Upload folder using huggingface_hub

21_128_e5_3e-5/checkpoint-1320/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

21_128_e5_3e-5/checkpoint-1320/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": [
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "v_proj",
+    "o_proj",
+    "k_proj",
+    "q_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

21_128_e5_3e-5/checkpoint-1320/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b969f4691c91e3e900054cb8eec44c1422120a34ee6c6c3ad0f026b346c13c50
+size 791751704

21_128_e5_3e-5/checkpoint-1320/latest

@@ -0,0 +1 @@
+global_step1320

21_128_e5_3e-5/checkpoint-1320/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0ee661e2eb299dc13e65b03c8856f33d8e1e4b5f8a70cb72a53fa838f5891d1
+size 15920

21_128_e5_3e-5/checkpoint-1320/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92404d4381131cb471171fd0ac861958420cdfb4f1b5dce85740c4c74984d2df
+size 15920

21_128_e5_3e-5/checkpoint-1320/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:430a02a44d64c268b905abe1b51cd36723b35d71f3777fab4423fdfd0b78f552
+size 15920

21_128_e5_3e-5/checkpoint-1320/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:823ce73381d6fd549a7e2bc895385b2932434e37dc4fb093c621934db4631d29
+size 15920

21_128_e5_3e-5/checkpoint-1320/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0f05827267f23b2a27d9c7187c7eb42fa8e50c2f2878e48dd8cf05bd80f3a1e
+size 15920

21_128_e5_3e-5/checkpoint-1320/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87ba0937445a4942ae179d6f2fbe2e5c466b6932cebb26d74b111389e3261b66
+size 15920

21_128_e5_3e-5/checkpoint-1320/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:118015773e9a9bfa1cc2fedb9638a1c91b4af84c4e8ff61787cc415bb2f12a18
+size 15920

21_128_e5_3e-5/checkpoint-1320/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e0a73d424223332f5cd0e5185173eb8a92380b5ff8a74b4905e5ed7dae8160a
+size 15920

21_128_e5_3e-5/checkpoint-1320/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a61ce24d3e6e9dea57c744f2e899f33c98348af746dec8b8eab3156e45392ac
+size 1064

21_128_e5_3e-5/checkpoint-1320/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
+  }
+}

21_128_e5_3e-5/checkpoint-1320/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
+}

21_128_e5_3e-5/checkpoint-1320/trainer_state.json

@@ -0,0 +1,1882 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 1320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.015174506828528073,
+      "grad_norm": 1.1642332077026367,
+      "learning_rate": 1.4457831325301207e-06,
+      "loss": 1.3119,
+      "step": 5
+    },
+    {
+      "epoch": 0.030349013657056147,
+      "grad_norm": 0.857893705368042,
+      "learning_rate": 3.2530120481927713e-06,
+      "loss": 1.2575,
+      "step": 10
+    },
+    {
+      "epoch": 0.04552352048558422,
+      "grad_norm": 0.6741542220115662,
+      "learning_rate": 5.060240963855422e-06,
+      "loss": 1.2179,
+      "step": 15
+    },
+    {
+      "epoch": 0.06069802731411229,
+      "grad_norm": 0.5497764348983765,
+      "learning_rate": 6.867469879518072e-06,
+      "loss": 1.2422,
+      "step": 20
+    },
+    {
+      "epoch": 0.07587253414264036,
+      "grad_norm": 0.5862205624580383,
+      "learning_rate": 8.674698795180722e-06,
+      "loss": 1.2837,
+      "step": 25
+    },
+    {
+      "epoch": 0.09104704097116843,
+      "grad_norm": 0.5110287070274353,
+      "learning_rate": 1.0481927710843374e-05,
+      "loss": 1.1704,
+      "step": 30
+    },
+    {
+      "epoch": 0.1062215477996965,
+      "grad_norm": 0.551201343536377,
+      "learning_rate": 1.2289156626506024e-05,
+      "loss": 1.1677,
+      "step": 35
+    },
+    {
+      "epoch": 0.12139605462822459,
+      "grad_norm": 0.5148503184318542,
+      "learning_rate": 1.4096385542168676e-05,
+      "loss": 1.116,
+      "step": 40
+    },
+    {
+      "epoch": 0.13657056145675264,
+      "grad_norm": 0.5068604946136475,
+      "learning_rate": 1.5903614457831326e-05,
+      "loss": 1.1516,
+      "step": 45
+    },
+    {
+      "epoch": 0.15174506828528073,
+      "grad_norm": 0.5807440280914307,
+      "learning_rate": 1.7710843373493978e-05,
+      "loss": 1.1822,
+      "step": 50
+    },
+    {
+      "epoch": 0.1669195751138088,
+      "grad_norm": 0.48767879605293274,
+      "learning_rate": 1.9518072289156627e-05,
+      "loss": 1.1189,
+      "step": 55
+    },
+    {
+      "epoch": 0.18209408194233687,
+      "grad_norm": 0.5024222135543823,
+      "learning_rate": 2.1325301204819275e-05,
+      "loss": 1.1291,
+      "step": 60
+    },
+    {
+      "epoch": 0.19726858877086495,
+      "grad_norm": 0.5667535662651062,
+      "learning_rate": 2.313253012048193e-05,
+      "loss": 1.1279,
+      "step": 65
+    },
+    {
+      "epoch": 0.212443095599393,
+      "grad_norm": 0.5600463151931763,
+      "learning_rate": 2.493975903614458e-05,
+      "loss": 1.1465,
+      "step": 70
+    },
+    {
+      "epoch": 0.2276176024279211,
+      "grad_norm": 0.5471869111061096,
+      "learning_rate": 2.674698795180723e-05,
+      "loss": 1.0742,
+      "step": 75
+    },
+    {
+      "epoch": 0.24279210925644917,
+      "grad_norm": 0.5258970856666565,
+      "learning_rate": 2.855421686746988e-05,
+      "loss": 1.0863,
+      "step": 80
+    },
+    {
+      "epoch": 0.25796661608497723,
+      "grad_norm": 0.5476792454719543,
+      "learning_rate": 2.9999969854473704e-05,
+      "loss": 1.0845,
+      "step": 85
+    },
+    {
+      "epoch": 0.2731411229135053,
+      "grad_norm": 0.5853098034858704,
+      "learning_rate": 2.9998914773775897e-05,
+      "loss": 1.0725,
+      "step": 90
+    },
+    {
+      "epoch": 0.2883156297420334,
+      "grad_norm": 0.5045304894447327,
+      "learning_rate": 2.9996352537928224e-05,
+      "loss": 1.0087,
+      "step": 95
+    },
+    {
+      "epoch": 0.30349013657056145,
+      "grad_norm": 0.5672217607498169,
+      "learning_rate": 2.9992283404395112e-05,
+      "loss": 1.0539,
+      "step": 100
+    },
+    {
+      "epoch": 0.3186646433990895,
+      "grad_norm": 0.5382909178733826,
+      "learning_rate": 2.9986707782060517e-05,
+      "loss": 1.0188,
+      "step": 105
+    },
+    {
+      "epoch": 0.3338391502276176,
+      "grad_norm": 0.7915878891944885,
+      "learning_rate": 2.997962623118683e-05,
+      "loss": 0.9988,
+      "step": 110
+    },
+    {
+      "epoch": 0.3490136570561457,
+      "grad_norm": 0.5800904035568237,
+      "learning_rate": 2.9971039463358594e-05,
+      "loss": 1.0071,
+      "step": 115
+    },
+    {
+      "epoch": 0.36418816388467373,
+      "grad_norm": 0.6986746191978455,
+      "learning_rate": 2.9960948341410993e-05,
+      "loss": 1.012,
+      "step": 120
+    },
+    {
+      "epoch": 0.37936267071320184,
+      "grad_norm": 0.6434286832809448,
+      "learning_rate": 2.994935387934314e-05,
+      "loss": 0.9365,
+      "step": 125
+    },
+    {
+      "epoch": 0.3945371775417299,
+      "grad_norm": 0.6894506216049194,
+      "learning_rate": 2.9936257242216216e-05,
+      "loss": 0.9749,
+      "step": 130
+    },
+    {
+      "epoch": 0.40971168437025796,
+      "grad_norm": 0.6694482564926147,
+      "learning_rate": 2.9921659746036364e-05,
+      "loss": 0.9807,
+      "step": 135
+    },
+    {
+      "epoch": 0.424886191198786,
+      "grad_norm": 0.5935041904449463,
+      "learning_rate": 2.9905562857622482e-05,
+      "loss": 0.9575,
+      "step": 140
+    },
+    {
+      "epoch": 0.4400606980273141,
+      "grad_norm": 0.8535542488098145,
+      "learning_rate": 2.9887968194458807e-05,
+      "loss": 0.9219,
+      "step": 145
+    },
+    {
+      "epoch": 0.4552352048558422,
+      "grad_norm": 0.775467038154602,
+      "learning_rate": 2.986887752453239e-05,
+      "loss": 0.9602,
+      "step": 150
+    },
+    {
+      "epoch": 0.47040971168437024,
+      "grad_norm": 0.7054561376571655,
+      "learning_rate": 2.984829276615546e-05,
+      "loss": 0.8944,
+      "step": 155
+    },
+    {
+      "epoch": 0.48558421851289835,
+      "grad_norm": 0.6600781679153442,
+      "learning_rate": 2.9826215987772642e-05,
+      "loss": 0.9139,
+      "step": 160
+    },
+    {
+      "epoch": 0.5007587253414264,
+      "grad_norm": 0.8322820663452148,
+      "learning_rate": 2.9802649407753105e-05,
+      "loss": 0.8685,
+      "step": 165
+    },
+    {
+      "epoch": 0.5159332321699545,
+      "grad_norm": 0.6224384903907776,
+      "learning_rate": 2.9777595394167674e-05,
+      "loss": 0.9406,
+      "step": 170
+    },
+    {
+      "epoch": 0.5311077389984825,
+      "grad_norm": 0.7838391065597534,
+      "learning_rate": 2.9751056464550863e-05,
+      "loss": 0.9418,
+      "step": 175
+    },
+    {
+      "epoch": 0.5462822458270106,
+      "grad_norm": 0.7811512351036072,
+      "learning_rate": 2.97230352856479e-05,
+      "loss": 0.8876,
+      "step": 180
+    },
+    {
+      "epoch": 0.5614567526555387,
+      "grad_norm": 0.7145566940307617,
+      "learning_rate": 2.9693534673146772e-05,
+      "loss": 0.851,
+      "step": 185
+    },
+    {
+      "epoch": 0.5766312594840668,
+      "grad_norm": 0.7759535908699036,
+      "learning_rate": 2.9662557591395282e-05,
+      "loss": 0.8149,
+      "step": 190
+    },
+    {
+      "epoch": 0.5918057663125948,
+      "grad_norm": 0.8379637002944946,
+      "learning_rate": 2.9630107153103176e-05,
+      "loss": 0.8479,
+      "step": 195
+    },
+    {
+      "epoch": 0.6069802731411229,
+      "grad_norm": 0.7673034071922302,
+      "learning_rate": 2.9596186619029382e-05,
+      "loss": 0.8367,
+      "step": 200
+    },
+    {
+      "epoch": 0.622154779969651,
+      "grad_norm": 0.7668901085853577,
+      "learning_rate": 2.9560799397654342e-05,
+      "loss": 0.8165,
+      "step": 205
+    },
+    {
+      "epoch": 0.637329286798179,
+      "grad_norm": 0.8146044015884399,
+      "learning_rate": 2.952394904483751e-05,
+      "loss": 0.823,
+      "step": 210
+    },
+    {
+      "epoch": 0.6525037936267072,
+      "grad_norm": 0.7815675735473633,
+      "learning_rate": 2.9485639263460045e-05,
+      "loss": 0.804,
+      "step": 215
+    },
+    {
+      "epoch": 0.6676783004552352,
+      "grad_norm": 0.8267041444778442,
+      "learning_rate": 2.944587390305275e-05,
+      "loss": 0.8068,
+      "step": 220
+    },
+    {
+      "epoch": 0.6828528072837633,
+      "grad_norm": 0.8476167321205139,
+      "learning_rate": 2.9404656959409228e-05,
+      "loss": 0.7598,
+      "step": 225
+    },
+    {
+      "epoch": 0.6980273141122914,
+      "grad_norm": 0.8841318488121033,
+      "learning_rate": 2.9361992574184376e-05,
+      "loss": 0.8277,
+      "step": 230
+    },
+    {
+      "epoch": 0.7132018209408194,
+      "grad_norm": 0.962478756904602,
+      "learning_rate": 2.931788503447822e-05,
+      "loss": 0.7416,
+      "step": 235
+    },
+    {
+      "epoch": 0.7283763277693475,
+      "grad_norm": 0.8502833247184753,
+      "learning_rate": 2.9272338772405128e-05,
+      "loss": 0.7805,
+      "step": 240
+    },
+    {
+      "epoch": 0.7435508345978755,
+      "grad_norm": 1.0657333135604858,
+      "learning_rate": 2.9225358364648438e-05,
+      "loss": 0.7696,
+      "step": 245
+    },
+    {
+      "epoch": 0.7587253414264037,
+      "grad_norm": 0.9457326531410217,
+      "learning_rate": 2.9176948532000594e-05,
+      "loss": 0.7713,
+      "step": 250
+    },
+    {
+      "epoch": 0.7738998482549317,
+      "grad_norm": 0.9325288534164429,
+      "learning_rate": 2.9127114138888778e-05,
+      "loss": 0.7448,
+      "step": 255
+    },
+    {
+      "epoch": 0.7890743550834598,
+      "grad_norm": 0.9588562250137329,
+      "learning_rate": 2.907586019288608e-05,
+      "loss": 0.7063,
+      "step": 260
+    },
+    {
+      "epoch": 0.8042488619119879,
+      "grad_norm": 0.8736801743507385,
+      "learning_rate": 2.9023191844208362e-05,
+      "loss": 0.7246,
+      "step": 265
+    },
+    {
+      "epoch": 0.8194233687405159,
+      "grad_norm": 1.0334984064102173,
+      "learning_rate": 2.896911438519671e-05,
+      "loss": 0.726,
+      "step": 270
+    },
+    {
+      "epoch": 0.834597875569044,
+      "grad_norm": 0.9704434275627136,
+      "learning_rate": 2.8913633249785653e-05,
+      "loss": 0.775,
+      "step": 275
+    },
+    {
+      "epoch": 0.849772382397572,
+      "grad_norm": 0.9025390148162842,
+      "learning_rate": 2.8856754012957126e-05,
+      "loss": 0.7014,
+      "step": 280
+    },
+    {
+      "epoch": 0.8649468892261002,
+      "grad_norm": 0.9079306721687317,
+      "learning_rate": 2.879848239018029e-05,
+      "loss": 0.681,
+      "step": 285
+    },
+    {
+      "epoch": 0.8801213960546282,
+      "grad_norm": 0.9602410793304443,
+      "learning_rate": 2.873882423683718e-05,
+      "loss": 0.7152,
+      "step": 290
+    },
+    {
+      "epoch": 0.8952959028831563,
+      "grad_norm": 0.9534751176834106,
+      "learning_rate": 2.8677785547634384e-05,
+      "loss": 0.7009,
+      "step": 295
+    },
+    {
+      "epoch": 0.9104704097116844,
+      "grad_norm": 0.9962713122367859,
+      "learning_rate": 2.861537245600063e-05,
+      "loss": 0.7087,
+      "step": 300
+    },
+    {
+      "epoch": 0.9256449165402124,
+      "grad_norm": 0.9465916752815247,
+      "learning_rate": 2.8551591233470488e-05,
+      "loss": 0.6897,
+      "step": 305
+    },
+    {
+      "epoch": 0.9408194233687405,
+      "grad_norm": 0.9778639078140259,
+      "learning_rate": 2.8486448289054164e-05,
+      "loss": 0.6658,
+      "step": 310
+    },
+    {
+      "epoch": 0.9559939301972686,
+      "grad_norm": 0.9425243139266968,
+      "learning_rate": 2.8419950168593528e-05,
+      "loss": 0.6454,
+      "step": 315
+    },
+    {
+      "epoch": 0.9711684370257967,
+      "grad_norm": 0.9817103743553162,
+      "learning_rate": 2.8352103554104313e-05,
+      "loss": 0.6343,
+      "step": 320
+    },
+    {
+      "epoch": 0.9863429438543247,
+      "grad_norm": 0.9938346147537231,
+      "learning_rate": 2.8282915263104728e-05,
+      "loss": 0.645,
+      "step": 325
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.365831971168518,
+      "learning_rate": 2.8212392247930368e-05,
+      "loss": 0.6416,
+      "step": 330
+    },
+    {
+      "epoch": 1.015174506828528,
+      "grad_norm": 0.9671252965927124,
+      "learning_rate": 2.814054159503563e-05,
+      "loss": 0.575,
+      "step": 335
+    },
+    {
+      "epoch": 1.0303490136570561,
+      "grad_norm": 1.005592942237854,
+      "learning_rate": 2.8067370524281617e-05,
+      "loss": 0.5576,
+      "step": 340
+    },
+    {
+      "epoch": 1.0455235204855842,
+      "grad_norm": 1.00735342502594,
+      "learning_rate": 2.7992886388210693e-05,
+      "loss": 0.5577,
+      "step": 345
+    },
+    {
+      "epoch": 1.0606980273141122,
+      "grad_norm": 1.090733528137207,
+      "learning_rate": 2.7917096671307624e-05,
+      "loss": 0.5428,
+      "step": 350
+    },
+    {
+      "epoch": 1.0758725341426403,
+      "grad_norm": 1.0599536895751953,
+      "learning_rate": 2.784000898924754e-05,
+      "loss": 0.519,
+      "step": 355
+    },
+    {
+      "epoch": 1.0910470409711683,
+      "grad_norm": 1.1048762798309326,
+      "learning_rate": 2.7761631088130654e-05,
+      "loss": 0.5437,
+      "step": 360
+    },
+    {
+      "epoch": 1.1062215477996964,
+      "grad_norm": 1.1115154027938843,
+      "learning_rate": 2.7681970843703926e-05,
+      "loss": 0.6217,
+      "step": 365
+    },
+    {
+      "epoch": 1.1213960546282247,
+      "grad_norm": 1.0769774913787842,
+      "learning_rate": 2.7601036260569646e-05,
+      "loss": 0.5095,
+      "step": 370
+    },
+    {
+      "epoch": 1.1365705614567527,
+      "grad_norm": 1.0201396942138672,
+      "learning_rate": 2.7518835471381117e-05,
+      "loss": 0.5247,
+      "step": 375
+    },
+    {
+      "epoch": 1.1517450682852808,
+      "grad_norm": 1.0590314865112305,
+      "learning_rate": 2.7435376736025447e-05,
+      "loss": 0.5516,
+      "step": 380
+    },
+    {
+      "epoch": 1.1669195751138088,
+      "grad_norm": 1.0749924182891846,
+      "learning_rate": 2.735066844079355e-05,
+      "loss": 0.5079,
+      "step": 385
+    },
+    {
+      "epoch": 1.182094081942337,
+      "grad_norm": 1.10120689868927,
+      "learning_rate": 2.7264719097537482e-05,
+      "loss": 0.539,
+      "step": 390
+    },
+    {
+      "epoch": 1.197268588770865,
+      "grad_norm": 1.3611030578613281,
+      "learning_rate": 2.7177537342815098e-05,
+      "loss": 0.4849,
+      "step": 395
+    },
+    {
+      "epoch": 1.212443095599393,
+      "grad_norm": 1.0392005443572998,
+      "learning_rate": 2.7089131937022242e-05,
+      "loss": 0.4807,
+      "step": 400
+    },
+    {
+      "epoch": 1.227617602427921,
+      "grad_norm": 1.0147825479507446,
+      "learning_rate": 2.699951176351245e-05,
+      "loss": 0.4962,
+      "step": 405
+    },
+    {
+      "epoch": 1.2427921092564491,
+      "grad_norm": 1.1136548519134521,
+      "learning_rate": 2.6908685827704324e-05,
+      "loss": 0.5106,
+      "step": 410
+    },
+    {
+      "epoch": 1.2579666160849772,
+      "grad_norm": 0.9843658804893494,
+      "learning_rate": 2.681666325617661e-05,
+      "loss": 0.4979,
+      "step": 415
+    },
+    {
+      "epoch": 1.2731411229135052,
+      "grad_norm": 1.0182534456253052,
+      "learning_rate": 2.6723453295751143e-05,
+      "loss": 0.4764,
+      "step": 420
+    },
+    {
+      "epoch": 1.2883156297420335,
+      "grad_norm": 1.0736995935440063,
+      "learning_rate": 2.6629065312563672e-05,
+      "loss": 0.4811,
+      "step": 425
+    },
+    {
+      "epoch": 1.3034901365705616,
+      "grad_norm": 1.0721853971481323,
+      "learning_rate": 2.6533508791122716e-05,
+      "loss": 0.4757,
+      "step": 430
+    },
+    {
+      "epoch": 1.3186646433990896,
+      "grad_norm": 1.0161422491073608,
+      "learning_rate": 2.6436793333356523e-05,
+      "loss": 0.4683,
+      "step": 435
+    },
+    {
+      "epoch": 1.3338391502276177,
+      "grad_norm": 1.159873127937317,
+      "learning_rate": 2.633892865764821e-05,
+      "loss": 0.4907,
+      "step": 440
+    },
+    {
+      "epoch": 1.3490136570561457,
+      "grad_norm": 1.148019790649414,
+      "learning_rate": 2.623992459785925e-05,
+      "loss": 0.4757,
+      "step": 445
+    },
+    {
+      "epoch": 1.3641881638846738,
+      "grad_norm": 1.0696481466293335,
+      "learning_rate": 2.6139791102341287e-05,
+      "loss": 0.4815,
+      "step": 450
+    },
+    {
+      "epoch": 1.3793626707132018,
+      "grad_norm": 1.1690481901168823,
+      "learning_rate": 2.6038538232936516e-05,
+      "loss": 0.4361,
+      "step": 455
+    },
+    {
+      "epoch": 1.39453717754173,
+      "grad_norm": 1.0492829084396362,
+      "learning_rate": 2.5936176163966597e-05,
+      "loss": 0.4764,
+      "step": 460
+    },
+    {
+      "epoch": 1.409711684370258,
+      "grad_norm": 1.0407378673553467,
+      "learning_rate": 2.583271518121032e-05,
+      "loss": 0.4543,
+      "step": 465
+    },
+    {
+      "epoch": 1.424886191198786,
+      "grad_norm": 1.1021435260772705,
+      "learning_rate": 2.572816568087003e-05,
+      "loss": 0.4631,
+      "step": 470
+    },
+    {
+      "epoch": 1.440060698027314,
+      "grad_norm": 1.0365526676177979,
+      "learning_rate": 2.5622538168526978e-05,
+      "loss": 0.4496,
+      "step": 475
+    },
+    {
+      "epoch": 1.4552352048558421,
+      "grad_norm": 1.0675585269927979,
+      "learning_rate": 2.5515843258085686e-05,
+      "loss": 0.4605,
+      "step": 480
+    },
+    {
+      "epoch": 1.4704097116843702,
+      "grad_norm": 1.1329057216644287,
+      "learning_rate": 2.5408091670707386e-05,
+      "loss": 0.4065,
+      "step": 485
+    },
+    {
+      "epoch": 1.4855842185128982,
+      "grad_norm": 0.9651331305503845,
+      "learning_rate": 2.5299294233732742e-05,
+      "loss": 0.4249,
+      "step": 490
+    },
+    {
+      "epoch": 1.5007587253414263,
+      "grad_norm": 1.067569375038147,
+      "learning_rate": 2.518946187959386e-05,
+      "loss": 0.4262,
+      "step": 495
+    },
+    {
+      "epoch": 1.5159332321699543,
+      "grad_norm": 1.1014666557312012,
+      "learning_rate": 2.507860564471575e-05,
+      "loss": 0.4463,
+      "step": 500
+    },
+    {
+      "epoch": 1.5311077389984824,
+      "grad_norm": 1.0636093616485596,
+      "learning_rate": 2.496673666840735e-05,
+      "loss": 0.4159,
+      "step": 505
+    },
+    {
+      "epoch": 1.5462822458270105,
+      "grad_norm": 0.9945455193519592,
+      "learning_rate": 2.4853866191742177e-05,
+      "loss": 0.4201,
+      "step": 510
+    },
+    {
+      "epoch": 1.5614567526555387,
+      "grad_norm": 1.0836797952651978,
+      "learning_rate": 2.47400055564288e-05,
+      "loss": 0.4312,
+      "step": 515
+    },
+    {
+      "epoch": 1.5766312594840668,
+      "grad_norm": 1.0805171728134155,
+      "learning_rate": 2.4625166203671166e-05,
+      "loss": 0.43,
+      "step": 520
+    },
+    {
+      "epoch": 1.5918057663125948,
+      "grad_norm": 1.305701732635498,
+      "learning_rate": 2.4509359673018933e-05,
+      "loss": 0.3803,
+      "step": 525
+    },
+    {
+      "epoch": 1.606980273141123,
+      "grad_norm": 1.0899828672409058,
+      "learning_rate": 2.439259760120794e-05,
+      "loss": 0.3797,
+      "step": 530
+    },
+    {
+      "epoch": 1.622154779969651,
+      "grad_norm": 1.0438095331192017,
+      "learning_rate": 2.4274891720990886e-05,
+      "loss": 0.384,
+      "step": 535
+    },
+    {
+      "epoch": 1.637329286798179,
+      "grad_norm": 1.0920283794403076,
+      "learning_rate": 2.4156253859958388e-05,
+      "loss": 0.415,
+      "step": 540
+    },
+    {
+      "epoch": 1.6525037936267073,
+      "grad_norm": 1.2114492654800415,
+      "learning_rate": 2.403669593935047e-05,
+      "loss": 0.4028,
+      "step": 545
+    },
+    {
+      "epoch": 1.6676783004552354,
+      "grad_norm": 1.0579054355621338,
+      "learning_rate": 2.3916229972858695e-05,
+      "loss": 0.388,
+      "step": 550
+    },
+    {
+      "epoch": 1.6828528072837634,
+      "grad_norm": 0.9982993006706238,
+      "learning_rate": 2.3794868065418944e-05,
+      "loss": 0.3883,
+      "step": 555
+    },
+    {
+      "epoch": 1.6980273141122915,
+      "grad_norm": 1.06562340259552,
+      "learning_rate": 2.3672622411995095e-05,
+      "loss": 0.3829,
+      "step": 560
+    },
+    {
+      "epoch": 1.7132018209408195,
+      "grad_norm": 1.1990240812301636,
+      "learning_rate": 2.3549505296353605e-05,
+      "loss": 0.4015,
+      "step": 565
+    },
+    {
+      "epoch": 1.7283763277693476,
+      "grad_norm": 1.1103229522705078,
+      "learning_rate": 2.3425529089829166e-05,
+      "loss": 0.3946,
+      "step": 570
+    },
+    {
+      "epoch": 1.7435508345978756,
+      "grad_norm": 1.086736798286438,
+      "learning_rate": 2.330070625008162e-05,
+      "loss": 0.4003,
+      "step": 575
+    },
+    {
+      "epoch": 1.7587253414264037,
+      "grad_norm": 1.0179426670074463,
+      "learning_rate": 2.3175049319844132e-05,
+      "loss": 0.3829,
+      "step": 580
+    },
+    {
+      "epoch": 1.7738998482549317,
+      "grad_norm": 1.1410490274429321,
+      "learning_rate": 2.3048570925662846e-05,
+      "loss": 0.3963,
+      "step": 585
+    },
+    {
+      "epoch": 1.7890743550834598,
+      "grad_norm": 1.1393828392028809,
+      "learning_rate": 2.2921283776628118e-05,
+      "loss": 0.3513,
+      "step": 590
+    },
+    {
+      "epoch": 1.8042488619119879,
+      "grad_norm": 1.034442663192749,
+      "learning_rate": 2.2793200663097453e-05,
+      "loss": 0.3373,
+      "step": 595
+    },
+    {
+      "epoch": 1.819423368740516,
+      "grad_norm": 1.1215357780456543,
+      "learning_rate": 2.266433445541028e-05,
+      "loss": 0.3689,
+      "step": 600
+    },
+    {
+      "epoch": 1.834597875569044,
+      "grad_norm": 1.1622790098190308,
+      "learning_rate": 2.253469810259468e-05,
+      "loss": 0.3734,
+      "step": 605
+    },
+    {
+      "epoch": 1.849772382397572,
+      "grad_norm": 1.1042276620864868,
+      "learning_rate": 2.240430463106621e-05,
+      "loss": 0.3145,
+      "step": 610
+    },
+    {
+      "epoch": 1.8649468892261,
+      "grad_norm": 1.176214575767517,
+      "learning_rate": 2.2273167143318956e-05,
+      "loss": 0.374,
+      "step": 615
+    },
+    {
+      "epoch": 1.8801213960546281,
+      "grad_norm": 1.162898302078247,
+      "learning_rate": 2.2141298816608935e-05,
+      "loss": 0.3398,
+      "step": 620
+    },
+    {
+      "epoch": 1.8952959028831562,
+      "grad_norm": 1.0589183568954468,
+      "learning_rate": 2.200871290163e-05,
+      "loss": 0.3475,
+      "step": 625
+    },
+    {
+      "epoch": 1.9104704097116842,
+      "grad_norm": 1.41371488571167,
+      "learning_rate": 2.1875422721182334e-05,
+      "loss": 0.3689,
+      "step": 630
+    },
+    {
+      "epoch": 1.9256449165402123,
+      "grad_norm": 1.095626711845398,
+      "learning_rate": 2.1741441668833733e-05,
+      "loss": 0.3618,
+      "step": 635
+    },
+    {
+      "epoch": 1.9408194233687404,
+      "grad_norm": 1.1281839609146118,
+      "learning_rate": 2.1606783207573757e-05,
+      "loss": 0.3358,
+      "step": 640
+    },
+    {
+      "epoch": 1.9559939301972686,
+      "grad_norm": 1.1524876356124878,
+      "learning_rate": 2.1471460868460922e-05,
+      "loss": 0.3513,
+      "step": 645
+    },
+    {
+      "epoch": 1.9711684370257967,
+      "grad_norm": 1.1511465311050415,
+      "learning_rate": 2.1335488249263012e-05,
+      "loss": 0.3236,
+      "step": 650
+    },
+    {
+      "epoch": 1.9863429438543247,
+      "grad_norm": 1.0634710788726807,
+      "learning_rate": 2.1198879013090763e-05,
+      "loss": 0.3491,
+      "step": 655
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.505767583847046,
+      "learning_rate": 2.106164688702488e-05,
+      "loss": 0.3141,
+      "step": 660
+    },
+    {
+      "epoch": 2.015174506828528,
+      "grad_norm": 1.161057472229004,
+      "learning_rate": 2.0923805660736732e-05,
+      "loss": 0.2651,
+      "step": 665
+    },
+    {
+      "epoch": 2.030349013657056,
+      "grad_norm": 1.1279088258743286,
+      "learning_rate": 2.0785369185102684e-05,
+      "loss": 0.2398,
+      "step": 670
+    },
+    {
+      "epoch": 2.045523520485584,
+      "grad_norm": 0.9957617521286011,
+      "learning_rate": 2.0646351370812293e-05,
+      "loss": 0.2783,
+      "step": 675
+    },
+    {
+      "epoch": 2.0606980273141122,
+      "grad_norm": 1.161773681640625,
+      "learning_rate": 2.050676618697052e-05,
+      "loss": 0.2599,
+      "step": 680
+    },
+    {
+      "epoch": 2.0758725341426403,
+      "grad_norm": 1.0346007347106934,
+      "learning_rate": 2.0366627659694043e-05,
+      "loss": 0.232,
+      "step": 685
+    },
+    {
+      "epoch": 2.0910470409711683,
+      "grad_norm": 1.2492696046829224,
+      "learning_rate": 2.022594987070185e-05,
+      "loss": 0.2544,
+      "step": 690
+    },
+    {
+      "epoch": 2.1062215477996964,
+      "grad_norm": 1.0093607902526855,
+      "learning_rate": 2.0084746955900266e-05,
+      "loss": 0.2732,
+      "step": 695
+    },
+    {
+      "epoch": 2.1213960546282244,
+      "grad_norm": 1.421979546546936,
+      "learning_rate": 1.994303310396251e-05,
+      "loss": 0.2437,
+      "step": 700
+    },
+    {
+      "epoch": 2.1365705614567525,
+      "grad_norm": 1.0597381591796875,
+      "learning_rate": 1.9800822554902924e-05,
+      "loss": 0.257,
+      "step": 705
+    },
+    {
+      "epoch": 2.1517450682852806,
+      "grad_norm": 1.0941869020462036,
+      "learning_rate": 1.9658129598646136e-05,
+      "loss": 0.2558,
+      "step": 710
+    },
+    {
+      "epoch": 2.1669195751138086,
+      "grad_norm": 1.3752810955047607,
+      "learning_rate": 1.9514968573591082e-05,
+      "loss": 0.2574,
+      "step": 715
+    },
+    {
+      "epoch": 2.1820940819423367,
+      "grad_norm": 0.9760404825210571,
+      "learning_rate": 1.9371353865170285e-05,
+      "loss": 0.2558,
+      "step": 720
+    },
+    {
+      "epoch": 2.197268588770865,
+      "grad_norm": 1.097494125366211,
+      "learning_rate": 1.9227299904404293e-05,
+      "loss": 0.2873,
+      "step": 725
+    },
+    {
+      "epoch": 2.212443095599393,
+      "grad_norm": 1.0384601354599,
+      "learning_rate": 1.9082821166451614e-05,
+      "loss": 0.2524,
+      "step": 730
+    },
+    {
+      "epoch": 2.2276176024279213,
+      "grad_norm": 1.1527771949768066,
+      "learning_rate": 1.8937932169154192e-05,
+      "loss": 0.2453,
+      "step": 735
+    },
+    {
+      "epoch": 2.2427921092564493,
+      "grad_norm": 1.0804898738861084,
+      "learning_rate": 1.8792647471578575e-05,
+      "loss": 0.2267,
+      "step": 740
+    },
+    {
+      "epoch": 2.2579666160849774,
+      "grad_norm": 1.0630391836166382,
+      "learning_rate": 1.8646981672552984e-05,
+      "loss": 0.2463,
+      "step": 745
+    },
+    {
+      "epoch": 2.2731411229135055,
+      "grad_norm": 1.1511931419372559,
+      "learning_rate": 1.8500949409200328e-05,
+      "loss": 0.2264,
+      "step": 750
+    },
+    {
+      "epoch": 2.2883156297420335,
+      "grad_norm": 1.0880099534988403,
+      "learning_rate": 1.835456535546743e-05,
+      "loss": 0.2229,
+      "step": 755
+    },
+    {
+      "epoch": 2.3034901365705616,
+      "grad_norm": 1.0572727918624878,
+      "learning_rate": 1.8207844220650513e-05,
+      "loss": 0.2103,
+      "step": 760
+    },
+    {
+      "epoch": 2.3186646433990896,
+      "grad_norm": 0.9872831702232361,
+      "learning_rate": 1.806080074791715e-05,
+      "loss": 0.2102,
+      "step": 765
+    },
+    {
+      "epoch": 2.3338391502276177,
+      "grad_norm": 1.1545583009719849,
+      "learning_rate": 1.7913449712824813e-05,
+      "loss": 0.2388,
+      "step": 770
+    },
+    {
+      "epoch": 2.3490136570561457,
+      "grad_norm": 1.0736984014511108,
+      "learning_rate": 1.7765805921836147e-05,
+      "loss": 0.2146,
+      "step": 775
+    },
+    {
+      "epoch": 2.364188163884674,
+      "grad_norm": 0.9836410284042358,
+      "learning_rate": 1.7617884210831165e-05,
+      "loss": 0.2078,
+      "step": 780
+    },
+    {
+      "epoch": 2.379362670713202,
+      "grad_norm": 1.0019104480743408,
+      "learning_rate": 1.7469699443616478e-05,
+      "loss": 0.2232,
+      "step": 785
+    },
+    {
+      "epoch": 2.39453717754173,
+      "grad_norm": 1.1876050233840942,
+      "learning_rate": 1.7321266510431707e-05,
+      "loss": 0.2428,
+      "step": 790
+    },
+    {
+      "epoch": 2.409711684370258,
+      "grad_norm": 1.09200918674469,
+      "learning_rate": 1.7172600326453253e-05,
+      "loss": 0.2331,
+      "step": 795
+    },
+    {
+      "epoch": 2.424886191198786,
+      "grad_norm": 1.0480817556381226,
+      "learning_rate": 1.702371583029555e-05,
+      "loss": 0.209,
+      "step": 800
+    },
+    {
+      "epoch": 2.440060698027314,
+      "grad_norm": 1.287725567817688,
+      "learning_rate": 1.687462798250998e-05,
+      "loss": 0.225,
+      "step": 805
+    },
+    {
+      "epoch": 2.455235204855842,
+      "grad_norm": 1.162737488746643,
+      "learning_rate": 1.672535176408156e-05,
+      "loss": 0.2031,
+      "step": 810
+    },
+    {
+      "epoch": 2.47040971168437,
+      "grad_norm": 0.9590432047843933,
+      "learning_rate": 1.657590217492361e-05,
+      "loss": 0.2049,
+      "step": 815
+    },
+    {
+      "epoch": 2.4855842185128982,
+      "grad_norm": 1.1492325067520142,
+      "learning_rate": 1.6426294232370468e-05,
+      "loss": 0.2181,
+      "step": 820
+    },
+    {
+      "epoch": 2.5007587253414263,
+      "grad_norm": 1.1130379438400269,
+      "learning_rate": 1.6276542969668502e-05,
+      "loss": 0.2078,
+      "step": 825
+    },
+    {
+      "epoch": 2.5159332321699543,
+      "grad_norm": 0.9611683487892151,
+      "learning_rate": 1.612666343446551e-05,
+      "loss": 0.2099,
+      "step": 830
+    },
+    {
+      "epoch": 2.5311077389984824,
+      "grad_norm": 1.2248095273971558,
+      "learning_rate": 1.597667068729865e-05,
+      "loss": 0.1976,
+      "step": 835
+    },
+    {
+      "epoch": 2.5462822458270105,
+      "grad_norm": 1.0842514038085938,
+      "learning_rate": 1.582657980008111e-05,
+      "loss": 0.1937,
+      "step": 840
+    },
+    {
+      "epoch": 2.5614567526555385,
+      "grad_norm": 1.0177435874938965,
+      "learning_rate": 1.56764058545876e-05,
+      "loss": 0.2114,
+      "step": 845
+    },
+    {
+      "epoch": 2.576631259484067,
+      "grad_norm": 1.003584384918213,
+      "learning_rate": 1.5526163940938892e-05,
+      "loss": 0.2152,
+      "step": 850
+    },
+    {
+      "epoch": 2.5918057663125946,
+      "grad_norm": 1.078927993774414,
+      "learning_rate": 1.537586915608549e-05,
+      "loss": 0.2134,
+      "step": 855
+    },
+    {
+      "epoch": 2.606980273141123,
+      "grad_norm": 1.1085734367370605,
+      "learning_rate": 1.5225536602290614e-05,
+      "loss": 0.1987,
+      "step": 860
+    },
+    {
+      "epoch": 2.6221547799696507,
+      "grad_norm": 1.0498510599136353,
+      "learning_rate": 1.5075181385612671e-05,
+      "loss": 0.1907,
+      "step": 865
+    },
+    {
+      "epoch": 2.6373292867981792,
+      "grad_norm": 1.0778013467788696,
+      "learning_rate": 1.4924818614387334e-05,
+      "loss": 0.1903,
+      "step": 870
+    },
+    {
+      "epoch": 2.6525037936267073,
+      "grad_norm": 1.304302453994751,
+      "learning_rate": 1.477446339770939e-05,
+      "loss": 0.1916,
+      "step": 875
+    },
+    {
+      "epoch": 2.6676783004552354,
+      "grad_norm": 0.9650601148605347,
+      "learning_rate": 1.4624130843914512e-05,
+      "loss": 0.1788,
+      "step": 880
+    },
+    {
+      "epoch": 2.6828528072837634,
+      "grad_norm": 1.2137047052383423,
+      "learning_rate": 1.4473836059061107e-05,
+      "loss": 0.2087,
+      "step": 885
+    },
+    {
+      "epoch": 2.6980273141122915,
+      "grad_norm": 1.222299337387085,
+      "learning_rate": 1.43235941454124e-05,
+      "loss": 0.1956,
+      "step": 890
+    },
+    {
+      "epoch": 2.7132018209408195,
+      "grad_norm": 1.2385069131851196,
+      "learning_rate": 1.4173420199918893e-05,
+      "loss": 0.1995,
+      "step": 895
+    },
+    {
+      "epoch": 2.7283763277693476,
+      "grad_norm": 1.1916569471359253,
+      "learning_rate": 1.402332931270135e-05,
+      "loss": 0.1818,
+      "step": 900
+    },
+    {
+      "epoch": 2.7435508345978756,
+      "grad_norm": 0.9629146456718445,
+      "learning_rate": 1.387333656553449e-05,
+      "loss": 0.1813,
+      "step": 905
+    },
+    {
+      "epoch": 2.7587253414264037,
+      "grad_norm": 1.1919859647750854,
+      "learning_rate": 1.3723457030331498e-05,
+      "loss": 0.1877,
+      "step": 910
+    },
+    {
+      "epoch": 2.7738998482549317,
+      "grad_norm": 1.1693427562713623,
+      "learning_rate": 1.3573705767629536e-05,
+      "loss": 0.2139,
+      "step": 915
+    },
+    {
+      "epoch": 2.78907435508346,
+      "grad_norm": 1.0847316980361938,
+      "learning_rate": 1.3424097825076394e-05,
+      "loss": 0.1796,
+      "step": 920
+    },
+    {
+      "epoch": 2.804248861911988,
+      "grad_norm": 1.064889907836914,
+      "learning_rate": 1.3274648235918442e-05,
+      "loss": 0.1743,
+      "step": 925
+    },
+    {
+      "epoch": 2.819423368740516,
+      "grad_norm": 1.2099905014038086,
+      "learning_rate": 1.3125372017490026e-05,
+      "loss": 0.178,
+      "step": 930
+    },
+    {
+      "epoch": 2.834597875569044,
+      "grad_norm": 1.0362218618392944,
+      "learning_rate": 1.2976284169704455e-05,
+      "loss": 0.2107,
+      "step": 935
+    },
+    {
+      "epoch": 2.849772382397572,
+      "grad_norm": 1.3111770153045654,
+      "learning_rate": 1.282739967354675e-05,
+      "loss": 0.1867,
+      "step": 940
+    },
+    {
+      "epoch": 2.8649468892261,
+      "grad_norm": 1.3245251178741455,
+      "learning_rate": 1.2678733489568292e-05,
+      "loss": 0.1815,
+      "step": 945
+    },
+    {
+      "epoch": 2.880121396054628,
+      "grad_norm": 0.9746212959289551,
+      "learning_rate": 1.2530300556383521e-05,
+      "loss": 0.1794,
+      "step": 950
+    },
+    {
+      "epoch": 2.895295902883156,
+      "grad_norm": 1.1812666654586792,
+      "learning_rate": 1.2382115789168834e-05,
+      "loss": 0.1815,
+      "step": 955
+    },
+    {
+      "epoch": 2.9104704097116842,
+      "grad_norm": 0.9376991987228394,
+      "learning_rate": 1.223419407816386e-05,
+      "loss": 0.1714,
+      "step": 960
+    },
+    {
+      "epoch": 2.9256449165402123,
+      "grad_norm": 1.120914340019226,
+      "learning_rate": 1.208655028717519e-05,
+      "loss": 0.1703,
+      "step": 965
+    },
+    {
+      "epoch": 2.9408194233687404,
+      "grad_norm": 1.2715625762939453,
+      "learning_rate": 1.1939199252082851e-05,
+      "loss": 0.1631,
+      "step": 970
+    },
+    {
+      "epoch": 2.955993930197269,
+      "grad_norm": 1.1299712657928467,
+      "learning_rate": 1.179215577934949e-05,
+      "loss": 0.1771,
+      "step": 975
+    },
+    {
+      "epoch": 2.9711684370257965,
+      "grad_norm": 1.0479075908660889,
+      "learning_rate": 1.1645434644532572e-05,
+      "loss": 0.1855,
+      "step": 980
+    },
+    {
+      "epoch": 2.986342943854325,
+      "grad_norm": 1.1739304065704346,
+      "learning_rate": 1.1499050590799675e-05,
+      "loss": 0.1697,
+      "step": 985
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.351037859916687,
+      "learning_rate": 1.1353018327447017e-05,
+      "loss": 0.1856,
+      "step": 990
+    },
+    {
+      "epoch": 3.015174506828528,
+      "grad_norm": 1.0247342586517334,
+      "learning_rate": 1.1207352528421424e-05,
+      "loss": 0.1543,
+      "step": 995
+    },
+    {
+      "epoch": 3.030349013657056,
+      "grad_norm": 0.9216309189796448,
+      "learning_rate": 1.1062067830845807e-05,
+      "loss": 0.1261,
+      "step": 1000
+    },
+    {
+      "epoch": 3.045523520485584,
+      "grad_norm": 0.9509443640708923,
+      "learning_rate": 1.091717883354839e-05,
+      "loss": 0.1147,
+      "step": 1005
+    },
+    {
+      "epoch": 3.0606980273141122,
+      "grad_norm": 1.0392087697982788,
+      "learning_rate": 1.0772700095595713e-05,
+      "loss": 0.1212,
+      "step": 1010
+    },
+    {
+      "epoch": 3.0758725341426403,
+      "grad_norm": 0.8555918335914612,
+      "learning_rate": 1.062864613482972e-05,
+      "loss": 0.1107,
+      "step": 1015
+    },
+    {
+      "epoch": 3.0910470409711683,
+      "grad_norm": 0.873559296131134,
+      "learning_rate": 1.048503142640892e-05,
+      "loss": 0.1257,
+      "step": 1020
+    },
+    {
+      "epoch": 3.1062215477996964,
+      "grad_norm": 0.9564691185951233,
+      "learning_rate": 1.034187040135387e-05,
+      "loss": 0.1284,
+      "step": 1025
+    },
+    {
+      "epoch": 3.1213960546282244,
+      "grad_norm": 1.031790852546692,
+      "learning_rate": 1.0199177445097075e-05,
+      "loss": 0.1249,
+      "step": 1030
+    },
+    {
+      "epoch": 3.1365705614567525,
+      "grad_norm": 1.0104645490646362,
+      "learning_rate": 1.0056966896037494e-05,
+      "loss": 0.125,
+      "step": 1035
+    },
+    {
+      "epoch": 3.1517450682852806,
+      "grad_norm": 1.0347521305084229,
+      "learning_rate": 9.915253044099731e-06,
+      "loss": 0.1366,
+      "step": 1040
+    },
+    {
+      "epoch": 3.1669195751138086,
+      "grad_norm": 0.9634484648704529,
+      "learning_rate": 9.77405012929815e-06,
+      "loss": 0.1225,
+      "step": 1045
+    },
+    {
+      "epoch": 3.1820940819423367,
+      "grad_norm": 1.00877046585083,
+      "learning_rate": 9.633372340305966e-06,
+      "loss": 0.1312,
+      "step": 1050
+    },
+    {
+      "epoch": 3.197268588770865,
+      "grad_norm": 1.2125271558761597,
+      "learning_rate": 9.493233813029485e-06,
+      "loss": 0.1188,
+      "step": 1055
+    },
+    {
+      "epoch": 3.212443095599393,
+      "grad_norm": 1.0163218975067139,
+      "learning_rate": 9.353648629187709e-06,
+      "loss": 0.1231,
+      "step": 1060
+    },
+    {
+      "epoch": 3.2276176024279213,
+      "grad_norm": 0.9734249711036682,
+      "learning_rate": 9.214630814897317e-06,
+      "loss": 0.1193,
+      "step": 1065
+    },
+    {
+      "epoch": 3.2427921092564493,
+      "grad_norm": 0.9149922728538513,
+      "learning_rate": 9.076194339263268e-06,
+      "loss": 0.1325,
+      "step": 1070
+    },
+    {
+      "epoch": 3.2579666160849774,
+      "grad_norm": 0.9013402462005615,
+      "learning_rate": 8.938353112975126e-06,
+      "loss": 0.1308,
+      "step": 1075
+    },
+    {
+      "epoch": 3.2731411229135055,
+      "grad_norm": 1.0030301809310913,
+      "learning_rate": 8.801120986909242e-06,
+      "loss": 0.1167,
+      "step": 1080
+    },
+    {
+      "epoch": 3.2883156297420335,
+      "grad_norm": 0.8948475122451782,
+      "learning_rate": 8.664511750736988e-06,
+      "loss": 0.1346,
+      "step": 1085
+    },
+    {
+      "epoch": 3.3034901365705616,
+      "grad_norm": 0.9730858206748962,
+      "learning_rate": 8.52853913153908e-06,
+      "loss": 0.1194,
+      "step": 1090
+    },
+    {
+      "epoch": 3.3186646433990896,
+      "grad_norm": 1.0079647302627563,
+      "learning_rate": 8.393216792426243e-06,
+      "loss": 0.1435,
+      "step": 1095
+    },
+    {
+      "epoch": 3.3338391502276177,
+      "grad_norm": 0.6983161568641663,
+      "learning_rate": 8.258558331166271e-06,
+      "loss": 0.11,
+      "step": 1100
+    },
+    {
+      "epoch": 3.3490136570561457,
+      "grad_norm": 0.9095783233642578,
+      "learning_rate": 8.124577278817668e-06,
+      "loss": 0.1032,
+      "step": 1105
+    },
+    {
+      "epoch": 3.364188163884674,
+      "grad_norm": 0.9054252505302429,
+      "learning_rate": 7.991287098370004e-06,
+      "loss": 0.1208,
+      "step": 1110
+    },
+    {
+      "epoch": 3.379362670713202,
+      "grad_norm": 1.133405327796936,
+      "learning_rate": 7.858701183391064e-06,
+      "loss": 0.1105,
+      "step": 1115
+    },
+    {
+      "epoch": 3.39453717754173,
+      "grad_norm": 0.9282613396644592,
+      "learning_rate": 7.726832856681047e-06,
+      "loss": 0.1132,
+      "step": 1120
+    },
+    {
+      "epoch": 3.409711684370258,
+      "grad_norm": 0.8912308812141418,
+      "learning_rate": 7.59569536893379e-06,
+      "loss": 0.116,
+      "step": 1125
+    },
+    {
+      "epoch": 3.424886191198786,
+      "grad_norm": 1.1447117328643799,
+      "learning_rate": 7.465301897405322e-06,
+      "loss": 0.1,
+      "step": 1130
+    },
+    {
+      "epoch": 3.440060698027314,
+      "grad_norm": 0.8710254430770874,
+      "learning_rate": 7.33566554458972e-06,
+      "loss": 0.1139,
+      "step": 1135
+    },
+    {
+      "epoch": 3.455235204855842,
+      "grad_norm": 0.9299088716506958,
+      "learning_rate": 7.206799336902558e-06,
+      "loss": 0.1067,
+      "step": 1140
+    },
+    {
+      "epoch": 3.47040971168437,
+      "grad_norm": 1.0533580780029297,
+      "learning_rate": 7.07871622337189e-06,
+      "loss": 0.1041,
+      "step": 1145
+    },
+    {
+      "epoch": 3.4855842185128982,
+      "grad_norm": 0.8250032067298889,
+      "learning_rate": 6.9514290743371575e-06,
+      "loss": 0.1082,
+      "step": 1150
+    },
+    {
+      "epoch": 3.5007587253414263,
+      "grad_norm": 0.8535242676734924,
+      "learning_rate": 6.824950680155871e-06,
+      "loss": 0.1141,
+      "step": 1155
+    },
+    {
+      "epoch": 3.5159332321699543,
+      "grad_norm": 0.9096361994743347,
+      "learning_rate": 6.69929374991838e-06,
+      "loss": 0.1144,
+      "step": 1160
+    },
+    {
+      "epoch": 3.5311077389984824,
+      "grad_norm": 0.9085047245025635,
+      "learning_rate": 6.5744709101708364e-06,
+      "loss": 0.1005,
+      "step": 1165
+    },
+    {
+      "epoch": 3.5462822458270105,
+      "grad_norm": 0.8639160990715027,
+      "learning_rate": 6.450494703646398e-06,
+      "loss": 0.102,
+      "step": 1170
+    },
+    {
+      "epoch": 3.5614567526555385,
+      "grad_norm": 0.8191717267036438,
+      "learning_rate": 6.327377588004907e-06,
+      "loss": 0.0963,
+      "step": 1175
+    },
+    {
+      "epoch": 3.576631259484067,
+      "grad_norm": 0.8398914933204651,
+      "learning_rate": 6.205131934581056e-06,
+      "loss": 0.1162,
+      "step": 1180
+    },
+    {
+      "epoch": 3.5918057663125946,
+      "grad_norm": 0.848587691783905,
+      "learning_rate": 6.083770027141313e-06,
+      "loss": 0.1177,
+      "step": 1185
+    },
+    {
+      "epoch": 3.606980273141123,
+      "grad_norm": 0.9553687572479248,
+      "learning_rate": 5.963304060649531e-06,
+      "loss": 0.1069,
+      "step": 1190
+    },
+    {
+      "epoch": 3.6221547799696507,
+      "grad_norm": 0.8801453113555908,
+      "learning_rate": 5.843746140041616e-06,
+      "loss": 0.1211,
+      "step": 1195
+    },
+    {
+      "epoch": 3.6373292867981792,
+      "grad_norm": 1.2324217557907104,
+      "learning_rate": 5.725108279009115e-06,
+      "loss": 0.1221,
+      "step": 1200
+    },
+    {
+      "epoch": 3.6525037936267073,
+      "grad_norm": 0.994218111038208,
+      "learning_rate": 5.60740239879206e-06,
+      "loss": 0.0946,
+      "step": 1205
+    },
+    {
+      "epoch": 3.6676783004552354,
+      "grad_norm": 0.9351308941841125,
+      "learning_rate": 5.490640326981069e-06,
+      "loss": 0.1059,
+      "step": 1210
+    },
+    {
+      "epoch": 3.6828528072837634,
+      "grad_norm": 0.95694500207901,
+      "learning_rate": 5.374833796328833e-06,
+      "loss": 0.1016,
+      "step": 1215
+    },
+    {
+      "epoch": 3.6980273141122915,
+      "grad_norm": 0.7896385192871094,
+      "learning_rate": 5.2599944435712e-06,
+      "loss": 0.1004,
+      "step": 1220
+    },
+    {
+      "epoch": 3.7132018209408195,
+      "grad_norm": 0.7855644226074219,
+      "learning_rate": 5.14613380825782e-06,
+      "loss": 0.0958,
+      "step": 1225
+    },
+    {
+      "epoch": 3.7283763277693476,
+      "grad_norm": 0.7871450781822205,
+      "learning_rate": 5.033263331592655e-06,
+      "loss": 0.1076,
+      "step": 1230
+    },
+    {
+      "epoch": 3.7435508345978756,
+      "grad_norm": 0.8118149042129517,
+      "learning_rate": 4.921394355284251e-06,
+      "loss": 0.1245,
+      "step": 1235
+    },
+    {
+      "epoch": 3.7587253414264037,
+      "grad_norm": 0.8458611369132996,
+      "learning_rate": 4.8105381204061465e-06,
+      "loss": 0.0993,
+      "step": 1240
+    },
+    {
+      "epoch": 3.7738998482549317,
+      "grad_norm": 1.4438375234603882,
+      "learning_rate": 4.700705766267262e-06,
+      "loss": 0.1019,
+      "step": 1245
+    },
+    {
+      "epoch": 3.78907435508346,
+      "grad_norm": 0.7945970892906189,
+      "learning_rate": 4.591908329292619e-06,
+      "loss": 0.1123,
+      "step": 1250
+    },
+    {
+      "epoch": 3.804248861911988,
+      "grad_norm": 0.8581335544586182,
+      "learning_rate": 4.484156741914315e-06,
+      "loss": 0.0937,
+      "step": 1255
+    },
+    {
+      "epoch": 3.819423368740516,
+      "grad_norm": 0.8385892510414124,
+      "learning_rate": 4.377461831473022e-06,
+      "loss": 0.1047,
+      "step": 1260
+    },
+    {
+      "epoch": 3.834597875569044,
+      "grad_norm": 0.9302660226821899,
+      "learning_rate": 4.271834319129972e-06,
+      "loss": 0.1081,
+      "step": 1265
+    },
+    {
+      "epoch": 3.849772382397572,
+      "grad_norm": 0.8506367206573486,
+      "learning_rate": 4.167284818789681e-06,
+      "loss": 0.099,
+      "step": 1270
+    },
+    {
+      "epoch": 3.8649468892261,
+      "grad_norm": 0.802346408367157,
+      "learning_rate": 4.063823836033407e-06,
+      "loss": 0.0998,
+      "step": 1275
+    },
+    {
+      "epoch": 3.880121396054628,
+      "grad_norm": 0.8249574303627014,
+      "learning_rate": 3.9614617670634885e-06,
+      "loss": 0.101,
+      "step": 1280
+    },
+    {
+      "epoch": 3.895295902883156,
+      "grad_norm": 1.0073894262313843,
+      "learning_rate": 3.8602088976587175e-06,
+      "loss": 0.1037,
+      "step": 1285
+    },
+    {
+      "epoch": 3.9104704097116842,
+      "grad_norm": 0.8201082348823547,
+      "learning_rate": 3.7600754021407537e-06,
+      "loss": 0.1002,
+      "step": 1290
+    },
+    {
+      "epoch": 3.9256449165402123,
+      "grad_norm": 0.9376416802406311,
+      "learning_rate": 3.6610713423517927e-06,
+      "loss": 0.0853,
+      "step": 1295
+    },
+    {
+      "epoch": 3.9408194233687404,
+      "grad_norm": 0.7816200256347656,
+      "learning_rate": 3.5632066666434794e-06,
+      "loss": 0.0953,
+      "step": 1300
+    },
+    {
+      "epoch": 3.955993930197269,
+      "grad_norm": 0.8962005376815796,
+      "learning_rate": 3.4664912088772865e-06,
+      "loss": 0.0943,
+      "step": 1305
+    },
+    {
+      "epoch": 3.9711684370257965,
+      "grad_norm": 0.894721269607544,
+      "learning_rate": 3.3709346874363295e-06,
+      "loss": 0.0959,
+      "step": 1310
+    },
+    {
+      "epoch": 3.986342943854325,
+      "grad_norm": 0.7250496745109558,
+      "learning_rate": 3.2765467042488607e-06,
+      "loss": 0.0927,
+      "step": 1315
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.2268779277801514,
+      "learning_rate": 3.18333674382339e-06,
+      "loss": 0.0973,
+      "step": 1320
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1650,
+  "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": 1.8977110903058596e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

21_128_e5_3e-5/checkpoint-1320/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:749f33b30b2b8c6910d60865d6784495303d813ca8f33cd1ddafcfa2fe3c03bf
+size 7736

21_128_e5_3e-5/checkpoint-1320/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)

21_128_e5_3e-5/checkpoint-1650/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

21_128_e5_3e-5/checkpoint-1650/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": [
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "v_proj",
+    "o_proj",
+    "k_proj",
+    "q_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

21_128_e5_3e-5/checkpoint-1650/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe52a5631e3060909590aefdf244241909bc9bebb365020cf7c9c1b4e6117e0f
+size 791751704

21_128_e5_3e-5/checkpoint-1650/latest

@@ -0,0 +1 @@
+global_step1650

21_128_e5_3e-5/checkpoint-1650/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:355defe31d0fcd8a4b05f9d1daa623062af62be4d7d2f936dfdb40e3f6b9b907
+size 15920

21_128_e5_3e-5/checkpoint-1650/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:148780ce4cfbf6ee7c7ee2c0edc674a9d4fe41a9df6bdfa15a4ce20466210dc1
+size 15920

21_128_e5_3e-5/checkpoint-1650/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d2cafc020176491532b96906d14b78cd37e87f5a6f14b036d4c5f8860a807e4
+size 15920

21_128_e5_3e-5/checkpoint-1650/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:187f3da01f2e98c7e7d71a10e9e80aeca9e954d5e5e935b8224a3377f73a14ee
+size 15920

21_128_e5_3e-5/checkpoint-1650/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c781d794eb5564a795dd53cc876b32983d4ebb5088ca73ffd23f4ff2e6bee5e
+size 15920

21_128_e5_3e-5/checkpoint-1650/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e942a2288ca9ccabdada8e93d2980bae38b454fa6ac122554bb42e84cd9819e
+size 15920

21_128_e5_3e-5/checkpoint-1650/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78c229e67926a7d26aead1c1afb587bf2aabef4e2fc5cbc7b6a31f2e3d49bffd
+size 15920

21_128_e5_3e-5/checkpoint-1650/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c5bc9cf11f1769fa2b3f18674d810608d23922f80a2f9b31772b9f7c464513d
+size 15920

21_128_e5_3e-5/checkpoint-1650/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:595680128de7f248a8bb25b29aeda8beaf5355577ded1be288be446a3292c735
+size 1064

21_128_e5_3e-5/checkpoint-1650/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
+  }
+}

21_128_e5_3e-5/checkpoint-1650/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
+}

21_128_e5_3e-5/checkpoint-1650/trainer_state.json

@@ -0,0 +1,2344 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.0,
+  "eval_steps": 500,
+  "global_step": 1650,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.015174506828528073,
+      "grad_norm": 1.1642332077026367,
+      "learning_rate": 1.4457831325301207e-06,
+      "loss": 1.3119,
+      "step": 5
+    },
+    {
+      "epoch": 0.030349013657056147,
+      "grad_norm": 0.857893705368042,
+      "learning_rate": 3.2530120481927713e-06,
+      "loss": 1.2575,
+      "step": 10
+    },
+    {
+      "epoch": 0.04552352048558422,
+      "grad_norm": 0.6741542220115662,
+      "learning_rate": 5.060240963855422e-06,
+      "loss": 1.2179,
+      "step": 15
+    },
+    {
+      "epoch": 0.06069802731411229,
+      "grad_norm": 0.5497764348983765,
+      "learning_rate": 6.867469879518072e-06,
+      "loss": 1.2422,
+      "step": 20
+    },
+    {
+      "epoch": 0.07587253414264036,
+      "grad_norm": 0.5862205624580383,
+      "learning_rate": 8.674698795180722e-06,
+      "loss": 1.2837,
+      "step": 25
+    },
+    {
+      "epoch": 0.09104704097116843,
+      "grad_norm": 0.5110287070274353,
+      "learning_rate": 1.0481927710843374e-05,
+      "loss": 1.1704,
+      "step": 30
+    },
+    {
+      "epoch": 0.1062215477996965,
+      "grad_norm": 0.551201343536377,
+      "learning_rate": 1.2289156626506024e-05,
+      "loss": 1.1677,
+      "step": 35
+    },
+    {
+      "epoch": 0.12139605462822459,
+      "grad_norm": 0.5148503184318542,
+      "learning_rate": 1.4096385542168676e-05,
+      "loss": 1.116,
+      "step": 40
+    },
+    {
+      "epoch": 0.13657056145675264,
+      "grad_norm": 0.5068604946136475,
+      "learning_rate": 1.5903614457831326e-05,
+      "loss": 1.1516,
+      "step": 45
+    },
+    {
+      "epoch": 0.15174506828528073,
+      "grad_norm": 0.5807440280914307,
+      "learning_rate": 1.7710843373493978e-05,
+      "loss": 1.1822,
+      "step": 50
+    },
+    {
+      "epoch": 0.1669195751138088,
+      "grad_norm": 0.48767879605293274,
+      "learning_rate": 1.9518072289156627e-05,
+      "loss": 1.1189,
+      "step": 55
+    },
+    {
+      "epoch": 0.18209408194233687,
+      "grad_norm": 0.5024222135543823,
+      "learning_rate": 2.1325301204819275e-05,
+      "loss": 1.1291,
+      "step": 60
+    },
+    {
+      "epoch": 0.19726858877086495,
+      "grad_norm": 0.5667535662651062,
+      "learning_rate": 2.313253012048193e-05,
+      "loss": 1.1279,
+      "step": 65
+    },
+    {
+      "epoch": 0.212443095599393,
+      "grad_norm": 0.5600463151931763,
+      "learning_rate": 2.493975903614458e-05,
+      "loss": 1.1465,
+      "step": 70
+    },
+    {
+      "epoch": 0.2276176024279211,
+      "grad_norm": 0.5471869111061096,
+      "learning_rate": 2.674698795180723e-05,
+      "loss": 1.0742,
+      "step": 75
+    },
+    {
+      "epoch": 0.24279210925644917,
+      "grad_norm": 0.5258970856666565,
+      "learning_rate": 2.855421686746988e-05,
+      "loss": 1.0863,
+      "step": 80
+    },
+    {
+      "epoch": 0.25796661608497723,
+      "grad_norm": 0.5476792454719543,
+      "learning_rate": 2.9999969854473704e-05,
+      "loss": 1.0845,
+      "step": 85
+    },
+    {
+      "epoch": 0.2731411229135053,
+      "grad_norm": 0.5853098034858704,
+      "learning_rate": 2.9998914773775897e-05,
+      "loss": 1.0725,
+      "step": 90
+    },
+    {
+      "epoch": 0.2883156297420334,
+      "grad_norm": 0.5045304894447327,
+      "learning_rate": 2.9996352537928224e-05,
+      "loss": 1.0087,
+      "step": 95
+    },
+    {
+      "epoch": 0.30349013657056145,
+      "grad_norm": 0.5672217607498169,
+      "learning_rate": 2.9992283404395112e-05,
+      "loss": 1.0539,
+      "step": 100
+    },
+    {
+      "epoch": 0.3186646433990895,
+      "grad_norm": 0.5382909178733826,
+      "learning_rate": 2.9986707782060517e-05,
+      "loss": 1.0188,
+      "step": 105
+    },
+    {
+      "epoch": 0.3338391502276176,
+      "grad_norm": 0.7915878891944885,
+      "learning_rate": 2.997962623118683e-05,
+      "loss": 0.9988,
+      "step": 110
+    },
+    {
+      "epoch": 0.3490136570561457,
+      "grad_norm": 0.5800904035568237,
+      "learning_rate": 2.9971039463358594e-05,
+      "loss": 1.0071,
+      "step": 115
+    },
+    {
+      "epoch": 0.36418816388467373,
+      "grad_norm": 0.6986746191978455,
+      "learning_rate": 2.9960948341410993e-05,
+      "loss": 1.012,
+      "step": 120
+    },
+    {
+      "epoch": 0.37936267071320184,
+      "grad_norm": 0.6434286832809448,
+      "learning_rate": 2.994935387934314e-05,
+      "loss": 0.9365,
+      "step": 125
+    },
+    {
+      "epoch": 0.3945371775417299,
+      "grad_norm": 0.6894506216049194,
+      "learning_rate": 2.9936257242216216e-05,
+      "loss": 0.9749,
+      "step": 130
+    },
+    {
+      "epoch": 0.40971168437025796,
+      "grad_norm": 0.6694482564926147,
+      "learning_rate": 2.9921659746036364e-05,
+      "loss": 0.9807,
+      "step": 135
+    },
+    {
+      "epoch": 0.424886191198786,
+      "grad_norm": 0.5935041904449463,
+      "learning_rate": 2.9905562857622482e-05,
+      "loss": 0.9575,
+      "step": 140
+    },
+    {
+      "epoch": 0.4400606980273141,
+      "grad_norm": 0.8535542488098145,
+      "learning_rate": 2.9887968194458807e-05,
+      "loss": 0.9219,
+      "step": 145
+    },
+    {
+      "epoch": 0.4552352048558422,
+      "grad_norm": 0.775467038154602,
+      "learning_rate": 2.986887752453239e-05,
+      "loss": 0.9602,
+      "step": 150
+    },
+    {
+      "epoch": 0.47040971168437024,
+      "grad_norm": 0.7054561376571655,
+      "learning_rate": 2.984829276615546e-05,
+      "loss": 0.8944,
+      "step": 155
+    },
+    {
+      "epoch": 0.48558421851289835,
+      "grad_norm": 0.6600781679153442,
+      "learning_rate": 2.9826215987772642e-05,
+      "loss": 0.9139,
+      "step": 160
+    },
+    {
+      "epoch": 0.5007587253414264,
+      "grad_norm": 0.8322820663452148,
+      "learning_rate": 2.9802649407753105e-05,
+      "loss": 0.8685,
+      "step": 165
+    },
+    {
+      "epoch": 0.5159332321699545,
+      "grad_norm": 0.6224384903907776,
+      "learning_rate": 2.9777595394167674e-05,
+      "loss": 0.9406,
+      "step": 170
+    },
+    {
+      "epoch": 0.5311077389984825,
+      "grad_norm": 0.7838391065597534,
+      "learning_rate": 2.9751056464550863e-05,
+      "loss": 0.9418,
+      "step": 175
+    },
+    {
+      "epoch": 0.5462822458270106,
+      "grad_norm": 0.7811512351036072,
+      "learning_rate": 2.97230352856479e-05,
+      "loss": 0.8876,
+      "step": 180
+    },
+    {
+      "epoch": 0.5614567526555387,
+      "grad_norm": 0.7145566940307617,
+      "learning_rate": 2.9693534673146772e-05,
+      "loss": 0.851,
+      "step": 185
+    },
+    {
+      "epoch": 0.5766312594840668,
+      "grad_norm": 0.7759535908699036,
+      "learning_rate": 2.9662557591395282e-05,
+      "loss": 0.8149,
+      "step": 190
+    },
+    {
+      "epoch": 0.5918057663125948,
+      "grad_norm": 0.8379637002944946,
+      "learning_rate": 2.9630107153103176e-05,
+      "loss": 0.8479,
+      "step": 195
+    },
+    {
+      "epoch": 0.6069802731411229,
+      "grad_norm": 0.7673034071922302,
+      "learning_rate": 2.9596186619029382e-05,
+      "loss": 0.8367,
+      "step": 200
+    },
+    {
+      "epoch": 0.622154779969651,
+      "grad_norm": 0.7668901085853577,
+      "learning_rate": 2.9560799397654342e-05,
+      "loss": 0.8165,
+      "step": 205
+    },
+    {
+      "epoch": 0.637329286798179,
+      "grad_norm": 0.8146044015884399,
+      "learning_rate": 2.952394904483751e-05,
+      "loss": 0.823,
+      "step": 210
+    },
+    {
+      "epoch": 0.6525037936267072,
+      "grad_norm": 0.7815675735473633,
+      "learning_rate": 2.9485639263460045e-05,
+      "loss": 0.804,
+      "step": 215
+    },
+    {
+      "epoch": 0.6676783004552352,
+      "grad_norm": 0.8267041444778442,
+      "learning_rate": 2.944587390305275e-05,
+      "loss": 0.8068,
+      "step": 220
+    },
+    {
+      "epoch": 0.6828528072837633,
+      "grad_norm": 0.8476167321205139,
+      "learning_rate": 2.9404656959409228e-05,
+      "loss": 0.7598,
+      "step": 225
+    },
+    {
+      "epoch": 0.6980273141122914,
+      "grad_norm": 0.8841318488121033,
+      "learning_rate": 2.9361992574184376e-05,
+      "loss": 0.8277,
+      "step": 230
+    },
+    {
+      "epoch": 0.7132018209408194,
+      "grad_norm": 0.962478756904602,
+      "learning_rate": 2.931788503447822e-05,
+      "loss": 0.7416,
+      "step": 235
+    },
+    {
+      "epoch": 0.7283763277693475,
+      "grad_norm": 0.8502833247184753,
+      "learning_rate": 2.9272338772405128e-05,
+      "loss": 0.7805,
+      "step": 240
+    },
+    {
+      "epoch": 0.7435508345978755,
+      "grad_norm": 1.0657333135604858,
+      "learning_rate": 2.9225358364648438e-05,
+      "loss": 0.7696,
+      "step": 245
+    },
+    {
+      "epoch": 0.7587253414264037,
+      "grad_norm": 0.9457326531410217,
+      "learning_rate": 2.9176948532000594e-05,
+      "loss": 0.7713,
+      "step": 250
+    },
+    {
+      "epoch": 0.7738998482549317,
+      "grad_norm": 0.9325288534164429,
+      "learning_rate": 2.9127114138888778e-05,
+      "loss": 0.7448,
+      "step": 255
+    },
+    {
+      "epoch": 0.7890743550834598,
+      "grad_norm": 0.9588562250137329,
+      "learning_rate": 2.907586019288608e-05,
+      "loss": 0.7063,
+      "step": 260
+    },
+    {
+      "epoch": 0.8042488619119879,
+      "grad_norm": 0.8736801743507385,
+      "learning_rate": 2.9023191844208362e-05,
+      "loss": 0.7246,
+      "step": 265
+    },
+    {
+      "epoch": 0.8194233687405159,
+      "grad_norm": 1.0334984064102173,
+      "learning_rate": 2.896911438519671e-05,
+      "loss": 0.726,
+      "step": 270
+    },
+    {
+      "epoch": 0.834597875569044,
+      "grad_norm": 0.9704434275627136,
+      "learning_rate": 2.8913633249785653e-05,
+      "loss": 0.775,
+      "step": 275
+    },
+    {
+      "epoch": 0.849772382397572,
+      "grad_norm": 0.9025390148162842,
+      "learning_rate": 2.8856754012957126e-05,
+      "loss": 0.7014,
+      "step": 280
+    },
+    {
+      "epoch": 0.8649468892261002,
+      "grad_norm": 0.9079306721687317,
+      "learning_rate": 2.879848239018029e-05,
+      "loss": 0.681,
+      "step": 285
+    },
+    {
+      "epoch": 0.8801213960546282,
+      "grad_norm": 0.9602410793304443,
+      "learning_rate": 2.873882423683718e-05,
+      "loss": 0.7152,
+      "step": 290
+    },
+    {
+      "epoch": 0.8952959028831563,
+      "grad_norm": 0.9534751176834106,
+      "learning_rate": 2.8677785547634384e-05,
+      "loss": 0.7009,
+      "step": 295
+    },
+    {
+      "epoch": 0.9104704097116844,
+      "grad_norm": 0.9962713122367859,
+      "learning_rate": 2.861537245600063e-05,
+      "loss": 0.7087,
+      "step": 300
+    },
+    {
+      "epoch": 0.9256449165402124,
+      "grad_norm": 0.9465916752815247,
+      "learning_rate": 2.8551591233470488e-05,
+      "loss": 0.6897,
+      "step": 305
+    },
+    {
+      "epoch": 0.9408194233687405,
+      "grad_norm": 0.9778639078140259,
+      "learning_rate": 2.8486448289054164e-05,
+      "loss": 0.6658,
+      "step": 310
+    },
+    {
+      "epoch": 0.9559939301972686,
+      "grad_norm": 0.9425243139266968,
+      "learning_rate": 2.8419950168593528e-05,
+      "loss": 0.6454,
+      "step": 315
+    },
+    {
+      "epoch": 0.9711684370257967,
+      "grad_norm": 0.9817103743553162,
+      "learning_rate": 2.8352103554104313e-05,
+      "loss": 0.6343,
+      "step": 320
+    },
+    {
+      "epoch": 0.9863429438543247,
+      "grad_norm": 0.9938346147537231,
+      "learning_rate": 2.8282915263104728e-05,
+      "loss": 0.645,
+      "step": 325
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.365831971168518,
+      "learning_rate": 2.8212392247930368e-05,
+      "loss": 0.6416,
+      "step": 330
+    },
+    {
+      "epoch": 1.015174506828528,
+      "grad_norm": 0.9671252965927124,
+      "learning_rate": 2.814054159503563e-05,
+      "loss": 0.575,
+      "step": 335
+    },
+    {
+      "epoch": 1.0303490136570561,
+      "grad_norm": 1.005592942237854,
+      "learning_rate": 2.8067370524281617e-05,
+      "loss": 0.5576,
+      "step": 340
+    },
+    {
+      "epoch": 1.0455235204855842,
+      "grad_norm": 1.00735342502594,
+      "learning_rate": 2.7992886388210693e-05,
+      "loss": 0.5577,
+      "step": 345
+    },
+    {
+      "epoch": 1.0606980273141122,
+      "grad_norm": 1.090733528137207,
+      "learning_rate": 2.7917096671307624e-05,
+      "loss": 0.5428,
+      "step": 350
+    },
+    {
+      "epoch": 1.0758725341426403,
+      "grad_norm": 1.0599536895751953,
+      "learning_rate": 2.784000898924754e-05,
+      "loss": 0.519,
+      "step": 355
+    },
+    {
+      "epoch": 1.0910470409711683,
+      "grad_norm": 1.1048762798309326,
+      "learning_rate": 2.7761631088130654e-05,
+      "loss": 0.5437,
+      "step": 360
+    },
+    {
+      "epoch": 1.1062215477996964,
+      "grad_norm": 1.1115154027938843,
+      "learning_rate": 2.7681970843703926e-05,
+      "loss": 0.6217,
+      "step": 365
+    },
+    {
+      "epoch": 1.1213960546282247,
+      "grad_norm": 1.0769774913787842,
+      "learning_rate": 2.7601036260569646e-05,
+      "loss": 0.5095,
+      "step": 370
+    },
+    {
+      "epoch": 1.1365705614567527,
+      "grad_norm": 1.0201396942138672,
+      "learning_rate": 2.7518835471381117e-05,
+      "loss": 0.5247,
+      "step": 375
+    },
+    {
+      "epoch": 1.1517450682852808,
+      "grad_norm": 1.0590314865112305,
+      "learning_rate": 2.7435376736025447e-05,
+      "loss": 0.5516,
+      "step": 380
+    },
+    {
+      "epoch": 1.1669195751138088,
+      "grad_norm": 1.0749924182891846,
+      "learning_rate": 2.735066844079355e-05,
+      "loss": 0.5079,
+      "step": 385
+    },
+    {
+      "epoch": 1.182094081942337,
+      "grad_norm": 1.10120689868927,
+      "learning_rate": 2.7264719097537482e-05,
+      "loss": 0.539,
+      "step": 390
+    },
+    {
+      "epoch": 1.197268588770865,
+      "grad_norm": 1.3611030578613281,
+      "learning_rate": 2.7177537342815098e-05,
+      "loss": 0.4849,
+      "step": 395
+    },
+    {
+      "epoch": 1.212443095599393,
+      "grad_norm": 1.0392005443572998,
+      "learning_rate": 2.7089131937022242e-05,
+      "loss": 0.4807,
+      "step": 400
+    },
+    {
+      "epoch": 1.227617602427921,
+      "grad_norm": 1.0147825479507446,
+      "learning_rate": 2.699951176351245e-05,
+      "loss": 0.4962,
+      "step": 405
+    },
+    {
+      "epoch": 1.2427921092564491,
+      "grad_norm": 1.1136548519134521,
+      "learning_rate": 2.6908685827704324e-05,
+      "loss": 0.5106,
+      "step": 410
+    },
+    {
+      "epoch": 1.2579666160849772,
+      "grad_norm": 0.9843658804893494,
+      "learning_rate": 2.681666325617661e-05,
+      "loss": 0.4979,
+      "step": 415
+    },
+    {
+      "epoch": 1.2731411229135052,
+      "grad_norm": 1.0182534456253052,
+      "learning_rate": 2.6723453295751143e-05,
+      "loss": 0.4764,
+      "step": 420
+    },
+    {
+      "epoch": 1.2883156297420335,
+      "grad_norm": 1.0736995935440063,
+      "learning_rate": 2.6629065312563672e-05,
+      "loss": 0.4811,
+      "step": 425
+    },
+    {
+      "epoch": 1.3034901365705616,
+      "grad_norm": 1.0721853971481323,
+      "learning_rate": 2.6533508791122716e-05,
+      "loss": 0.4757,
+      "step": 430
+    },
+    {
+      "epoch": 1.3186646433990896,
+      "grad_norm": 1.0161422491073608,
+      "learning_rate": 2.6436793333356523e-05,
+      "loss": 0.4683,
+      "step": 435
+    },
+    {
+      "epoch": 1.3338391502276177,
+      "grad_norm": 1.159873127937317,
+      "learning_rate": 2.633892865764821e-05,
+      "loss": 0.4907,
+      "step": 440
+    },
+    {
+      "epoch": 1.3490136570561457,
+      "grad_norm": 1.148019790649414,
+      "learning_rate": 2.623992459785925e-05,
+      "loss": 0.4757,
+      "step": 445
+    },
+    {
+      "epoch": 1.3641881638846738,
+      "grad_norm": 1.0696481466293335,
+      "learning_rate": 2.6139791102341287e-05,
+      "loss": 0.4815,
+      "step": 450
+    },
+    {
+      "epoch": 1.3793626707132018,
+      "grad_norm": 1.1690481901168823,
+      "learning_rate": 2.6038538232936516e-05,
+      "loss": 0.4361,
+      "step": 455
+    },
+    {
+      "epoch": 1.39453717754173,
+      "grad_norm": 1.0492829084396362,
+      "learning_rate": 2.5936176163966597e-05,
+      "loss": 0.4764,
+      "step": 460
+    },
+    {
+      "epoch": 1.409711684370258,
+      "grad_norm": 1.0407378673553467,
+      "learning_rate": 2.583271518121032e-05,
+      "loss": 0.4543,
+      "step": 465
+    },
+    {
+      "epoch": 1.424886191198786,
+      "grad_norm": 1.1021435260772705,
+      "learning_rate": 2.572816568087003e-05,
+      "loss": 0.4631,
+      "step": 470
+    },
+    {
+      "epoch": 1.440060698027314,
+      "grad_norm": 1.0365526676177979,
+      "learning_rate": 2.5622538168526978e-05,
+      "loss": 0.4496,
+      "step": 475
+    },
+    {
+      "epoch": 1.4552352048558421,
+      "grad_norm": 1.0675585269927979,
+      "learning_rate": 2.5515843258085686e-05,
+      "loss": 0.4605,
+      "step": 480
+    },
+    {
+      "epoch": 1.4704097116843702,
+      "grad_norm": 1.1329057216644287,
+      "learning_rate": 2.5408091670707386e-05,
+      "loss": 0.4065,
+      "step": 485
+    },
+    {
+      "epoch": 1.4855842185128982,
+      "grad_norm": 0.9651331305503845,
+      "learning_rate": 2.5299294233732742e-05,
+      "loss": 0.4249,
+      "step": 490
+    },
+    {
+      "epoch": 1.5007587253414263,
+      "grad_norm": 1.067569375038147,
+      "learning_rate": 2.518946187959386e-05,
+      "loss": 0.4262,
+      "step": 495
+    },
+    {
+      "epoch": 1.5159332321699543,
+      "grad_norm": 1.1014666557312012,
+      "learning_rate": 2.507860564471575e-05,
+      "loss": 0.4463,
+      "step": 500
+    },
+    {
+      "epoch": 1.5311077389984824,
+      "grad_norm": 1.0636093616485596,
+      "learning_rate": 2.496673666840735e-05,
+      "loss": 0.4159,
+      "step": 505
+    },
+    {
+      "epoch": 1.5462822458270105,
+      "grad_norm": 0.9945455193519592,
+      "learning_rate": 2.4853866191742177e-05,
+      "loss": 0.4201,
+      "step": 510
+    },
+    {
+      "epoch": 1.5614567526555387,
+      "grad_norm": 1.0836797952651978,
+      "learning_rate": 2.47400055564288e-05,
+      "loss": 0.4312,
+      "step": 515
+    },
+    {
+      "epoch": 1.5766312594840668,
+      "grad_norm": 1.0805171728134155,
+      "learning_rate": 2.4625166203671166e-05,
+      "loss": 0.43,
+      "step": 520
+    },
+    {
+      "epoch": 1.5918057663125948,
+      "grad_norm": 1.305701732635498,
+      "learning_rate": 2.4509359673018933e-05,
+      "loss": 0.3803,
+      "step": 525
+    },
+    {
+      "epoch": 1.606980273141123,
+      "grad_norm": 1.0899828672409058,
+      "learning_rate": 2.439259760120794e-05,
+      "loss": 0.3797,
+      "step": 530
+    },
+    {
+      "epoch": 1.622154779969651,
+      "grad_norm": 1.0438095331192017,
+      "learning_rate": 2.4274891720990886e-05,
+      "loss": 0.384,
+      "step": 535
+    },
+    {
+      "epoch": 1.637329286798179,
+      "grad_norm": 1.0920283794403076,
+      "learning_rate": 2.4156253859958388e-05,
+      "loss": 0.415,
+      "step": 540
+    },
+    {
+      "epoch": 1.6525037936267073,
+      "grad_norm": 1.2114492654800415,
+      "learning_rate": 2.403669593935047e-05,
+      "loss": 0.4028,
+      "step": 545
+    },
+    {
+      "epoch": 1.6676783004552354,
+      "grad_norm": 1.0579054355621338,
+      "learning_rate": 2.3916229972858695e-05,
+      "loss": 0.388,
+      "step": 550
+    },
+    {
+      "epoch": 1.6828528072837634,
+      "grad_norm": 0.9982993006706238,
+      "learning_rate": 2.3794868065418944e-05,
+      "loss": 0.3883,
+      "step": 555
+    },
+    {
+      "epoch": 1.6980273141122915,
+      "grad_norm": 1.06562340259552,
+      "learning_rate": 2.3672622411995095e-05,
+      "loss": 0.3829,
+      "step": 560
+    },
+    {
+      "epoch": 1.7132018209408195,
+      "grad_norm": 1.1990240812301636,
+      "learning_rate": 2.3549505296353605e-05,
+      "loss": 0.4015,
+      "step": 565
+    },
+    {
+      "epoch": 1.7283763277693476,
+      "grad_norm": 1.1103229522705078,
+      "learning_rate": 2.3425529089829166e-05,
+      "loss": 0.3946,
+      "step": 570
+    },
+    {
+      "epoch": 1.7435508345978756,
+      "grad_norm": 1.086736798286438,
+      "learning_rate": 2.330070625008162e-05,
+      "loss": 0.4003,
+      "step": 575
+    },
+    {
+      "epoch": 1.7587253414264037,
+      "grad_norm": 1.0179426670074463,
+      "learning_rate": 2.3175049319844132e-05,
+      "loss": 0.3829,
+      "step": 580
+    },
+    {
+      "epoch": 1.7738998482549317,
+      "grad_norm": 1.1410490274429321,
+      "learning_rate": 2.3048570925662846e-05,
+      "loss": 0.3963,
+      "step": 585
+    },
+    {
+      "epoch": 1.7890743550834598,
+      "grad_norm": 1.1393828392028809,
+      "learning_rate": 2.2921283776628118e-05,
+      "loss": 0.3513,
+      "step": 590
+    },
+    {
+      "epoch": 1.8042488619119879,
+      "grad_norm": 1.034442663192749,
+      "learning_rate": 2.2793200663097453e-05,
+      "loss": 0.3373,
+      "step": 595
+    },
+    {
+      "epoch": 1.819423368740516,
+      "grad_norm": 1.1215357780456543,
+      "learning_rate": 2.266433445541028e-05,
+      "loss": 0.3689,
+      "step": 600
+    },
+    {
+      "epoch": 1.834597875569044,
+      "grad_norm": 1.1622790098190308,
+      "learning_rate": 2.253469810259468e-05,
+      "loss": 0.3734,
+      "step": 605
+    },
+    {
+      "epoch": 1.849772382397572,
+      "grad_norm": 1.1042276620864868,
+      "learning_rate": 2.240430463106621e-05,
+      "loss": 0.3145,
+      "step": 610
+    },
+    {
+      "epoch": 1.8649468892261,
+      "grad_norm": 1.176214575767517,
+      "learning_rate": 2.2273167143318956e-05,
+      "loss": 0.374,
+      "step": 615
+    },
+    {
+      "epoch": 1.8801213960546281,
+      "grad_norm": 1.162898302078247,
+      "learning_rate": 2.2141298816608935e-05,
+      "loss": 0.3398,
+      "step": 620
+    },
+    {
+      "epoch": 1.8952959028831562,
+      "grad_norm": 1.0589183568954468,
+      "learning_rate": 2.200871290163e-05,
+      "loss": 0.3475,
+      "step": 625
+    },
+    {
+      "epoch": 1.9104704097116842,
+      "grad_norm": 1.41371488571167,
+      "learning_rate": 2.1875422721182334e-05,
+      "loss": 0.3689,
+      "step": 630
+    },
+    {
+      "epoch": 1.9256449165402123,
+      "grad_norm": 1.095626711845398,
+      "learning_rate": 2.1741441668833733e-05,
+      "loss": 0.3618,
+      "step": 635
+    },
+    {
+      "epoch": 1.9408194233687404,
+      "grad_norm": 1.1281839609146118,
+      "learning_rate": 2.1606783207573757e-05,
+      "loss": 0.3358,
+      "step": 640
+    },
+    {
+      "epoch": 1.9559939301972686,
+      "grad_norm": 1.1524876356124878,
+      "learning_rate": 2.1471460868460922e-05,
+      "loss": 0.3513,
+      "step": 645
+    },
+    {
+      "epoch": 1.9711684370257967,
+      "grad_norm": 1.1511465311050415,
+      "learning_rate": 2.1335488249263012e-05,
+      "loss": 0.3236,
+      "step": 650
+    },
+    {
+      "epoch": 1.9863429438543247,
+      "grad_norm": 1.0634710788726807,
+      "learning_rate": 2.1198879013090763e-05,
+      "loss": 0.3491,
+      "step": 655
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.505767583847046,
+      "learning_rate": 2.106164688702488e-05,
+      "loss": 0.3141,
+      "step": 660
+    },
+    {
+      "epoch": 2.015174506828528,
+      "grad_norm": 1.161057472229004,
+      "learning_rate": 2.0923805660736732e-05,
+      "loss": 0.2651,
+      "step": 665
+    },
+    {
+      "epoch": 2.030349013657056,
+      "grad_norm": 1.1279088258743286,
+      "learning_rate": 2.0785369185102684e-05,
+      "loss": 0.2398,
+      "step": 670
+    },
+    {
+      "epoch": 2.045523520485584,
+      "grad_norm": 0.9957617521286011,
+      "learning_rate": 2.0646351370812293e-05,
+      "loss": 0.2783,
+      "step": 675
+    },
+    {
+      "epoch": 2.0606980273141122,
+      "grad_norm": 1.161773681640625,
+      "learning_rate": 2.050676618697052e-05,
+      "loss": 0.2599,
+      "step": 680
+    },
+    {
+      "epoch": 2.0758725341426403,
+      "grad_norm": 1.0346007347106934,
+      "learning_rate": 2.0366627659694043e-05,
+      "loss": 0.232,
+      "step": 685
+    },
+    {
+      "epoch": 2.0910470409711683,
+      "grad_norm": 1.2492696046829224,
+      "learning_rate": 2.022594987070185e-05,
+      "loss": 0.2544,
+      "step": 690
+    },
+    {
+      "epoch": 2.1062215477996964,
+      "grad_norm": 1.0093607902526855,
+      "learning_rate": 2.0084746955900266e-05,
+      "loss": 0.2732,
+      "step": 695
+    },
+    {
+      "epoch": 2.1213960546282244,
+      "grad_norm": 1.421979546546936,
+      "learning_rate": 1.994303310396251e-05,
+      "loss": 0.2437,
+      "step": 700
+    },
+    {
+      "epoch": 2.1365705614567525,
+      "grad_norm": 1.0597381591796875,
+      "learning_rate": 1.9800822554902924e-05,
+      "loss": 0.257,
+      "step": 705
+    },
+    {
+      "epoch": 2.1517450682852806,
+      "grad_norm": 1.0941869020462036,
+      "learning_rate": 1.9658129598646136e-05,
+      "loss": 0.2558,
+      "step": 710
+    },
+    {
+      "epoch": 2.1669195751138086,
+      "grad_norm": 1.3752810955047607,
+      "learning_rate": 1.9514968573591082e-05,
+      "loss": 0.2574,
+      "step": 715
+    },
+    {
+      "epoch": 2.1820940819423367,
+      "grad_norm": 0.9760404825210571,
+      "learning_rate": 1.9371353865170285e-05,
+      "loss": 0.2558,
+      "step": 720
+    },
+    {
+      "epoch": 2.197268588770865,
+      "grad_norm": 1.097494125366211,
+      "learning_rate": 1.9227299904404293e-05,
+      "loss": 0.2873,
+      "step": 725
+    },
+    {
+      "epoch": 2.212443095599393,
+      "grad_norm": 1.0384601354599,
+      "learning_rate": 1.9082821166451614e-05,
+      "loss": 0.2524,
+      "step": 730
+    },
+    {
+      "epoch": 2.2276176024279213,
+      "grad_norm": 1.1527771949768066,
+      "learning_rate": 1.8937932169154192e-05,
+      "loss": 0.2453,
+      "step": 735
+    },
+    {
+      "epoch": 2.2427921092564493,
+      "grad_norm": 1.0804898738861084,
+      "learning_rate": 1.8792647471578575e-05,
+      "loss": 0.2267,
+      "step": 740
+    },
+    {
+      "epoch": 2.2579666160849774,
+      "grad_norm": 1.0630391836166382,
+      "learning_rate": 1.8646981672552984e-05,
+      "loss": 0.2463,
+      "step": 745
+    },
+    {
+      "epoch": 2.2731411229135055,
+      "grad_norm": 1.1511931419372559,
+      "learning_rate": 1.8500949409200328e-05,
+      "loss": 0.2264,
+      "step": 750
+    },
+    {
+      "epoch": 2.2883156297420335,
+      "grad_norm": 1.0880099534988403,
+      "learning_rate": 1.835456535546743e-05,
+      "loss": 0.2229,
+      "step": 755
+    },
+    {
+      "epoch": 2.3034901365705616,
+      "grad_norm": 1.0572727918624878,
+      "learning_rate": 1.8207844220650513e-05,
+      "loss": 0.2103,
+      "step": 760
+    },
+    {
+      "epoch": 2.3186646433990896,
+      "grad_norm": 0.9872831702232361,
+      "learning_rate": 1.806080074791715e-05,
+      "loss": 0.2102,
+      "step": 765
+    },
+    {
+      "epoch": 2.3338391502276177,
+      "grad_norm": 1.1545583009719849,
+      "learning_rate": 1.7913449712824813e-05,
+      "loss": 0.2388,
+      "step": 770
+    },
+    {
+      "epoch": 2.3490136570561457,
+      "grad_norm": 1.0736984014511108,
+      "learning_rate": 1.7765805921836147e-05,
+      "loss": 0.2146,
+      "step": 775
+    },
+    {
+      "epoch": 2.364188163884674,
+      "grad_norm": 0.9836410284042358,
+      "learning_rate": 1.7617884210831165e-05,
+      "loss": 0.2078,
+      "step": 780
+    },
+    {
+      "epoch": 2.379362670713202,
+      "grad_norm": 1.0019104480743408,
+      "learning_rate": 1.7469699443616478e-05,
+      "loss": 0.2232,
+      "step": 785
+    },
+    {
+      "epoch": 2.39453717754173,
+      "grad_norm": 1.1876050233840942,
+      "learning_rate": 1.7321266510431707e-05,
+      "loss": 0.2428,
+      "step": 790
+    },
+    {
+      "epoch": 2.409711684370258,
+      "grad_norm": 1.09200918674469,
+      "learning_rate": 1.7172600326453253e-05,
+      "loss": 0.2331,
+      "step": 795
+    },
+    {
+      "epoch": 2.424886191198786,
+      "grad_norm": 1.0480817556381226,
+      "learning_rate": 1.702371583029555e-05,
+      "loss": 0.209,
+      "step": 800
+    },
+    {
+      "epoch": 2.440060698027314,
+      "grad_norm": 1.287725567817688,
+      "learning_rate": 1.687462798250998e-05,
+      "loss": 0.225,
+      "step": 805
+    },
+    {
+      "epoch": 2.455235204855842,
+      "grad_norm": 1.162737488746643,
+      "learning_rate": 1.672535176408156e-05,
+      "loss": 0.2031,
+      "step": 810
+    },
+    {
+      "epoch": 2.47040971168437,
+      "grad_norm": 0.9590432047843933,
+      "learning_rate": 1.657590217492361e-05,
+      "loss": 0.2049,
+      "step": 815
+    },
+    {
+      "epoch": 2.4855842185128982,
+      "grad_norm": 1.1492325067520142,
+      "learning_rate": 1.6426294232370468e-05,
+      "loss": 0.2181,
+      "step": 820
+    },
+    {
+      "epoch": 2.5007587253414263,
+      "grad_norm": 1.1130379438400269,
+      "learning_rate": 1.6276542969668502e-05,
+      "loss": 0.2078,
+      "step": 825
+    },
+    {
+      "epoch": 2.5159332321699543,
+      "grad_norm": 0.9611683487892151,
+      "learning_rate": 1.612666343446551e-05,
+      "loss": 0.2099,
+      "step": 830
+    },
+    {
+      "epoch": 2.5311077389984824,
+      "grad_norm": 1.2248095273971558,
+      "learning_rate": 1.597667068729865e-05,
+      "loss": 0.1976,
+      "step": 835
+    },
+    {
+      "epoch": 2.5462822458270105,
+      "grad_norm": 1.0842514038085938,
+      "learning_rate": 1.582657980008111e-05,
+      "loss": 0.1937,
+      "step": 840
+    },
+    {
+      "epoch": 2.5614567526555385,
+      "grad_norm": 1.0177435874938965,
+      "learning_rate": 1.56764058545876e-05,
+      "loss": 0.2114,
+      "step": 845
+    },
+    {
+      "epoch": 2.576631259484067,
+      "grad_norm": 1.003584384918213,
+      "learning_rate": 1.5526163940938892e-05,
+      "loss": 0.2152,
+      "step": 850
+    },
+    {
+      "epoch": 2.5918057663125946,
+      "grad_norm": 1.078927993774414,
+      "learning_rate": 1.537586915608549e-05,
+      "loss": 0.2134,
+      "step": 855
+    },
+    {
+      "epoch": 2.606980273141123,
+      "grad_norm": 1.1085734367370605,
+      "learning_rate": 1.5225536602290614e-05,
+      "loss": 0.1987,
+      "step": 860
+    },
+    {
+      "epoch": 2.6221547799696507,
+      "grad_norm": 1.0498510599136353,
+      "learning_rate": 1.5075181385612671e-05,
+      "loss": 0.1907,
+      "step": 865
+    },
+    {
+      "epoch": 2.6373292867981792,
+      "grad_norm": 1.0778013467788696,
+      "learning_rate": 1.4924818614387334e-05,
+      "loss": 0.1903,
+      "step": 870
+    },
+    {
+      "epoch": 2.6525037936267073,
+      "grad_norm": 1.304302453994751,
+      "learning_rate": 1.477446339770939e-05,
+      "loss": 0.1916,
+      "step": 875
+    },
+    {
+      "epoch": 2.6676783004552354,
+      "grad_norm": 0.9650601148605347,
+      "learning_rate": 1.4624130843914512e-05,
+      "loss": 0.1788,
+      "step": 880
+    },
+    {
+      "epoch": 2.6828528072837634,
+      "grad_norm": 1.2137047052383423,
+      "learning_rate": 1.4473836059061107e-05,
+      "loss": 0.2087,
+      "step": 885
+    },
+    {
+      "epoch": 2.6980273141122915,
+      "grad_norm": 1.222299337387085,
+      "learning_rate": 1.43235941454124e-05,
+      "loss": 0.1956,
+      "step": 890
+    },
+    {
+      "epoch": 2.7132018209408195,
+      "grad_norm": 1.2385069131851196,
+      "learning_rate": 1.4173420199918893e-05,
+      "loss": 0.1995,
+      "step": 895
+    },
+    {
+      "epoch": 2.7283763277693476,
+      "grad_norm": 1.1916569471359253,
+      "learning_rate": 1.402332931270135e-05,
+      "loss": 0.1818,
+      "step": 900
+    },
+    {
+      "epoch": 2.7435508345978756,
+      "grad_norm": 0.9629146456718445,
+      "learning_rate": 1.387333656553449e-05,
+      "loss": 0.1813,
+      "step": 905
+    },
+    {
+      "epoch": 2.7587253414264037,
+      "grad_norm": 1.1919859647750854,
+      "learning_rate": 1.3723457030331498e-05,
+      "loss": 0.1877,
+      "step": 910
+    },
+    {
+      "epoch": 2.7738998482549317,
+      "grad_norm": 1.1693427562713623,
+      "learning_rate": 1.3573705767629536e-05,
+      "loss": 0.2139,
+      "step": 915
+    },
+    {
+      "epoch": 2.78907435508346,
+      "grad_norm": 1.0847316980361938,
+      "learning_rate": 1.3424097825076394e-05,
+      "loss": 0.1796,
+      "step": 920
+    },
+    {
+      "epoch": 2.804248861911988,
+      "grad_norm": 1.064889907836914,
+      "learning_rate": 1.3274648235918442e-05,
+      "loss": 0.1743,
+      "step": 925
+    },
+    {
+      "epoch": 2.819423368740516,
+      "grad_norm": 1.2099905014038086,
+      "learning_rate": 1.3125372017490026e-05,
+      "loss": 0.178,
+      "step": 930
+    },
+    {
+      "epoch": 2.834597875569044,
+      "grad_norm": 1.0362218618392944,
+      "learning_rate": 1.2976284169704455e-05,
+      "loss": 0.2107,
+      "step": 935
+    },
+    {
+      "epoch": 2.849772382397572,
+      "grad_norm": 1.3111770153045654,
+      "learning_rate": 1.282739967354675e-05,
+      "loss": 0.1867,
+      "step": 940
+    },
+    {
+      "epoch": 2.8649468892261,
+      "grad_norm": 1.3245251178741455,
+      "learning_rate": 1.2678733489568292e-05,
+      "loss": 0.1815,
+      "step": 945
+    },
+    {
+      "epoch": 2.880121396054628,
+      "grad_norm": 0.9746212959289551,
+      "learning_rate": 1.2530300556383521e-05,
+      "loss": 0.1794,
+      "step": 950
+    },
+    {
+      "epoch": 2.895295902883156,
+      "grad_norm": 1.1812666654586792,
+      "learning_rate": 1.2382115789168834e-05,
+      "loss": 0.1815,
+      "step": 955
+    },
+    {
+      "epoch": 2.9104704097116842,
+      "grad_norm": 0.9376991987228394,
+      "learning_rate": 1.223419407816386e-05,
+      "loss": 0.1714,
+      "step": 960
+    },
+    {
+      "epoch": 2.9256449165402123,
+      "grad_norm": 1.120914340019226,
+      "learning_rate": 1.208655028717519e-05,
+      "loss": 0.1703,
+      "step": 965
+    },
+    {
+      "epoch": 2.9408194233687404,
+      "grad_norm": 1.2715625762939453,
+      "learning_rate": 1.1939199252082851e-05,
+      "loss": 0.1631,
+      "step": 970
+    },
+    {
+      "epoch": 2.955993930197269,
+      "grad_norm": 1.1299712657928467,
+      "learning_rate": 1.179215577934949e-05,
+      "loss": 0.1771,
+      "step": 975
+    },
+    {
+      "epoch": 2.9711684370257965,
+      "grad_norm": 1.0479075908660889,
+      "learning_rate": 1.1645434644532572e-05,
+      "loss": 0.1855,
+      "step": 980
+    },
+    {
+      "epoch": 2.986342943854325,
+      "grad_norm": 1.1739304065704346,
+      "learning_rate": 1.1499050590799675e-05,
+      "loss": 0.1697,
+      "step": 985
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.351037859916687,
+      "learning_rate": 1.1353018327447017e-05,
+      "loss": 0.1856,
+      "step": 990
+    },
+    {
+      "epoch": 3.015174506828528,
+      "grad_norm": 1.0247342586517334,
+      "learning_rate": 1.1207352528421424e-05,
+      "loss": 0.1543,
+      "step": 995
+    },
+    {
+      "epoch": 3.030349013657056,
+      "grad_norm": 0.9216309189796448,
+      "learning_rate": 1.1062067830845807e-05,
+      "loss": 0.1261,
+      "step": 1000
+    },
+    {
+      "epoch": 3.045523520485584,
+      "grad_norm": 0.9509443640708923,
+      "learning_rate": 1.091717883354839e-05,
+      "loss": 0.1147,
+      "step": 1005
+    },
+    {
+      "epoch": 3.0606980273141122,
+      "grad_norm": 1.0392087697982788,
+      "learning_rate": 1.0772700095595713e-05,
+      "loss": 0.1212,
+      "step": 1010
+    },
+    {
+      "epoch": 3.0758725341426403,
+      "grad_norm": 0.8555918335914612,
+      "learning_rate": 1.062864613482972e-05,
+      "loss": 0.1107,
+      "step": 1015
+    },
+    {
+      "epoch": 3.0910470409711683,
+      "grad_norm": 0.873559296131134,
+      "learning_rate": 1.048503142640892e-05,
+      "loss": 0.1257,
+      "step": 1020
+    },
+    {
+      "epoch": 3.1062215477996964,
+      "grad_norm": 0.9564691185951233,
+      "learning_rate": 1.034187040135387e-05,
+      "loss": 0.1284,
+      "step": 1025
+    },
+    {
+      "epoch": 3.1213960546282244,
+      "grad_norm": 1.031790852546692,
+      "learning_rate": 1.0199177445097075e-05,
+      "loss": 0.1249,
+      "step": 1030
+    },
+    {
+      "epoch": 3.1365705614567525,
+      "grad_norm": 1.0104645490646362,
+      "learning_rate": 1.0056966896037494e-05,
+      "loss": 0.125,
+      "step": 1035
+    },
+    {
+      "epoch": 3.1517450682852806,
+      "grad_norm": 1.0347521305084229,
+      "learning_rate": 9.915253044099731e-06,
+      "loss": 0.1366,
+      "step": 1040
+    },
+    {
+      "epoch": 3.1669195751138086,
+      "grad_norm": 0.9634484648704529,
+      "learning_rate": 9.77405012929815e-06,
+      "loss": 0.1225,
+      "step": 1045
+    },
+    {
+      "epoch": 3.1820940819423367,
+      "grad_norm": 1.00877046585083,
+      "learning_rate": 9.633372340305966e-06,
+      "loss": 0.1312,
+      "step": 1050
+    },
+    {
+      "epoch": 3.197268588770865,
+      "grad_norm": 1.2125271558761597,
+      "learning_rate": 9.493233813029485e-06,
+      "loss": 0.1188,
+      "step": 1055
+    },
+    {
+      "epoch": 3.212443095599393,
+      "grad_norm": 1.0163218975067139,
+      "learning_rate": 9.353648629187709e-06,
+      "loss": 0.1231,
+      "step": 1060
+    },
+    {
+      "epoch": 3.2276176024279213,
+      "grad_norm": 0.9734249711036682,
+      "learning_rate": 9.214630814897317e-06,
+      "loss": 0.1193,
+      "step": 1065
+    },
+    {
+      "epoch": 3.2427921092564493,
+      "grad_norm": 0.9149922728538513,
+      "learning_rate": 9.076194339263268e-06,
+      "loss": 0.1325,
+      "step": 1070
+    },
+    {
+      "epoch": 3.2579666160849774,
+      "grad_norm": 0.9013402462005615,
+      "learning_rate": 8.938353112975126e-06,
+      "loss": 0.1308,
+      "step": 1075
+    },
+    {
+      "epoch": 3.2731411229135055,
+      "grad_norm": 1.0030301809310913,
+      "learning_rate": 8.801120986909242e-06,
+      "loss": 0.1167,
+      "step": 1080
+    },
+    {
+      "epoch": 3.2883156297420335,
+      "grad_norm": 0.8948475122451782,
+      "learning_rate": 8.664511750736988e-06,
+      "loss": 0.1346,
+      "step": 1085
+    },
+    {
+      "epoch": 3.3034901365705616,
+      "grad_norm": 0.9730858206748962,
+      "learning_rate": 8.52853913153908e-06,
+      "loss": 0.1194,
+      "step": 1090
+    },
+    {
+      "epoch": 3.3186646433990896,
+      "grad_norm": 1.0079647302627563,
+      "learning_rate": 8.393216792426243e-06,
+      "loss": 0.1435,
+      "step": 1095
+    },
+    {
+      "epoch": 3.3338391502276177,
+      "grad_norm": 0.6983161568641663,
+      "learning_rate": 8.258558331166271e-06,
+      "loss": 0.11,
+      "step": 1100
+    },
+    {
+      "epoch": 3.3490136570561457,
+      "grad_norm": 0.9095783233642578,
+      "learning_rate": 8.124577278817668e-06,
+      "loss": 0.1032,
+      "step": 1105
+    },
+    {
+      "epoch": 3.364188163884674,
+      "grad_norm": 0.9054252505302429,
+      "learning_rate": 7.991287098370004e-06,
+      "loss": 0.1208,
+      "step": 1110
+    },
+    {
+      "epoch": 3.379362670713202,
+      "grad_norm": 1.133405327796936,
+      "learning_rate": 7.858701183391064e-06,
+      "loss": 0.1105,
+      "step": 1115
+    },
+    {
+      "epoch": 3.39453717754173,
+      "grad_norm": 0.9282613396644592,
+      "learning_rate": 7.726832856681047e-06,
+      "loss": 0.1132,
+      "step": 1120
+    },
+    {
+      "epoch": 3.409711684370258,
+      "grad_norm": 0.8912308812141418,
+      "learning_rate": 7.59569536893379e-06,
+      "loss": 0.116,
+      "step": 1125
+    },
+    {
+      "epoch": 3.424886191198786,
+      "grad_norm": 1.1447117328643799,
+      "learning_rate": 7.465301897405322e-06,
+      "loss": 0.1,
+      "step": 1130
+    },
+    {
+      "epoch": 3.440060698027314,
+      "grad_norm": 0.8710254430770874,
+      "learning_rate": 7.33566554458972e-06,
+      "loss": 0.1139,
+      "step": 1135
+    },
+    {
+      "epoch": 3.455235204855842,
+      "grad_norm": 0.9299088716506958,
+      "learning_rate": 7.206799336902558e-06,
+      "loss": 0.1067,
+      "step": 1140
+    },
+    {
+      "epoch": 3.47040971168437,
+      "grad_norm": 1.0533580780029297,
+      "learning_rate": 7.07871622337189e-06,
+      "loss": 0.1041,
+      "step": 1145
+    },
+    {
+      "epoch": 3.4855842185128982,
+      "grad_norm": 0.8250032067298889,
+      "learning_rate": 6.9514290743371575e-06,
+      "loss": 0.1082,
+      "step": 1150
+    },
+    {
+      "epoch": 3.5007587253414263,
+      "grad_norm": 0.8535242676734924,
+      "learning_rate": 6.824950680155871e-06,
+      "loss": 0.1141,
+      "step": 1155
+    },
+    {
+      "epoch": 3.5159332321699543,
+      "grad_norm": 0.9096361994743347,
+      "learning_rate": 6.69929374991838e-06,
+      "loss": 0.1144,
+      "step": 1160
+    },
+    {
+      "epoch": 3.5311077389984824,
+      "grad_norm": 0.9085047245025635,
+      "learning_rate": 6.5744709101708364e-06,
+      "loss": 0.1005,
+      "step": 1165
+    },
+    {
+      "epoch": 3.5462822458270105,
+      "grad_norm": 0.8639160990715027,
+      "learning_rate": 6.450494703646398e-06,
+      "loss": 0.102,
+      "step": 1170
+    },
+    {
+      "epoch": 3.5614567526555385,
+      "grad_norm": 0.8191717267036438,
+      "learning_rate": 6.327377588004907e-06,
+      "loss": 0.0963,
+      "step": 1175
+    },
+    {
+      "epoch": 3.576631259484067,
+      "grad_norm": 0.8398914933204651,
+      "learning_rate": 6.205131934581056e-06,
+      "loss": 0.1162,
+      "step": 1180
+    },
+    {
+      "epoch": 3.5918057663125946,
+      "grad_norm": 0.848587691783905,
+      "learning_rate": 6.083770027141313e-06,
+      "loss": 0.1177,
+      "step": 1185
+    },
+    {
+      "epoch": 3.606980273141123,
+      "grad_norm": 0.9553687572479248,
+      "learning_rate": 5.963304060649531e-06,
+      "loss": 0.1069,
+      "step": 1190
+    },
+    {
+      "epoch": 3.6221547799696507,
+      "grad_norm": 0.8801453113555908,
+      "learning_rate": 5.843746140041616e-06,
+      "loss": 0.1211,
+      "step": 1195
+    },
+    {
+      "epoch": 3.6373292867981792,
+      "grad_norm": 1.2324217557907104,
+      "learning_rate": 5.725108279009115e-06,
+      "loss": 0.1221,
+      "step": 1200
+    },
+    {
+      "epoch": 3.6525037936267073,
+      "grad_norm": 0.994218111038208,
+      "learning_rate": 5.60740239879206e-06,
+      "loss": 0.0946,
+      "step": 1205
+    },
+    {
+      "epoch": 3.6676783004552354,
+      "grad_norm": 0.9351308941841125,
+      "learning_rate": 5.490640326981069e-06,
+      "loss": 0.1059,
+      "step": 1210
+    },
+    {
+      "epoch": 3.6828528072837634,
+      "grad_norm": 0.95694500207901,
+      "learning_rate": 5.374833796328833e-06,
+      "loss": 0.1016,
+      "step": 1215
+    },
+    {
+      "epoch": 3.6980273141122915,
+      "grad_norm": 0.7896385192871094,
+      "learning_rate": 5.2599944435712e-06,
+      "loss": 0.1004,
+      "step": 1220
+    },
+    {
+      "epoch": 3.7132018209408195,
+      "grad_norm": 0.7855644226074219,
+      "learning_rate": 5.14613380825782e-06,
+      "loss": 0.0958,
+      "step": 1225
+    },
+    {
+      "epoch": 3.7283763277693476,
+      "grad_norm": 0.7871450781822205,
+      "learning_rate": 5.033263331592655e-06,
+      "loss": 0.1076,
+      "step": 1230
+    },
+    {
+      "epoch": 3.7435508345978756,
+      "grad_norm": 0.8118149042129517,
+      "learning_rate": 4.921394355284251e-06,
+      "loss": 0.1245,
+      "step": 1235
+    },
+    {
+      "epoch": 3.7587253414264037,
+      "grad_norm": 0.8458611369132996,
+      "learning_rate": 4.8105381204061465e-06,
+      "loss": 0.0993,
+      "step": 1240
+    },
+    {
+      "epoch": 3.7738998482549317,
+      "grad_norm": 1.4438375234603882,
+      "learning_rate": 4.700705766267262e-06,
+      "loss": 0.1019,
+      "step": 1245
+    },
+    {
+      "epoch": 3.78907435508346,
+      "grad_norm": 0.7945970892906189,
+      "learning_rate": 4.591908329292619e-06,
+      "loss": 0.1123,
+      "step": 1250
+    },
+    {
+      "epoch": 3.804248861911988,
+      "grad_norm": 0.8581335544586182,
+      "learning_rate": 4.484156741914315e-06,
+      "loss": 0.0937,
+      "step": 1255
+    },
+    {
+      "epoch": 3.819423368740516,
+      "grad_norm": 0.8385892510414124,
+      "learning_rate": 4.377461831473022e-06,
+      "loss": 0.1047,
+      "step": 1260
+    },
+    {
+      "epoch": 3.834597875569044,
+      "grad_norm": 0.9302660226821899,
+      "learning_rate": 4.271834319129972e-06,
+      "loss": 0.1081,
+      "step": 1265
+    },
+    {
+      "epoch": 3.849772382397572,
+      "grad_norm": 0.8506367206573486,
+      "learning_rate": 4.167284818789681e-06,
+      "loss": 0.099,
+      "step": 1270
+    },
+    {
+      "epoch": 3.8649468892261,
+      "grad_norm": 0.802346408367157,
+      "learning_rate": 4.063823836033407e-06,
+      "loss": 0.0998,
+      "step": 1275
+    },
+    {
+      "epoch": 3.880121396054628,
+      "grad_norm": 0.8249574303627014,
+      "learning_rate": 3.9614617670634885e-06,
+      "loss": 0.101,
+      "step": 1280
+    },
+    {
+      "epoch": 3.895295902883156,
+      "grad_norm": 1.0073894262313843,
+      "learning_rate": 3.8602088976587175e-06,
+      "loss": 0.1037,
+      "step": 1285
+    },
+    {
+      "epoch": 3.9104704097116842,
+      "grad_norm": 0.8201082348823547,
+      "learning_rate": 3.7600754021407537e-06,
+      "loss": 0.1002,
+      "step": 1290
+    },
+    {
+      "epoch": 3.9256449165402123,
+      "grad_norm": 0.9376416802406311,
+      "learning_rate": 3.6610713423517927e-06,
+      "loss": 0.0853,
+      "step": 1295
+    },
+    {
+      "epoch": 3.9408194233687404,
+      "grad_norm": 0.7816200256347656,
+      "learning_rate": 3.5632066666434794e-06,
+      "loss": 0.0953,
+      "step": 1300
+    },
+    {
+      "epoch": 3.955993930197269,
+      "grad_norm": 0.8962005376815796,
+      "learning_rate": 3.4664912088772865e-06,
+      "loss": 0.0943,
+      "step": 1305
+    },
+    {
+      "epoch": 3.9711684370257965,
+      "grad_norm": 0.894721269607544,
+      "learning_rate": 3.3709346874363295e-06,
+      "loss": 0.0959,
+      "step": 1310
+    },
+    {
+      "epoch": 3.986342943854325,
+      "grad_norm": 0.7250496745109558,
+      "learning_rate": 3.2765467042488607e-06,
+      "loss": 0.0927,
+      "step": 1315
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.2268779277801514,
+      "learning_rate": 3.18333674382339e-06,
+      "loss": 0.0973,
+      "step": 1320
+    },
+    {
+      "epoch": 4.0151745068285285,
+      "grad_norm": 0.7619444727897644,
+      "learning_rate": 3.0913141722956786e-06,
+      "loss": 0.0854,
+      "step": 1325
+    },
+    {
+      "epoch": 4.030349013657056,
+      "grad_norm": 0.7040801644325256,
+      "learning_rate": 3.0004882364875517e-06,
+      "loss": 0.0754,
+      "step": 1330
+    },
+    {
+      "epoch": 4.045523520485585,
+      "grad_norm": 0.6949658393859863,
+      "learning_rate": 2.9108680629777584e-06,
+      "loss": 0.0727,
+      "step": 1335
+    },
+    {
+      "epoch": 4.060698027314112,
+      "grad_norm": 0.7198961973190308,
+      "learning_rate": 2.8224626571849056e-06,
+      "loss": 0.0755,
+      "step": 1340
+    },
+    {
+      "epoch": 4.075872534142641,
+      "grad_norm": 0.6946797966957092,
+      "learning_rate": 2.7352809024625192e-06,
+      "loss": 0.0849,
+      "step": 1345
+    },
+    {
+      "epoch": 4.091047040971168,
+      "grad_norm": 0.6197686791419983,
+      "learning_rate": 2.6493315592064516e-06,
+      "loss": 0.0673,
+      "step": 1350
+    },
+    {
+      "epoch": 4.106221547799697,
+      "grad_norm": 0.7615764141082764,
+      "learning_rate": 2.5646232639745565e-06,
+      "loss": 0.0815,
+      "step": 1355
+    },
+    {
+      "epoch": 4.1213960546282244,
+      "grad_norm": 0.6945781707763672,
+      "learning_rate": 2.4811645286188873e-06,
+      "loss": 0.0724,
+      "step": 1360
+    },
+    {
+      "epoch": 4.136570561456753,
+      "grad_norm": 0.6930224895477295,
+      "learning_rate": 2.3989637394303555e-06,
+      "loss": 0.0764,
+      "step": 1365
+    },
+    {
+      "epoch": 4.151745068285281,
+      "grad_norm": 0.7800251245498657,
+      "learning_rate": 2.3180291562960753e-06,
+      "loss": 0.069,
+      "step": 1370
+    },
+    {
+      "epoch": 4.166919575113809,
+      "grad_norm": 0.7373736500740051,
+      "learning_rate": 2.238368911869348e-06,
+      "loss": 0.0822,
+      "step": 1375
+    },
+    {
+      "epoch": 4.182094081942337,
+      "grad_norm": 0.7133133411407471,
+      "learning_rate": 2.159991010752464e-06,
+      "loss": 0.0787,
+      "step": 1380
+    },
+    {
+      "epoch": 4.197268588770865,
+      "grad_norm": 0.7388117909431458,
+      "learning_rate": 2.0829033286923798e-06,
+      "loss": 0.0709,
+      "step": 1385
+    },
+    {
+      "epoch": 4.212443095599393,
+      "grad_norm": 0.7258192896842957,
+      "learning_rate": 2.0071136117893104e-06,
+      "loss": 0.08,
+      "step": 1390
+    },
+    {
+      "epoch": 4.227617602427921,
+      "grad_norm": 0.6834006309509277,
+      "learning_rate": 1.9326294757183854e-06,
+      "loss": 0.0715,
+      "step": 1395
+    },
+    {
+      "epoch": 4.242792109256449,
+      "grad_norm": 0.6493126749992371,
+      "learning_rate": 1.8594584049643727e-06,
+      "loss": 0.0708,
+      "step": 1400
+    },
+    {
+      "epoch": 4.257966616084977,
+      "grad_norm": 0.6763819456100464,
+      "learning_rate": 1.7876077520696316e-06,
+      "loss": 0.0761,
+      "step": 1405
+    },
+    {
+      "epoch": 4.273141122913505,
+      "grad_norm": 0.7636880278587341,
+      "learning_rate": 1.7170847368952697e-06,
+      "loss": 0.0834,
+      "step": 1410
+    },
+    {
+      "epoch": 4.2883156297420335,
+      "grad_norm": 0.711479663848877,
+      "learning_rate": 1.6478964458956868e-06,
+      "loss": 0.0741,
+      "step": 1415
+    },
+    {
+      "epoch": 4.303490136570561,
+      "grad_norm": 0.8034400343894958,
+      "learning_rate": 1.580049831406477e-06,
+      "loss": 0.0864,
+      "step": 1420
+    },
+    {
+      "epoch": 4.31866464339909,
+      "grad_norm": 0.7478986382484436,
+      "learning_rate": 1.513551710945837e-06,
+      "loss": 0.0804,
+      "step": 1425
+    },
+    {
+      "epoch": 4.333839150227617,
+      "grad_norm": 0.8546178340911865,
+      "learning_rate": 1.4484087665295155e-06,
+      "loss": 0.0768,
+      "step": 1430
+    },
+    {
+      "epoch": 4.349013657056146,
+      "grad_norm": 0.6924677491188049,
+      "learning_rate": 1.3846275439993694e-06,
+      "loss": 0.0745,
+      "step": 1435
+    },
+    {
+      "epoch": 4.364188163884673,
+      "grad_norm": 0.8895645141601562,
+      "learning_rate": 1.3222144523656166e-06,
+      "loss": 0.0823,
+      "step": 1440
+    },
+    {
+      "epoch": 4.379362670713202,
+      "grad_norm": 0.6914752721786499,
+      "learning_rate": 1.2611757631628217e-06,
+      "loss": 0.0709,
+      "step": 1445
+    },
+    {
+      "epoch": 4.39453717754173,
+      "grad_norm": 0.7775288224220276,
+      "learning_rate": 1.2015176098197166e-06,
+      "loss": 0.0775,
+      "step": 1450
+    },
+    {
+      "epoch": 4.409711684370258,
+      "grad_norm": 0.6630114912986755,
+      "learning_rate": 1.1432459870428723e-06,
+      "loss": 0.0835,
+      "step": 1455
+    },
+    {
+      "epoch": 4.424886191198786,
+      "grad_norm": 0.602451503276825,
+      "learning_rate": 1.0863667502143482e-06,
+      "loss": 0.0726,
+      "step": 1460
+    },
+    {
+      "epoch": 4.440060698027314,
+      "grad_norm": 0.7203993201255798,
+      "learning_rate": 1.0308856148032918e-06,
+      "loss": 0.0834,
+      "step": 1465
+    },
+    {
+      "epoch": 4.455235204855843,
+      "grad_norm": 0.6996979713439941,
+      "learning_rate": 9.768081557916397e-07,
+      "loss": 0.0794,
+      "step": 1470
+    },
+    {
+      "epoch": 4.47040971168437,
+      "grad_norm": 0.6748331189155579,
+      "learning_rate": 9.241398071139223e-07,
+      "loss": 0.0852,
+      "step": 1475
+    },
+    {
+      "epoch": 4.485584218512899,
+      "grad_norm": 0.6650934219360352,
+      "learning_rate": 8.728858611112234e-07,
+      "loss": 0.0685,
+      "step": 1480
+    },
+    {
+      "epoch": 4.500758725341426,
+      "grad_norm": 0.6824582815170288,
+      "learning_rate": 8.230514679994033e-07,
+      "loss": 0.0661,
+      "step": 1485
+    },
+    {
+      "epoch": 4.515933232169955,
+      "grad_norm": 0.7158738970756531,
+      "learning_rate": 7.746416353515618e-07,
+      "loss": 0.0801,
+      "step": 1490
+    },
+    {
+      "epoch": 4.531107738998482,
+      "grad_norm": 0.7681190371513367,
+      "learning_rate": 7.276612275948746e-07,
+      "loss": 0.0754,
+      "step": 1495
+    },
+    {
+      "epoch": 4.546282245827011,
+      "grad_norm": 0.7390575408935547,
+      "learning_rate": 6.821149655217789e-07,
+      "loss": 0.0803,
+      "step": 1500
+    },
+    {
+      "epoch": 4.5614567526555385,
+      "grad_norm": 0.6839151382446289,
+      "learning_rate": 6.380074258156254e-07,
+      "loss": 0.0703,
+      "step": 1505
+    },
+    {
+      "epoch": 4.576631259484067,
+      "grad_norm": 0.8700344562530518,
+      "learning_rate": 5.953430405907729e-07,
+      "loss": 0.0727,
+      "step": 1510
+    },
+    {
+      "epoch": 4.591805766312595,
+      "grad_norm": 0.6669519543647766,
+      "learning_rate": 5.541260969472478e-07,
+      "loss": 0.0724,
+      "step": 1515
+    },
+    {
+      "epoch": 4.606980273141123,
+      "grad_norm": 0.6359818577766418,
+      "learning_rate": 5.143607365399544e-07,
+      "loss": 0.0678,
+      "step": 1520
+    },
+    {
+      "epoch": 4.622154779969651,
+      "grad_norm": 0.691038966178894,
+      "learning_rate": 4.76050955162492e-07,
+      "loss": 0.067,
+      "step": 1525
+    },
+    {
+      "epoch": 4.637329286798179,
+      "grad_norm": 0.7088958621025085,
+      "learning_rate": 4.392006023456596e-07,
+      "loss": 0.0734,
+      "step": 1530
+    },
+    {
+      "epoch": 4.652503793626707,
+      "grad_norm": 0.8911525011062622,
+      "learning_rate": 4.038133809706174e-07,
+      "loss": 0.0728,
+      "step": 1535
+    },
+    {
+      "epoch": 4.667678300455235,
+      "grad_norm": 0.6613883376121521,
+      "learning_rate": 3.6989284689682545e-07,
+      "loss": 0.0671,
+      "step": 1540
+    },
+    {
+      "epoch": 4.682852807283763,
+      "grad_norm": 0.7440558075904846,
+      "learning_rate": 3.374424086047201e-07,
+      "loss": 0.0819,
+      "step": 1545
+    },
+    {
+      "epoch": 4.6980273141122915,
+      "grad_norm": 0.6607381105422974,
+      "learning_rate": 3.064653268532308e-07,
+      "loss": 0.0799,
+      "step": 1550
+    },
+    {
+      "epoch": 4.713201820940819,
+      "grad_norm": 0.6847361326217651,
+      "learning_rate": 2.769647143521009e-07,
+      "loss": 0.0783,
+      "step": 1555
+    },
+    {
+      "epoch": 4.728376327769348,
+      "grad_norm": 0.6499977707862854,
+      "learning_rate": 2.489435354491393e-07,
+      "loss": 0.0658,
+      "step": 1560
+    },
+    {
+      "epoch": 4.743550834597875,
+      "grad_norm": 0.6675688624382019,
+      "learning_rate": 2.2240460583232703e-07,
+      "loss": 0.0745,
+      "step": 1565
+    },
+    {
+      "epoch": 4.758725341426404,
+      "grad_norm": 0.6232402324676514,
+      "learning_rate": 1.9735059224689545e-07,
+      "loss": 0.0675,
+      "step": 1570
+    },
+    {
+      "epoch": 4.773899848254931,
+      "grad_norm": 0.6561621427536011,
+      "learning_rate": 1.7378401222735874e-07,
+      "loss": 0.0651,
+      "step": 1575
+    },
+    {
+      "epoch": 4.78907435508346,
+      "grad_norm": 0.6981615424156189,
+      "learning_rate": 1.5170723384453634e-07,
+      "loss": 0.0801,
+      "step": 1580
+    },
+    {
+      "epoch": 4.804248861911988,
+      "grad_norm": 0.6096828579902649,
+      "learning_rate": 1.3112247546760992e-07,
+      "loss": 0.0672,
+      "step": 1585
+    },
+    {
+      "epoch": 4.819423368740516,
+      "grad_norm": 0.6685919165611267,
+      "learning_rate": 1.1203180554119675e-07,
+      "loss": 0.0666,
+      "step": 1590
+    },
+    {
+      "epoch": 4.8345978755690435,
+      "grad_norm": 0.7524701356887817,
+      "learning_rate": 9.443714237752088e-08,
+      "loss": 0.0791,
+      "step": 1595
+    },
+    {
+      "epoch": 4.849772382397572,
+      "grad_norm": 0.7297736406326294,
+      "learning_rate": 7.834025396363743e-08,
+      "loss": 0.0725,
+      "step": 1600
+    },
+    {
+      "epoch": 4.8649468892261005,
+      "grad_norm": 0.7565823197364807,
+      "learning_rate": 6.374275778378624e-08,
+      "loss": 0.0712,
+      "step": 1605
+    },
+    {
+      "epoch": 4.880121396054628,
+      "grad_norm": 0.7099778056144714,
+      "learning_rate": 5.064612065686036e-08,
+      "loss": 0.0749,
+      "step": 1610
+    },
+    {
+      "epoch": 4.895295902883157,
+      "grad_norm": 0.6616671085357666,
+      "learning_rate": 3.90516585890105e-08,
+      "loss": 0.0758,
+      "step": 1615
+    },
+    {
+      "epoch": 4.910470409711684,
+      "grad_norm": 0.7530394196510315,
+      "learning_rate": 2.8960536641405854e-08,
+      "loss": 0.0701,
+      "step": 1620
+    },
+    {
+      "epoch": 4.925644916540213,
+      "grad_norm": 0.681731641292572,
+      "learning_rate": 2.0373768813169437e-08,
+      "loss": 0.0679,
+      "step": 1625
+    },
+    {
+      "epoch": 4.94081942336874,
+      "grad_norm": 0.612575888633728,
+      "learning_rate": 1.3292217939484541e-08,
+      "loss": 0.0666,
+      "step": 1630
+    },
+    {
+      "epoch": 4.955993930197269,
+      "grad_norm": 0.6688991785049438,
+      "learning_rate": 7.716595604887467e-09,
+      "loss": 0.0719,
+      "step": 1635
+    },
+    {
+      "epoch": 4.9711684370257965,
+      "grad_norm": 0.6995802521705627,
+      "learning_rate": 3.647462071778018e-09,
+      "loss": 0.0775,
+      "step": 1640
+    },
+    {
+      "epoch": 4.986342943854325,
+      "grad_norm": 0.6894600987434387,
+      "learning_rate": 1.085226224104563e-09,
+      "loss": 0.0745,
+      "step": 1645
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 1.0343323945999146,
+      "learning_rate": 3.014552629354572e-11,
+      "loss": 0.0737,
+      "step": 1650
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1650,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.3725261684033454e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

21_128_e5_3e-5/checkpoint-1650/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:749f33b30b2b8c6910d60865d6784495303d813ca8f33cd1ddafcfa2fe3c03bf
+size 7736

21_128_e5_3e-5/checkpoint-1650/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)

21_128_e5_3e-5/checkpoint-330/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

21_128_e5_3e-5/checkpoint-330/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": [
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "v_proj",
+    "o_proj",
+    "k_proj",
+    "q_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

21_128_e5_3e-5/checkpoint-330/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afac547bc942c2f80c55adaea1d7fc934575fe012497c1b07a998269857b44c1
+size 791751704

21_128_e5_3e-5/checkpoint-330/latest

@@ -0,0 +1 @@
+global_step330

21_128_e5_3e-5/checkpoint-330/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6425307d12d08ad71968e4eaf8da33d7187677e9c6946037aac833da1d8cde7
+size 15920

21_128_e5_3e-5/checkpoint-330/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ad6a7a534b2e92826897c56d0ea22035cf9dceb79a95b0f6907e9850dd3b2bc0
+size 15920

21_128_e5_3e-5/checkpoint-330/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:674f6ef8e1e3995ecd946fd4a2386bddad35a2d569c29214eb29bb6d1b840755
+size 15920