Upload folder using huggingface_hub

2_128_e5_3e-5/checkpoint-1005/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

2_128_e5_3e-5/checkpoint-1005/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": [
+    "down_proj",
+    "v_proj",
+    "k_proj",
+    "o_proj",
+    "up_proj",
+    "q_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

2_128_e5_3e-5/checkpoint-1005/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3e2a2a6a30625480dfd9bcd467694f4f9bb5f8e43b90382d0574f28f122060d
+size 791751704

2_128_e5_3e-5/checkpoint-1005/latest

@@ -0,0 +1 @@
+global_step1005

2_128_e5_3e-5/checkpoint-1005/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75828851ddda11dc576a6d2555a8c6bc52e7cdacbc2cf04a4e97cd943a4b6543
+size 15920

2_128_e5_3e-5/checkpoint-1005/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac2cc63a54d4cbb0ad22092105e7c0e6109354c75bdf15583c9e3ba791a1ba53
+size 15920

2_128_e5_3e-5/checkpoint-1005/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11225730ca0a28ae5616a8060b29db06131f66b11b4995d342c92726e196ea09
+size 15920

2_128_e5_3e-5/checkpoint-1005/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24dd598d62d85c433f0f1ac7bee34d91dab5cf94a6b8362ad52c50accdfd833d
+size 15920

2_128_e5_3e-5/checkpoint-1005/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e8beb304fcd4fd448cc5d8d9337660bd0bdb06a2c016a4dc434e8212dcd835a
+size 15920

2_128_e5_3e-5/checkpoint-1005/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb31bd6bd32d64c5cf68030b87e5684123c657292bd51f513c8c7bc2f883e891
+size 15920

2_128_e5_3e-5/checkpoint-1005/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9c7b1dc3ff269940cab5e81b06001d13cb1889ef185752fd9df06924310b1c0
+size 15920

2_128_e5_3e-5/checkpoint-1005/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2e8cbbb0ef2a965b09775ea837394de663a6200303be514fc166b90b3d73791
+size 15920

2_128_e5_3e-5/checkpoint-1005/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:caad05429a75a65ce78613022a8e89b94f4b098f0e7ebdcfbf73314595396152
+size 1064

2_128_e5_3e-5/checkpoint-1005/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
+  }
+}

2_128_e5_3e-5/checkpoint-1005/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
+}

2_128_e5_3e-5/checkpoint-1005/trainer_state.json

@@ -0,0 +1,1441 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1005,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.014925373134328358,
+      "grad_norm": 1.2194169759750366,
+      "learning_rate": 1.4285714285714286e-06,
+      "loss": 1.3181,
+      "step": 5
+    },
+    {
+      "epoch": 0.029850746268656716,
+      "grad_norm": 0.9168437123298645,
+      "learning_rate": 3.2142857142857143e-06,
+      "loss": 1.3037,
+      "step": 10
+    },
+    {
+      "epoch": 0.04477611940298507,
+      "grad_norm": 0.6358347535133362,
+      "learning_rate": 4.9999999999999996e-06,
+      "loss": 1.3083,
+      "step": 15
+    },
+    {
+      "epoch": 0.05970149253731343,
+      "grad_norm": 1.2521260976791382,
+      "learning_rate": 6.785714285714286e-06,
+      "loss": 1.2654,
+      "step": 20
+    },
+    {
+      "epoch": 0.07462686567164178,
+      "grad_norm": 0.6080345511436462,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 1.2305,
+      "step": 25
+    },
+    {
+      "epoch": 0.08955223880597014,
+      "grad_norm": 0.4800509214401245,
+      "learning_rate": 1.0357142857142857e-05,
+      "loss": 1.2006,
+      "step": 30
+    },
+    {
+      "epoch": 0.1044776119402985,
+      "grad_norm": 0.4233926236629486,
+      "learning_rate": 1.2142857142857144e-05,
+      "loss": 1.2602,
+      "step": 35
+    },
+    {
+      "epoch": 0.11940298507462686,
+      "grad_norm": 0.5414039492607117,
+      "learning_rate": 1.3928571428571429e-05,
+      "loss": 1.2663,
+      "step": 40
+    },
+    {
+      "epoch": 0.13432835820895522,
+      "grad_norm": 0.5675264000892639,
+      "learning_rate": 1.5714285714285715e-05,
+      "loss": 1.1681,
+      "step": 45
+    },
+    {
+      "epoch": 0.14925373134328357,
+      "grad_norm": 0.4302530884742737,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 1.2069,
+      "step": 50
+    },
+    {
+      "epoch": 0.16417910447761194,
+      "grad_norm": 0.4482235610485077,
+      "learning_rate": 1.928571428571429e-05,
+      "loss": 1.1548,
+      "step": 55
+    },
+    {
+      "epoch": 0.1791044776119403,
+      "grad_norm": 0.5453466773033142,
+      "learning_rate": 2.107142857142857e-05,
+      "loss": 1.151,
+      "step": 60
+    },
+    {
+      "epoch": 0.19402985074626866,
+      "grad_norm": 0.5155040621757507,
+      "learning_rate": 2.2857142857142858e-05,
+      "loss": 1.0988,
+      "step": 65
+    },
+    {
+      "epoch": 0.208955223880597,
+      "grad_norm": 0.5083202719688416,
+      "learning_rate": 2.464285714285714e-05,
+      "loss": 1.0979,
+      "step": 70
+    },
+    {
+      "epoch": 0.22388059701492538,
+      "grad_norm": 0.5227262377738953,
+      "learning_rate": 2.6428571428571428e-05,
+      "loss": 1.1366,
+      "step": 75
+    },
+    {
+      "epoch": 0.23880597014925373,
+      "grad_norm": 0.5148192048072815,
+      "learning_rate": 2.8214285714285714e-05,
+      "loss": 1.1455,
+      "step": 80
+    },
+    {
+      "epoch": 0.2537313432835821,
+      "grad_norm": 0.4345020651817322,
+      "learning_rate": 3e-05,
+      "loss": 1.1615,
+      "step": 85
+    },
+    {
+      "epoch": 0.26865671641791045,
+      "grad_norm": 0.5491366386413574,
+      "learning_rate": 2.999926893308441e-05,
+      "loss": 1.091,
+      "step": 90
+    },
+    {
+      "epoch": 0.2835820895522388,
+      "grad_norm": 0.5117303133010864,
+      "learning_rate": 2.9997075803598805e-05,
+      "loss": 1.1199,
+      "step": 95
+    },
+    {
+      "epoch": 0.29850746268656714,
+      "grad_norm": 0.5951908826828003,
+      "learning_rate": 2.9993420825319784e-05,
+      "loss": 1.0835,
+      "step": 100
+    },
+    {
+      "epoch": 0.31343283582089554,
+      "grad_norm": 0.5786913633346558,
+      "learning_rate": 2.9988304354518498e-05,
+      "loss": 1.0623,
+      "step": 105
+    },
+    {
+      "epoch": 0.3283582089552239,
+      "grad_norm": 0.5928995609283447,
+      "learning_rate": 2.9981726889925956e-05,
+      "loss": 1.0168,
+      "step": 110
+    },
+    {
+      "epoch": 0.34328358208955223,
+      "grad_norm": 0.505250871181488,
+      "learning_rate": 2.997368907268439e-05,
+      "loss": 1.0542,
+      "step": 115
+    },
+    {
+      "epoch": 0.3582089552238806,
+      "grad_norm": 0.5257184505462646,
+      "learning_rate": 2.9964191686284768e-05,
+      "loss": 1.0217,
+      "step": 120
+    },
+    {
+      "epoch": 0.373134328358209,
+      "grad_norm": 0.641715943813324,
+      "learning_rate": 2.995323565649042e-05,
+      "loss": 1.0544,
+      "step": 125
+    },
+    {
+      "epoch": 0.3880597014925373,
+      "grad_norm": 0.6116381883621216,
+      "learning_rate": 2.99408220512468e-05,
+      "loss": 1.0089,
+      "step": 130
+    },
+    {
+      "epoch": 0.40298507462686567,
+      "grad_norm": 0.8415507078170776,
+      "learning_rate": 2.9926952080577392e-05,
+      "loss": 1.0024,
+      "step": 135
+    },
+    {
+      "epoch": 0.417910447761194,
+      "grad_norm": 0.5881995558738708,
+      "learning_rate": 2.991162709646575e-05,
+      "loss": 0.9376,
+      "step": 140
+    },
+    {
+      "epoch": 0.43283582089552236,
+      "grad_norm": 0.6294875144958496,
+      "learning_rate": 2.9894848592723724e-05,
+      "loss": 0.939,
+      "step": 145
+    },
+    {
+      "epoch": 0.44776119402985076,
+      "grad_norm": 0.6948407292366028,
+      "learning_rate": 2.9876618204845835e-05,
+      "loss": 0.9616,
+      "step": 150
+    },
+    {
+      "epoch": 0.4626865671641791,
+      "grad_norm": 0.7214933633804321,
+      "learning_rate": 2.985693770984989e-05,
+      "loss": 0.9484,
+      "step": 155
+    },
+    {
+      "epoch": 0.47761194029850745,
+      "grad_norm": 0.7646703124046326,
+      "learning_rate": 2.9835809026103717e-05,
+      "loss": 0.8907,
+      "step": 160
+    },
+    {
+      "epoch": 0.4925373134328358,
+      "grad_norm": 0.8051431775093079,
+      "learning_rate": 2.98132342131382e-05,
+      "loss": 0.9145,
+      "step": 165
+    },
+    {
+      "epoch": 0.5074626865671642,
+      "grad_norm": 0.8606762290000916,
+      "learning_rate": 2.9789215471446528e-05,
+      "loss": 0.9351,
+      "step": 170
+    },
+    {
+      "epoch": 0.5223880597014925,
+      "grad_norm": 0.6801915168762207,
+      "learning_rate": 2.9763755142269695e-05,
+      "loss": 0.8833,
+      "step": 175
+    },
+    {
+      "epoch": 0.5373134328358209,
+      "grad_norm": 0.7298025488853455,
+      "learning_rate": 2.9736855707368266e-05,
+      "loss": 0.8845,
+      "step": 180
+    },
+    {
+      "epoch": 0.5522388059701493,
+      "grad_norm": 0.7883251309394836,
+      "learning_rate": 2.97085197887805e-05,
+      "loss": 0.8407,
+      "step": 185
+    },
+    {
+      "epoch": 0.5671641791044776,
+      "grad_norm": 0.8319213390350342,
+      "learning_rate": 2.9678750148566748e-05,
+      "loss": 0.8268,
+      "step": 190
+    },
+    {
+      "epoch": 0.582089552238806,
+      "grad_norm": 0.7365439534187317,
+      "learning_rate": 2.9647549688540206e-05,
+      "loss": 0.8979,
+      "step": 195
+    },
+    {
+      "epoch": 0.5970149253731343,
+      "grad_norm": 0.8525620102882385,
+      "learning_rate": 2.9614921449984092e-05,
+      "loss": 0.8602,
+      "step": 200
+    },
+    {
+      "epoch": 0.6119402985074627,
+      "grad_norm": 0.7475487589836121,
+      "learning_rate": 2.958086861335517e-05,
+      "loss": 0.8242,
+      "step": 205
+    },
+    {
+      "epoch": 0.6268656716417911,
+      "grad_norm": 0.9775540828704834,
+      "learning_rate": 2.9545394497973736e-05,
+      "loss": 0.8501,
+      "step": 210
+    },
+    {
+      "epoch": 0.6417910447761194,
+      "grad_norm": 0.7747155427932739,
+      "learning_rate": 2.950850256170007e-05,
+      "loss": 0.8561,
+      "step": 215
+    },
+    {
+      "epoch": 0.6567164179104478,
+      "grad_norm": 0.8525500893592834,
+      "learning_rate": 2.9470196400597385e-05,
+      "loss": 0.7717,
+      "step": 220
+    },
+    {
+      "epoch": 0.6716417910447762,
+      "grad_norm": 0.8092201352119446,
+      "learning_rate": 2.9430479748581286e-05,
+      "loss": 0.8772,
+      "step": 225
+    },
+    {
+      "epoch": 0.6865671641791045,
+      "grad_norm": 0.8191531300544739,
+      "learning_rate": 2.9389356477055807e-05,
+      "loss": 0.8057,
+      "step": 230
+    },
+    {
+      "epoch": 0.7014925373134329,
+      "grad_norm": 0.888316810131073,
+      "learning_rate": 2.934683059453605e-05,
+      "loss": 0.7826,
+      "step": 235
+    },
+    {
+      "epoch": 0.7164179104477612,
+      "grad_norm": 0.8837902545928955,
+      "learning_rate": 2.9302906246257462e-05,
+      "loss": 0.7865,
+      "step": 240
+    },
+    {
+      "epoch": 0.7313432835820896,
+      "grad_norm": 0.7749768495559692,
+      "learning_rate": 2.925758771377174e-05,
+      "loss": 0.7745,
+      "step": 245
+    },
+    {
+      "epoch": 0.746268656716418,
+      "grad_norm": 0.8685063719749451,
+      "learning_rate": 2.9210879414529526e-05,
+      "loss": 0.7487,
+      "step": 250
+    },
+    {
+      "epoch": 0.7611940298507462,
+      "grad_norm": 0.8156796097755432,
+      "learning_rate": 2.9162785901449783e-05,
+      "loss": 0.7301,
+      "step": 255
+    },
+    {
+      "epoch": 0.7761194029850746,
+      "grad_norm": 0.8999550342559814,
+      "learning_rate": 2.9113311862476023e-05,
+      "loss": 0.7865,
+      "step": 260
+    },
+    {
+      "epoch": 0.7910447761194029,
+      "grad_norm": 0.912211537361145,
+      "learning_rate": 2.906246212011931e-05,
+      "loss": 0.7367,
+      "step": 265
+    },
+    {
+      "epoch": 0.8059701492537313,
+      "grad_norm": 0.790409505367279,
+      "learning_rate": 2.901024163098822e-05,
+      "loss": 0.7197,
+      "step": 270
+    },
+    {
+      "epoch": 0.8208955223880597,
+      "grad_norm": 0.8294647932052612,
+      "learning_rate": 2.8956655485305684e-05,
+      "loss": 0.7105,
+      "step": 275
+    },
+    {
+      "epoch": 0.835820895522388,
+      "grad_norm": 1.0282889604568481,
+      "learning_rate": 2.8901708906412793e-05,
+      "loss": 0.7443,
+      "step": 280
+    },
+    {
+      "epoch": 0.8507462686567164,
+      "grad_norm": 0.988297164440155,
+      "learning_rate": 2.8845407250259678e-05,
+      "loss": 0.681,
+      "step": 285
+    },
+    {
+      "epoch": 0.8656716417910447,
+      "grad_norm": 0.9914687275886536,
+      "learning_rate": 2.8787756004883418e-05,
+      "loss": 0.6855,
+      "step": 290
+    },
+    {
+      "epoch": 0.8805970149253731,
+      "grad_norm": 0.8625020980834961,
+      "learning_rate": 2.8728760789873096e-05,
+      "loss": 0.6977,
+      "step": 295
+    },
+    {
+      "epoch": 0.8955223880597015,
+      "grad_norm": 0.9143536686897278,
+      "learning_rate": 2.8668427355822036e-05,
+      "loss": 0.6866,
+      "step": 300
+    },
+    {
+      "epoch": 0.9104477611940298,
+      "grad_norm": 0.9402806162834167,
+      "learning_rate": 2.860676158376724e-05,
+      "loss": 0.6707,
+      "step": 305
+    },
+    {
+      "epoch": 0.9253731343283582,
+      "grad_norm": 0.9718786478042603,
+      "learning_rate": 2.8543769484616136e-05,
+      "loss": 0.7404,
+      "step": 310
+    },
+    {
+      "epoch": 0.9402985074626866,
+      "grad_norm": 1.1747087240219116,
+      "learning_rate": 2.8479457198560684e-05,
+      "loss": 0.6379,
+      "step": 315
+    },
+    {
+      "epoch": 0.9552238805970149,
+      "grad_norm": 1.171994924545288,
+      "learning_rate": 2.841383099447883e-05,
+      "loss": 0.693,
+      "step": 320
+    },
+    {
+      "epoch": 0.9701492537313433,
+      "grad_norm": 1.037919044494629,
+      "learning_rate": 2.8346897269323456e-05,
+      "loss": 0.6481,
+      "step": 325
+    },
+    {
+      "epoch": 0.9850746268656716,
+      "grad_norm": 1.0355802774429321,
+      "learning_rate": 2.827866254749882e-05,
+      "loss": 0.632,
+      "step": 330
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.9163682460784912,
+      "learning_rate": 2.8209133480224612e-05,
+      "loss": 0.6208,
+      "step": 335
+    },
+    {
+      "epoch": 1.0149253731343284,
+      "grad_norm": 1.093385934829712,
+      "learning_rate": 2.81383168448876e-05,
+      "loss": 0.6189,
+      "step": 340
+    },
+    {
+      "epoch": 1.0298507462686568,
+      "grad_norm": 1.2190536260604858,
+      "learning_rate": 2.8066219544381002e-05,
+      "loss": 0.5895,
+      "step": 345
+    },
+    {
+      "epoch": 1.044776119402985,
+      "grad_norm": 0.9977616667747498,
+      "learning_rate": 2.7992848606431635e-05,
+      "loss": 0.5669,
+      "step": 350
+    },
+    {
+      "epoch": 1.0597014925373134,
+      "grad_norm": 1.1409471035003662,
+      "learning_rate": 2.7918211182914864e-05,
+      "loss": 0.5753,
+      "step": 355
+    },
+    {
+      "epoch": 1.0746268656716418,
+      "grad_norm": 1.215071439743042,
+      "learning_rate": 2.7842314549157505e-05,
+      "loss": 0.5559,
+      "step": 360
+    },
+    {
+      "epoch": 1.0895522388059702,
+      "grad_norm": 1.0842992067337036,
+      "learning_rate": 2.7765166103228607e-05,
+      "loss": 0.5272,
+      "step": 365
+    },
+    {
+      "epoch": 1.1044776119402986,
+      "grad_norm": 1.1455897092819214,
+      "learning_rate": 2.7686773365218354e-05,
+      "loss": 0.5764,
+      "step": 370
+    },
+    {
+      "epoch": 1.1194029850746268,
+      "grad_norm": 0.9704266786575317,
+      "learning_rate": 2.7607143976505047e-05,
+      "loss": 0.5814,
+      "step": 375
+    },
+    {
+      "epoch": 1.1343283582089552,
+      "grad_norm": 1.0804754495620728,
+      "learning_rate": 2.7526285699010223e-05,
+      "loss": 0.5184,
+      "step": 380
+    },
+    {
+      "epoch": 1.1492537313432836,
+      "grad_norm": 0.9866484999656677,
+      "learning_rate": 2.7444206414442095e-05,
+      "loss": 0.5613,
+      "step": 385
+    },
+    {
+      "epoch": 1.164179104477612,
+      "grad_norm": 0.930899441242218,
+      "learning_rate": 2.7360914123527242e-05,
+      "loss": 0.4945,
+      "step": 390
+    },
+    {
+      "epoch": 1.1791044776119404,
+      "grad_norm": 1.0647799968719482,
+      "learning_rate": 2.7276416945230763e-05,
+      "loss": 0.5209,
+      "step": 395
+    },
+    {
+      "epoch": 1.1940298507462686,
+      "grad_norm": 1.0100319385528564,
+      "learning_rate": 2.719072311596486e-05,
+      "loss": 0.5011,
+      "step": 400
+    },
+    {
+      "epoch": 1.208955223880597,
+      "grad_norm": 1.1145292520523071,
+      "learning_rate": 2.710384098878599e-05,
+      "loss": 0.528,
+      "step": 405
+    },
+    {
+      "epoch": 1.2238805970149254,
+      "grad_norm": 1.0882689952850342,
+      "learning_rate": 2.7015779032580654e-05,
+      "loss": 0.4835,
+      "step": 410
+    },
+    {
+      "epoch": 1.2388059701492538,
+      "grad_norm": 1.1221665143966675,
+      "learning_rate": 2.692654583123988e-05,
+      "loss": 0.5101,
+      "step": 415
+    },
+    {
+      "epoch": 1.2537313432835822,
+      "grad_norm": 1.1663085222244263,
+      "learning_rate": 2.6836150082822498e-05,
+      "loss": 0.5282,
+      "step": 420
+    },
+    {
+      "epoch": 1.2686567164179103,
+      "grad_norm": 1.2804216146469116,
+      "learning_rate": 2.674460059870731e-05,
+      "loss": 0.4894,
+      "step": 425
+    },
+    {
+      "epoch": 1.2835820895522387,
+      "grad_norm": 1.0360511541366577,
+      "learning_rate": 2.665190630273418e-05,
+      "loss": 0.5039,
+      "step": 430
+    },
+    {
+      "epoch": 1.2985074626865671,
+      "grad_norm": 1.1996111869812012,
+      "learning_rate": 2.6558076230334178e-05,
+      "loss": 0.5003,
+      "step": 435
+    },
+    {
+      "epoch": 1.3134328358208955,
+      "grad_norm": 1.0713297128677368,
+      "learning_rate": 2.6463119527648864e-05,
+      "loss": 0.5172,
+      "step": 440
+    },
+    {
+      "epoch": 1.328358208955224,
+      "grad_norm": 1.0644522905349731,
+      "learning_rate": 2.636704545063872e-05,
+      "loss": 0.4664,
+      "step": 445
+    },
+    {
+      "epoch": 1.3432835820895521,
+      "grad_norm": 1.044643759727478,
+      "learning_rate": 2.6269863364180978e-05,
+      "loss": 0.4551,
+      "step": 450
+    },
+    {
+      "epoch": 1.3582089552238805,
+      "grad_norm": 1.1682442426681519,
+      "learning_rate": 2.617158274115673e-05,
+      "loss": 0.4607,
+      "step": 455
+    },
+    {
+      "epoch": 1.373134328358209,
+      "grad_norm": 1.1079002618789673,
+      "learning_rate": 2.6072213161527563e-05,
+      "loss": 0.4773,
+      "step": 460
+    },
+    {
+      "epoch": 1.3880597014925373,
+      "grad_norm": 1.1027525663375854,
+      "learning_rate": 2.5971764311401755e-05,
+      "loss": 0.4478,
+      "step": 465
+    },
+    {
+      "epoch": 1.4029850746268657,
+      "grad_norm": 1.1082680225372314,
+      "learning_rate": 2.587024598209012e-05,
+      "loss": 0.4771,
+      "step": 470
+    },
+    {
+      "epoch": 1.417910447761194,
+      "grad_norm": 1.2115520238876343,
+      "learning_rate": 2.5767668069151566e-05,
+      "loss": 0.4897,
+      "step": 475
+    },
+    {
+      "epoch": 1.4328358208955223,
+      "grad_norm": 1.154123306274414,
+      "learning_rate": 2.5664040571428548e-05,
+      "loss": 0.4519,
+      "step": 480
+    },
+    {
+      "epoch": 1.4477611940298507,
+      "grad_norm": 1.2055610418319702,
+      "learning_rate": 2.555937359007243e-05,
+      "loss": 0.4554,
+      "step": 485
+    },
+    {
+      "epoch": 1.462686567164179,
+      "grad_norm": 1.02617609500885,
+      "learning_rate": 2.5453677327558825e-05,
+      "loss": 0.4553,
+      "step": 490
+    },
+    {
+      "epoch": 1.4776119402985075,
+      "grad_norm": 1.068691611289978,
+      "learning_rate": 2.534696208669316e-05,
+      "loss": 0.4251,
+      "step": 495
+    },
+    {
+      "epoch": 1.4925373134328357,
+      "grad_norm": 1.0976392030715942,
+      "learning_rate": 2.5239238269606367e-05,
+      "loss": 0.4592,
+      "step": 500
+    },
+    {
+      "epoch": 1.5074626865671643,
+      "grad_norm": 1.240228295326233,
+      "learning_rate": 2.5130516376740936e-05,
+      "loss": 0.4446,
+      "step": 505
+    },
+    {
+      "epoch": 1.5223880597014925,
+      "grad_norm": 1.0058823823928833,
+      "learning_rate": 2.502080700582738e-05,
+      "loss": 0.4171,
+      "step": 510
+    },
+    {
+      "epoch": 1.537313432835821,
+      "grad_norm": 1.0588983297348022,
+      "learning_rate": 2.491012085085122e-05,
+      "loss": 0.4509,
+      "step": 515
+    },
+    {
+      "epoch": 1.5522388059701493,
+      "grad_norm": 1.0546432733535767,
+      "learning_rate": 2.4798468701010592e-05,
+      "loss": 0.4383,
+      "step": 520
+    },
+    {
+      "epoch": 1.5671641791044775,
+      "grad_norm": 1.0632411241531372,
+      "learning_rate": 2.4685861439664513e-05,
+      "loss": 0.4136,
+      "step": 525
+    },
+    {
+      "epoch": 1.582089552238806,
+      "grad_norm": 1.0170810222625732,
+      "learning_rate": 2.45723100432721e-05,
+      "loss": 0.4165,
+      "step": 530
+    },
+    {
+      "epoch": 1.5970149253731343,
+      "grad_norm": 1.1334420442581177,
+      "learning_rate": 2.4457825580322564e-05,
+      "loss": 0.4142,
+      "step": 535
+    },
+    {
+      "epoch": 1.6119402985074627,
+      "grad_norm": 1.2818471193313599,
+      "learning_rate": 2.4342419210256327e-05,
+      "loss": 0.4279,
+      "step": 540
+    },
+    {
+      "epoch": 1.626865671641791,
+      "grad_norm": 1.061077356338501,
+      "learning_rate": 2.4226102182377263e-05,
+      "loss": 0.3976,
+      "step": 545
+    },
+    {
+      "epoch": 1.6417910447761193,
+      "grad_norm": 1.2361209392547607,
+      "learning_rate": 2.4108885834756144e-05,
+      "loss": 0.3832,
+      "step": 550
+    },
+    {
+      "epoch": 1.6567164179104479,
+      "grad_norm": 1.0615969896316528,
+      "learning_rate": 2.399078159312546e-05,
+      "loss": 0.4109,
+      "step": 555
+    },
+    {
+      "epoch": 1.671641791044776,
+      "grad_norm": 1.1073319911956787,
+      "learning_rate": 2.38718009697657e-05,
+      "loss": 0.4479,
+      "step": 560
+    },
+    {
+      "epoch": 1.6865671641791045,
+      "grad_norm": 1.0212117433547974,
+      "learning_rate": 2.3751955562383184e-05,
+      "loss": 0.3754,
+      "step": 565
+    },
+    {
+      "epoch": 1.7014925373134329,
+      "grad_norm": 1.1604119539260864,
+      "learning_rate": 2.3631257052979543e-05,
+      "loss": 0.3981,
+      "step": 570
+    },
+    {
+      "epoch": 1.716417910447761,
+      "grad_norm": 1.202643632888794,
+      "learning_rate": 2.3509717206713047e-05,
+      "loss": 0.3882,
+      "step": 575
+    },
+    {
+      "epoch": 1.7313432835820897,
+      "grad_norm": 1.1154124736785889,
+      "learning_rate": 2.3387347870751768e-05,
+      "loss": 0.4,
+      "step": 580
+    },
+    {
+      "epoch": 1.7462686567164178,
+      "grad_norm": 1.0910370349884033,
+      "learning_rate": 2.3264160973118772e-05,
+      "loss": 0.3797,
+      "step": 585
+    },
+    {
+      "epoch": 1.7611940298507462,
+      "grad_norm": 1.1347483396530151,
+      "learning_rate": 2.314016852152943e-05,
+      "loss": 0.3952,
+      "step": 590
+    },
+    {
+      "epoch": 1.7761194029850746,
+      "grad_norm": 1.3554637432098389,
+      "learning_rate": 2.301538260222096e-05,
+      "loss": 0.4012,
+      "step": 595
+    },
+    {
+      "epoch": 1.7910447761194028,
+      "grad_norm": 1.1606727838516235,
+      "learning_rate": 2.2889815378774316e-05,
+      "loss": 0.408,
+      "step": 600
+    },
+    {
+      "epoch": 1.8059701492537314,
+      "grad_norm": 1.2994942665100098,
+      "learning_rate": 2.2763479090928535e-05,
+      "loss": 0.3578,
+      "step": 605
+    },
+    {
+      "epoch": 1.8208955223880596,
+      "grad_norm": 1.1260433197021484,
+      "learning_rate": 2.263638605338765e-05,
+      "loss": 0.3824,
+      "step": 610
+    },
+    {
+      "epoch": 1.835820895522388,
+      "grad_norm": 1.2883944511413574,
+      "learning_rate": 2.2508548654620317e-05,
+      "loss": 0.3669,
+      "step": 615
+    },
+    {
+      "epoch": 1.8507462686567164,
+      "grad_norm": 1.1746373176574707,
+      "learning_rate": 2.237997935565225e-05,
+      "loss": 0.3918,
+      "step": 620
+    },
+    {
+      "epoch": 1.8656716417910446,
+      "grad_norm": 8.744837760925293,
+      "learning_rate": 2.2250690688851556e-05,
+      "loss": 0.3819,
+      "step": 625
+    },
+    {
+      "epoch": 1.8805970149253732,
+      "grad_norm": 1.2316313982009888,
+      "learning_rate": 2.2120695256707158e-05,
+      "loss": 0.3576,
+      "step": 630
+    },
+    {
+      "epoch": 1.8955223880597014,
+      "grad_norm": 1.0147652626037598,
+      "learning_rate": 2.1990005730600332e-05,
+      "loss": 0.369,
+      "step": 635
+    },
+    {
+      "epoch": 1.9104477611940298,
+      "grad_norm": 1.0189952850341797,
+      "learning_rate": 2.1858634849569578e-05,
+      "loss": 0.4188,
+      "step": 640
+    },
+    {
+      "epoch": 1.9253731343283582,
+      "grad_norm": 1.1541681289672852,
+      "learning_rate": 2.1726595419068872e-05,
+      "loss": 0.3433,
+      "step": 645
+    },
+    {
+      "epoch": 1.9402985074626866,
+      "grad_norm": 1.0665652751922607,
+      "learning_rate": 2.1593900309719437e-05,
+      "loss": 0.3466,
+      "step": 650
+    },
+    {
+      "epoch": 1.955223880597015,
+      "grad_norm": 1.1769914627075195,
+      "learning_rate": 2.146056245605518e-05,
+      "loss": 0.3296,
+      "step": 655
+    },
+    {
+      "epoch": 1.9701492537313432,
+      "grad_norm": 1.0341894626617432,
+      "learning_rate": 2.132659485526189e-05,
+      "loss": 0.3264,
+      "step": 660
+    },
+    {
+      "epoch": 1.9850746268656716,
+      "grad_norm": 1.1811909675598145,
+      "learning_rate": 2.1192010565910326e-05,
+      "loss": 0.3394,
+      "step": 665
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.179243803024292,
+      "learning_rate": 2.1056822706683333e-05,
+      "loss": 0.319,
+      "step": 670
+    },
+    {
+      "epoch": 2.014925373134328,
+      "grad_norm": 1.177918791770935,
+      "learning_rate": 2.0921044455097076e-05,
+      "loss": 0.2634,
+      "step": 675
+    },
+    {
+      "epoch": 2.029850746268657,
+      "grad_norm": 1.1220797300338745,
+      "learning_rate": 2.0784689046216568e-05,
+      "loss": 0.2622,
+      "step": 680
+    },
+    {
+      "epoch": 2.044776119402985,
+      "grad_norm": 1.0185974836349487,
+      "learning_rate": 2.0647769771365564e-05,
+      "loss": 0.2984,
+      "step": 685
+    },
+    {
+      "epoch": 2.0597014925373136,
+      "grad_norm": 1.002383828163147,
+      "learning_rate": 2.0510299976831002e-05,
+      "loss": 0.2555,
+      "step": 690
+    },
+    {
+      "epoch": 2.074626865671642,
+      "grad_norm": 1.3278491497039795,
+      "learning_rate": 2.0372293062562032e-05,
+      "loss": 0.2591,
+      "step": 695
+    },
+    {
+      "epoch": 2.08955223880597,
+      "grad_norm": 1.1084357500076294,
+      "learning_rate": 2.0233762480863876e-05,
+      "loss": 0.2784,
+      "step": 700
+    },
+    {
+      "epoch": 2.1044776119402986,
+      "grad_norm": 1.2520588636398315,
+      "learning_rate": 2.0094721735086535e-05,
+      "loss": 0.2704,
+      "step": 705
+    },
+    {
+      "epoch": 2.1194029850746268,
+      "grad_norm": 1.1068615913391113,
+      "learning_rate": 1.9955184378308575e-05,
+      "loss": 0.2859,
+      "step": 710
+    },
+    {
+      "epoch": 2.1343283582089554,
+      "grad_norm": 1.205536127090454,
+      "learning_rate": 1.9815164012015994e-05,
+      "loss": 0.2546,
+      "step": 715
+    },
+    {
+      "epoch": 2.1492537313432836,
+      "grad_norm": 1.0301965475082397,
+      "learning_rate": 1.9674674284776436e-05,
+      "loss": 0.2552,
+      "step": 720
+    },
+    {
+      "epoch": 2.1641791044776117,
+      "grad_norm": 1.3132309913635254,
+      "learning_rate": 1.9533728890908772e-05,
+      "loss": 0.2383,
+      "step": 725
+    },
+    {
+      "epoch": 2.1791044776119404,
+      "grad_norm": 1.168063759803772,
+      "learning_rate": 1.9392341569148254e-05,
+      "loss": 0.2681,
+      "step": 730
+    },
+    {
+      "epoch": 2.1940298507462686,
+      "grad_norm": 1.1290392875671387,
+      "learning_rate": 1.9250526101307318e-05,
+      "loss": 0.267,
+      "step": 735
+    },
+    {
+      "epoch": 2.208955223880597,
+      "grad_norm": 1.1353788375854492,
+      "learning_rate": 1.9108296310932168e-05,
+      "loss": 0.22,
+      "step": 740
+    },
+    {
+      "epoch": 2.2238805970149254,
+      "grad_norm": 1.0472898483276367,
+      "learning_rate": 1.896566606195537e-05,
+      "loss": 0.2552,
+      "step": 745
+    },
+    {
+      "epoch": 2.2388059701492535,
+      "grad_norm": 1.207857370376587,
+      "learning_rate": 1.8822649257344414e-05,
+      "loss": 0.2515,
+      "step": 750
+    },
+    {
+      "epoch": 2.253731343283582,
+      "grad_norm": 1.133759617805481,
+      "learning_rate": 1.867925983774653e-05,
+      "loss": 0.2708,
+      "step": 755
+    },
+    {
+      "epoch": 2.2686567164179103,
+      "grad_norm": 1.4060381650924683,
+      "learning_rate": 1.853551178012982e-05,
+      "loss": 0.2415,
+      "step": 760
+    },
+    {
+      "epoch": 2.283582089552239,
+      "grad_norm": 1.005429744720459,
+      "learning_rate": 1.8391419096420813e-05,
+      "loss": 0.2294,
+      "step": 765
+    },
+    {
+      "epoch": 2.298507462686567,
+      "grad_norm": 1.328147530555725,
+      "learning_rate": 1.8246995832138703e-05,
+      "loss": 0.2498,
+      "step": 770
+    },
+    {
+      "epoch": 2.3134328358208958,
+      "grad_norm": 1.0649443864822388,
+      "learning_rate": 1.8102256065026202e-05,
+      "loss": 0.2203,
+      "step": 775
+    },
+    {
+      "epoch": 2.328358208955224,
+      "grad_norm": 1.0904861688613892,
+      "learning_rate": 1.7957213903677317e-05,
+      "loss": 0.2401,
+      "step": 780
+    },
+    {
+      "epoch": 2.343283582089552,
+      "grad_norm": 1.0852935314178467,
+      "learning_rate": 1.781188348616213e-05,
+      "loss": 0.2464,
+      "step": 785
+    },
+    {
+      "epoch": 2.3582089552238807,
+      "grad_norm": 1.0837558507919312,
+      "learning_rate": 1.7666278978648644e-05,
+      "loss": 0.2319,
+      "step": 790
+    },
+    {
+      "epoch": 2.373134328358209,
+      "grad_norm": 1.2392449378967285,
+      "learning_rate": 1.752041457402195e-05,
+      "loss": 0.2349,
+      "step": 795
+    },
+    {
+      "epoch": 2.388059701492537,
+      "grad_norm": 1.0948951244354248,
+      "learning_rate": 1.7374304490500776e-05,
+      "loss": 0.2366,
+      "step": 800
+    },
+    {
+      "epoch": 2.4029850746268657,
+      "grad_norm": 1.1904592514038086,
+      "learning_rate": 1.722796297025152e-05,
+      "loss": 0.2393,
+      "step": 805
+    },
+    {
+      "epoch": 2.417910447761194,
+      "grad_norm": 1.0641899108886719,
+      "learning_rate": 1.708140427800004e-05,
+      "loss": 0.2044,
+      "step": 810
+    },
+    {
+      "epoch": 2.4328358208955225,
+      "grad_norm": 1.12578284740448,
+      "learning_rate": 1.6934642699641144e-05,
+      "loss": 0.2435,
+      "step": 815
+    },
+    {
+      "epoch": 2.4477611940298507,
+      "grad_norm": 1.0488451719284058,
+      "learning_rate": 1.6787692540846092e-05,
+      "loss": 0.2272,
+      "step": 820
+    },
+    {
+      "epoch": 2.4626865671641793,
+      "grad_norm": 1.061360239982605,
+      "learning_rate": 1.664056812566812e-05,
+      "loss": 0.2476,
+      "step": 825
+    },
+    {
+      "epoch": 2.4776119402985075,
+      "grad_norm": 1.2904311418533325,
+      "learning_rate": 1.649328379514622e-05,
+      "loss": 0.261,
+      "step": 830
+    },
+    {
+      "epoch": 2.4925373134328357,
+      "grad_norm": 1.061362385749817,
+      "learning_rate": 1.6345853905907223e-05,
+      "loss": 0.2309,
+      "step": 835
+    },
+    {
+      "epoch": 2.5074626865671643,
+      "grad_norm": 1.009899616241455,
+      "learning_rate": 1.6198292828766382e-05,
+      "loss": 0.2142,
+      "step": 840
+    },
+    {
+      "epoch": 2.5223880597014925,
+      "grad_norm": 1.2374935150146484,
+      "learning_rate": 1.6050614947326568e-05,
+      "loss": 0.2337,
+      "step": 845
+    },
+    {
+      "epoch": 2.5373134328358207,
+      "grad_norm": 1.1741938591003418,
+      "learning_rate": 1.5902834656576223e-05,
+      "loss": 0.2342,
+      "step": 850
+    },
+    {
+      "epoch": 2.5522388059701493,
+      "grad_norm": 1.0697554349899292,
+      "learning_rate": 1.5754966361486182e-05,
+      "loss": 0.2045,
+      "step": 855
+    },
+    {
+      "epoch": 2.5671641791044775,
+      "grad_norm": 1.1109567880630493,
+      "learning_rate": 1.5607024475605575e-05,
+      "loss": 0.2097,
+      "step": 860
+    },
+    {
+      "epoch": 2.582089552238806,
+      "grad_norm": 1.240860939025879,
+      "learning_rate": 1.5459023419656824e-05,
+      "loss": 0.2262,
+      "step": 865
+    },
+    {
+      "epoch": 2.5970149253731343,
+      "grad_norm": 1.0825039148330688,
+      "learning_rate": 1.5310977620129992e-05,
+      "loss": 0.2083,
+      "step": 870
+    },
+    {
+      "epoch": 2.611940298507463,
+      "grad_norm": 1.149062156677246,
+      "learning_rate": 1.5162901507876552e-05,
+      "loss": 0.2022,
+      "step": 875
+    },
+    {
+      "epoch": 2.626865671641791,
+      "grad_norm": 1.0815327167510986,
+      "learning_rate": 1.5014809516702727e-05,
+      "loss": 0.2267,
+      "step": 880
+    },
+    {
+      "epoch": 2.6417910447761193,
+      "grad_norm": 1.0375422239303589,
+      "learning_rate": 1.4866716081962542e-05,
+      "loss": 0.1734,
+      "step": 885
+    },
+    {
+      "epoch": 2.656716417910448,
+      "grad_norm": 1.094551682472229,
+      "learning_rate": 1.471863563915073e-05,
+      "loss": 0.2004,
+      "step": 890
+    },
+    {
+      "epoch": 2.671641791044776,
+      "grad_norm": 1.1885653734207153,
+      "learning_rate": 1.4570582622495653e-05,
+      "loss": 0.2025,
+      "step": 895
+    },
+    {
+      "epoch": 2.6865671641791042,
+      "grad_norm": 1.1636579036712646,
+      "learning_rate": 1.4422571463552267e-05,
+      "loss": 0.2116,
+      "step": 900
+    },
+    {
+      "epoch": 2.701492537313433,
+      "grad_norm": 1.226163387298584,
+      "learning_rate": 1.4274616589795424e-05,
+      "loss": 0.2063,
+      "step": 905
+    },
+    {
+      "epoch": 2.716417910447761,
+      "grad_norm": 1.2392303943634033,
+      "learning_rate": 1.412673242321356e-05,
+      "loss": 0.1976,
+      "step": 910
+    },
+    {
+      "epoch": 2.7313432835820897,
+      "grad_norm": 1.0528442859649658,
+      "learning_rate": 1.3978933378902877e-05,
+      "loss": 0.2086,
+      "step": 915
+    },
+    {
+      "epoch": 2.746268656716418,
+      "grad_norm": 1.0389113426208496,
+      "learning_rate": 1.3831233863662228e-05,
+      "loss": 0.1652,
+      "step": 920
+    },
+    {
+      "epoch": 2.7611940298507465,
+      "grad_norm": 1.0428828001022339,
+      "learning_rate": 1.368364827458883e-05,
+      "loss": 0.1931,
+      "step": 925
+    },
+    {
+      "epoch": 2.7761194029850746,
+      "grad_norm": 1.058120608329773,
+      "learning_rate": 1.3536190997674859e-05,
+      "loss": 0.1734,
+      "step": 930
+    },
+    {
+      "epoch": 2.791044776119403,
+      "grad_norm": 1.2238277196884155,
+      "learning_rate": 1.3388876406405206e-05,
+      "loss": 0.2117,
+      "step": 935
+    },
+    {
+      "epoch": 2.8059701492537314,
+      "grad_norm": 1.2835285663604736,
+      "learning_rate": 1.324171886035638e-05,
+      "loss": 0.1802,
+      "step": 940
+    },
+    {
+      "epoch": 2.8208955223880596,
+      "grad_norm": 1.141916036605835,
+      "learning_rate": 1.3094732703796819e-05,
+      "loss": 0.1846,
+      "step": 945
+    },
+    {
+      "epoch": 2.835820895522388,
+      "grad_norm": 1.187520146369934,
+      "learning_rate": 1.2947932264288677e-05,
+      "loss": 0.2056,
+      "step": 950
+    },
+    {
+      "epoch": 2.8507462686567164,
+      "grad_norm": 1.2382432222366333,
+      "learning_rate": 1.280133185129122e-05,
+      "loss": 0.1951,
+      "step": 955
+    },
+    {
+      "epoch": 2.8656716417910446,
+      "grad_norm": 1.0451785326004028,
+      "learning_rate": 1.2654945754766007e-05,
+      "loss": 0.1759,
+      "step": 960
+    },
+    {
+      "epoch": 2.8805970149253732,
+      "grad_norm": 1.2064586877822876,
+      "learning_rate": 1.2508788243783997e-05,
+      "loss": 0.2066,
+      "step": 965
+    },
+    {
+      "epoch": 2.8955223880597014,
+      "grad_norm": 1.071471929550171,
+      "learning_rate": 1.236287356513461e-05,
+      "loss": 0.2146,
+      "step": 970
+    },
+    {
+      "epoch": 2.91044776119403,
+      "grad_norm": 1.2358578443527222,
+      "learning_rate": 1.2217215941937061e-05,
+      "loss": 0.1642,
+      "step": 975
+    },
+    {
+      "epoch": 2.925373134328358,
+      "grad_norm": 1.0519459247589111,
+      "learning_rate": 1.2071829572253926e-05,
+      "loss": 0.1688,
+      "step": 980
+    },
+    {
+      "epoch": 2.9402985074626864,
+      "grad_norm": 1.163450837135315,
+      "learning_rate": 1.1926728627707176e-05,
+      "loss": 0.189,
+      "step": 985
+    },
+    {
+      "epoch": 2.955223880597015,
+      "grad_norm": 1.1002942323684692,
+      "learning_rate": 1.1781927252096824e-05,
+      "loss": 0.1758,
+      "step": 990
+    },
+    {
+      "epoch": 2.970149253731343,
+      "grad_norm": 1.026760220527649,
+      "learning_rate": 1.1637439560022196e-05,
+      "loss": 0.188,
+      "step": 995
+    },
+    {
+      "epoch": 2.9850746268656714,
+      "grad_norm": 1.1044259071350098,
+      "learning_rate": 1.149327963550615e-05,
+      "loss": 0.1618,
+      "step": 1000
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.0454556941986084,
+      "learning_rate": 1.1349461530622203e-05,
+      "loss": 0.1607,
+      "step": 1005
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1675,
+  "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": 1.482127184633004e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

2_128_e5_3e-5/checkpoint-1005/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecfa82b92b45b87d589ccd5ed731a54d01e94a8d5dc857ed4c36d4edefc4970a
+size 7736

2_128_e5_3e-5/checkpoint-1005/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)

2_128_e5_3e-5/checkpoint-1340/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

2_128_e5_3e-5/checkpoint-1340/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": [
+    "down_proj",
+    "v_proj",
+    "k_proj",
+    "o_proj",
+    "up_proj",
+    "q_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

2_128_e5_3e-5/checkpoint-1340/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5de7a881d98b07c708f9677918f654e30d2b7426761c69027bdb6c1948592afc
+size 791751704

2_128_e5_3e-5/checkpoint-1340/latest

@@ -0,0 +1 @@
+global_step1340

2_128_e5_3e-5/checkpoint-1340/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02ac7cf91ed4354f33cfcea9647396eaba951378ce815200f1b3445f6ef87201
+size 15920

2_128_e5_3e-5/checkpoint-1340/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97e361b983afd6084b4890a59c4662fcfc942011a4cd4dab323597c246de75d9
+size 15920

2_128_e5_3e-5/checkpoint-1340/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd27cb76747e00c26e754e2d4f4ed7ca81261dcf859e359a9203008d97b7cd85
+size 15920

2_128_e5_3e-5/checkpoint-1340/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00e84e858f26107d7323b46cfdfec5e95240d07b8d2d1e773fbf46b1c9bac66b
+size 15920

2_128_e5_3e-5/checkpoint-1340/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49c848c7a3347bf462a906cc6ecbc6d9b13aa2b5ef0324776d755e8d8f5bb226
+size 15920

2_128_e5_3e-5/checkpoint-1340/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a8515920ddc3cb26f400f7c323fe657a18c201eef16d6d149e41cb6a4557e7e
+size 15920

2_128_e5_3e-5/checkpoint-1340/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50ca51a290b614a99c0f89f914cde009de38da58f26c7c8e43da3c666dc79993
+size 15920

2_128_e5_3e-5/checkpoint-1340/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2beaf9212dc089d639660b21a48d66613b8de37c102e9f2218d3525999553fa
+size 15920

2_128_e5_3e-5/checkpoint-1340/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d531cda4036eea8dcab35baf4c34805a2cb5c514d32f9afbfd43053b265b4e8d
+size 1064

2_128_e5_3e-5/checkpoint-1340/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
+  }
+}

2_128_e5_3e-5/checkpoint-1340/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
+}

2_128_e5_3e-5/checkpoint-1340/trainer_state.json

@@ -0,0 +1,1910 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 1340,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.014925373134328358,
+      "grad_norm": 1.2194169759750366,
+      "learning_rate": 1.4285714285714286e-06,
+      "loss": 1.3181,
+      "step": 5
+    },
+    {
+      "epoch": 0.029850746268656716,
+      "grad_norm": 0.9168437123298645,
+      "learning_rate": 3.2142857142857143e-06,
+      "loss": 1.3037,
+      "step": 10
+    },
+    {
+      "epoch": 0.04477611940298507,
+      "grad_norm": 0.6358347535133362,
+      "learning_rate": 4.9999999999999996e-06,
+      "loss": 1.3083,
+      "step": 15
+    },
+    {
+      "epoch": 0.05970149253731343,
+      "grad_norm": 1.2521260976791382,
+      "learning_rate": 6.785714285714286e-06,
+      "loss": 1.2654,
+      "step": 20
+    },
+    {
+      "epoch": 0.07462686567164178,
+      "grad_norm": 0.6080345511436462,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 1.2305,
+      "step": 25
+    },
+    {
+      "epoch": 0.08955223880597014,
+      "grad_norm": 0.4800509214401245,
+      "learning_rate": 1.0357142857142857e-05,
+      "loss": 1.2006,
+      "step": 30
+    },
+    {
+      "epoch": 0.1044776119402985,
+      "grad_norm": 0.4233926236629486,
+      "learning_rate": 1.2142857142857144e-05,
+      "loss": 1.2602,
+      "step": 35
+    },
+    {
+      "epoch": 0.11940298507462686,
+      "grad_norm": 0.5414039492607117,
+      "learning_rate": 1.3928571428571429e-05,
+      "loss": 1.2663,
+      "step": 40
+    },
+    {
+      "epoch": 0.13432835820895522,
+      "grad_norm": 0.5675264000892639,
+      "learning_rate": 1.5714285714285715e-05,
+      "loss": 1.1681,
+      "step": 45
+    },
+    {
+      "epoch": 0.14925373134328357,
+      "grad_norm": 0.4302530884742737,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 1.2069,
+      "step": 50
+    },
+    {
+      "epoch": 0.16417910447761194,
+      "grad_norm": 0.4482235610485077,
+      "learning_rate": 1.928571428571429e-05,
+      "loss": 1.1548,
+      "step": 55
+    },
+    {
+      "epoch": 0.1791044776119403,
+      "grad_norm": 0.5453466773033142,
+      "learning_rate": 2.107142857142857e-05,
+      "loss": 1.151,
+      "step": 60
+    },
+    {
+      "epoch": 0.19402985074626866,
+      "grad_norm": 0.5155040621757507,
+      "learning_rate": 2.2857142857142858e-05,
+      "loss": 1.0988,
+      "step": 65
+    },
+    {
+      "epoch": 0.208955223880597,
+      "grad_norm": 0.5083202719688416,
+      "learning_rate": 2.464285714285714e-05,
+      "loss": 1.0979,
+      "step": 70
+    },
+    {
+      "epoch": 0.22388059701492538,
+      "grad_norm": 0.5227262377738953,
+      "learning_rate": 2.6428571428571428e-05,
+      "loss": 1.1366,
+      "step": 75
+    },
+    {
+      "epoch": 0.23880597014925373,
+      "grad_norm": 0.5148192048072815,
+      "learning_rate": 2.8214285714285714e-05,
+      "loss": 1.1455,
+      "step": 80
+    },
+    {
+      "epoch": 0.2537313432835821,
+      "grad_norm": 0.4345020651817322,
+      "learning_rate": 3e-05,
+      "loss": 1.1615,
+      "step": 85
+    },
+    {
+      "epoch": 0.26865671641791045,
+      "grad_norm": 0.5491366386413574,
+      "learning_rate": 2.999926893308441e-05,
+      "loss": 1.091,
+      "step": 90
+    },
+    {
+      "epoch": 0.2835820895522388,
+      "grad_norm": 0.5117303133010864,
+      "learning_rate": 2.9997075803598805e-05,
+      "loss": 1.1199,
+      "step": 95
+    },
+    {
+      "epoch": 0.29850746268656714,
+      "grad_norm": 0.5951908826828003,
+      "learning_rate": 2.9993420825319784e-05,
+      "loss": 1.0835,
+      "step": 100
+    },
+    {
+      "epoch": 0.31343283582089554,
+      "grad_norm": 0.5786913633346558,
+      "learning_rate": 2.9988304354518498e-05,
+      "loss": 1.0623,
+      "step": 105
+    },
+    {
+      "epoch": 0.3283582089552239,
+      "grad_norm": 0.5928995609283447,
+      "learning_rate": 2.9981726889925956e-05,
+      "loss": 1.0168,
+      "step": 110
+    },
+    {
+      "epoch": 0.34328358208955223,
+      "grad_norm": 0.505250871181488,
+      "learning_rate": 2.997368907268439e-05,
+      "loss": 1.0542,
+      "step": 115
+    },
+    {
+      "epoch": 0.3582089552238806,
+      "grad_norm": 0.5257184505462646,
+      "learning_rate": 2.9964191686284768e-05,
+      "loss": 1.0217,
+      "step": 120
+    },
+    {
+      "epoch": 0.373134328358209,
+      "grad_norm": 0.641715943813324,
+      "learning_rate": 2.995323565649042e-05,
+      "loss": 1.0544,
+      "step": 125
+    },
+    {
+      "epoch": 0.3880597014925373,
+      "grad_norm": 0.6116381883621216,
+      "learning_rate": 2.99408220512468e-05,
+      "loss": 1.0089,
+      "step": 130
+    },
+    {
+      "epoch": 0.40298507462686567,
+      "grad_norm": 0.8415507078170776,
+      "learning_rate": 2.9926952080577392e-05,
+      "loss": 1.0024,
+      "step": 135
+    },
+    {
+      "epoch": 0.417910447761194,
+      "grad_norm": 0.5881995558738708,
+      "learning_rate": 2.991162709646575e-05,
+      "loss": 0.9376,
+      "step": 140
+    },
+    {
+      "epoch": 0.43283582089552236,
+      "grad_norm": 0.6294875144958496,
+      "learning_rate": 2.9894848592723724e-05,
+      "loss": 0.939,
+      "step": 145
+    },
+    {
+      "epoch": 0.44776119402985076,
+      "grad_norm": 0.6948407292366028,
+      "learning_rate": 2.9876618204845835e-05,
+      "loss": 0.9616,
+      "step": 150
+    },
+    {
+      "epoch": 0.4626865671641791,
+      "grad_norm": 0.7214933633804321,
+      "learning_rate": 2.985693770984989e-05,
+      "loss": 0.9484,
+      "step": 155
+    },
+    {
+      "epoch": 0.47761194029850745,
+      "grad_norm": 0.7646703124046326,
+      "learning_rate": 2.9835809026103717e-05,
+      "loss": 0.8907,
+      "step": 160
+    },
+    {
+      "epoch": 0.4925373134328358,
+      "grad_norm": 0.8051431775093079,
+      "learning_rate": 2.98132342131382e-05,
+      "loss": 0.9145,
+      "step": 165
+    },
+    {
+      "epoch": 0.5074626865671642,
+      "grad_norm": 0.8606762290000916,
+      "learning_rate": 2.9789215471446528e-05,
+      "loss": 0.9351,
+      "step": 170
+    },
+    {
+      "epoch": 0.5223880597014925,
+      "grad_norm": 0.6801915168762207,
+      "learning_rate": 2.9763755142269695e-05,
+      "loss": 0.8833,
+      "step": 175
+    },
+    {
+      "epoch": 0.5373134328358209,
+      "grad_norm": 0.7298025488853455,
+      "learning_rate": 2.9736855707368266e-05,
+      "loss": 0.8845,
+      "step": 180
+    },
+    {
+      "epoch": 0.5522388059701493,
+      "grad_norm": 0.7883251309394836,
+      "learning_rate": 2.97085197887805e-05,
+      "loss": 0.8407,
+      "step": 185
+    },
+    {
+      "epoch": 0.5671641791044776,
+      "grad_norm": 0.8319213390350342,
+      "learning_rate": 2.9678750148566748e-05,
+      "loss": 0.8268,
+      "step": 190
+    },
+    {
+      "epoch": 0.582089552238806,
+      "grad_norm": 0.7365439534187317,
+      "learning_rate": 2.9647549688540206e-05,
+      "loss": 0.8979,
+      "step": 195
+    },
+    {
+      "epoch": 0.5970149253731343,
+      "grad_norm": 0.8525620102882385,
+      "learning_rate": 2.9614921449984092e-05,
+      "loss": 0.8602,
+      "step": 200
+    },
+    {
+      "epoch": 0.6119402985074627,
+      "grad_norm": 0.7475487589836121,
+      "learning_rate": 2.958086861335517e-05,
+      "loss": 0.8242,
+      "step": 205
+    },
+    {
+      "epoch": 0.6268656716417911,
+      "grad_norm": 0.9775540828704834,
+      "learning_rate": 2.9545394497973736e-05,
+      "loss": 0.8501,
+      "step": 210
+    },
+    {
+      "epoch": 0.6417910447761194,
+      "grad_norm": 0.7747155427932739,
+      "learning_rate": 2.950850256170007e-05,
+      "loss": 0.8561,
+      "step": 215
+    },
+    {
+      "epoch": 0.6567164179104478,
+      "grad_norm": 0.8525500893592834,
+      "learning_rate": 2.9470196400597385e-05,
+      "loss": 0.7717,
+      "step": 220
+    },
+    {
+      "epoch": 0.6716417910447762,
+      "grad_norm": 0.8092201352119446,
+      "learning_rate": 2.9430479748581286e-05,
+      "loss": 0.8772,
+      "step": 225
+    },
+    {
+      "epoch": 0.6865671641791045,
+      "grad_norm": 0.8191531300544739,
+      "learning_rate": 2.9389356477055807e-05,
+      "loss": 0.8057,
+      "step": 230
+    },
+    {
+      "epoch": 0.7014925373134329,
+      "grad_norm": 0.888316810131073,
+      "learning_rate": 2.934683059453605e-05,
+      "loss": 0.7826,
+      "step": 235
+    },
+    {
+      "epoch": 0.7164179104477612,
+      "grad_norm": 0.8837902545928955,
+      "learning_rate": 2.9302906246257462e-05,
+      "loss": 0.7865,
+      "step": 240
+    },
+    {
+      "epoch": 0.7313432835820896,
+      "grad_norm": 0.7749768495559692,
+      "learning_rate": 2.925758771377174e-05,
+      "loss": 0.7745,
+      "step": 245
+    },
+    {
+      "epoch": 0.746268656716418,
+      "grad_norm": 0.8685063719749451,
+      "learning_rate": 2.9210879414529526e-05,
+      "loss": 0.7487,
+      "step": 250
+    },
+    {
+      "epoch": 0.7611940298507462,
+      "grad_norm": 0.8156796097755432,
+      "learning_rate": 2.9162785901449783e-05,
+      "loss": 0.7301,
+      "step": 255
+    },
+    {
+      "epoch": 0.7761194029850746,
+      "grad_norm": 0.8999550342559814,
+      "learning_rate": 2.9113311862476023e-05,
+      "loss": 0.7865,
+      "step": 260
+    },
+    {
+      "epoch": 0.7910447761194029,
+      "grad_norm": 0.912211537361145,
+      "learning_rate": 2.906246212011931e-05,
+      "loss": 0.7367,
+      "step": 265
+    },
+    {
+      "epoch": 0.8059701492537313,
+      "grad_norm": 0.790409505367279,
+      "learning_rate": 2.901024163098822e-05,
+      "loss": 0.7197,
+      "step": 270
+    },
+    {
+      "epoch": 0.8208955223880597,
+      "grad_norm": 0.8294647932052612,
+      "learning_rate": 2.8956655485305684e-05,
+      "loss": 0.7105,
+      "step": 275
+    },
+    {
+      "epoch": 0.835820895522388,
+      "grad_norm": 1.0282889604568481,
+      "learning_rate": 2.8901708906412793e-05,
+      "loss": 0.7443,
+      "step": 280
+    },
+    {
+      "epoch": 0.8507462686567164,
+      "grad_norm": 0.988297164440155,
+      "learning_rate": 2.8845407250259678e-05,
+      "loss": 0.681,
+      "step": 285
+    },
+    {
+      "epoch": 0.8656716417910447,
+      "grad_norm": 0.9914687275886536,
+      "learning_rate": 2.8787756004883418e-05,
+      "loss": 0.6855,
+      "step": 290
+    },
+    {
+      "epoch": 0.8805970149253731,
+      "grad_norm": 0.8625020980834961,
+      "learning_rate": 2.8728760789873096e-05,
+      "loss": 0.6977,
+      "step": 295
+    },
+    {
+      "epoch": 0.8955223880597015,
+      "grad_norm": 0.9143536686897278,
+      "learning_rate": 2.8668427355822036e-05,
+      "loss": 0.6866,
+      "step": 300
+    },
+    {
+      "epoch": 0.9104477611940298,
+      "grad_norm": 0.9402806162834167,
+      "learning_rate": 2.860676158376724e-05,
+      "loss": 0.6707,
+      "step": 305
+    },
+    {
+      "epoch": 0.9253731343283582,
+      "grad_norm": 0.9718786478042603,
+      "learning_rate": 2.8543769484616136e-05,
+      "loss": 0.7404,
+      "step": 310
+    },
+    {
+      "epoch": 0.9402985074626866,
+      "grad_norm": 1.1747087240219116,
+      "learning_rate": 2.8479457198560684e-05,
+      "loss": 0.6379,
+      "step": 315
+    },
+    {
+      "epoch": 0.9552238805970149,
+      "grad_norm": 1.171994924545288,
+      "learning_rate": 2.841383099447883e-05,
+      "loss": 0.693,
+      "step": 320
+    },
+    {
+      "epoch": 0.9701492537313433,
+      "grad_norm": 1.037919044494629,
+      "learning_rate": 2.8346897269323456e-05,
+      "loss": 0.6481,
+      "step": 325
+    },
+    {
+      "epoch": 0.9850746268656716,
+      "grad_norm": 1.0355802774429321,
+      "learning_rate": 2.827866254749882e-05,
+      "loss": 0.632,
+      "step": 330
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.9163682460784912,
+      "learning_rate": 2.8209133480224612e-05,
+      "loss": 0.6208,
+      "step": 335
+    },
+    {
+      "epoch": 1.0149253731343284,
+      "grad_norm": 1.093385934829712,
+      "learning_rate": 2.81383168448876e-05,
+      "loss": 0.6189,
+      "step": 340
+    },
+    {
+      "epoch": 1.0298507462686568,
+      "grad_norm": 1.2190536260604858,
+      "learning_rate": 2.8066219544381002e-05,
+      "loss": 0.5895,
+      "step": 345
+    },
+    {
+      "epoch": 1.044776119402985,
+      "grad_norm": 0.9977616667747498,
+      "learning_rate": 2.7992848606431635e-05,
+      "loss": 0.5669,
+      "step": 350
+    },
+    {
+      "epoch": 1.0597014925373134,
+      "grad_norm": 1.1409471035003662,
+      "learning_rate": 2.7918211182914864e-05,
+      "loss": 0.5753,
+      "step": 355
+    },
+    {
+      "epoch": 1.0746268656716418,
+      "grad_norm": 1.215071439743042,
+      "learning_rate": 2.7842314549157505e-05,
+      "loss": 0.5559,
+      "step": 360
+    },
+    {
+      "epoch": 1.0895522388059702,
+      "grad_norm": 1.0842992067337036,
+      "learning_rate": 2.7765166103228607e-05,
+      "loss": 0.5272,
+      "step": 365
+    },
+    {
+      "epoch": 1.1044776119402986,
+      "grad_norm": 1.1455897092819214,
+      "learning_rate": 2.7686773365218354e-05,
+      "loss": 0.5764,
+      "step": 370
+    },
+    {
+      "epoch": 1.1194029850746268,
+      "grad_norm": 0.9704266786575317,
+      "learning_rate": 2.7607143976505047e-05,
+      "loss": 0.5814,
+      "step": 375
+    },
+    {
+      "epoch": 1.1343283582089552,
+      "grad_norm": 1.0804754495620728,
+      "learning_rate": 2.7526285699010223e-05,
+      "loss": 0.5184,
+      "step": 380
+    },
+    {
+      "epoch": 1.1492537313432836,
+      "grad_norm": 0.9866484999656677,
+      "learning_rate": 2.7444206414442095e-05,
+      "loss": 0.5613,
+      "step": 385
+    },
+    {
+      "epoch": 1.164179104477612,
+      "grad_norm": 0.930899441242218,
+      "learning_rate": 2.7360914123527242e-05,
+      "loss": 0.4945,
+      "step": 390
+    },
+    {
+      "epoch": 1.1791044776119404,
+      "grad_norm": 1.0647799968719482,
+      "learning_rate": 2.7276416945230763e-05,
+      "loss": 0.5209,
+      "step": 395
+    },
+    {
+      "epoch": 1.1940298507462686,
+      "grad_norm": 1.0100319385528564,
+      "learning_rate": 2.719072311596486e-05,
+      "loss": 0.5011,
+      "step": 400
+    },
+    {
+      "epoch": 1.208955223880597,
+      "grad_norm": 1.1145292520523071,
+      "learning_rate": 2.710384098878599e-05,
+      "loss": 0.528,
+      "step": 405
+    },
+    {
+      "epoch": 1.2238805970149254,
+      "grad_norm": 1.0882689952850342,
+      "learning_rate": 2.7015779032580654e-05,
+      "loss": 0.4835,
+      "step": 410
+    },
+    {
+      "epoch": 1.2388059701492538,
+      "grad_norm": 1.1221665143966675,
+      "learning_rate": 2.692654583123988e-05,
+      "loss": 0.5101,
+      "step": 415
+    },
+    {
+      "epoch": 1.2537313432835822,
+      "grad_norm": 1.1663085222244263,
+      "learning_rate": 2.6836150082822498e-05,
+      "loss": 0.5282,
+      "step": 420
+    },
+    {
+      "epoch": 1.2686567164179103,
+      "grad_norm": 1.2804216146469116,
+      "learning_rate": 2.674460059870731e-05,
+      "loss": 0.4894,
+      "step": 425
+    },
+    {
+      "epoch": 1.2835820895522387,
+      "grad_norm": 1.0360511541366577,
+      "learning_rate": 2.665190630273418e-05,
+      "loss": 0.5039,
+      "step": 430
+    },
+    {
+      "epoch": 1.2985074626865671,
+      "grad_norm": 1.1996111869812012,
+      "learning_rate": 2.6558076230334178e-05,
+      "loss": 0.5003,
+      "step": 435
+    },
+    {
+      "epoch": 1.3134328358208955,
+      "grad_norm": 1.0713297128677368,
+      "learning_rate": 2.6463119527648864e-05,
+      "loss": 0.5172,
+      "step": 440
+    },
+    {
+      "epoch": 1.328358208955224,
+      "grad_norm": 1.0644522905349731,
+      "learning_rate": 2.636704545063872e-05,
+      "loss": 0.4664,
+      "step": 445
+    },
+    {
+      "epoch": 1.3432835820895521,
+      "grad_norm": 1.044643759727478,
+      "learning_rate": 2.6269863364180978e-05,
+      "loss": 0.4551,
+      "step": 450
+    },
+    {
+      "epoch": 1.3582089552238805,
+      "grad_norm": 1.1682442426681519,
+      "learning_rate": 2.617158274115673e-05,
+      "loss": 0.4607,
+      "step": 455
+    },
+    {
+      "epoch": 1.373134328358209,
+      "grad_norm": 1.1079002618789673,
+      "learning_rate": 2.6072213161527563e-05,
+      "loss": 0.4773,
+      "step": 460
+    },
+    {
+      "epoch": 1.3880597014925373,
+      "grad_norm": 1.1027525663375854,
+      "learning_rate": 2.5971764311401755e-05,
+      "loss": 0.4478,
+      "step": 465
+    },
+    {
+      "epoch": 1.4029850746268657,
+      "grad_norm": 1.1082680225372314,
+      "learning_rate": 2.587024598209012e-05,
+      "loss": 0.4771,
+      "step": 470
+    },
+    {
+      "epoch": 1.417910447761194,
+      "grad_norm": 1.2115520238876343,
+      "learning_rate": 2.5767668069151566e-05,
+      "loss": 0.4897,
+      "step": 475
+    },
+    {
+      "epoch": 1.4328358208955223,
+      "grad_norm": 1.154123306274414,
+      "learning_rate": 2.5664040571428548e-05,
+      "loss": 0.4519,
+      "step": 480
+    },
+    {
+      "epoch": 1.4477611940298507,
+      "grad_norm": 1.2055610418319702,
+      "learning_rate": 2.555937359007243e-05,
+      "loss": 0.4554,
+      "step": 485
+    },
+    {
+      "epoch": 1.462686567164179,
+      "grad_norm": 1.02617609500885,
+      "learning_rate": 2.5453677327558825e-05,
+      "loss": 0.4553,
+      "step": 490
+    },
+    {
+      "epoch": 1.4776119402985075,
+      "grad_norm": 1.068691611289978,
+      "learning_rate": 2.534696208669316e-05,
+      "loss": 0.4251,
+      "step": 495
+    },
+    {
+      "epoch": 1.4925373134328357,
+      "grad_norm": 1.0976392030715942,
+      "learning_rate": 2.5239238269606367e-05,
+      "loss": 0.4592,
+      "step": 500
+    },
+    {
+      "epoch": 1.5074626865671643,
+      "grad_norm": 1.240228295326233,
+      "learning_rate": 2.5130516376740936e-05,
+      "loss": 0.4446,
+      "step": 505
+    },
+    {
+      "epoch": 1.5223880597014925,
+      "grad_norm": 1.0058823823928833,
+      "learning_rate": 2.502080700582738e-05,
+      "loss": 0.4171,
+      "step": 510
+    },
+    {
+      "epoch": 1.537313432835821,
+      "grad_norm": 1.0588983297348022,
+      "learning_rate": 2.491012085085122e-05,
+      "loss": 0.4509,
+      "step": 515
+    },
+    {
+      "epoch": 1.5522388059701493,
+      "grad_norm": 1.0546432733535767,
+      "learning_rate": 2.4798468701010592e-05,
+      "loss": 0.4383,
+      "step": 520
+    },
+    {
+      "epoch": 1.5671641791044775,
+      "grad_norm": 1.0632411241531372,
+      "learning_rate": 2.4685861439664513e-05,
+      "loss": 0.4136,
+      "step": 525
+    },
+    {
+      "epoch": 1.582089552238806,
+      "grad_norm": 1.0170810222625732,
+      "learning_rate": 2.45723100432721e-05,
+      "loss": 0.4165,
+      "step": 530
+    },
+    {
+      "epoch": 1.5970149253731343,
+      "grad_norm": 1.1334420442581177,
+      "learning_rate": 2.4457825580322564e-05,
+      "loss": 0.4142,
+      "step": 535
+    },
+    {
+      "epoch": 1.6119402985074627,
+      "grad_norm": 1.2818471193313599,
+      "learning_rate": 2.4342419210256327e-05,
+      "loss": 0.4279,
+      "step": 540
+    },
+    {
+      "epoch": 1.626865671641791,
+      "grad_norm": 1.061077356338501,
+      "learning_rate": 2.4226102182377263e-05,
+      "loss": 0.3976,
+      "step": 545
+    },
+    {
+      "epoch": 1.6417910447761193,
+      "grad_norm": 1.2361209392547607,
+      "learning_rate": 2.4108885834756144e-05,
+      "loss": 0.3832,
+      "step": 550
+    },
+    {
+      "epoch": 1.6567164179104479,
+      "grad_norm": 1.0615969896316528,
+      "learning_rate": 2.399078159312546e-05,
+      "loss": 0.4109,
+      "step": 555
+    },
+    {
+      "epoch": 1.671641791044776,
+      "grad_norm": 1.1073319911956787,
+      "learning_rate": 2.38718009697657e-05,
+      "loss": 0.4479,
+      "step": 560
+    },
+    {
+      "epoch": 1.6865671641791045,
+      "grad_norm": 1.0212117433547974,
+      "learning_rate": 2.3751955562383184e-05,
+      "loss": 0.3754,
+      "step": 565
+    },
+    {
+      "epoch": 1.7014925373134329,
+      "grad_norm": 1.1604119539260864,
+      "learning_rate": 2.3631257052979543e-05,
+      "loss": 0.3981,
+      "step": 570
+    },
+    {
+      "epoch": 1.716417910447761,
+      "grad_norm": 1.202643632888794,
+      "learning_rate": 2.3509717206713047e-05,
+      "loss": 0.3882,
+      "step": 575
+    },
+    {
+      "epoch": 1.7313432835820897,
+      "grad_norm": 1.1154124736785889,
+      "learning_rate": 2.3387347870751768e-05,
+      "loss": 0.4,
+      "step": 580
+    },
+    {
+      "epoch": 1.7462686567164178,
+      "grad_norm": 1.0910370349884033,
+      "learning_rate": 2.3264160973118772e-05,
+      "loss": 0.3797,
+      "step": 585
+    },
+    {
+      "epoch": 1.7611940298507462,
+      "grad_norm": 1.1347483396530151,
+      "learning_rate": 2.314016852152943e-05,
+      "loss": 0.3952,
+      "step": 590
+    },
+    {
+      "epoch": 1.7761194029850746,
+      "grad_norm": 1.3554637432098389,
+      "learning_rate": 2.301538260222096e-05,
+      "loss": 0.4012,
+      "step": 595
+    },
+    {
+      "epoch": 1.7910447761194028,
+      "grad_norm": 1.1606727838516235,
+      "learning_rate": 2.2889815378774316e-05,
+      "loss": 0.408,
+      "step": 600
+    },
+    {
+      "epoch": 1.8059701492537314,
+      "grad_norm": 1.2994942665100098,
+      "learning_rate": 2.2763479090928535e-05,
+      "loss": 0.3578,
+      "step": 605
+    },
+    {
+      "epoch": 1.8208955223880596,
+      "grad_norm": 1.1260433197021484,
+      "learning_rate": 2.263638605338765e-05,
+      "loss": 0.3824,
+      "step": 610
+    },
+    {
+      "epoch": 1.835820895522388,
+      "grad_norm": 1.2883944511413574,
+      "learning_rate": 2.2508548654620317e-05,
+      "loss": 0.3669,
+      "step": 615
+    },
+    {
+      "epoch": 1.8507462686567164,
+      "grad_norm": 1.1746373176574707,
+      "learning_rate": 2.237997935565225e-05,
+      "loss": 0.3918,
+      "step": 620
+    },
+    {
+      "epoch": 1.8656716417910446,
+      "grad_norm": 8.744837760925293,
+      "learning_rate": 2.2250690688851556e-05,
+      "loss": 0.3819,
+      "step": 625
+    },
+    {
+      "epoch": 1.8805970149253732,
+      "grad_norm": 1.2316313982009888,
+      "learning_rate": 2.2120695256707158e-05,
+      "loss": 0.3576,
+      "step": 630
+    },
+    {
+      "epoch": 1.8955223880597014,
+      "grad_norm": 1.0147652626037598,
+      "learning_rate": 2.1990005730600332e-05,
+      "loss": 0.369,
+      "step": 635
+    },
+    {
+      "epoch": 1.9104477611940298,
+      "grad_norm": 1.0189952850341797,
+      "learning_rate": 2.1858634849569578e-05,
+      "loss": 0.4188,
+      "step": 640
+    },
+    {
+      "epoch": 1.9253731343283582,
+      "grad_norm": 1.1541681289672852,
+      "learning_rate": 2.1726595419068872e-05,
+      "loss": 0.3433,
+      "step": 645
+    },
+    {
+      "epoch": 1.9402985074626866,
+      "grad_norm": 1.0665652751922607,
+      "learning_rate": 2.1593900309719437e-05,
+      "loss": 0.3466,
+      "step": 650
+    },
+    {
+      "epoch": 1.955223880597015,
+      "grad_norm": 1.1769914627075195,
+      "learning_rate": 2.146056245605518e-05,
+      "loss": 0.3296,
+      "step": 655
+    },
+    {
+      "epoch": 1.9701492537313432,
+      "grad_norm": 1.0341894626617432,
+      "learning_rate": 2.132659485526189e-05,
+      "loss": 0.3264,
+      "step": 660
+    },
+    {
+      "epoch": 1.9850746268656716,
+      "grad_norm": 1.1811909675598145,
+      "learning_rate": 2.1192010565910326e-05,
+      "loss": 0.3394,
+      "step": 665
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.179243803024292,
+      "learning_rate": 2.1056822706683333e-05,
+      "loss": 0.319,
+      "step": 670
+    },
+    {
+      "epoch": 2.014925373134328,
+      "grad_norm": 1.177918791770935,
+      "learning_rate": 2.0921044455097076e-05,
+      "loss": 0.2634,
+      "step": 675
+    },
+    {
+      "epoch": 2.029850746268657,
+      "grad_norm": 1.1220797300338745,
+      "learning_rate": 2.0784689046216568e-05,
+      "loss": 0.2622,
+      "step": 680
+    },
+    {
+      "epoch": 2.044776119402985,
+      "grad_norm": 1.0185974836349487,
+      "learning_rate": 2.0647769771365564e-05,
+      "loss": 0.2984,
+      "step": 685
+    },
+    {
+      "epoch": 2.0597014925373136,
+      "grad_norm": 1.002383828163147,
+      "learning_rate": 2.0510299976831002e-05,
+      "loss": 0.2555,
+      "step": 690
+    },
+    {
+      "epoch": 2.074626865671642,
+      "grad_norm": 1.3278491497039795,
+      "learning_rate": 2.0372293062562032e-05,
+      "loss": 0.2591,
+      "step": 695
+    },
+    {
+      "epoch": 2.08955223880597,
+      "grad_norm": 1.1084357500076294,
+      "learning_rate": 2.0233762480863876e-05,
+      "loss": 0.2784,
+      "step": 700
+    },
+    {
+      "epoch": 2.1044776119402986,
+      "grad_norm": 1.2520588636398315,
+      "learning_rate": 2.0094721735086535e-05,
+      "loss": 0.2704,
+      "step": 705
+    },
+    {
+      "epoch": 2.1194029850746268,
+      "grad_norm": 1.1068615913391113,
+      "learning_rate": 1.9955184378308575e-05,
+      "loss": 0.2859,
+      "step": 710
+    },
+    {
+      "epoch": 2.1343283582089554,
+      "grad_norm": 1.205536127090454,
+      "learning_rate": 1.9815164012015994e-05,
+      "loss": 0.2546,
+      "step": 715
+    },
+    {
+      "epoch": 2.1492537313432836,
+      "grad_norm": 1.0301965475082397,
+      "learning_rate": 1.9674674284776436e-05,
+      "loss": 0.2552,
+      "step": 720
+    },
+    {
+      "epoch": 2.1641791044776117,
+      "grad_norm": 1.3132309913635254,
+      "learning_rate": 1.9533728890908772e-05,
+      "loss": 0.2383,
+      "step": 725
+    },
+    {
+      "epoch": 2.1791044776119404,
+      "grad_norm": 1.168063759803772,
+      "learning_rate": 1.9392341569148254e-05,
+      "loss": 0.2681,
+      "step": 730
+    },
+    {
+      "epoch": 2.1940298507462686,
+      "grad_norm": 1.1290392875671387,
+      "learning_rate": 1.9250526101307318e-05,
+      "loss": 0.267,
+      "step": 735
+    },
+    {
+      "epoch": 2.208955223880597,
+      "grad_norm": 1.1353788375854492,
+      "learning_rate": 1.9108296310932168e-05,
+      "loss": 0.22,
+      "step": 740
+    },
+    {
+      "epoch": 2.2238805970149254,
+      "grad_norm": 1.0472898483276367,
+      "learning_rate": 1.896566606195537e-05,
+      "loss": 0.2552,
+      "step": 745
+    },
+    {
+      "epoch": 2.2388059701492535,
+      "grad_norm": 1.207857370376587,
+      "learning_rate": 1.8822649257344414e-05,
+      "loss": 0.2515,
+      "step": 750
+    },
+    {
+      "epoch": 2.253731343283582,
+      "grad_norm": 1.133759617805481,
+      "learning_rate": 1.867925983774653e-05,
+      "loss": 0.2708,
+      "step": 755
+    },
+    {
+      "epoch": 2.2686567164179103,
+      "grad_norm": 1.4060381650924683,
+      "learning_rate": 1.853551178012982e-05,
+      "loss": 0.2415,
+      "step": 760
+    },
+    {
+      "epoch": 2.283582089552239,
+      "grad_norm": 1.005429744720459,
+      "learning_rate": 1.8391419096420813e-05,
+      "loss": 0.2294,
+      "step": 765
+    },
+    {
+      "epoch": 2.298507462686567,
+      "grad_norm": 1.328147530555725,
+      "learning_rate": 1.8246995832138703e-05,
+      "loss": 0.2498,
+      "step": 770
+    },
+    {
+      "epoch": 2.3134328358208958,
+      "grad_norm": 1.0649443864822388,
+      "learning_rate": 1.8102256065026202e-05,
+      "loss": 0.2203,
+      "step": 775
+    },
+    {
+      "epoch": 2.328358208955224,
+      "grad_norm": 1.0904861688613892,
+      "learning_rate": 1.7957213903677317e-05,
+      "loss": 0.2401,
+      "step": 780
+    },
+    {
+      "epoch": 2.343283582089552,
+      "grad_norm": 1.0852935314178467,
+      "learning_rate": 1.781188348616213e-05,
+      "loss": 0.2464,
+      "step": 785
+    },
+    {
+      "epoch": 2.3582089552238807,
+      "grad_norm": 1.0837558507919312,
+      "learning_rate": 1.7666278978648644e-05,
+      "loss": 0.2319,
+      "step": 790
+    },
+    {
+      "epoch": 2.373134328358209,
+      "grad_norm": 1.2392449378967285,
+      "learning_rate": 1.752041457402195e-05,
+      "loss": 0.2349,
+      "step": 795
+    },
+    {
+      "epoch": 2.388059701492537,
+      "grad_norm": 1.0948951244354248,
+      "learning_rate": 1.7374304490500776e-05,
+      "loss": 0.2366,
+      "step": 800
+    },
+    {
+      "epoch": 2.4029850746268657,
+      "grad_norm": 1.1904592514038086,
+      "learning_rate": 1.722796297025152e-05,
+      "loss": 0.2393,
+      "step": 805
+    },
+    {
+      "epoch": 2.417910447761194,
+      "grad_norm": 1.0641899108886719,
+      "learning_rate": 1.708140427800004e-05,
+      "loss": 0.2044,
+      "step": 810
+    },
+    {
+      "epoch": 2.4328358208955225,
+      "grad_norm": 1.12578284740448,
+      "learning_rate": 1.6934642699641144e-05,
+      "loss": 0.2435,
+      "step": 815
+    },
+    {
+      "epoch": 2.4477611940298507,
+      "grad_norm": 1.0488451719284058,
+      "learning_rate": 1.6787692540846092e-05,
+      "loss": 0.2272,
+      "step": 820
+    },
+    {
+      "epoch": 2.4626865671641793,
+      "grad_norm": 1.061360239982605,
+      "learning_rate": 1.664056812566812e-05,
+      "loss": 0.2476,
+      "step": 825
+    },
+    {
+      "epoch": 2.4776119402985075,
+      "grad_norm": 1.2904311418533325,
+      "learning_rate": 1.649328379514622e-05,
+      "loss": 0.261,
+      "step": 830
+    },
+    {
+      "epoch": 2.4925373134328357,
+      "grad_norm": 1.061362385749817,
+      "learning_rate": 1.6345853905907223e-05,
+      "loss": 0.2309,
+      "step": 835
+    },
+    {
+      "epoch": 2.5074626865671643,
+      "grad_norm": 1.009899616241455,
+      "learning_rate": 1.6198292828766382e-05,
+      "loss": 0.2142,
+      "step": 840
+    },
+    {
+      "epoch": 2.5223880597014925,
+      "grad_norm": 1.2374935150146484,
+      "learning_rate": 1.6050614947326568e-05,
+      "loss": 0.2337,
+      "step": 845
+    },
+    {
+      "epoch": 2.5373134328358207,
+      "grad_norm": 1.1741938591003418,
+      "learning_rate": 1.5902834656576223e-05,
+      "loss": 0.2342,
+      "step": 850
+    },
+    {
+      "epoch": 2.5522388059701493,
+      "grad_norm": 1.0697554349899292,
+      "learning_rate": 1.5754966361486182e-05,
+      "loss": 0.2045,
+      "step": 855
+    },
+    {
+      "epoch": 2.5671641791044775,
+      "grad_norm": 1.1109567880630493,
+      "learning_rate": 1.5607024475605575e-05,
+      "loss": 0.2097,
+      "step": 860
+    },
+    {
+      "epoch": 2.582089552238806,
+      "grad_norm": 1.240860939025879,
+      "learning_rate": 1.5459023419656824e-05,
+      "loss": 0.2262,
+      "step": 865
+    },
+    {
+      "epoch": 2.5970149253731343,
+      "grad_norm": 1.0825039148330688,
+      "learning_rate": 1.5310977620129992e-05,
+      "loss": 0.2083,
+      "step": 870
+    },
+    {
+      "epoch": 2.611940298507463,
+      "grad_norm": 1.149062156677246,
+      "learning_rate": 1.5162901507876552e-05,
+      "loss": 0.2022,
+      "step": 875
+    },
+    {
+      "epoch": 2.626865671641791,
+      "grad_norm": 1.0815327167510986,
+      "learning_rate": 1.5014809516702727e-05,
+      "loss": 0.2267,
+      "step": 880
+    },
+    {
+      "epoch": 2.6417910447761193,
+      "grad_norm": 1.0375422239303589,
+      "learning_rate": 1.4866716081962542e-05,
+      "loss": 0.1734,
+      "step": 885
+    },
+    {
+      "epoch": 2.656716417910448,
+      "grad_norm": 1.094551682472229,
+      "learning_rate": 1.471863563915073e-05,
+      "loss": 0.2004,
+      "step": 890
+    },
+    {
+      "epoch": 2.671641791044776,
+      "grad_norm": 1.1885653734207153,
+      "learning_rate": 1.4570582622495653e-05,
+      "loss": 0.2025,
+      "step": 895
+    },
+    {
+      "epoch": 2.6865671641791042,
+      "grad_norm": 1.1636579036712646,
+      "learning_rate": 1.4422571463552267e-05,
+      "loss": 0.2116,
+      "step": 900
+    },
+    {
+      "epoch": 2.701492537313433,
+      "grad_norm": 1.226163387298584,
+      "learning_rate": 1.4274616589795424e-05,
+      "loss": 0.2063,
+      "step": 905
+    },
+    {
+      "epoch": 2.716417910447761,
+      "grad_norm": 1.2392303943634033,
+      "learning_rate": 1.412673242321356e-05,
+      "loss": 0.1976,
+      "step": 910
+    },
+    {
+      "epoch": 2.7313432835820897,
+      "grad_norm": 1.0528442859649658,
+      "learning_rate": 1.3978933378902877e-05,
+      "loss": 0.2086,
+      "step": 915
+    },
+    {
+      "epoch": 2.746268656716418,
+      "grad_norm": 1.0389113426208496,
+      "learning_rate": 1.3831233863662228e-05,
+      "loss": 0.1652,
+      "step": 920
+    },
+    {
+      "epoch": 2.7611940298507465,
+      "grad_norm": 1.0428828001022339,
+      "learning_rate": 1.368364827458883e-05,
+      "loss": 0.1931,
+      "step": 925
+    },
+    {
+      "epoch": 2.7761194029850746,
+      "grad_norm": 1.058120608329773,
+      "learning_rate": 1.3536190997674859e-05,
+      "loss": 0.1734,
+      "step": 930
+    },
+    {
+      "epoch": 2.791044776119403,
+      "grad_norm": 1.2238277196884155,
+      "learning_rate": 1.3388876406405206e-05,
+      "loss": 0.2117,
+      "step": 935
+    },
+    {
+      "epoch": 2.8059701492537314,
+      "grad_norm": 1.2835285663604736,
+      "learning_rate": 1.324171886035638e-05,
+      "loss": 0.1802,
+      "step": 940
+    },
+    {
+      "epoch": 2.8208955223880596,
+      "grad_norm": 1.141916036605835,
+      "learning_rate": 1.3094732703796819e-05,
+      "loss": 0.1846,
+      "step": 945
+    },
+    {
+      "epoch": 2.835820895522388,
+      "grad_norm": 1.187520146369934,
+      "learning_rate": 1.2947932264288677e-05,
+      "loss": 0.2056,
+      "step": 950
+    },
+    {
+      "epoch": 2.8507462686567164,
+      "grad_norm": 1.2382432222366333,
+      "learning_rate": 1.280133185129122e-05,
+      "loss": 0.1951,
+      "step": 955
+    },
+    {
+      "epoch": 2.8656716417910446,
+      "grad_norm": 1.0451785326004028,
+      "learning_rate": 1.2654945754766007e-05,
+      "loss": 0.1759,
+      "step": 960
+    },
+    {
+      "epoch": 2.8805970149253732,
+      "grad_norm": 1.2064586877822876,
+      "learning_rate": 1.2508788243783997e-05,
+      "loss": 0.2066,
+      "step": 965
+    },
+    {
+      "epoch": 2.8955223880597014,
+      "grad_norm": 1.071471929550171,
+      "learning_rate": 1.236287356513461e-05,
+      "loss": 0.2146,
+      "step": 970
+    },
+    {
+      "epoch": 2.91044776119403,
+      "grad_norm": 1.2358578443527222,
+      "learning_rate": 1.2217215941937061e-05,
+      "loss": 0.1642,
+      "step": 975
+    },
+    {
+      "epoch": 2.925373134328358,
+      "grad_norm": 1.0519459247589111,
+      "learning_rate": 1.2071829572253926e-05,
+      "loss": 0.1688,
+      "step": 980
+    },
+    {
+      "epoch": 2.9402985074626864,
+      "grad_norm": 1.163450837135315,
+      "learning_rate": 1.1926728627707176e-05,
+      "loss": 0.189,
+      "step": 985
+    },
+    {
+      "epoch": 2.955223880597015,
+      "grad_norm": 1.1002942323684692,
+      "learning_rate": 1.1781927252096824e-05,
+      "loss": 0.1758,
+      "step": 990
+    },
+    {
+      "epoch": 2.970149253731343,
+      "grad_norm": 1.026760220527649,
+      "learning_rate": 1.1637439560022196e-05,
+      "loss": 0.188,
+      "step": 995
+    },
+    {
+      "epoch": 2.9850746268656714,
+      "grad_norm": 1.1044259071350098,
+      "learning_rate": 1.149327963550615e-05,
+      "loss": 0.1618,
+      "step": 1000
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.0454556941986084,
+      "learning_rate": 1.1349461530622203e-05,
+      "loss": 0.1607,
+      "step": 1005
+    },
+    {
+      "epoch": 3.014925373134328,
+      "grad_norm": 1.0062533617019653,
+      "learning_rate": 1.1205999264124788e-05,
+      "loss": 0.1399,
+      "step": 1010
+    },
+    {
+      "epoch": 3.029850746268657,
+      "grad_norm": 1.2436738014221191,
+      "learning_rate": 1.1062906820082812e-05,
+      "loss": 0.1226,
+      "step": 1015
+    },
+    {
+      "epoch": 3.044776119402985,
+      "grad_norm": 0.9873276948928833,
+      "learning_rate": 1.0920198146516497e-05,
+      "loss": 0.1255,
+      "step": 1020
+    },
+    {
+      "epoch": 3.0597014925373136,
+      "grad_norm": 1.0272220373153687,
+      "learning_rate": 1.0777887154037807e-05,
+      "loss": 0.1137,
+      "step": 1025
+    },
+    {
+      "epoch": 3.074626865671642,
+      "grad_norm": 1.1051563024520874,
+      "learning_rate": 1.0635987714494541e-05,
+      "loss": 0.1276,
+      "step": 1030
+    },
+    {
+      "epoch": 3.08955223880597,
+      "grad_norm": 0.9112171530723572,
+      "learning_rate": 1.0494513659618086e-05,
+      "loss": 0.1263,
+      "step": 1035
+    },
+    {
+      "epoch": 3.1044776119402986,
+      "grad_norm": 0.9920433759689331,
+      "learning_rate": 1.0353478779675256e-05,
+      "loss": 0.1286,
+      "step": 1040
+    },
+    {
+      "epoch": 3.1194029850746268,
+      "grad_norm": 1.1244879961013794,
+      "learning_rate": 1.0212896822123994e-05,
+      "loss": 0.1353,
+      "step": 1045
+    },
+    {
+      "epoch": 3.1343283582089554,
+      "grad_norm": 1.1649903059005737,
+      "learning_rate": 1.0072781490273377e-05,
+      "loss": 0.1601,
+      "step": 1050
+    },
+    {
+      "epoch": 3.1492537313432836,
+      "grad_norm": 0.9812712669372559,
+      "learning_rate": 9.93314644194788e-06,
+      "loss": 0.1095,
+      "step": 1055
+    },
+    {
+      "epoch": 3.1641791044776117,
+      "grad_norm": 1.2677569389343262,
+      "learning_rate": 9.79400528815604e-06,
+      "loss": 0.1364,
+      "step": 1060
+    },
+    {
+      "epoch": 3.1791044776119404,
+      "grad_norm": 0.9897525906562805,
+      "learning_rate": 9.65537159176375e-06,
+      "loss": 0.1269,
+      "step": 1065
+    },
+    {
+      "epoch": 3.1940298507462686,
+      "grad_norm": 1.0115331411361694,
+      "learning_rate": 9.517258866172176e-06,
+      "loss": 0.1367,
+      "step": 1070
+    },
+    {
+      "epoch": 3.208955223880597,
+      "grad_norm": 0.982448935508728,
+      "learning_rate": 9.379680574000558e-06,
+      "loss": 0.124,
+      "step": 1075
+    },
+    {
+      "epoch": 3.2238805970149254,
+      "grad_norm": 1.2059563398361206,
+      "learning_rate": 9.242650125773936e-06,
+      "loss": 0.1347,
+      "step": 1080
+    },
+    {
+      "epoch": 3.2388059701492535,
+      "grad_norm": 1.1463255882263184,
+      "learning_rate": 9.106180878615919e-06,
+      "loss": 0.1325,
+      "step": 1085
+    },
+    {
+      "epoch": 3.253731343283582,
+      "grad_norm": 0.986930251121521,
+      "learning_rate": 8.970286134946717e-06,
+      "loss": 0.1311,
+      "step": 1090
+    },
+    {
+      "epoch": 3.2686567164179103,
+      "grad_norm": 0.9221780300140381,
+      "learning_rate": 8.83497914118648e-06,
+      "loss": 0.1225,
+      "step": 1095
+    },
+    {
+      "epoch": 3.283582089552239,
+      "grad_norm": 1.1156846284866333,
+      "learning_rate": 8.700273086464083e-06,
+      "loss": 0.1147,
+      "step": 1100
+    },
+    {
+      "epoch": 3.298507462686567,
+      "grad_norm": 0.9974888563156128,
+      "learning_rate": 8.566181101331528e-06,
+      "loss": 0.1323,
+      "step": 1105
+    },
+    {
+      "epoch": 3.3134328358208958,
+      "grad_norm": 1.019337773323059,
+      "learning_rate": 8.432716256484e-06,
+      "loss": 0.1179,
+      "step": 1110
+    },
+    {
+      "epoch": 3.328358208955224,
+      "grad_norm": 0.9972241520881653,
+      "learning_rate": 8.29989156148583e-06,
+      "loss": 0.1202,
+      "step": 1115
+    },
+    {
+      "epoch": 3.343283582089552,
+      "grad_norm": 1.1177517175674438,
+      "learning_rate": 8.16771996350237e-06,
+      "loss": 0.1326,
+      "step": 1120
+    },
+    {
+      "epoch": 3.3582089552238807,
+      "grad_norm": 1.0671625137329102,
+      "learning_rate": 8.03621434603794e-06,
+      "loss": 0.1251,
+      "step": 1125
+    },
+    {
+      "epoch": 3.373134328358209,
+      "grad_norm": 0.8058292269706726,
+      "learning_rate": 7.905387527680028e-06,
+      "loss": 0.12,
+      "step": 1130
+    },
+    {
+      "epoch": 3.388059701492537,
+      "grad_norm": 1.0809980630874634,
+      "learning_rate": 7.775252260849787e-06,
+      "loss": 0.1168,
+      "step": 1135
+    },
+    {
+      "epoch": 3.4029850746268657,
+      "grad_norm": 0.9724788069725037,
+      "learning_rate": 7.645821230558951e-06,
+      "loss": 0.1125,
+      "step": 1140
+    },
+    {
+      "epoch": 3.417910447761194,
+      "grad_norm": 1.1242866516113281,
+      "learning_rate": 7.517107053173411e-06,
+      "loss": 0.121,
+      "step": 1145
+    },
+    {
+      "epoch": 3.4328358208955225,
+      "grad_norm": 0.9678345918655396,
+      "learning_rate": 7.389122275183384e-06,
+      "loss": 0.1188,
+      "step": 1150
+    },
+    {
+      "epoch": 3.4477611940298507,
+      "grad_norm": 1.1586215496063232,
+      "learning_rate": 7.261879371980451e-06,
+      "loss": 0.1301,
+      "step": 1155
+    },
+    {
+      "epoch": 3.4626865671641793,
+      "grad_norm": 0.9388768672943115,
+      "learning_rate": 7.135390746641527e-06,
+      "loss": 0.1297,
+      "step": 1160
+    },
+    {
+      "epoch": 3.4776119402985075,
+      "grad_norm": 1.0194765329360962,
+      "learning_rate": 7.009668728719824e-06,
+      "loss": 0.1175,
+      "step": 1165
+    },
+    {
+      "epoch": 3.4925373134328357,
+      "grad_norm": 1.005406141281128,
+      "learning_rate": 6.8847255730430655e-06,
+      "loss": 0.1258,
+      "step": 1170
+    },
+    {
+      "epoch": 3.5074626865671643,
+      "grad_norm": 0.8349246978759766,
+      "learning_rate": 6.760573458518911e-06,
+      "loss": 0.1116,
+      "step": 1175
+    },
+    {
+      "epoch": 3.5223880597014925,
+      "grad_norm": 0.9510399699211121,
+      "learning_rate": 6.637224486947809e-06,
+      "loss": 0.1189,
+      "step": 1180
+    },
+    {
+      "epoch": 3.5373134328358207,
+      "grad_norm": 1.1545650959014893,
+      "learning_rate": 6.514690681843396e-06,
+      "loss": 0.1111,
+      "step": 1185
+    },
+    {
+      "epoch": 3.5522388059701493,
+      "grad_norm": 0.9536241292953491,
+      "learning_rate": 6.392983987260459e-06,
+      "loss": 0.1099,
+      "step": 1190
+    },
+    {
+      "epoch": 3.5671641791044775,
+      "grad_norm": 0.9180517792701721,
+      "learning_rate": 6.27211626663071e-06,
+      "loss": 0.101,
+      "step": 1195
+    },
+    {
+      "epoch": 3.582089552238806,
+      "grad_norm": 0.895145833492279,
+      "learning_rate": 6.152099301606375e-06,
+      "loss": 0.1182,
+      "step": 1200
+    },
+    {
+      "epoch": 3.5970149253731343,
+      "grad_norm": 1.1857234239578247,
+      "learning_rate": 6.032944790911775e-06,
+      "loss": 0.1234,
+      "step": 1205
+    },
+    {
+      "epoch": 3.611940298507463,
+      "grad_norm": 0.8817716836929321,
+      "learning_rate": 5.914664349203002e-06,
+      "loss": 0.1089,
+      "step": 1210
+    },
+    {
+      "epoch": 3.626865671641791,
+      "grad_norm": 0.9345617294311523,
+      "learning_rate": 5.797269505935743e-06,
+      "loss": 0.1084,
+      "step": 1215
+    },
+    {
+      "epoch": 3.6417910447761193,
+      "grad_norm": 0.9709360599517822,
+      "learning_rate": 5.680771704241458e-06,
+      "loss": 0.118,
+      "step": 1220
+    },
+    {
+      "epoch": 3.656716417910448,
+      "grad_norm": 0.8643797636032104,
+      "learning_rate": 5.565182299811963e-06,
+      "loss": 0.0972,
+      "step": 1225
+    },
+    {
+      "epoch": 3.671641791044776,
+      "grad_norm": 0.9221701622009277,
+      "learning_rate": 5.450512559792497e-06,
+      "loss": 0.1077,
+      "step": 1230
+    },
+    {
+      "epoch": 3.6865671641791042,
+      "grad_norm": 1.0147606134414673,
+      "learning_rate": 5.336773661683491e-06,
+      "loss": 0.1205,
+      "step": 1235
+    },
+    {
+      "epoch": 3.701492537313433,
+      "grad_norm": 0.9583662748336792,
+      "learning_rate": 5.223976692250995e-06,
+      "loss": 0.0985,
+      "step": 1240
+    },
+    {
+      "epoch": 3.716417910447761,
+      "grad_norm": 0.9088737368583679,
+      "learning_rate": 5.112132646446011e-06,
+      "loss": 0.1093,
+      "step": 1245
+    },
+    {
+      "epoch": 3.7313432835820897,
+      "grad_norm": 0.9676306247711182,
+      "learning_rate": 5.00125242633276e-06,
+      "loss": 0.1066,
+      "step": 1250
+    },
+    {
+      "epoch": 3.746268656716418,
+      "grad_norm": 0.8011495471000671,
+      "learning_rate": 4.891346840025971e-06,
+      "loss": 0.1018,
+      "step": 1255
+    },
+    {
+      "epoch": 3.7611940298507465,
+      "grad_norm": 1.1067790985107422,
+      "learning_rate": 4.782426600637379e-06,
+      "loss": 0.1151,
+      "step": 1260
+    },
+    {
+      "epoch": 3.7761194029850746,
+      "grad_norm": 1.0292284488677979,
+      "learning_rate": 4.674502325231442e-06,
+      "loss": 0.1016,
+      "step": 1265
+    },
+    {
+      "epoch": 3.791044776119403,
+      "grad_norm": 0.9458780884742737,
+      "learning_rate": 4.5675845337904425e-06,
+      "loss": 0.0952,
+      "step": 1270
+    },
+    {
+      "epoch": 3.8059701492537314,
+      "grad_norm": 1.020756483078003,
+      "learning_rate": 4.461683648189056e-06,
+      "loss": 0.1138,
+      "step": 1275
+    },
+    {
+      "epoch": 3.8208955223880596,
+      "grad_norm": 0.9671748280525208,
+      "learning_rate": 4.356809991178451e-06,
+      "loss": 0.1179,
+      "step": 1280
+    },
+    {
+      "epoch": 3.835820895522388,
+      "grad_norm": 1.0346472263336182,
+      "learning_rate": 4.252973785380086e-06,
+      "loss": 0.1111,
+      "step": 1285
+    },
+    {
+      "epoch": 3.8507462686567164,
+      "grad_norm": 0.9963144659996033,
+      "learning_rate": 4.150185152289262e-06,
+      "loss": 0.1043,
+      "step": 1290
+    },
+    {
+      "epoch": 3.8656716417910446,
+      "grad_norm": 0.8412414789199829,
+      "learning_rate": 4.0484541112885015e-06,
+      "loss": 0.0872,
+      "step": 1295
+    },
+    {
+      "epoch": 3.8805970149253732,
+      "grad_norm": 0.8710880279541016,
+      "learning_rate": 3.947790578670914e-06,
+      "loss": 0.1059,
+      "step": 1300
+    },
+    {
+      "epoch": 3.8955223880597014,
+      "grad_norm": 1.1415584087371826,
+      "learning_rate": 3.848204366673616e-06,
+      "loss": 0.0969,
+      "step": 1305
+    },
+    {
+      "epoch": 3.91044776119403,
+      "grad_norm": 0.9186813235282898,
+      "learning_rate": 3.7497051825212503e-06,
+      "loss": 0.0902,
+      "step": 1310
+    },
+    {
+      "epoch": 3.925373134328358,
+      "grad_norm": 0.9267393946647644,
+      "learning_rate": 3.6523026274797832e-06,
+      "loss": 0.1018,
+      "step": 1315
+    },
+    {
+      "epoch": 3.9402985074626864,
+      "grad_norm": 0.8800696730613708,
+      "learning_rate": 3.5560061959206097e-06,
+      "loss": 0.107,
+      "step": 1320
+    },
+    {
+      "epoch": 3.955223880597015,
+      "grad_norm": 0.8548220992088318,
+      "learning_rate": 3.4608252743950875e-06,
+      "loss": 0.1073,
+      "step": 1325
+    },
+    {
+      "epoch": 3.970149253731343,
+      "grad_norm": 0.9937451481819153,
+      "learning_rate": 3.3667691407195883e-06,
+      "loss": 0.1014,
+      "step": 1330
+    },
+    {
+      "epoch": 3.9850746268656714,
+      "grad_norm": 0.8701783418655396,
+      "learning_rate": 3.27384696307111e-06,
+      "loss": 0.0943,
+      "step": 1335
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 0.8932617902755737,
+      "learning_rate": 3.1820677990936337e-06,
+      "loss": 0.0933,
+      "step": 1340
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1675,
+  "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": 1.9757498391571988e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

2_128_e5_3e-5/checkpoint-1340/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecfa82b92b45b87d589ccd5ed731a54d01e94a8d5dc857ed4c36d4edefc4970a
+size 7736

2_128_e5_3e-5/checkpoint-1340/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)

2_128_e5_3e-5/checkpoint-1675/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

2_128_e5_3e-5/checkpoint-1675/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": [
+    "down_proj",
+    "v_proj",
+    "k_proj",
+    "o_proj",
+    "up_proj",
+    "q_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

2_128_e5_3e-5/checkpoint-1675/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a11bf9ee29ad785a5b648d31eb196f6c5e6888f4eaacab9be584ddedde8fcb4f
+size 791751704

2_128_e5_3e-5/checkpoint-1675/latest

@@ -0,0 +1 @@
+global_step1675

2_128_e5_3e-5/checkpoint-1675/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1362b13d6fe3858637f9194fdb713cb5d89b025166385607e93dff2637d6e2da
+size 15920

2_128_e5_3e-5/checkpoint-1675/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eee744a2a1e8a84bbb70ae50cfb6e2abdb834c6564052b58549192137c56d783
+size 15920

2_128_e5_3e-5/checkpoint-1675/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7635b412de5d267d11cd6d5b85efe5dfd962f7a69677aa2c3d4a28869b3f873c
+size 15920