      "loss": 0.118,
+      "step": 895
+    },
+    {
+      "epoch": 3.4615384615384617,
+      "grad_norm": 1.1877714395523071,
+      "learning_rate": 7.150324776870125e-06,
+      "loss": 0.1127,
+      "step": 900
+    },
+    {
+      "epoch": 3.480769230769231,
+      "grad_norm": 1.0862516164779663,
+      "learning_rate": 6.988388314195793e-06,
+      "loss": 0.1221,
+      "step": 905
+    },
+    {
+      "epoch": 3.5,
+      "grad_norm": 1.2270557880401611,
+      "learning_rate": 6.827747895908212e-06,
+      "loss": 0.1194,
+      "step": 910
+    },
+    {
+      "epoch": 3.519230769230769,
+      "grad_norm": 1.0056644678115845,
+      "learning_rate": 6.6684295089274565e-06,
+      "loss": 0.1217,
+      "step": 915
+    },
+    {
+      "epoch": 3.5384615384615383,
+      "grad_norm": 1.038124918937683,
+      "learning_rate": 6.510458926307577e-06,
+      "loss": 0.1194,
+      "step": 920
+    },
+    {
+      "epoch": 3.5576923076923075,
+      "grad_norm": 0.970857560634613,
+      "learning_rate": 6.353861703067332e-06,
+      "loss": 0.1127,
+      "step": 925
+    },
+    {
+      "epoch": 3.5769230769230766,
+      "grad_norm": 1.0015406608581543,
+      "learning_rate": 6.198663172056109e-06,
+      "loss": 0.1263,
+      "step": 930
+    },
+    {
+      "epoch": 3.5961538461538463,
+      "grad_norm": 0.9921681880950928,
+      "learning_rate": 6.044888439855806e-06,
+      "loss": 0.1341,
+      "step": 935
+    },
+    {
+      "epoch": 3.6153846153846154,
+      "grad_norm": 1.0572803020477295,
+      "learning_rate": 5.892562382719333e-06,
+      "loss": 0.1299,
+      "step": 940
+    },
+    {
+      "epoch": 3.6346153846153846,
+      "grad_norm": 1.028430700302124,
+      "learning_rate": 5.741709642546352e-06,
+      "loss": 0.1112,
+      "step": 945
+    },
+    {
+      "epoch": 3.6538461538461537,
+      "grad_norm": 1.1680800914764404,
+      "learning_rate": 5.592354622896944e-06,
+      "loss": 0.1118,
+      "step": 950
+    },
+    {
+      "epoch": 3.6730769230769234,
+      "grad_norm": 0.914523184299469,
+      "learning_rate": 5.444521485043831e-06,
+      "loss": 0.1073,
+      "step": 955
+    },
+    {
+      "epoch": 3.6923076923076925,
+      "grad_norm": 0.8823187351226807,
+      "learning_rate": 5.2982341440637804e-06,
+      "loss": 0.1217,
+      "step": 960
+    },
+    {
+      "epoch": 3.7115384615384617,
+      "grad_norm": 1.0132150650024414,
+      "learning_rate": 5.153516264968841e-06,
+      "loss": 0.1198,
+      "step": 965
+    },
+    {
+      "epoch": 3.730769230769231,
+      "grad_norm": 1.0348647832870483,
+      "learning_rate": 5.010391258878049e-06,
+      "loss": 0.1189,
+      "step": 970
+    },
+    {
+      "epoch": 3.75,
+      "grad_norm": 1.1547722816467285,
+      "learning_rate": 4.868882279230208e-06,
+      "loss": 0.1255,
+      "step": 975
+    },
+    {
+      "epoch": 3.769230769230769,
+      "grad_norm": 0.9065052270889282,
+      "learning_rate": 4.729012218038287e-06,
+      "loss": 0.1161,
+      "step": 980
+    },
+    {
+      "epoch": 3.7884615384615383,
+      "grad_norm": 0.8526612520217896,
+      "learning_rate": 4.590803702186219e-06,
+      "loss": 0.0981,
+      "step": 985
+    },
+    {
+      "epoch": 3.8076923076923075,
+      "grad_norm": 0.9093031883239746,
+      "learning_rate": 4.454279089768503e-06,
+      "loss": 0.1082,
+      "step": 990
+    },
+    {
+      "epoch": 3.8269230769230766,
+      "grad_norm": 0.9617008566856384,
+      "learning_rate": 4.319460466473328e-06,
+      "loss": 0.1013,
+      "step": 995
+    },
+    {
+      "epoch": 3.8461538461538463,
+      "grad_norm": 1.0160218477249146,
+      "learning_rate": 4.186369642009753e-06,
+      "loss": 0.116,
+      "step": 1000
+    },
+    {
+      "epoch": 3.8653846153846154,
+      "grad_norm": 0.9817861318588257,
+      "learning_rate": 4.055028146579532e-06,
+      "loss": 0.1139,
+      "step": 1005
+    },
+    {
+      "epoch": 3.8846153846153846,
+      "grad_norm": 0.9847229719161987,
+      "learning_rate": 3.925457227394158e-06,
+      "loss": 0.1148,
+      "step": 1010
+    },
+    {
+      "epoch": 3.9038461538461537,
+      "grad_norm": 1.0290164947509766,
+      "learning_rate": 3.7976778452376966e-06,
+      "loss": 0.1196,
+      "step": 1015
+    },
+    {
+      "epoch": 3.9230769230769234,
+      "grad_norm": 0.7677153944969177,
+      "learning_rate": 3.671710671075913e-06,
+      "loss": 0.1116,
+      "step": 1020
+    },
+    {
+      "epoch": 3.9423076923076925,
+      "grad_norm": 0.9820197820663452,
+      "learning_rate": 3.547576082712351e-06,
+      "loss": 0.0969,
+      "step": 1025
+    },
+    {
+      "epoch": 3.9615384615384617,
+      "grad_norm": 0.8637183904647827,
+      "learning_rate": 3.425294161491796e-06,
+      "loss": 0.1065,
+      "step": 1030
+    },
+    {
+      "epoch": 3.980769230769231,
+      "grad_norm": 0.9800280928611755,
+      "learning_rate": 3.3048846890516658e-06,
+      "loss": 0.1113,
+      "step": 1035
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 0.785662055015564,
+      "learning_rate": 3.1863671441219634e-06,
+      "loss": 0.1047,
+      "step": 1040
+    },
+    {
+      "epoch": 4.019230769230769,
+      "grad_norm": 0.9184972643852234,
+      "learning_rate": 3.0697606993741616e-06,
+      "loss": 0.102,
+      "step": 1045
+    },
+    {
+      "epoch": 4.038461538461538,
+      "grad_norm": 0.8287720084190369,
+      "learning_rate": 2.955084218319636e-06,
+      "loss": 0.0896,
+      "step": 1050
+    },
+    {
+      "epoch": 4.0576923076923075,
+      "grad_norm": 0.9467900395393372,
+      "learning_rate": 2.842356252258102e-06,
+      "loss": 0.0893,
+      "step": 1055
+    },
+    {
+      "epoch": 4.076923076923077,
+      "grad_norm": 0.8027834892272949,
+      "learning_rate": 2.7315950372765473e-06,
+      "loss": 0.0881,
+      "step": 1060
+    },
+    {
+      "epoch": 4.096153846153846,
+      "grad_norm": 0.7623370289802551,
+      "learning_rate": 2.6228184912991813e-06,
+      "loss": 0.0883,
+      "step": 1065
+    },
+    {
+      "epoch": 4.115384615384615,
+      "grad_norm": 0.8034529685974121,
+      "learning_rate": 2.516044211188823e-06,
+      "loss": 0.083,
+      "step": 1070
+    },
+    {
+      "epoch": 4.134615384615385,
+      "grad_norm": 0.9694064259529114,
+      "learning_rate": 2.411289469900261e-06,
+      "loss": 0.0776,
+      "step": 1075
+    },
+    {
+      "epoch": 4.153846153846154,
+      "grad_norm": 0.9289525151252747,
+      "learning_rate": 2.308571213685967e-06,
+      "loss": 0.0779,
+      "step": 1080
+    },
+    {
+      "epoch": 4.173076923076923,
+      "grad_norm": 0.7800962924957275,
+      "learning_rate": 2.2079060593547324e-06,
+      "loss": 0.0795,
+      "step": 1085
+    },
+    {
+      "epoch": 4.1923076923076925,
+      "grad_norm": 0.7357608079910278,
+      "learning_rate": 2.1093102915835117e-06,
+      "loss": 0.0815,
+      "step": 1090
+    },
+    {
+      "epoch": 4.211538461538462,
+      "grad_norm": 0.7938529253005981,
+      "learning_rate": 2.0127998602830538e-06,
+      "loss": 0.0834,
+      "step": 1095
+    },
+    {
+      "epoch": 4.230769230769231,
+      "grad_norm": 0.9307594895362854,
+      "learning_rate": 1.918390378017675e-06,
+      "loss": 0.0824,
+      "step": 1100
+    },
+    {
+      "epoch": 4.25,
+      "grad_norm": 0.7176527976989746,
+      "learning_rate": 1.8260971174796203e-06,
+      "loss": 0.0794,
+      "step": 1105
+    },
+    {
+      "epoch": 4.269230769230769,
+      "grad_norm": 0.7794285416603088,
+      "learning_rate": 1.7359350090183423e-06,
+      "loss": 0.0813,
+      "step": 1110
+    },
+    {
+      "epoch": 4.288461538461538,
+      "grad_norm": 0.9609277844429016,
+      "learning_rate": 1.6479186382252525e-06,
+      "loss": 0.0838,
+      "step": 1115
+    },
+    {
+      "epoch": 4.3076923076923075,
+      "grad_norm": 0.7197667956352234,
+      "learning_rate": 1.562062243574181e-06,
+      "loss": 0.0922,
+      "step": 1120
+    },
+    {
+      "epoch": 4.326923076923077,
+      "grad_norm": 0.8626694679260254,
+      "learning_rate": 1.4783797141180071e-06,
+      "loss": 0.0709,
+      "step": 1125
+    },
+    {
+      "epoch": 4.346153846153846,
+      "grad_norm": 0.7555365562438965,
+      "learning_rate": 1.396884587241819e-06,
+      "loss": 0.0828,
+      "step": 1130
+    },
+    {
+      "epoch": 4.365384615384615,
+      "grad_norm": 0.861768901348114,
+      "learning_rate": 1.3175900464729595e-06,
+      "loss": 0.0838,
+      "step": 1135
+    },
+    {
+      "epoch": 4.384615384615385,
+      "grad_norm": 0.7930833697319031,
+      "learning_rate": 1.2405089193483233e-06,
+      "loss": 0.0816,
+      "step": 1140
+    },
+    {
+      "epoch": 4.403846153846154,
+      "grad_norm": 0.8239856958389282,
+      "learning_rate": 1.1656536753392288e-06,
+      "loss": 0.0913,
+      "step": 1145
+    },
+    {
+      "epoch": 4.423076923076923,
+      "grad_norm": 0.9674453139305115,
+      "learning_rate": 1.0930364238342284e-06,
+      "loss": 0.0778,
+      "step": 1150
+    },
+    {
+      "epoch": 4.4423076923076925,
+      "grad_norm": 0.7881841659545898,
+      "learning_rate": 1.0226689121801536e-06,
+      "loss": 0.0741,
+      "step": 1155
+    },
+    {
+      "epoch": 4.461538461538462,
+      "grad_norm": 0.7504427433013916,
+      "learning_rate": 9.5456252378175e-07,
+      "loss": 0.081,
+      "step": 1160
+    },
+    {
+      "epoch": 4.480769230769231,
+      "grad_norm": 0.7033846974372864,
+      "learning_rate": 8.887282762601801e-07,
+      "loss": 0.0747,
+      "step": 1165
+    },
+    {
+      "epoch": 4.5,
+      "grad_norm": 0.7966051697731018,
+      "learning_rate": 8.251768196706694e-07,
+      "loss": 0.0892,
+      "step": 1170
+    },
+    {
+      "epoch": 4.519230769230769,
+      "grad_norm": 0.8197062611579895,
+      "learning_rate": 7.639184347796646e-07,
+      "loss": 0.0837,
+      "step": 1175
+    },
+    {
+      "epoch": 4.538461538461538,
+      "grad_norm": 0.7823281288146973,
+      "learning_rate": 7.049630314016931e-07,
+      "loss": 0.0831,
+      "step": 1180
+    },
+    {
+      "epoch": 4.5576923076923075,
+      "grad_norm": 0.7187005877494812,
+      "learning_rate": 6.483201467962485e-07,
+      "loss": 0.0707,
+      "step": 1185
+    },
+    {
+      "epoch": 4.576923076923077,
+      "grad_norm": 0.829845130443573,
+      "learning_rate": 5.939989441249471e-07,
+      "loss": 0.0771,
+      "step": 1190
+    },
+    {
+      "epoch": 4.596153846153846,
+      "grad_norm": 0.7916256189346313,
+      "learning_rate": 5.420082109691766e-07,
+      "loss": 0.0844,
+      "step": 1195
+    },
+    {
+      "epoch": 4.615384615384615,
+      "grad_norm": 0.8681178689002991,
+      "learning_rate": 4.923563579085511e-07,
+      "loss": 0.0813,
+      "step": 1200
+    },
+    {
+      "epoch": 4.634615384615385,
+      "grad_norm": 0.7274530529975891,
+      "learning_rate": 4.450514171603037e-07,
+      "loss": 0.0824,
+      "step": 1205
+    },
+    {
+      "epoch": 4.653846153846154,
+      "grad_norm": 0.6939621567726135,
+      "learning_rate": 4.0010104127991385e-07,
+      "loss": 0.0783,
+      "step": 1210
+    },
+    {
+      "epoch": 4.673076923076923,
+      "grad_norm": 0.7913796305656433,
+      "learning_rate": 3.5751250192315066e-07,
+      "loss": 0.0819,
+      "step": 1215
+    },
+    {
+      "epoch": 4.6923076923076925,
+      "grad_norm": 0.8079668879508972,
+      "learning_rate": 3.1729268866973584e-07,
+      "loss": 0.0677,
+      "step": 1220
+    },
+    {
+      "epoch": 4.711538461538462,
+      "grad_norm": 0.7310797572135925,
+      "learning_rate": 2.794481079087985e-07,
+      "loss": 0.088,
+      "step": 1225
+    },
+    {
+      "epoch": 4.730769230769231,
+      "grad_norm": 0.7207171320915222,
+      "learning_rate": 2.4398488178634215e-07,
+      "loss": 0.0774,
+      "step": 1230
+    },
+    {
+      "epoch": 4.75,
+      "grad_norm": 0.7653047442436218,
+      "learning_rate": 2.1090874721485854e-07,
+      "loss": 0.0814,
+      "step": 1235
+    },
+    {
+      "epoch": 4.769230769230769,
+      "grad_norm": 0.8429785370826721,
+      "learning_rate": 1.8022505494526164e-07,
+      "loss": 0.0798,
+      "step": 1240
+    },
+    {
+      "epoch": 4.788461538461538,
+      "grad_norm": 0.782753586769104,
+      "learning_rate": 1.5193876870129188e-07,
+      "loss": 0.0809,
+      "step": 1245
+    },
+    {
+      "epoch": 4.8076923076923075,
+      "grad_norm": 0.8216264843940735,
+      "learning_rate": 1.2605446437653157e-07,
+      "loss": 0.0789,
+      "step": 1250
+    },
+    {
+      "epoch": 4.826923076923077,
+      "grad_norm": 0.8806825280189514,
+      "learning_rate": 1.0257632929416394e-07,
+      "loss": 0.0921,
+      "step": 1255
+    },
+    {
+      "epoch": 4.846153846153846,
+      "grad_norm": 0.7621264457702637,
+      "learning_rate": 8.150816152957819e-08,
+      "loss": 0.0772,
+      "step": 1260
+    },
+    {
+      "epoch": 4.865384615384615,
+      "grad_norm": 0.7992205619812012,
+      "learning_rate": 6.285336929595931e-08,
+      "loss": 0.0911,
+      "step": 1265
+    },
+    {
+      "epoch": 4.884615384615385,
+      "grad_norm": 0.9502942562103271,
+      "learning_rate": 4.6614970392934165e-08,
+      "loss": 0.0968,
+      "step": 1270
+    },
+    {
+      "epoch": 4.903846153846154,
+      "grad_norm": 0.7410289645195007,
+      "learning_rate": 3.279559171838198e-08,
+      "loss": 0.0799,
+      "step": 1275
+    },
+    {
+      "epoch": 4.923076923076923,
+      "grad_norm": 0.7301416993141174,
+      "learning_rate": 2.1397468843479263e-08,
+      "loss": 0.0688,
+      "step": 1280
+    },
+    {
+      "epoch": 4.9423076923076925,
+      "grad_norm": 0.8698253631591797,
+      "learning_rate": 1.2422445651047443e-08,
+      "loss": 0.0754,
+      "step": 1285
+    },
+    {
+      "epoch": 4.961538461538462,
+      "grad_norm": 0.7773427367210388,
+      "learning_rate": 5.871974037271466e-09,
+      "loss": 0.0816,
+      "step": 1290
+    },
+    {
+      "epoch": 4.980769230769231,
+      "grad_norm": 0.7329732775688171,
+      "learning_rate": 1.7471136768210105e-09,
+      "loss": 0.0733,
+      "step": 1295
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 0.6405507326126099,
+      "learning_rate": 4.853185142594807e-11,
+      "loss": 0.0715,
+      "step": 1300
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1300,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.9316274796009882e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

14_128_e5_3e-5/checkpoint-1300/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dbfb4ffc8d0db1e939d7417e2f349ee624f0febeeb1e1b8c00f77c5c3770afcd
+size 7736

14_128_e5_3e-5/checkpoint-1300/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/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, exclude_frozen_parameters):
+    """
+    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,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+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 _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+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() if _has_callable(shape, 'numel') else math.prod(shape)
+            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,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _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,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _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, exclude_frozen_parameters=False):
+    """
+    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``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    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, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    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``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    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("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    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,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

14_128_e5_3e-5/checkpoint-260/README.md

@@ -0,0 +1,202 @@
+---
+base_model: ibm-granite/granite-3.3-8b-base
+library_name: peft
+---
+
+# 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]
+### Framework versions
+
+- PEFT 0.15.2

14_128_e5_3e-5/checkpoint-260/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "v_proj",
+    "q_proj",
+    "k_proj",
+    "up_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

14_128_e5_3e-5/checkpoint-260/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3f5bc130b5413e5f524772e16e1a0b312bcc12d455960a3adc47b82a0124bd7
+size 791751704

14_128_e5_3e-5/checkpoint-260/latest

@@ -0,0 +1 @@
+global_step260

14_128_e5_3e-5/checkpoint-260/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5c49f91ae65cfe9c278cb87eb61340c2aa1b0916fc6447bfd7cc79d5d8d1907
+size 15920

14_128_e5_3e-5/checkpoint-260/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e45c7ef3cebb0073f0b7564ae10cfb87ab6ff7a51e303416b504cd569a198cc
+size 15920

14_128_e5_3e-5/checkpoint-260/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e331c436c0892b89e5112f4894a20aaaa98657cdb31fa998f584a3f560e25603
+size 15920