Upload folder using huggingface_hub

28_128_e5_3e-5/checkpoint-1244/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

28_128_e5_3e-5/checkpoint-1244/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",
+    "q_proj",
+    "v_proj",
+    "o_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

28_128_e5_3e-5/checkpoint-1244/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28a97346b080baca448a14b3b19e82610ee4f5f7f4be14e769a1ad86c2a53b9e
+size 791751704

28_128_e5_3e-5/checkpoint-1244/latest

@@ -0,0 +1 @@
+global_step1244

28_128_e5_3e-5/checkpoint-1244/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ea231a5f3c76cf3e378635b5fe9da55820f91a1e2a1ecf07ffa6356ef4fb815
+size 15920

28_128_e5_3e-5/checkpoint-1244/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92cc4ed17cb8bb40c61b30eb7a6c423ebfa83c36cb2ebd4007aaab705a131929
+size 15920

28_128_e5_3e-5/checkpoint-1244/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c56d64a480dda4c5f7ebac7784b85c17ff132dcb6775a3df4d179cdf1f7aee41
+size 15920

28_128_e5_3e-5/checkpoint-1244/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe7af9a1c1ac197c0112bc924d333954b13d57aa68d5b8115d9e54c74afc9f90
+size 15920

28_128_e5_3e-5/checkpoint-1244/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78adc7b3de25b2c712d771242c33122acb124cbc25fa61afbf352f0527c76a15
+size 15920

28_128_e5_3e-5/checkpoint-1244/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b49a34f720750a837c612e96922763d476e8f1fefcc9a3e15b49abf81f49a2fc
+size 15920

28_128_e5_3e-5/checkpoint-1244/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87651e4d1b2c5f8702238ca44e4a5738abf900524cc940f13b54b9d6cbc00151
+size 15920

28_128_e5_3e-5/checkpoint-1244/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45f45bbb0a47d6d9d3079a1383b5ad516509f402f788a4db29cb03d88412865b
+size 15920

28_128_e5_3e-5/checkpoint-1244/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1108af116c80e53b1fbb3f4a7634ef7202a3eae0b1e77a1748e5281338198a33
+size 1064

28_128_e5_3e-5/checkpoint-1244/special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "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": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

28_128_e5_3e-5/checkpoint-1244/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": "<|endoftext|>",
+  "padding_side": "left",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

28_128_e5_3e-5/checkpoint-1244/trainer_state.json

@@ -0,0 +1,1770 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 1244,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01610305958132045,
+      "grad_norm": 1.1391489505767822,
+      "learning_rate": 1.5384615384615385e-06,
+      "loss": 1.2885,
+      "step": 5
+    },
+    {
+      "epoch": 0.0322061191626409,
+      "grad_norm": 0.9224602580070496,
+      "learning_rate": 3.4615384615384617e-06,
+      "loss": 1.3364,
+      "step": 10
+    },
+    {
+      "epoch": 0.04830917874396135,
+      "grad_norm": 0.6716117262840271,
+      "learning_rate": 5.384615384615385e-06,
+      "loss": 1.2729,
+      "step": 15
+    },
+    {
+      "epoch": 0.0644122383252818,
+      "grad_norm": 0.6429063081741333,
+      "learning_rate": 7.307692307692308e-06,
+      "loss": 1.2407,
+      "step": 20
+    },
+    {
+      "epoch": 0.08051529790660225,
+      "grad_norm": 0.8855071663856506,
+      "learning_rate": 9.230769230769232e-06,
+      "loss": 1.2438,
+      "step": 25
+    },
+    {
+      "epoch": 0.0966183574879227,
+      "grad_norm": 0.5981923937797546,
+      "learning_rate": 1.1153846153846154e-05,
+      "loss": 1.2773,
+      "step": 30
+    },
+    {
+      "epoch": 0.11272141706924316,
+      "grad_norm": 0.5501493215560913,
+      "learning_rate": 1.3076923076923078e-05,
+      "loss": 1.2786,
+      "step": 35
+    },
+    {
+      "epoch": 0.1288244766505636,
+      "grad_norm": 0.5783489942550659,
+      "learning_rate": 1.5e-05,
+      "loss": 1.1728,
+      "step": 40
+    },
+    {
+      "epoch": 0.14492753623188406,
+      "grad_norm": 0.492718368768692,
+      "learning_rate": 1.6923076923076924e-05,
+      "loss": 1.1959,
+      "step": 45
+    },
+    {
+      "epoch": 0.1610305958132045,
+      "grad_norm": 0.5287221670150757,
+      "learning_rate": 1.8846153846153846e-05,
+      "loss": 1.1704,
+      "step": 50
+    },
+    {
+      "epoch": 0.17713365539452497,
+      "grad_norm": 0.5613968968391418,
+      "learning_rate": 2.076923076923077e-05,
+      "loss": 1.1734,
+      "step": 55
+    },
+    {
+      "epoch": 0.1932367149758454,
+      "grad_norm": 0.46896591782569885,
+      "learning_rate": 2.269230769230769e-05,
+      "loss": 1.1008,
+      "step": 60
+    },
+    {
+      "epoch": 0.20933977455716588,
+      "grad_norm": 0.6196975111961365,
+      "learning_rate": 2.4615384615384616e-05,
+      "loss": 1.1522,
+      "step": 65
+    },
+    {
+      "epoch": 0.22544283413848631,
+      "grad_norm": 0.4796932339668274,
+      "learning_rate": 2.6538461538461538e-05,
+      "loss": 1.133,
+      "step": 70
+    },
+    {
+      "epoch": 0.24154589371980675,
+      "grad_norm": 0.5374883413314819,
+      "learning_rate": 2.846153846153846e-05,
+      "loss": 1.1675,
+      "step": 75
+    },
+    {
+      "epoch": 0.2576489533011272,
+      "grad_norm": 0.4975369870662689,
+      "learning_rate": 2.999996606875036e-05,
+      "loss": 1.1127,
+      "step": 80
+    },
+    {
+      "epoch": 0.27375201288244766,
+      "grad_norm": 0.48042240738868713,
+      "learning_rate": 2.9998778491131415e-05,
+      "loss": 1.0703,
+      "step": 85
+    },
+    {
+      "epoch": 0.2898550724637681,
+      "grad_norm": 0.6405603885650635,
+      "learning_rate": 2.9995894504537867e-05,
+      "loss": 1.0789,
+      "step": 90
+    },
+    {
+      "epoch": 0.3059581320450886,
+      "grad_norm": 0.585081934928894,
+      "learning_rate": 2.999131443515766e-05,
+      "loss": 1.0105,
+      "step": 95
+    },
+    {
+      "epoch": 0.322061191626409,
+      "grad_norm": 0.7342278957366943,
+      "learning_rate": 2.998503880101102e-05,
+      "loss": 1.0567,
+      "step": 100
+    },
+    {
+      "epoch": 0.33816425120772947,
+      "grad_norm": 0.6049452424049377,
+      "learning_rate": 2.99770683118919e-05,
+      "loss": 1.0566,
+      "step": 105
+    },
+    {
+      "epoch": 0.35426731078904994,
+      "grad_norm": 0.5845102667808533,
+      "learning_rate": 2.996740386928766e-05,
+      "loss": 1.0511,
+      "step": 110
+    },
+    {
+      "epoch": 0.37037037037037035,
+      "grad_norm": 0.6237640976905823,
+      "learning_rate": 2.9956046566277126e-05,
+      "loss": 0.9973,
+      "step": 115
+    },
+    {
+      "epoch": 0.3864734299516908,
+      "grad_norm": 0.6733880043029785,
+      "learning_rate": 2.994299768740695e-05,
+      "loss": 0.9932,
+      "step": 120
+    },
+    {
+      "epoch": 0.4025764895330113,
+      "grad_norm": 0.668197751045227,
+      "learning_rate": 2.9928258708546335e-05,
+      "loss": 1.0146,
+      "step": 125
+    },
+    {
+      "epoch": 0.41867954911433175,
+      "grad_norm": 0.6295420527458191,
+      "learning_rate": 2.99118312967201e-05,
+      "loss": 0.9231,
+      "step": 130
+    },
+    {
+      "epoch": 0.43478260869565216,
+      "grad_norm": 0.624722421169281,
+      "learning_rate": 2.9893717309920134e-05,
+      "loss": 0.9942,
+      "step": 135
+    },
+    {
+      "epoch": 0.45088566827697263,
+      "grad_norm": 0.9356432557106018,
+      "learning_rate": 2.9873918796895273e-05,
+      "loss": 0.9984,
+      "step": 140
+    },
+    {
+      "epoch": 0.4669887278582931,
+      "grad_norm": 0.7457837462425232,
+      "learning_rate": 2.9852437996919537e-05,
+      "loss": 0.9668,
+      "step": 145
+    },
+    {
+      "epoch": 0.4830917874396135,
+      "grad_norm": 0.8209810853004456,
+      "learning_rate": 2.9829277339538903e-05,
+      "loss": 0.9771,
+      "step": 150
+    },
+    {
+      "epoch": 0.499194847020934,
+      "grad_norm": 0.6697631478309631,
+      "learning_rate": 2.9804439444296495e-05,
+      "loss": 0.9911,
+      "step": 155
+    },
+    {
+      "epoch": 0.5152979066022544,
+      "grad_norm": 0.6588212847709656,
+      "learning_rate": 2.9777927120436293e-05,
+      "loss": 0.9403,
+      "step": 160
+    },
+    {
+      "epoch": 0.5314009661835749,
+      "grad_norm": 0.7619159817695618,
+      "learning_rate": 2.974974336658545e-05,
+      "loss": 0.9147,
+      "step": 165
+    },
+    {
+      "epoch": 0.5475040257648953,
+      "grad_norm": 0.8137719631195068,
+      "learning_rate": 2.9719891370415072e-05,
+      "loss": 0.9449,
+      "step": 170
+    },
+    {
+      "epoch": 0.5636070853462157,
+      "grad_norm": 0.7924593091011047,
+      "learning_rate": 2.9688374508279715e-05,
+      "loss": 0.8465,
+      "step": 175
+    },
+    {
+      "epoch": 0.5797101449275363,
+      "grad_norm": 0.8301030993461609,
+      "learning_rate": 2.9655196344835528e-05,
+      "loss": 0.9196,
+      "step": 180
+    },
+    {
+      "epoch": 0.5958132045088567,
+      "grad_norm": 0.7267712950706482,
+      "learning_rate": 2.962036063263704e-05,
+      "loss": 0.846,
+      "step": 185
+    },
+    {
+      "epoch": 0.6119162640901772,
+      "grad_norm": 0.7453832030296326,
+      "learning_rate": 2.9583871311712753e-05,
+      "loss": 0.8713,
+      "step": 190
+    },
+    {
+      "epoch": 0.6280193236714976,
+      "grad_norm": 0.8772758841514587,
+      "learning_rate": 2.9545732509119526e-05,
+      "loss": 0.863,
+      "step": 195
+    },
+    {
+      "epoch": 0.644122383252818,
+      "grad_norm": 0.7424098253250122,
+      "learning_rate": 2.9505948538475754e-05,
+      "loss": 0.8345,
+      "step": 200
+    },
+    {
+      "epoch": 0.6602254428341385,
+      "grad_norm": 0.8788964748382568,
+      "learning_rate": 2.946452389947353e-05,
+      "loss": 0.8497,
+      "step": 205
+    },
+    {
+      "epoch": 0.6763285024154589,
+      "grad_norm": 1.0086477994918823,
+      "learning_rate": 2.9421463277369678e-05,
+      "loss": 0.7906,
+      "step": 210
+    },
+    {
+      "epoch": 0.6924315619967794,
+      "grad_norm": 0.9472551941871643,
+      "learning_rate": 2.9376771542455868e-05,
+      "loss": 0.8153,
+      "step": 215
+    },
+    {
+      "epoch": 0.7085346215780999,
+      "grad_norm": 0.8603758215904236,
+      "learning_rate": 2.9330453749507733e-05,
+      "loss": 0.7529,
+      "step": 220
+    },
+    {
+      "epoch": 0.7246376811594203,
+      "grad_norm": 0.8331507444381714,
+      "learning_rate": 2.9282515137213192e-05,
+      "loss": 0.7793,
+      "step": 225
+    },
+    {
+      "epoch": 0.7407407407407407,
+      "grad_norm": 0.8622682690620422,
+      "learning_rate": 2.9232961127579922e-05,
+      "loss": 0.7718,
+      "step": 230
+    },
+    {
+      "epoch": 0.7568438003220612,
+      "grad_norm": 0.9470008611679077,
+      "learning_rate": 2.9181797325322123e-05,
+      "loss": 0.7664,
+      "step": 235
+    },
+    {
+      "epoch": 0.7729468599033816,
+      "grad_norm": 0.9765321016311646,
+      "learning_rate": 2.912902951722658e-05,
+      "loss": 0.785,
+      "step": 240
+    },
+    {
+      "epoch": 0.789049919484702,
+      "grad_norm": 0.976305365562439,
+      "learning_rate": 2.907466367149819e-05,
+      "loss": 0.7656,
+      "step": 245
+    },
+    {
+      "epoch": 0.8051529790660226,
+      "grad_norm": 1.0335639715194702,
+      "learning_rate": 2.901870593708493e-05,
+      "loss": 0.7198,
+      "step": 250
+    },
+    {
+      "epoch": 0.821256038647343,
+      "grad_norm": 0.9155466556549072,
+      "learning_rate": 2.896116264298239e-05,
+      "loss": 0.7483,
+      "step": 255
+    },
+    {
+      "epoch": 0.8373590982286635,
+      "grad_norm": 0.8999483585357666,
+      "learning_rate": 2.890204029751793e-05,
+      "loss": 0.7637,
+      "step": 260
+    },
+    {
+      "epoch": 0.8534621578099839,
+      "grad_norm": 0.9181938171386719,
+      "learning_rate": 2.8841345587614587e-05,
+      "loss": 0.7486,
+      "step": 265
+    },
+    {
+      "epoch": 0.8695652173913043,
+      "grad_norm": 1.064903974533081,
+      "learning_rate": 2.8779085378034768e-05,
+      "loss": 0.7434,
+      "step": 270
+    },
+    {
+      "epoch": 0.8856682769726248,
+      "grad_norm": 0.9250743389129639,
+      "learning_rate": 2.87152667106038e-05,
+      "loss": 0.737,
+      "step": 275
+    },
+    {
+      "epoch": 0.9017713365539453,
+      "grad_norm": 0.9378864169120789,
+      "learning_rate": 2.864989680341348e-05,
+      "loss": 0.7047,
+      "step": 280
+    },
+    {
+      "epoch": 0.9178743961352657,
+      "grad_norm": 0.8972018957138062,
+      "learning_rate": 2.858298305000573e-05,
+      "loss": 0.6843,
+      "step": 285
+    },
+    {
+      "epoch": 0.9339774557165862,
+      "grad_norm": 0.9521335959434509,
+      "learning_rate": 2.8514533018536286e-05,
+      "loss": 0.7043,
+      "step": 290
+    },
+    {
+      "epoch": 0.9500805152979066,
+      "grad_norm": 1.0538114309310913,
+      "learning_rate": 2.84445544509188e-05,
+      "loss": 0.6254,
+      "step": 295
+    },
+    {
+      "epoch": 0.966183574879227,
+      "grad_norm": 0.8278699517250061,
+      "learning_rate": 2.8373055261949138e-05,
+      "loss": 0.6942,
+      "step": 300
+    },
+    {
+      "epoch": 0.9822866344605475,
+      "grad_norm": 0.9784626960754395,
+      "learning_rate": 2.830004353841023e-05,
+      "loss": 0.6504,
+      "step": 305
+    },
+    {
+      "epoch": 0.998389694041868,
+      "grad_norm": 1.0742645263671875,
+      "learning_rate": 2.8225527538157413e-05,
+      "loss": 0.6466,
+      "step": 310
+    },
+    {
+      "epoch": 1.0128824476650564,
+      "grad_norm": 1.3430242538452148,
+      "learning_rate": 2.814951568918446e-05,
+      "loss": 0.6389,
+      "step": 315
+    },
+    {
+      "epoch": 1.0289855072463767,
+      "grad_norm": 1.1638330221176147,
+      "learning_rate": 2.8072016588670316e-05,
+      "loss": 0.5618,
+      "step": 320
+    },
+    {
+      "epoch": 1.0450885668276972,
+      "grad_norm": 1.0941189527511597,
+      "learning_rate": 2.799303900200676e-05,
+      "loss": 0.5797,
+      "step": 325
+    },
+    {
+      "epoch": 1.0611916264090178,
+      "grad_norm": 0.9851027727127075,
+      "learning_rate": 2.7912591861806994e-05,
+      "loss": 0.5489,
+      "step": 330
+    },
+    {
+      "epoch": 1.077294685990338,
+      "grad_norm": 1.0907893180847168,
+      "learning_rate": 2.7830684266895352e-05,
+      "loss": 0.5627,
+      "step": 335
+    },
+    {
+      "epoch": 1.0933977455716586,
+      "grad_norm": 1.039016604423523,
+      "learning_rate": 2.774732548127817e-05,
+      "loss": 0.5682,
+      "step": 340
+    },
+    {
+      "epoch": 1.109500805152979,
+      "grad_norm": 1.169583797454834,
+      "learning_rate": 2.766252493309603e-05,
+      "loss": 0.5668,
+      "step": 345
+    },
+    {
+      "epoch": 1.1256038647342996,
+      "grad_norm": 1.1832174062728882,
+      "learning_rate": 2.7576292213557372e-05,
+      "loss": 0.5808,
+      "step": 350
+    },
+    {
+      "epoch": 1.14170692431562,
+      "grad_norm": 1.1079707145690918,
+      "learning_rate": 2.748863707585373e-05,
+      "loss": 0.5511,
+      "step": 355
+    },
+    {
+      "epoch": 1.1578099838969405,
+      "grad_norm": 1.1411006450653076,
+      "learning_rate": 2.7399569434056595e-05,
+      "loss": 0.5355,
+      "step": 360
+    },
+    {
+      "epoch": 1.1739130434782608,
+      "grad_norm": 1.250238299369812,
+      "learning_rate": 2.7309099361996115e-05,
+      "loss": 0.5373,
+      "step": 365
+    },
+    {
+      "epoch": 1.1900161030595813,
+      "grad_norm": 1.157636046409607,
+      "learning_rate": 2.7217237092121703e-05,
+      "loss": 0.539,
+      "step": 370
+    },
+    {
+      "epoch": 1.2061191626409018,
+      "grad_norm": 1.102122187614441,
+      "learning_rate": 2.7123993014344714e-05,
+      "loss": 0.5551,
+      "step": 375
+    },
+    {
+      "epoch": 1.2222222222222223,
+      "grad_norm": 1.134345293045044,
+      "learning_rate": 2.7029377674863332e-05,
+      "loss": 0.5424,
+      "step": 380
+    },
+    {
+      "epoch": 1.2383252818035426,
+      "grad_norm": 1.1444910764694214,
+      "learning_rate": 2.6933401774969734e-05,
+      "loss": 0.5118,
+      "step": 385
+    },
+    {
+      "epoch": 1.2544283413848631,
+      "grad_norm": 1.1335145235061646,
+      "learning_rate": 2.6836076169839765e-05,
+      "loss": 0.5002,
+      "step": 390
+    },
+    {
+      "epoch": 1.2705314009661834,
+      "grad_norm": 1.1598331928253174,
+      "learning_rate": 2.673741186730517e-05,
+      "loss": 0.4804,
+      "step": 395
+    },
+    {
+      "epoch": 1.286634460547504,
+      "grad_norm": 1.0515048503875732,
+      "learning_rate": 2.6637420026608573e-05,
+      "loss": 0.5247,
+      "step": 400
+    },
+    {
+      "epoch": 1.3027375201288245,
+      "grad_norm": 0.957242488861084,
+      "learning_rate": 2.653611195714134e-05,
+      "loss": 0.453,
+      "step": 405
+    },
+    {
+      "epoch": 1.318840579710145,
+      "grad_norm": 1.0974361896514893,
+      "learning_rate": 2.6433499117164444e-05,
+      "loss": 0.5344,
+      "step": 410
+    },
+    {
+      "epoch": 1.3349436392914653,
+      "grad_norm": 0.9843302965164185,
+      "learning_rate": 2.6329593112512508e-05,
+      "loss": 0.5074,
+      "step": 415
+    },
+    {
+      "epoch": 1.3510466988727858,
+      "grad_norm": 1.0893441438674927,
+      "learning_rate": 2.6224405695281136e-05,
+      "loss": 0.4771,
+      "step": 420
+    },
+    {
+      "epoch": 1.3671497584541064,
+      "grad_norm": 1.155216097831726,
+      "learning_rate": 2.611794876249772e-05,
+      "loss": 0.5031,
+      "step": 425
+    },
+    {
+      "epoch": 1.3832528180354267,
+      "grad_norm": 0.9953705072402954,
+      "learning_rate": 2.601023435477586e-05,
+      "loss": 0.491,
+      "step": 430
+    },
+    {
+      "epoch": 1.3993558776167472,
+      "grad_norm": 1.152533769607544,
+      "learning_rate": 2.5901274654953506e-05,
+      "loss": 0.4655,
+      "step": 435
+    },
+    {
+      "epoch": 1.4154589371980677,
+      "grad_norm": 1.1234312057495117,
+      "learning_rate": 2.5791081986715076e-05,
+      "loss": 0.4415,
+      "step": 440
+    },
+    {
+      "epoch": 1.431561996779388,
+      "grad_norm": 1.0570783615112305,
+      "learning_rate": 2.5679668813197583e-05,
+      "loss": 0.4547,
+      "step": 445
+    },
+    {
+      "epoch": 1.4476650563607085,
+      "grad_norm": 1.0669089555740356,
+      "learning_rate": 2.556704773558101e-05,
+      "loss": 0.4658,
+      "step": 450
+    },
+    {
+      "epoch": 1.463768115942029,
+      "grad_norm": 1.142380714416504,
+      "learning_rate": 2.5453231491663104e-05,
+      "loss": 0.4596,
+      "step": 455
+    },
+    {
+      "epoch": 1.4798711755233493,
+      "grad_norm": 1.0438144207000732,
+      "learning_rate": 2.5338232954418675e-05,
+      "loss": 0.4214,
+      "step": 460
+    },
+    {
+      "epoch": 1.4959742351046699,
+      "grad_norm": 1.1825107336044312,
+      "learning_rate": 2.522206513054361e-05,
+      "loss": 0.4783,
+      "step": 465
+    },
+    {
+      "epoch": 1.5120772946859904,
+      "grad_norm": 1.420599102973938,
+      "learning_rate": 2.5104741158983814e-05,
+      "loss": 0.4716,
+      "step": 470
+    },
+    {
+      "epoch": 1.528180354267311,
+      "grad_norm": 1.1626514196395874,
+      "learning_rate": 2.4986274309449088e-05,
+      "loss": 0.4689,
+      "step": 475
+    },
+    {
+      "epoch": 1.5442834138486312,
+      "grad_norm": 1.133529543876648,
+      "learning_rate": 2.4866677980912363e-05,
+      "loss": 0.4341,
+      "step": 480
+    },
+    {
+      "epoch": 1.5603864734299517,
+      "grad_norm": 1.196297526359558,
+      "learning_rate": 2.4745965700094172e-05,
+      "loss": 0.5102,
+      "step": 485
+    },
+    {
+      "epoch": 1.576489533011272,
+      "grad_norm": 1.0283925533294678,
+      "learning_rate": 2.462415111993276e-05,
+      "loss": 0.4533,
+      "step": 490
+    },
+    {
+      "epoch": 1.5925925925925926,
+      "grad_norm": 1.2448772192001343,
+      "learning_rate": 2.4501248018039897e-05,
+      "loss": 0.4215,
+      "step": 495
+    },
+    {
+      "epoch": 1.608695652173913,
+      "grad_norm": 1.2707397937774658,
+      "learning_rate": 2.4377270295142582e-05,
+      "loss": 0.4395,
+      "step": 500
+    },
+    {
+      "epoch": 1.6247987117552336,
+      "grad_norm": 1.6315351724624634,
+      "learning_rate": 2.425223197351081e-05,
+      "loss": 0.427,
+      "step": 505
+    },
+    {
+      "epoch": 1.640901771336554,
+      "grad_norm": 1.1135799884796143,
+      "learning_rate": 2.4126147195371647e-05,
+      "loss": 0.3926,
+      "step": 510
+    },
+    {
+      "epoch": 1.6570048309178744,
+      "grad_norm": 1.2687816619873047,
+      "learning_rate": 2.3999030221309655e-05,
+      "loss": 0.3848,
+      "step": 515
+    },
+    {
+      "epoch": 1.6731078904991947,
+      "grad_norm": 1.1300817728042603,
+      "learning_rate": 2.3870895428654035e-05,
+      "loss": 0.3926,
+      "step": 520
+    },
+    {
+      "epoch": 1.6892109500805152,
+      "grad_norm": 1.067254662513733,
+      "learning_rate": 2.374175730985242e-05,
+      "loss": 0.3987,
+      "step": 525
+    },
+    {
+      "epoch": 1.7053140096618358,
+      "grad_norm": 1.2030500173568726,
+      "learning_rate": 2.3611630470831824e-05,
+      "loss": 0.3844,
+      "step": 530
+    },
+    {
+      "epoch": 1.7214170692431563,
+      "grad_norm": 1.1542716026306152,
+      "learning_rate": 2.3480529629346597e-05,
+      "loss": 0.383,
+      "step": 535
+    },
+    {
+      "epoch": 1.7375201288244766,
+      "grad_norm": 1.099084496498108,
+      "learning_rate": 2.3348469613313854e-05,
+      "loss": 0.422,
+      "step": 540
+    },
+    {
+      "epoch": 1.7536231884057971,
+      "grad_norm": 1.084632158279419,
+      "learning_rate": 2.3215465359136338e-05,
+      "loss": 0.3831,
+      "step": 545
+    },
+    {
+      "epoch": 1.7697262479871174,
+      "grad_norm": 1.0466201305389404,
+      "learning_rate": 2.3081531910013104e-05,
+      "loss": 0.4008,
+      "step": 550
+    },
+    {
+      "epoch": 1.785829307568438,
+      "grad_norm": 1.0963324308395386,
+      "learning_rate": 2.2946684414238093e-05,
+      "loss": 0.3731,
+      "step": 555
+    },
+    {
+      "epoch": 1.8019323671497585,
+      "grad_norm": 1.297102928161621,
+      "learning_rate": 2.281093812348676e-05,
+      "loss": 0.3656,
+      "step": 560
+    },
+    {
+      "epoch": 1.818035426731079,
+      "grad_norm": 1.0965611934661865,
+      "learning_rate": 2.2674308391091127e-05,
+      "loss": 0.3477,
+      "step": 565
+    },
+    {
+      "epoch": 1.8341384863123995,
+      "grad_norm": 1.2567371129989624,
+      "learning_rate": 2.2536810670303214e-05,
+      "loss": 0.3658,
+      "step": 570
+    },
+    {
+      "epoch": 1.8502415458937198,
+      "grad_norm": 1.1441166400909424,
+      "learning_rate": 2.239846051254728e-05,
+      "loss": 0.3489,
+      "step": 575
+    },
+    {
+      "epoch": 1.86634460547504,
+      "grad_norm": 1.2327871322631836,
+      "learning_rate": 2.2259273565660876e-05,
+      "loss": 0.3866,
+      "step": 580
+    },
+    {
+      "epoch": 1.8824476650563606,
+      "grad_norm": 1.0107234716415405,
+      "learning_rate": 2.2119265572125027e-05,
+      "loss": 0.3854,
+      "step": 585
+    },
+    {
+      "epoch": 1.8985507246376812,
+      "grad_norm": 1.259210228919983,
+      "learning_rate": 2.197845236728371e-05,
+      "loss": 0.3592,
+      "step": 590
+    },
+    {
+      "epoch": 1.9146537842190017,
+      "grad_norm": 1.232729196548462,
+      "learning_rate": 2.1836849877552854e-05,
+      "loss": 0.3441,
+      "step": 595
+    },
+    {
+      "epoch": 1.9307568438003222,
+      "grad_norm": 1.1457910537719727,
+      "learning_rate": 2.1694474118618954e-05,
+      "loss": 0.344,
+      "step": 600
+    },
+    {
+      "epoch": 1.9468599033816425,
+      "grad_norm": 1.2113591432571411,
+      "learning_rate": 2.1551341193627717e-05,
+      "loss": 0.3344,
+      "step": 605
+    },
+    {
+      "epoch": 1.9629629629629628,
+      "grad_norm": 1.2011780738830566,
+      "learning_rate": 2.1407467291362686e-05,
+      "loss": 0.3739,
+      "step": 610
+    },
+    {
+      "epoch": 1.9790660225442833,
+      "grad_norm": 1.178965449333191,
+      "learning_rate": 2.1262868684414265e-05,
+      "loss": 0.3731,
+      "step": 615
+    },
+    {
+      "epoch": 1.9951690821256038,
+      "grad_norm": 1.1474207639694214,
+      "learning_rate": 2.111756172733925e-05,
+      "loss": 0.366,
+      "step": 620
+    },
+    {
+      "epoch": 2.0096618357487923,
+      "grad_norm": 1.1576290130615234,
+      "learning_rate": 2.0971562854811044e-05,
+      "loss": 0.2854,
+      "step": 625
+    },
+    {
+      "epoch": 2.025764895330113,
+      "grad_norm": 1.0710585117340088,
+      "learning_rate": 2.0824888579760874e-05,
+      "loss": 0.2722,
+      "step": 630
+    },
+    {
+      "epoch": 2.0418679549114334,
+      "grad_norm": 1.2322947978973389,
+      "learning_rate": 2.06775554915101e-05,
+      "loss": 0.2559,
+      "step": 635
+    },
+    {
+      "epoch": 2.0579710144927534,
+      "grad_norm": 1.1783329248428345,
+      "learning_rate": 2.0529580253893944e-05,
+      "loss": 0.2794,
+      "step": 640
+    },
+    {
+      "epoch": 2.074074074074074,
+      "grad_norm": 1.219993233680725,
+      "learning_rate": 2.0380979603376726e-05,
+      "loss": 0.2706,
+      "step": 645
+    },
+    {
+      "epoch": 2.0901771336553945,
+      "grad_norm": 1.0898008346557617,
+      "learning_rate": 2.0231770347158933e-05,
+      "loss": 0.2613,
+      "step": 650
+    },
+    {
+      "epoch": 2.106280193236715,
+      "grad_norm": 1.127703070640564,
+      "learning_rate": 2.0081969361276276e-05,
+      "loss": 0.2319,
+      "step": 655
+    },
+    {
+      "epoch": 2.1223832528180355,
+      "grad_norm": 1.1997917890548706,
+      "learning_rate": 1.9931593588690932e-05,
+      "loss": 0.2639,
+      "step": 660
+    },
+    {
+      "epoch": 2.138486312399356,
+      "grad_norm": 1.1341209411621094,
+      "learning_rate": 1.9780660037375284e-05,
+      "loss": 0.2476,
+      "step": 665
+    },
+    {
+      "epoch": 2.154589371980676,
+      "grad_norm": 1.404829502105713,
+      "learning_rate": 1.9629185778388215e-05,
+      "loss": 0.2549,
+      "step": 670
+    },
+    {
+      "epoch": 2.1706924315619966,
+      "grad_norm": 1.1979633569717407,
+      "learning_rate": 1.947718794394438e-05,
+      "loss": 0.2537,
+      "step": 675
+    },
+    {
+      "epoch": 2.186795491143317,
+      "grad_norm": 1.1800684928894043,
+      "learning_rate": 1.932468372547643e-05,
+      "loss": 0.2771,
+      "step": 680
+    },
+    {
+      "epoch": 2.2028985507246377,
+      "grad_norm": 1.140649437904358,
+      "learning_rate": 1.9171690371690655e-05,
+      "loss": 0.2173,
+      "step": 685
+    },
+    {
+      "epoch": 2.219001610305958,
+      "grad_norm": 1.2186757326126099,
+      "learning_rate": 1.9018225186616085e-05,
+      "loss": 0.2474,
+      "step": 690
+    },
+    {
+      "epoch": 2.2351046698872787,
+      "grad_norm": 1.1380257606506348,
+      "learning_rate": 1.886430552764735e-05,
+      "loss": 0.2718,
+      "step": 695
+    },
+    {
+      "epoch": 2.2512077294685993,
+      "grad_norm": 1.2019944190979004,
+      "learning_rate": 1.870994880358151e-05,
+      "loss": 0.2524,
+      "step": 700
+    },
+    {
+      "epoch": 2.2673107890499193,
+      "grad_norm": 1.1367535591125488,
+      "learning_rate": 1.8555172472649048e-05,
+      "loss": 0.2537,
+      "step": 705
+    },
+    {
+      "epoch": 2.28341384863124,
+      "grad_norm": 1.3347067832946777,
+      "learning_rate": 1.8399994040539322e-05,
+      "loss": 0.2532,
+      "step": 710
+    },
+    {
+      "epoch": 2.2995169082125604,
+      "grad_norm": 1.1048965454101562,
+      "learning_rate": 1.8244431058420576e-05,
+      "loss": 0.243,
+      "step": 715
+    },
+    {
+      "epoch": 2.315619967793881,
+      "grad_norm": 1.1416301727294922,
+      "learning_rate": 1.808850112095489e-05,
+      "loss": 0.2702,
+      "step": 720
+    },
+    {
+      "epoch": 2.3317230273752014,
+      "grad_norm": 1.139872431755066,
+      "learning_rate": 1.7932221864308118e-05,
+      "loss": 0.2278,
+      "step": 725
+    },
+    {
+      "epoch": 2.3478260869565215,
+      "grad_norm": 1.0932283401489258,
+      "learning_rate": 1.7775610964155246e-05,
+      "loss": 0.2578,
+      "step": 730
+    },
+    {
+      "epoch": 2.363929146537842,
+      "grad_norm": 1.0193161964416504,
+      "learning_rate": 1.761868613368115e-05,
+      "loss": 0.2336,
+      "step": 735
+    },
+    {
+      "epoch": 2.3800322061191626,
+      "grad_norm": 1.3200891017913818,
+      "learning_rate": 1.7461465121577234e-05,
+      "loss": 0.2587,
+      "step": 740
+    },
+    {
+      "epoch": 2.396135265700483,
+      "grad_norm": 1.067944884300232,
+      "learning_rate": 1.7303965710033957e-05,
+      "loss": 0.2546,
+      "step": 745
+    },
+    {
+      "epoch": 2.4122383252818036,
+      "grad_norm": 1.0622363090515137,
+      "learning_rate": 1.714620571272965e-05,
+      "loss": 0.2567,
+      "step": 750
+    },
+    {
+      "epoch": 2.428341384863124,
+      "grad_norm": 1.0919643640518188,
+      "learning_rate": 1.69882029728157e-05,
+      "loss": 0.2284,
+      "step": 755
+    },
+    {
+      "epoch": 2.4444444444444446,
+      "grad_norm": 1.2129340171813965,
+      "learning_rate": 1.682997536089844e-05,
+      "loss": 0.2458,
+      "step": 760
+    },
+    {
+      "epoch": 2.4605475040257647,
+      "grad_norm": 1.169572114944458,
+      "learning_rate": 1.6671540773017952e-05,
+      "loss": 0.2036,
+      "step": 765
+    },
+    {
+      "epoch": 2.4766505636070852,
+      "grad_norm": 1.082538366317749,
+      "learning_rate": 1.6512917128623927e-05,
+      "loss": 0.2383,
+      "step": 770
+    },
+    {
+      "epoch": 2.4927536231884058,
+      "grad_norm": 1.3170980215072632,
+      "learning_rate": 1.6354122368548965e-05,
+      "loss": 0.2345,
+      "step": 775
+    },
+    {
+      "epoch": 2.5088566827697263,
+      "grad_norm": 1.1367696523666382,
+      "learning_rate": 1.6195174452979365e-05,
+      "loss": 0.204,
+      "step": 780
+    },
+    {
+      "epoch": 2.524959742351047,
+      "grad_norm": 1.016545295715332,
+      "learning_rate": 1.6036091359423807e-05,
+      "loss": 0.2246,
+      "step": 785
+    },
+    {
+      "epoch": 2.541062801932367,
+      "grad_norm": 1.1528297662734985,
+      "learning_rate": 1.587689108068002e-05,
+      "loss": 0.2489,
+      "step": 790
+    },
+    {
+      "epoch": 2.5571658615136874,
+      "grad_norm": 1.1874185800552368,
+      "learning_rate": 1.571759162279975e-05,
+      "loss": 0.2267,
+      "step": 795
+    },
+    {
+      "epoch": 2.573268921095008,
+      "grad_norm": 1.1973860263824463,
+      "learning_rate": 1.5558211003052222e-05,
+      "loss": 0.2096,
+      "step": 800
+    },
+    {
+      "epoch": 2.5893719806763285,
+      "grad_norm": 1.1747304201126099,
+      "learning_rate": 1.539876724788631e-05,
+      "loss": 0.2161,
+      "step": 805
+    },
+    {
+      "epoch": 2.605475040257649,
+      "grad_norm": 1.0052316188812256,
+      "learning_rate": 1.5239278390891703e-05,
+      "loss": 0.1939,
+      "step": 810
+    },
+    {
+      "epoch": 2.6215780998389695,
+      "grad_norm": 1.1736071109771729,
+      "learning_rate": 1.5079762470759255e-05,
+      "loss": 0.1796,
+      "step": 815
+    },
+    {
+      "epoch": 2.63768115942029,
+      "grad_norm": 1.1553905010223389,
+      "learning_rate": 1.492023752924075e-05,
+      "loss": 0.2206,
+      "step": 820
+    },
+    {
+      "epoch": 2.6537842190016105,
+      "grad_norm": 1.1105087995529175,
+      "learning_rate": 1.4760721609108302e-05,
+      "loss": 0.1833,
+      "step": 825
+    },
+    {
+      "epoch": 2.6698872785829306,
+      "grad_norm": 1.209580659866333,
+      "learning_rate": 1.4601232752113696e-05,
+      "loss": 0.1896,
+      "step": 830
+    },
+    {
+      "epoch": 2.685990338164251,
+      "grad_norm": 1.0659430027008057,
+      "learning_rate": 1.444178899694778e-05,
+      "loss": 0.2153,
+      "step": 835
+    },
+    {
+      "epoch": 2.7020933977455717,
+      "grad_norm": 1.2064878940582275,
+      "learning_rate": 1.428240837720025e-05,
+      "loss": 0.222,
+      "step": 840
+    },
+    {
+      "epoch": 2.718196457326892,
+      "grad_norm": 0.9531728029251099,
+      "learning_rate": 1.4123108919319986e-05,
+      "loss": 0.2248,
+      "step": 845
+    },
+    {
+      "epoch": 2.7342995169082127,
+      "grad_norm": 1.1249150037765503,
+      "learning_rate": 1.39639086405762e-05,
+      "loss": 0.2273,
+      "step": 850
+    },
+    {
+      "epoch": 2.750402576489533,
+      "grad_norm": 1.3298691511154175,
+      "learning_rate": 1.3804825547020644e-05,
+      "loss": 0.1782,
+      "step": 855
+    },
+    {
+      "epoch": 2.7665056360708533,
+      "grad_norm": 0.9931784868240356,
+      "learning_rate": 1.364587763145104e-05,
+      "loss": 0.1956,
+      "step": 860
+    },
+    {
+      "epoch": 2.782608695652174,
+      "grad_norm": 1.1622772216796875,
+      "learning_rate": 1.3487082871376075e-05,
+      "loss": 0.2083,
+      "step": 865
+    },
+    {
+      "epoch": 2.7987117552334944,
+      "grad_norm": 1.6023427248001099,
+      "learning_rate": 1.332845922698205e-05,
+      "loss": 0.1982,
+      "step": 870
+    },
+    {
+      "epoch": 2.814814814814815,
+      "grad_norm": 1.1077145338058472,
+      "learning_rate": 1.3170024639101561e-05,
+      "loss": 0.2135,
+      "step": 875
+    },
+    {
+      "epoch": 2.8309178743961354,
+      "grad_norm": 1.1962898969650269,
+      "learning_rate": 1.3011797027184303e-05,
+      "loss": 0.1855,
+      "step": 880
+    },
+    {
+      "epoch": 2.847020933977456,
+      "grad_norm": 0.966684877872467,
+      "learning_rate": 1.2853794287270352e-05,
+      "loss": 0.1766,
+      "step": 885
+    },
+    {
+      "epoch": 2.863123993558776,
+      "grad_norm": 1.170991063117981,
+      "learning_rate": 1.2696034289966044e-05,
+      "loss": 0.2242,
+      "step": 890
+    },
+    {
+      "epoch": 2.8792270531400965,
+      "grad_norm": 1.1262538433074951,
+      "learning_rate": 1.2538534878422772e-05,
+      "loss": 0.1887,
+      "step": 895
+    },
+    {
+      "epoch": 2.895330112721417,
+      "grad_norm": 1.3256655931472778,
+      "learning_rate": 1.2381313866318854e-05,
+      "loss": 0.1948,
+      "step": 900
+    },
+    {
+      "epoch": 2.9114331723027376,
+      "grad_norm": 1.0575200319290161,
+      "learning_rate": 1.2224389035844756e-05,
+      "loss": 0.1989,
+      "step": 905
+    },
+    {
+      "epoch": 2.927536231884058,
+      "grad_norm": 1.126052975654602,
+      "learning_rate": 1.2067778135691881e-05,
+      "loss": 0.1872,
+      "step": 910
+    },
+    {
+      "epoch": 2.943639291465378,
+      "grad_norm": 1.2957228422164917,
+      "learning_rate": 1.1911498879045113e-05,
+      "loss": 0.1808,
+      "step": 915
+    },
+    {
+      "epoch": 2.9597423510466987,
+      "grad_norm": 1.0871576070785522,
+      "learning_rate": 1.1755568941579424e-05,
+      "loss": 0.1736,
+      "step": 920
+    },
+    {
+      "epoch": 2.975845410628019,
+      "grad_norm": 1.102060317993164,
+      "learning_rate": 1.160000595946068e-05,
+      "loss": 0.1767,
+      "step": 925
+    },
+    {
+      "epoch": 2.9919484702093397,
+      "grad_norm": 0.9936461448669434,
+      "learning_rate": 1.1444827527350954e-05,
+      "loss": 0.1695,
+      "step": 930
+    },
+    {
+      "epoch": 3.006441223832528,
+      "grad_norm": 0.9595783948898315,
+      "learning_rate": 1.1290051196418494e-05,
+      "loss": 0.1457,
+      "step": 935
+    },
+    {
+      "epoch": 3.0225442834138487,
+      "grad_norm": 1.0653377771377563,
+      "learning_rate": 1.1135694472352652e-05,
+      "loss": 0.1399,
+      "step": 940
+    },
+    {
+      "epoch": 3.0386473429951693,
+      "grad_norm": 1.34004807472229,
+      "learning_rate": 1.0981774813383916e-05,
+      "loss": 0.1538,
+      "step": 945
+    },
+    {
+      "epoch": 3.0547504025764893,
+      "grad_norm": 1.1128737926483154,
+      "learning_rate": 1.0828309628309347e-05,
+      "loss": 0.1252,
+      "step": 950
+    },
+    {
+      "epoch": 3.07085346215781,
+      "grad_norm": 1.0216444730758667,
+      "learning_rate": 1.0675316274523571e-05,
+      "loss": 0.1356,
+      "step": 955
+    },
+    {
+      "epoch": 3.0869565217391304,
+      "grad_norm": 1.013954997062683,
+      "learning_rate": 1.0522812056055617e-05,
+      "loss": 0.1334,
+      "step": 960
+    },
+    {
+      "epoch": 3.103059581320451,
+      "grad_norm": 1.003402590751648,
+      "learning_rate": 1.037081422161178e-05,
+      "loss": 0.1263,
+      "step": 965
+    },
+    {
+      "epoch": 3.1191626409017714,
+      "grad_norm": 1.1828265190124512,
+      "learning_rate": 1.0219339962624717e-05,
+      "loss": 0.1254,
+      "step": 970
+    },
+    {
+      "epoch": 3.135265700483092,
+      "grad_norm": 1.0109069347381592,
+      "learning_rate": 1.0068406411309068e-05,
+      "loss": 0.1307,
+      "step": 975
+    },
+    {
+      "epoch": 3.151368760064412,
+      "grad_norm": 1.0399396419525146,
+      "learning_rate": 9.918030638723725e-06,
+      "loss": 0.1195,
+      "step": 980
+    },
+    {
+      "epoch": 3.1674718196457325,
+      "grad_norm": 0.9721539616584778,
+      "learning_rate": 9.768229652841066e-06,
+      "loss": 0.1347,
+      "step": 985
+    },
+    {
+      "epoch": 3.183574879227053,
+      "grad_norm": 0.9482682943344116,
+      "learning_rate": 9.61902039662328e-06,
+      "loss": 0.147,
+      "step": 990
+    },
+    {
+      "epoch": 3.1996779388083736,
+      "grad_norm": 0.9318935871124268,
+      "learning_rate": 9.47041974610606e-06,
+      "loss": 0.1341,
+      "step": 995
+    },
+    {
+      "epoch": 3.215780998389694,
+      "grad_norm": 1.1986130475997925,
+      "learning_rate": 9.322444508489905e-06,
+      "loss": 0.1155,
+      "step": 1000
+    },
+    {
+      "epoch": 3.2318840579710146,
+      "grad_norm": 1.0879420042037964,
+      "learning_rate": 9.175111420239133e-06,
+      "loss": 0.1399,
+      "step": 1005
+    },
+    {
+      "epoch": 3.247987117552335,
+      "grad_norm": 0.9248384833335876,
+      "learning_rate": 9.028437145188963e-06,
+      "loss": 0.1218,
+      "step": 1010
+    },
+    {
+      "epoch": 3.2640901771336552,
+      "grad_norm": 1.1074427366256714,
+      "learning_rate": 8.882438272660755e-06,
+      "loss": 0.154,
+      "step": 1015
+    },
+    {
+      "epoch": 3.2801932367149758,
+      "grad_norm": 1.03411865234375,
+      "learning_rate": 8.737131315585736e-06,
+      "loss": 0.1329,
+      "step": 1020
+    },
+    {
+      "epoch": 3.2962962962962963,
+      "grad_norm": 0.847226619720459,
+      "learning_rate": 8.592532708637319e-06,
+      "loss": 0.1385,
+      "step": 1025
+    },
+    {
+      "epoch": 3.312399355877617,
+      "grad_norm": 0.9666579961776733,
+      "learning_rate": 8.448658806372289e-06,
+      "loss": 0.1359,
+      "step": 1030
+    },
+    {
+      "epoch": 3.3285024154589373,
+      "grad_norm": 1.055481195449829,
+      "learning_rate": 8.305525881381045e-06,
+      "loss": 0.1445,
+      "step": 1035
+    },
+    {
+      "epoch": 3.3446054750402574,
+      "grad_norm": 1.088436484336853,
+      "learning_rate": 8.16315012244715e-06,
+      "loss": 0.1073,
+      "step": 1040
+    },
+    {
+      "epoch": 3.360708534621578,
+      "grad_norm": 0.939185380935669,
+      "learning_rate": 8.021547632716291e-06,
+      "loss": 0.1106,
+      "step": 1045
+    },
+    {
+      "epoch": 3.3768115942028984,
+      "grad_norm": 1.077433705329895,
+      "learning_rate": 7.88073442787498e-06,
+      "loss": 0.1247,
+      "step": 1050
+    },
+    {
+      "epoch": 3.392914653784219,
+      "grad_norm": 1.1379334926605225,
+      "learning_rate": 7.740726434339125e-06,
+      "loss": 0.1339,
+      "step": 1055
+    },
+    {
+      "epoch": 3.4090177133655395,
+      "grad_norm": 0.8844690322875977,
+      "learning_rate": 7.60153948745272e-06,
+      "loss": 0.1169,
+      "step": 1060
+    },
+    {
+      "epoch": 3.42512077294686,
+      "grad_norm": 1.0570695400238037,
+      "learning_rate": 7.463189329696789e-06,
+      "loss": 0.1228,
+      "step": 1065
+    },
+    {
+      "epoch": 3.4412238325281805,
+      "grad_norm": 1.0314443111419678,
+      "learning_rate": 7.3256916089088825e-06,
+      "loss": 0.1339,
+      "step": 1070
+    },
+    {
+      "epoch": 3.4573268921095006,
+      "grad_norm": 0.9575313329696655,
+      "learning_rate": 7.189061876513241e-06,
+      "loss": 0.1418,
+      "step": 1075
+    },
+    {
+      "epoch": 3.473429951690821,
+      "grad_norm": 0.9746488332748413,
+      "learning_rate": 7.053315585761911e-06,
+      "loss": 0.1268,
+      "step": 1080
+    },
+    {
+      "epoch": 3.4895330112721417,
+      "grad_norm": 1.0186628103256226,
+      "learning_rate": 6.9184680899868955e-06,
+      "loss": 0.1346,
+      "step": 1085
+    },
+    {
+      "epoch": 3.505636070853462,
+      "grad_norm": 1.020210862159729,
+      "learning_rate": 6.784534640863668e-06,
+      "loss": 0.1264,
+      "step": 1090
+    },
+    {
+      "epoch": 3.5217391304347827,
+      "grad_norm": 0.9518364071846008,
+      "learning_rate": 6.651530386686148e-06,
+      "loss": 0.1332,
+      "step": 1095
+    },
+    {
+      "epoch": 3.537842190016103,
+      "grad_norm": 1.0923478603363037,
+      "learning_rate": 6.519470370653402e-06,
+      "loss": 0.1183,
+      "step": 1100
+    },
+    {
+      "epoch": 3.5539452495974233,
+      "grad_norm": 1.100982666015625,
+      "learning_rate": 6.388369529168181e-06,
+      "loss": 0.1185,
+      "step": 1105
+    },
+    {
+      "epoch": 3.570048309178744,
+      "grad_norm": 0.9394665956497192,
+      "learning_rate": 6.258242690147579e-06,
+      "loss": 0.1018,
+      "step": 1110
+    },
+    {
+      "epoch": 3.5861513687600644,
+      "grad_norm": 0.7709954380989075,
+      "learning_rate": 6.129104571345969e-06,
+      "loss": 0.1156,
+      "step": 1115
+    },
+    {
+      "epoch": 3.602254428341385,
+      "grad_norm": 0.9713561534881592,
+      "learning_rate": 6.000969778690342e-06,
+      "loss": 0.1164,
+      "step": 1120
+    },
+    {
+      "epoch": 3.6183574879227054,
+      "grad_norm": 0.8961610198020935,
+      "learning_rate": 5.873852804628358e-06,
+      "loss": 0.1068,
+      "step": 1125
+    },
+    {
+      "epoch": 3.634460547504026,
+      "grad_norm": 0.9026066064834595,
+      "learning_rate": 5.747768026489188e-06,
+      "loss": 0.1151,
+      "step": 1130
+    },
+    {
+      "epoch": 3.6505636070853464,
+      "grad_norm": 1.1281657218933105,
+      "learning_rate": 5.622729704857419e-06,
+      "loss": 0.1415,
+      "step": 1135
+    },
+    {
+      "epoch": 3.6666666666666665,
+      "grad_norm": 0.8988010883331299,
+      "learning_rate": 5.4987519819601e-06,
+      "loss": 0.0987,
+      "step": 1140
+    },
+    {
+      "epoch": 3.682769726247987,
+      "grad_norm": 1.0092145204544067,
+      "learning_rate": 5.37584888006724e-06,
+      "loss": 0.1018,
+      "step": 1145
+    },
+    {
+      "epoch": 3.6988727858293076,
+      "grad_norm": 0.853235125541687,
+      "learning_rate": 5.254034299905825e-06,
+      "loss": 0.1116,
+      "step": 1150
+    },
+    {
+      "epoch": 3.714975845410628,
+      "grad_norm": 1.030652642250061,
+      "learning_rate": 5.1333220190876366e-06,
+      "loss": 0.1178,
+      "step": 1155
+    },
+    {
+      "epoch": 3.7310789049919486,
+      "grad_norm": 1.0030436515808105,
+      "learning_rate": 5.0137256905509125e-06,
+      "loss": 0.0959,
+      "step": 1160
+    },
+    {
+      "epoch": 3.7471819645732687,
+      "grad_norm": 1.0387457609176636,
+      "learning_rate": 4.895258841016193e-06,
+      "loss": 0.1171,
+      "step": 1165
+    },
+    {
+      "epoch": 3.763285024154589,
+      "grad_norm": 0.9441067576408386,
+      "learning_rate": 4.777934869456392e-06,
+      "loss": 0.0964,
+      "step": 1170
+    },
+    {
+      "epoch": 3.7793880837359097,
+      "grad_norm": 1.0259944200515747,
+      "learning_rate": 4.661767045581333e-06,
+      "loss": 0.1132,
+      "step": 1175
+    },
+    {
+      "epoch": 3.7954911433172303,
+      "grad_norm": 0.8966212272644043,
+      "learning_rate": 4.546768508336897e-06,
+      "loss": 0.1011,
+      "step": 1180
+    },
+    {
+      "epoch": 3.8115942028985508,
+      "grad_norm": 0.9555193185806274,
+      "learning_rate": 4.432952264418995e-06,
+      "loss": 0.1013,
+      "step": 1185
+    },
+    {
+      "epoch": 3.8276972624798713,
+      "grad_norm": 1.0005531311035156,
+      "learning_rate": 4.320331186802423e-06,
+      "loss": 0.1013,
+      "step": 1190
+    },
+    {
+      "epoch": 3.843800322061192,
+      "grad_norm": 0.860589325428009,
+      "learning_rate": 4.208918013284927e-06,
+      "loss": 0.1059,
+      "step": 1195
+    },
+    {
+      "epoch": 3.8599033816425123,
+      "grad_norm": 1.0399781465530396,
+      "learning_rate": 4.098725345046495e-06,
+      "loss": 0.0976,
+      "step": 1200
+    },
+    {
+      "epoch": 3.8760064412238324,
+      "grad_norm": 0.7752050161361694,
+      "learning_rate": 3.9897656452241445e-06,
+      "loss": 0.1039,
+      "step": 1205
+    },
+    {
+      "epoch": 3.892109500805153,
+      "grad_norm": 0.708637535572052,
+      "learning_rate": 3.8820512375022834e-06,
+      "loss": 0.1156,
+      "step": 1210
+    },
+    {
+      "epoch": 3.9082125603864735,
+      "grad_norm": 0.9214046597480774,
+      "learning_rate": 3.7755943047188697e-06,
+      "loss": 0.1057,
+      "step": 1215
+    },
+    {
+      "epoch": 3.924315619967794,
+      "grad_norm": 0.935538649559021,
+      "learning_rate": 3.6704068874874917e-06,
+      "loss": 0.1086,
+      "step": 1220
+    },
+    {
+      "epoch": 3.940418679549114,
+      "grad_norm": 0.9492272138595581,
+      "learning_rate": 3.5665008828355566e-06,
+      "loss": 0.0996,
+      "step": 1225
+    },
+    {
+      "epoch": 3.9565217391304346,
+      "grad_norm": 0.919436514377594,
+      "learning_rate": 3.463888042858664e-06,
+      "loss": 0.1099,
+      "step": 1230
+    },
+    {
+      "epoch": 3.972624798711755,
+      "grad_norm": 0.9560986161231995,
+      "learning_rate": 3.3625799733914286e-06,
+      "loss": 0.0967,
+      "step": 1235
+    },
+    {
+      "epoch": 3.9887278582930756,
+      "grad_norm": 0.873535692691803,
+      "learning_rate": 3.2625881326948326e-06,
+      "loss": 0.1053,
+      "step": 1240
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1555,
+  "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": 1.7675301903904276e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

28_128_e5_3e-5/checkpoint-1244/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3839c69e7206b6c68ac17087343cb6c4560ba7b2b5157d7f3e1c366218d509a5
+size 7736

28_128_e5_3e-5/checkpoint-1244/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)

28_128_e5_3e-5/checkpoint-1555/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

28_128_e5_3e-5/checkpoint-1555/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",
+    "q_proj",
+    "v_proj",
+    "o_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

28_128_e5_3e-5/checkpoint-1555/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:330f02590ca9fdba2b41785cec6752f1e0b3abdd28470623cdc6a20b5038e515
+size 791751704

28_128_e5_3e-5/checkpoint-1555/latest

@@ -0,0 +1 @@
+global_step1555

28_128_e5_3e-5/checkpoint-1555/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78b6c38d2a7d80c306d3853cea5137aa0de7fb2561cb900f2a232bc558ad2953
+size 15920

28_128_e5_3e-5/checkpoint-1555/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c182df63041fc4499936561451ad035d9fa927f761429d52c85c0275b79e21a3
+size 15920

28_128_e5_3e-5/checkpoint-1555/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:077fff772200ce2a5a6db6d4d31d4544fd762623108a0f6d86cc5ab304a8eaa8
+size 15920

28_128_e5_3e-5/checkpoint-1555/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc39e5eb8191eea239b95451b942a0ab8ab143d176bb9ee97b3db30f256ff352
+size 15920

28_128_e5_3e-5/checkpoint-1555/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:862025bc4f56e062045210b41e625041584311dcf56403ea27beaf74bcaa3cd1
+size 15920

28_128_e5_3e-5/checkpoint-1555/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dab26b1d7b944d94f7d2550bf1bf149f044bf839ce28d44ee396bc3830b00dae
+size 15920

28_128_e5_3e-5/checkpoint-1555/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8e6618668df645054f89621b726871013e139f63c2378a7431d316f5b70a993
+size 15920

28_128_e5_3e-5/checkpoint-1555/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d12c852bb52b8340e2c93252865e31d870dad8716b3b097776c3ed2303dc94cb
+size 15920

28_128_e5_3e-5/checkpoint-1555/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33a30a98dd713f2e0a77c86d356ec450d8dcb6ad88a94d857bdd49a065c8b2a5
+size 1064

28_128_e5_3e-5/checkpoint-1555/special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "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": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

28_128_e5_3e-5/checkpoint-1555/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": "<|endoftext|>",
+  "padding_side": "left",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

28_128_e5_3e-5/checkpoint-1555/trainer_state.json

@@ -0,0 +1,2211 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.0,
+  "eval_steps": 500,
+  "global_step": 1555,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01610305958132045,
+      "grad_norm": 1.1391489505767822,
+      "learning_rate": 1.5384615384615385e-06,
+      "loss": 1.2885,
+      "step": 5
+    },
+    {
+      "epoch": 0.0322061191626409,
+      "grad_norm": 0.9224602580070496,
+      "learning_rate": 3.4615384615384617e-06,
+      "loss": 1.3364,
+      "step": 10
+    },
+    {
+      "epoch": 0.04830917874396135,
+      "grad_norm": 0.6716117262840271,
+      "learning_rate": 5.384615384615385e-06,
+      "loss": 1.2729,
+      "step": 15
+    },
+    {
+      "epoch": 0.0644122383252818,
+      "grad_norm": 0.6429063081741333,
+      "learning_rate": 7.307692307692308e-06,
+      "loss": 1.2407,
+      "step": 20
+    },
+    {
+      "epoch": 0.08051529790660225,
+      "grad_norm": 0.8855071663856506,
+      "learning_rate": 9.230769230769232e-06,
+      "loss": 1.2438,
+      "step": 25
+    },
+    {
+      "epoch": 0.0966183574879227,
+      "grad_norm": 0.5981923937797546,
+      "learning_rate": 1.1153846153846154e-05,
+      "loss": 1.2773,
+      "step": 30
+    },
+    {
+      "epoch": 0.11272141706924316,
+      "grad_norm": 0.5501493215560913,
+      "learning_rate": 1.3076923076923078e-05,
+      "loss": 1.2786,
+      "step": 35
+    },
+    {
+      "epoch": 0.1288244766505636,
+      "grad_norm": 0.5783489942550659,
+      "learning_rate": 1.5e-05,
+      "loss": 1.1728,
+      "step": 40
+    },
+    {
+      "epoch": 0.14492753623188406,
+      "grad_norm": 0.492718368768692,
+      "learning_rate": 1.6923076923076924e-05,
+      "loss": 1.1959,
+      "step": 45
+    },
+    {
+      "epoch": 0.1610305958132045,
+      "grad_norm": 0.5287221670150757,
+      "learning_rate": 1.8846153846153846e-05,
+      "loss": 1.1704,
+      "step": 50
+    },
+    {
+      "epoch": 0.17713365539452497,
+      "grad_norm": 0.5613968968391418,
+      "learning_rate": 2.076923076923077e-05,
+      "loss": 1.1734,
+      "step": 55
+    },
+    {
+      "epoch": 0.1932367149758454,
+      "grad_norm": 0.46896591782569885,
+      "learning_rate": 2.269230769230769e-05,
+      "loss": 1.1008,
+      "step": 60
+    },
+    {
+      "epoch": 0.20933977455716588,
+      "grad_norm": 0.6196975111961365,
+      "learning_rate": 2.4615384615384616e-05,
+      "loss": 1.1522,
+      "step": 65
+    },
+    {
+      "epoch": 0.22544283413848631,
+      "grad_norm": 0.4796932339668274,
+      "learning_rate": 2.6538461538461538e-05,
+      "loss": 1.133,
+      "step": 70
+    },
+    {
+      "epoch": 0.24154589371980675,
+      "grad_norm": 0.5374883413314819,
+      "learning_rate": 2.846153846153846e-05,
+      "loss": 1.1675,
+      "step": 75
+    },
+    {
+      "epoch": 0.2576489533011272,
+      "grad_norm": 0.4975369870662689,
+      "learning_rate": 2.999996606875036e-05,
+      "loss": 1.1127,
+      "step": 80
+    },
+    {
+      "epoch": 0.27375201288244766,
+      "grad_norm": 0.48042240738868713,
+      "learning_rate": 2.9998778491131415e-05,
+      "loss": 1.0703,
+      "step": 85
+    },
+    {
+      "epoch": 0.2898550724637681,
+      "grad_norm": 0.6405603885650635,
+      "learning_rate": 2.9995894504537867e-05,
+      "loss": 1.0789,
+      "step": 90
+    },
+    {
+      "epoch": 0.3059581320450886,
+      "grad_norm": 0.585081934928894,
+      "learning_rate": 2.999131443515766e-05,
+      "loss": 1.0105,
+      "step": 95
+    },
+    {
+      "epoch": 0.322061191626409,
+      "grad_norm": 0.7342278957366943,
+      "learning_rate": 2.998503880101102e-05,
+      "loss": 1.0567,
+      "step": 100
+    },
+    {
+      "epoch": 0.33816425120772947,
+      "grad_norm": 0.6049452424049377,
+      "learning_rate": 2.99770683118919e-05,
+      "loss": 1.0566,
+      "step": 105
+    },
+    {
+      "epoch": 0.35426731078904994,
+      "grad_norm": 0.5845102667808533,
+      "learning_rate": 2.996740386928766e-05,
+      "loss": 1.0511,
+      "step": 110
+    },
+    {
+      "epoch": 0.37037037037037035,
+      "grad_norm": 0.6237640976905823,
+      "learning_rate": 2.9956046566277126e-05,
+      "loss": 0.9973,
+      "step": 115
+    },
+    {
+      "epoch": 0.3864734299516908,
+      "grad_norm": 0.6733880043029785,
+      "learning_rate": 2.994299768740695e-05,
+      "loss": 0.9932,
+      "step": 120
+    },
+    {
+      "epoch": 0.4025764895330113,
+      "grad_norm": 0.668197751045227,
+      "learning_rate": 2.9928258708546335e-05,
+      "loss": 1.0146,
+      "step": 125
+    },
+    {
+      "epoch": 0.41867954911433175,
+      "grad_norm": 0.6295420527458191,
+      "learning_rate": 2.99118312967201e-05,
+      "loss": 0.9231,
+      "step": 130
+    },
+    {
+      "epoch": 0.43478260869565216,
+      "grad_norm": 0.624722421169281,
+      "learning_rate": 2.9893717309920134e-05,
+      "loss": 0.9942,
+      "step": 135
+    },
+    {
+      "epoch": 0.45088566827697263,
+      "grad_norm": 0.9356432557106018,
+      "learning_rate": 2.9873918796895273e-05,
+      "loss": 0.9984,
+      "step": 140
+    },
+    {
+      "epoch": 0.4669887278582931,
+      "grad_norm": 0.7457837462425232,
+      "learning_rate": 2.9852437996919537e-05,
+      "loss": 0.9668,
+      "step": 145
+    },
+    {
+      "epoch": 0.4830917874396135,
+      "grad_norm": 0.8209810853004456,
+      "learning_rate": 2.9829277339538903e-05,
+      "loss": 0.9771,
+      "step": 150
+    },
+    {
+      "epoch": 0.499194847020934,
+      "grad_norm": 0.6697631478309631,
+      "learning_rate": 2.9804439444296495e-05,
+      "loss": 0.9911,
+      "step": 155
+    },
+    {
+      "epoch": 0.5152979066022544,
+      "grad_norm": 0.6588212847709656,
+      "learning_rate": 2.9777927120436293e-05,
+      "loss": 0.9403,
+      "step": 160
+    },
+    {
+      "epoch": 0.5314009661835749,
+      "grad_norm": 0.7619159817695618,
+      "learning_rate": 2.974974336658545e-05,
+      "loss": 0.9147,
+      "step": 165
+    },
+    {
+      "epoch": 0.5475040257648953,
+      "grad_norm": 0.8137719631195068,
+      "learning_rate": 2.9719891370415072e-05,
+      "loss": 0.9449,
+      "step": 170
+    },
+    {
+      "epoch": 0.5636070853462157,
+      "grad_norm": 0.7924593091011047,
+      "learning_rate": 2.9688374508279715e-05,
+      "loss": 0.8465,
+      "step": 175
+    },
+    {
+      "epoch": 0.5797101449275363,
+      "grad_norm": 0.8301030993461609,
+      "learning_rate": 2.9655196344835528e-05,
+      "loss": 0.9196,
+      "step": 180
+    },
+    {
+      "epoch": 0.5958132045088567,
+      "grad_norm": 0.7267712950706482,
+      "learning_rate": 2.962036063263704e-05,
+      "loss": 0.846,
+      "step": 185
+    },
+    {
+      "epoch": 0.6119162640901772,
+      "grad_norm": 0.7453832030296326,
+      "learning_rate": 2.9583871311712753e-05,
+      "loss": 0.8713,
+      "step": 190
+    },
+    {
+      "epoch": 0.6280193236714976,
+      "grad_norm": 0.8772758841514587,
+      "learning_rate": 2.9545732509119526e-05,
+      "loss": 0.863,
+      "step": 195
+    },
+    {
+      "epoch": 0.644122383252818,
+      "grad_norm": 0.7424098253250122,
+      "learning_rate": 2.9505948538475754e-05,
+      "loss": 0.8345,
+      "step": 200
+    },
+    {
+      "epoch": 0.6602254428341385,
+      "grad_norm": 0.8788964748382568,
+      "learning_rate": 2.946452389947353e-05,
+      "loss": 0.8497,
+      "step": 205
+    },
+    {
+      "epoch": 0.6763285024154589,
+      "grad_norm": 1.0086477994918823,
+      "learning_rate": 2.9421463277369678e-05,
+      "loss": 0.7906,
+      "step": 210
+    },
+    {
+      "epoch": 0.6924315619967794,
+      "grad_norm": 0.9472551941871643,
+      "learning_rate": 2.9376771542455868e-05,
+      "loss": 0.8153,
+      "step": 215
+    },
+    {
+      "epoch": 0.7085346215780999,
+      "grad_norm": 0.8603758215904236,
+      "learning_rate": 2.9330453749507733e-05,
+      "loss": 0.7529,
+      "step": 220
+    },
+    {
+      "epoch": 0.7246376811594203,
+      "grad_norm": 0.8331507444381714,
+      "learning_rate": 2.9282515137213192e-05,
+      "loss": 0.7793,
+      "step": 225
+    },
+    {
+      "epoch": 0.7407407407407407,
+      "grad_norm": 0.8622682690620422,
+      "learning_rate": 2.9232961127579922e-05,
+      "loss": 0.7718,
+      "step": 230
+    },
+    {
+      "epoch": 0.7568438003220612,
+      "grad_norm": 0.9470008611679077,
+      "learning_rate": 2.9181797325322123e-05,
+      "loss": 0.7664,
+      "step": 235
+    },
+    {
+      "epoch": 0.7729468599033816,
+      "grad_norm": 0.9765321016311646,
+      "learning_rate": 2.912902951722658e-05,
+      "loss": 0.785,
+      "step": 240
+    },
+    {
+      "epoch": 0.789049919484702,
+      "grad_norm": 0.976305365562439,
+      "learning_rate": 2.907466367149819e-05,
+      "loss": 0.7656,
+      "step": 245
+    },
+    {
+      "epoch": 0.8051529790660226,
+      "grad_norm": 1.0335639715194702,
+      "learning_rate": 2.901870593708493e-05,
+      "loss": 0.7198,
+      "step": 250
+    },
+    {
+      "epoch": 0.821256038647343,
+      "grad_norm": 0.9155466556549072,
+      "learning_rate": 2.896116264298239e-05,
+      "loss": 0.7483,
+      "step": 255
+    },
+    {
+      "epoch": 0.8373590982286635,
+      "grad_norm": 0.8999483585357666,
+      "learning_rate": 2.890204029751793e-05,
+      "loss": 0.7637,
+      "step": 260
+    },
+    {
+      "epoch": 0.8534621578099839,
+      "grad_norm": 0.9181938171386719,
+      "learning_rate": 2.8841345587614587e-05,
+      "loss": 0.7486,
+      "step": 265
+    },
+    {
+      "epoch": 0.8695652173913043,
+      "grad_norm": 1.064903974533081,
+      "learning_rate": 2.8779085378034768e-05,
+      "loss": 0.7434,
+      "step": 270
+    },
+    {
+      "epoch": 0.8856682769726248,
+      "grad_norm": 0.9250743389129639,
+      "learning_rate": 2.87152667106038e-05,
+      "loss": 0.737,
+      "step": 275
+    },
+    {
+      "epoch": 0.9017713365539453,
+      "grad_norm": 0.9378864169120789,
+      "learning_rate": 2.864989680341348e-05,
+      "loss": 0.7047,
+      "step": 280
+    },
+    {
+      "epoch": 0.9178743961352657,
+      "grad_norm": 0.8972018957138062,
+      "learning_rate": 2.858298305000573e-05,
+      "loss": 0.6843,
+      "step": 285
+    },
+    {
+      "epoch": 0.9339774557165862,
+      "grad_norm": 0.9521335959434509,
+      "learning_rate": 2.8514533018536286e-05,
+      "loss": 0.7043,
+      "step": 290
+    },
+    {
+      "epoch": 0.9500805152979066,
+      "grad_norm": 1.0538114309310913,
+      "learning_rate": 2.84445544509188e-05,
+      "loss": 0.6254,
+      "step": 295
+    },
+    {
+      "epoch": 0.966183574879227,
+      "grad_norm": 0.8278699517250061,
+      "learning_rate": 2.8373055261949138e-05,
+      "loss": 0.6942,
+      "step": 300
+    },
+    {
+      "epoch": 0.9822866344605475,
+      "grad_norm": 0.9784626960754395,
+      "learning_rate": 2.830004353841023e-05,
+      "loss": 0.6504,
+      "step": 305
+    },
+    {
+      "epoch": 0.998389694041868,
+      "grad_norm": 1.0742645263671875,
+      "learning_rate": 2.8225527538157413e-05,
+      "loss": 0.6466,
+      "step": 310
+    },
+    {
+      "epoch": 1.0128824476650564,
+      "grad_norm": 1.3430242538452148,
+      "learning_rate": 2.814951568918446e-05,
+      "loss": 0.6389,
+      "step": 315
+    },
+    {
+      "epoch": 1.0289855072463767,
+      "grad_norm": 1.1638330221176147,
+      "learning_rate": 2.8072016588670316e-05,
+      "loss": 0.5618,
+      "step": 320
+    },
+    {
+      "epoch": 1.0450885668276972,
+      "grad_norm": 1.0941189527511597,
+      "learning_rate": 2.799303900200676e-05,
+      "loss": 0.5797,
+      "step": 325
+    },
+    {
+      "epoch": 1.0611916264090178,
+      "grad_norm": 0.9851027727127075,
+      "learning_rate": 2.7912591861806994e-05,
+      "loss": 0.5489,
+      "step": 330
+    },
+    {
+      "epoch": 1.077294685990338,
+      "grad_norm": 1.0907893180847168,
+      "learning_rate": 2.7830684266895352e-05,
+      "loss": 0.5627,
+      "step": 335
+    },
+    {
+      "epoch": 1.0933977455716586,
+      "grad_norm": 1.039016604423523,
+      "learning_rate": 2.774732548127817e-05,
+      "loss": 0.5682,
+      "step": 340
+    },
+    {
+      "epoch": 1.109500805152979,
+      "grad_norm": 1.169583797454834,
+      "learning_rate": 2.766252493309603e-05,
+      "loss": 0.5668,
+      "step": 345
+    },
+    {
+      "epoch": 1.1256038647342996,
+      "grad_norm": 1.1832174062728882,
+      "learning_rate": 2.7576292213557372e-05,
+      "loss": 0.5808,
+      "step": 350
+    },
+    {
+      "epoch": 1.14170692431562,
+      "grad_norm": 1.1079707145690918,
+      "learning_rate": 2.748863707585373e-05,
+      "loss": 0.5511,
+      "step": 355
+    },
+    {
+      "epoch": 1.1578099838969405,
+      "grad_norm": 1.1411006450653076,
+      "learning_rate": 2.7399569434056595e-05,
+      "loss": 0.5355,
+      "step": 360
+    },
+    {
+      "epoch": 1.1739130434782608,
+      "grad_norm": 1.250238299369812,
+      "learning_rate": 2.7309099361996115e-05,
+      "loss": 0.5373,
+      "step": 365
+    },
+    {
+      "epoch": 1.1900161030595813,
+      "grad_norm": 1.157636046409607,
+      "learning_rate": 2.7217237092121703e-05,
+      "loss": 0.539,
+      "step": 370
+    },
+    {
+      "epoch": 1.2061191626409018,
+      "grad_norm": 1.102122187614441,
+      "learning_rate": 2.7123993014344714e-05,
+      "loss": 0.5551,
+      "step": 375
+    },
+    {
+      "epoch": 1.2222222222222223,
+      "grad_norm": 1.134345293045044,
+      "learning_rate": 2.7029377674863332e-05,
+      "loss": 0.5424,
+      "step": 380
+    },
+    {
+      "epoch": 1.2383252818035426,
+      "grad_norm": 1.1444910764694214,
+      "learning_rate": 2.6933401774969734e-05,
+      "loss": 0.5118,
+      "step": 385
+    },
+    {
+      "epoch": 1.2544283413848631,
+      "grad_norm": 1.1335145235061646,
+      "learning_rate": 2.6836076169839765e-05,
+      "loss": 0.5002,
+      "step": 390
+    },
+    {
+      "epoch": 1.2705314009661834,
+      "grad_norm": 1.1598331928253174,
+      "learning_rate": 2.673741186730517e-05,
+      "loss": 0.4804,
+      "step": 395
+    },
+    {
+      "epoch": 1.286634460547504,
+      "grad_norm": 1.0515048503875732,
+      "learning_rate": 2.6637420026608573e-05,
+      "loss": 0.5247,
+      "step": 400
+    },
+    {
+      "epoch": 1.3027375201288245,
+      "grad_norm": 0.957242488861084,
+      "learning_rate": 2.653611195714134e-05,
+      "loss": 0.453,
+      "step": 405
+    },
+    {
+      "epoch": 1.318840579710145,
+      "grad_norm": 1.0974361896514893,
+      "learning_rate": 2.6433499117164444e-05,
+      "loss": 0.5344,
+      "step": 410
+    },
+    {
+      "epoch": 1.3349436392914653,
+      "grad_norm": 0.9843302965164185,
+      "learning_rate": 2.6329593112512508e-05,
+      "loss": 0.5074,
+      "step": 415
+    },
+    {
+      "epoch": 1.3510466988727858,
+      "grad_norm": 1.0893441438674927,
+      "learning_rate": 2.6224405695281136e-05,
+      "loss": 0.4771,
+      "step": 420
+    },
+    {
+      "epoch": 1.3671497584541064,
+      "grad_norm": 1.155216097831726,
+      "learning_rate": 2.611794876249772e-05,
+      "loss": 0.5031,
+      "step": 425
+    },
+    {
+      "epoch": 1.3832528180354267,
+      "grad_norm": 0.9953705072402954,
+      "learning_rate": 2.601023435477586e-05,
+      "loss": 0.491,
+      "step": 430
+    },
+    {
+      "epoch": 1.3993558776167472,
+      "grad_norm": 1.152533769607544,
+      "learning_rate": 2.5901274654953506e-05,
+      "loss": 0.4655,
+      "step": 435
+    },
+    {
+      "epoch": 1.4154589371980677,
+      "grad_norm": 1.1234312057495117,
+      "learning_rate": 2.5791081986715076e-05,
+      "loss": 0.4415,
+      "step": 440
+    },
+    {
+      "epoch": 1.431561996779388,
+      "grad_norm": 1.0570783615112305,
+      "learning_rate": 2.5679668813197583e-05,
+      "loss": 0.4547,
+      "step": 445
+    },
+    {
+      "epoch": 1.4476650563607085,
+      "grad_norm": 1.0669089555740356,
+      "learning_rate": 2.556704773558101e-05,
+      "loss": 0.4658,
+      "step": 450
+    },
+    {
+      "epoch": 1.463768115942029,
+      "grad_norm": 1.142380714416504,
+      "learning_rate": 2.5453231491663104e-05,
+      "loss": 0.4596,
+      "step": 455
+    },
+    {
+      "epoch": 1.4798711755233493,
+      "grad_norm": 1.0438144207000732,
+      "learning_rate": 2.5338232954418675e-05,
+      "loss": 0.4214,
+      "step": 460
+    },
+    {
+      "epoch": 1.4959742351046699,
+      "grad_norm": 1.1825107336044312,
+      "learning_rate": 2.522206513054361e-05,
+      "loss": 0.4783,
+      "step": 465
+    },
+    {
+      "epoch": 1.5120772946859904,
+      "grad_norm": 1.420599102973938,
+      "learning_rate": 2.5104741158983814e-05,
+      "loss": 0.4716,
+      "step": 470
+    },
+    {
+      "epoch": 1.528180354267311,
+      "grad_norm": 1.1626514196395874,
+      "learning_rate": 2.4986274309449088e-05,
+      "loss": 0.4689,
+      "step": 475
+    },
+    {
+      "epoch": 1.5442834138486312,
+      "grad_norm": 1.133529543876648,
+      "learning_rate": 2.4866677980912363e-05,
+      "loss": 0.4341,
+      "step": 480
+    },
+    {
+      "epoch": 1.5603864734299517,
+      "grad_norm": 1.196297526359558,
+      "learning_rate": 2.4745965700094172e-05,
+      "loss": 0.5102,
+      "step": 485
+    },
+    {
+      "epoch": 1.576489533011272,
+      "grad_norm": 1.0283925533294678,
+      "learning_rate": 2.462415111993276e-05,
+      "loss": 0.4533,
+      "step": 490
+    },
+    {
+      "epoch": 1.5925925925925926,
+      "grad_norm": 1.2448772192001343,
+      "learning_rate": 2.4501248018039897e-05,
+      "loss": 0.4215,
+      "step": 495
+    },
+    {
+      "epoch": 1.608695652173913,
+      "grad_norm": 1.2707397937774658,
+      "learning_rate": 2.4377270295142582e-05,
+      "loss": 0.4395,
+      "step": 500
+    },
+    {
+      "epoch": 1.6247987117552336,
+      "grad_norm": 1.6315351724624634,
+      "learning_rate": 2.425223197351081e-05,
+      "loss": 0.427,
+      "step": 505
+    },
+    {
+      "epoch": 1.640901771336554,
+      "grad_norm": 1.1135799884796143,
+      "learning_rate": 2.4126147195371647e-05,
+      "loss": 0.3926,
+      "step": 510
+    },
+    {
+      "epoch": 1.6570048309178744,
+      "grad_norm": 1.2687816619873047,
+      "learning_rate": 2.3999030221309655e-05,
+      "loss": 0.3848,
+      "step": 515
+    },
+    {
+      "epoch": 1.6731078904991947,
+      "grad_norm": 1.1300817728042603,
+      "learning_rate": 2.3870895428654035e-05,
+      "loss": 0.3926,
+      "step": 520
+    },
+    {
+      "epoch": 1.6892109500805152,
+      "grad_norm": 1.067254662513733,
+      "learning_rate": 2.374175730985242e-05,
+      "loss": 0.3987,
+      "step": 525
+    },
+    {
+      "epoch": 1.7053140096618358,
+      "grad_norm": 1.2030500173568726,
+      "learning_rate": 2.3611630470831824e-05,
+      "loss": 0.3844,
+      "step": 530
+    },
+    {
+      "epoch": 1.7214170692431563,
+      "grad_norm": 1.1542716026306152,
+      "learning_rate": 2.3480529629346597e-05,
+      "loss": 0.383,
+      "step": 535
+    },
+    {
+      "epoch": 1.7375201288244766,
+      "grad_norm": 1.099084496498108,
+      "learning_rate": 2.3348469613313854e-05,
+      "loss": 0.422,
+      "step": 540
+    },
+    {
+      "epoch": 1.7536231884057971,
+      "grad_norm": 1.084632158279419,
+      "learning_rate": 2.3215465359136338e-05,
+      "loss": 0.3831,
+      "step": 545
+    },
+    {
+      "epoch": 1.7697262479871174,
+      "grad_norm": 1.0466201305389404,
+      "learning_rate": 2.3081531910013104e-05,
+      "loss": 0.4008,
+      "step": 550
+    },
+    {
+      "epoch": 1.785829307568438,
+      "grad_norm": 1.0963324308395386,
+      "learning_rate": 2.2946684414238093e-05,
+      "loss": 0.3731,
+      "step": 555
+    },
+    {
+      "epoch": 1.8019323671497585,
+      "grad_norm": 1.297102928161621,
+      "learning_rate": 2.281093812348676e-05,
+      "loss": 0.3656,
+      "step": 560
+    },
+    {
+      "epoch": 1.818035426731079,
+      "grad_norm": 1.0965611934661865,
+      "learning_rate": 2.2674308391091127e-05,
+      "loss": 0.3477,
+      "step": 565
+    },
+    {
+      "epoch": 1.8341384863123995,
+      "grad_norm": 1.2567371129989624,
+      "learning_rate": 2.2536810670303214e-05,
+      "loss": 0.3658,
+      "step": 570
+    },
+    {
+      "epoch": 1.8502415458937198,
+      "grad_norm": 1.1441166400909424,
+      "learning_rate": 2.239846051254728e-05,
+      "loss": 0.3489,
+      "step": 575
+    },
+    {
+      "epoch": 1.86634460547504,
+      "grad_norm": 1.2327871322631836,
+      "learning_rate": 2.2259273565660876e-05,
+      "loss": 0.3866,
+      "step": 580
+    },
+    {
+      "epoch": 1.8824476650563606,
+      "grad_norm": 1.0107234716415405,
+      "learning_rate": 2.2119265572125027e-05,
+      "loss": 0.3854,
+      "step": 585
+    },
+    {
+      "epoch": 1.8985507246376812,
+      "grad_norm": 1.259210228919983,
+      "learning_rate": 2.197845236728371e-05,
+      "loss": 0.3592,
+      "step": 590
+    },
+    {
+      "epoch": 1.9146537842190017,
+      "grad_norm": 1.232729196548462,
+      "learning_rate": 2.1836849877552854e-05,
+      "loss": 0.3441,
+      "step": 595
+    },
+    {
+      "epoch": 1.9307568438003222,
+      "grad_norm": 1.1457910537719727,
+      "learning_rate": 2.1694474118618954e-05,
+      "loss": 0.344,
+      "step": 600
+    },
+    {
+      "epoch": 1.9468599033816425,
+      "grad_norm": 1.2113591432571411,
+      "learning_rate": 2.1551341193627717e-05,
+      "loss": 0.3344,
+      "step": 605
+    },
+    {
+      "epoch": 1.9629629629629628,
+      "grad_norm": 1.2011780738830566,
+      "learning_rate": 2.1407467291362686e-05,
+      "loss": 0.3739,
+      "step": 610
+    },
+    {
+      "epoch": 1.9790660225442833,
+      "grad_norm": 1.178965449333191,
+      "learning_rate": 2.1262868684414265e-05,
+      "loss": 0.3731,
+      "step": 615
+    },
+    {
+      "epoch": 1.9951690821256038,
+      "grad_norm": 1.1474207639694214,
+      "learning_rate": 2.111756172733925e-05,
+      "loss": 0.366,
+      "step": 620
+    },
+    {
+      "epoch": 2.0096618357487923,
+      "grad_norm": 1.1576290130615234,
+      "learning_rate": 2.0971562854811044e-05,
+      "loss": 0.2854,
+      "step": 625
+    },
+    {
+      "epoch": 2.025764895330113,
+      "grad_norm": 1.0710585117340088,
+      "learning_rate": 2.0824888579760874e-05,
+      "loss": 0.2722,
+      "step": 630
+    },
+    {
+      "epoch": 2.0418679549114334,
+      "grad_norm": 1.2322947978973389,
+      "learning_rate": 2.06775554915101e-05,
+      "loss": 0.2559,
+      "step": 635
+    },
+    {
+      "epoch": 2.0579710144927534,
+      "grad_norm": 1.1783329248428345,
+      "learning_rate": 2.0529580253893944e-05,
+      "loss": 0.2794,
+      "step": 640
+    },
+    {
+      "epoch": 2.074074074074074,
+      "grad_norm": 1.219993233680725,
+      "learning_rate": 2.0380979603376726e-05,
+      "loss": 0.2706,
+      "step": 645
+    },
+    {
+      "epoch": 2.0901771336553945,
+      "grad_norm": 1.0898008346557617,
+      "learning_rate": 2.0231770347158933e-05,
+      "loss": 0.2613,
+      "step": 650
+    },
+    {
+      "epoch": 2.106280193236715,
+      "grad_norm": 1.127703070640564,
+      "learning_rate": 2.0081969361276276e-05,
+      "loss": 0.2319,
+      "step": 655
+    },
+    {
+      "epoch": 2.1223832528180355,
+      "grad_norm": 1.1997917890548706,
+      "learning_rate": 1.9931593588690932e-05,
+      "loss": 0.2639,
+      "step": 660
+    },
+    {
+      "epoch": 2.138486312399356,
+      "grad_norm": 1.1341209411621094,
+      "learning_rate": 1.9780660037375284e-05,
+      "loss": 0.2476,
+      "step": 665
+    },
+    {
+      "epoch": 2.154589371980676,
+      "grad_norm": 1.404829502105713,
+      "learning_rate": 1.9629185778388215e-05,
+      "loss": 0.2549,
+      "step": 670
+    },
+    {
+      "epoch": 2.1706924315619966,
+      "grad_norm": 1.1979633569717407,
+      "learning_rate": 1.947718794394438e-05,
+      "loss": 0.2537,
+      "step": 675
+    },
+    {
+      "epoch": 2.186795491143317,
+      "grad_norm": 1.1800684928894043,
+      "learning_rate": 1.932468372547643e-05,
+      "loss": 0.2771,
+      "step": 680
+    },
+    {
+      "epoch": 2.2028985507246377,
+      "grad_norm": 1.140649437904358,
+      "learning_rate": 1.9171690371690655e-05,
+      "loss": 0.2173,
+      "step": 685
+    },
+    {
+      "epoch": 2.219001610305958,
+      "grad_norm": 1.2186757326126099,
+      "learning_rate": 1.9018225186616085e-05,
+      "loss": 0.2474,
+      "step": 690
+    },
+    {
+      "epoch": 2.2351046698872787,
+      "grad_norm": 1.1380257606506348,
+      "learning_rate": 1.886430552764735e-05,
+      "loss": 0.2718,
+      "step": 695
+    },
+    {
+      "epoch": 2.2512077294685993,
+      "grad_norm": 1.2019944190979004,
+      "learning_rate": 1.870994880358151e-05,
+      "loss": 0.2524,
+      "step": 700
+    },
+    {
+      "epoch": 2.2673107890499193,
+      "grad_norm": 1.1367535591125488,
+      "learning_rate": 1.8555172472649048e-05,
+      "loss": 0.2537,
+      "step": 705
+    },
+    {
+      "epoch": 2.28341384863124,
+      "grad_norm": 1.3347067832946777,
+      "learning_rate": 1.8399994040539322e-05,
+      "loss": 0.2532,
+      "step": 710
+    },
+    {
+      "epoch": 2.2995169082125604,
+      "grad_norm": 1.1048965454101562,
+      "learning_rate": 1.8244431058420576e-05,
+      "loss": 0.243,
+      "step": 715
+    },
+    {
+      "epoch": 2.315619967793881,
+      "grad_norm": 1.1416301727294922,
+      "learning_rate": 1.808850112095489e-05,
+      "loss": 0.2702,
+      "step": 720
+    },
+    {
+      "epoch": 2.3317230273752014,
+      "grad_norm": 1.139872431755066,
+      "learning_rate": 1.7932221864308118e-05,
+      "loss": 0.2278,
+      "step": 725
+    },
+    {
+      "epoch": 2.3478260869565215,
+      "grad_norm": 1.0932283401489258,
+      "learning_rate": 1.7775610964155246e-05,
+      "loss": 0.2578,
+      "step": 730
+    },
+    {
+      "epoch": 2.363929146537842,
+      "grad_norm": 1.0193161964416504,
+      "learning_rate": 1.761868613368115e-05,
+      "loss": 0.2336,
+      "step": 735
+    },
+    {
+      "epoch": 2.3800322061191626,
+      "grad_norm": 1.3200891017913818,
+      "learning_rate": 1.7461465121577234e-05,
+      "loss": 0.2587,
+      "step": 740
+    },
+    {
+      "epoch": 2.396135265700483,
+      "grad_norm": 1.067944884300232,
+      "learning_rate": 1.7303965710033957e-05,
+      "loss": 0.2546,
+      "step": 745
+    },
+    {
+      "epoch": 2.4122383252818036,
+      "grad_norm": 1.0622363090515137,
+      "learning_rate": 1.714620571272965e-05,
+      "loss": 0.2567,
+      "step": 750
+    },
+    {
+      "epoch": 2.428341384863124,
+      "grad_norm": 1.0919643640518188,
+      "learning_rate": 1.69882029728157e-05,
+      "loss": 0.2284,
+      "step": 755
+    },
+    {
+      "epoch": 2.4444444444444446,
+      "grad_norm": 1.2129340171813965,
+      "learning_rate": 1.682997536089844e-05,
+      "loss": 0.2458,
+      "step": 760
+    },
+    {
+      "epoch": 2.4605475040257647,
+      "grad_norm": 1.169572114944458,
+      "learning_rate": 1.6671540773017952e-05,
+      "loss": 0.2036,
+      "step": 765
+    },
+    {
+      "epoch": 2.4766505636070852,
+      "grad_norm": 1.082538366317749,
+      "learning_rate": 1.6512917128623927e-05,
+      "loss": 0.2383,
+      "step": 770
+    },
+    {
+      "epoch": 2.4927536231884058,
+      "grad_norm": 1.3170980215072632,
+      "learning_rate": 1.6354122368548965e-05,
+      "loss": 0.2345,
+      "step": 775
+    },
+    {
+      "epoch": 2.5088566827697263,
+      "grad_norm": 1.1367696523666382,
+      "learning_rate": 1.6195174452979365e-05,
+      "loss": 0.204,
+      "step": 780
+    },
+    {
+      "epoch": 2.524959742351047,
+      "grad_norm": 1.016545295715332,
+      "learning_rate": 1.6036091359423807e-05,
+      "loss": 0.2246,
+      "step": 785
+    },
+    {
+      "epoch": 2.541062801932367,
+      "grad_norm": 1.1528297662734985,
+      "learning_rate": 1.587689108068002e-05,
+      "loss": 0.2489,
+      "step": 790
+    },
+    {
+      "epoch": 2.5571658615136874,
+      "grad_norm": 1.1874185800552368,
+      "learning_rate": 1.571759162279975e-05,
+      "loss": 0.2267,
+      "step": 795
+    },
+    {
+      "epoch": 2.573268921095008,
+      "grad_norm": 1.1973860263824463,
+      "learning_rate": 1.5558211003052222e-05,
+      "loss": 0.2096,
+      "step": 800
+    },
+    {
+      "epoch": 2.5893719806763285,
+      "grad_norm": 1.1747304201126099,
+      "learning_rate": 1.539876724788631e-05,
+      "loss": 0.2161,
+      "step": 805
+    },
+    {
+      "epoch": 2.605475040257649,
+      "grad_norm": 1.0052316188812256,
+      "learning_rate": 1.5239278390891703e-05,
+      "loss": 0.1939,
+      "step": 810
+    },
+    {
+      "epoch": 2.6215780998389695,
+      "grad_norm": 1.1736071109771729,
+      "learning_rate": 1.5079762470759255e-05,
+      "loss": 0.1796,
+      "step": 815
+    },
+    {
+      "epoch": 2.63768115942029,
+      "grad_norm": 1.1553905010223389,
+      "learning_rate": 1.492023752924075e-05,
+      "loss": 0.2206,
+      "step": 820
+    },
+    {
+      "epoch": 2.6537842190016105,
+      "grad_norm": 1.1105087995529175,
+      "learning_rate": 1.4760721609108302e-05,
+      "loss": 0.1833,
+      "step": 825
+    },
+    {
+      "epoch": 2.6698872785829306,
+      "grad_norm": 1.209580659866333,
+      "learning_rate": 1.4601232752113696e-05,
+      "loss": 0.1896,
+      "step": 830
+    },
+    {
+      "epoch": 2.685990338164251,
+      "grad_norm": 1.0659430027008057,
+      "learning_rate": 1.444178899694778e-05,
+      "loss": 0.2153,
+      "step": 835
+    },
+    {
+      "epoch": 2.7020933977455717,
+      "grad_norm": 1.2064878940582275,
+      "learning_rate": 1.428240837720025e-05,
+      "loss": 0.222,
+      "step": 840
+    },
+    {
+      "epoch": 2.718196457326892,
+      "grad_norm": 0.9531728029251099,
+      "learning_rate": 1.4123108919319986e-05,
+      "loss": 0.2248,
+      "step": 845
+    },
+    {
+      "epoch": 2.7342995169082127,
+      "grad_norm": 1.1249150037765503,
+      "learning_rate": 1.39639086405762e-05,
+      "loss": 0.2273,
+      "step": 850
+    },
+    {
+      "epoch": 2.750402576489533,
+      "grad_norm": 1.3298691511154175,
+      "learning_rate": 1.3804825547020644e-05,
+      "loss": 0.1782,
+      "step": 855
+    },
+    {
+      "epoch": 2.7665056360708533,
+      "grad_norm": 0.9931784868240356,
+      "learning_rate": 1.364587763145104e-05,
+      "loss": 0.1956,
+      "step": 860
+    },
+    {
+      "epoch": 2.782608695652174,
+      "grad_norm": 1.1622772216796875,
+      "learning_rate": 1.3487082871376075e-05,
+      "loss": 0.2083,
+      "step": 865
+    },
+    {
+      "epoch": 2.7987117552334944,
+      "grad_norm": 1.6023427248001099,
+      "learning_rate": 1.332845922698205e-05,
+      "loss": 0.1982,
+      "step": 870
+    },
+    {
+      "epoch": 2.814814814814815,
+      "grad_norm": 1.1077145338058472,
+      "learning_rate": 1.3170024639101561e-05,
+      "loss": 0.2135,
+      "step": 875
+    },
+    {
+      "epoch": 2.8309178743961354,
+      "grad_norm": 1.1962898969650269,
+      "learning_rate": 1.3011797027184303e-05,
+      "loss": 0.1855,
+      "step": 880
+    },
+    {
+      "epoch": 2.847020933977456,
+      "grad_norm": 0.966684877872467,
+      "learning_rate": 1.2853794287270352e-05,
+      "loss": 0.1766,
+      "step": 885
+    },
+    {
+      "epoch": 2.863123993558776,
+      "grad_norm": 1.170991063117981,
+      "learning_rate": 1.2696034289966044e-05,
+      "loss": 0.2242,
+      "step": 890
+    },
+    {
+      "epoch": 2.8792270531400965,
+      "grad_norm": 1.1262538433074951,
+      "learning_rate": 1.2538534878422772e-05,
+      "loss": 0.1887,
+      "step": 895
+    },
+    {
+      "epoch": 2.895330112721417,
+      "grad_norm": 1.3256655931472778,
+      "learning_rate": 1.2381313866318854e-05,
+      "loss": 0.1948,
+      "step": 900
+    },
+    {
+      "epoch": 2.9114331723027376,
+      "grad_norm": 1.0575200319290161,
+      "learning_rate": 1.2224389035844756e-05,
+      "loss": 0.1989,
+      "step": 905
+    },
+    {
+      "epoch": 2.927536231884058,
+      "grad_norm": 1.126052975654602,
+      "learning_rate": 1.2067778135691881e-05,
+      "loss": 0.1872,
+      "step": 910
+    },
+    {
+      "epoch": 2.943639291465378,
+      "grad_norm": 1.2957228422164917,
+      "learning_rate": 1.1911498879045113e-05,
+      "loss": 0.1808,
+      "step": 915
+    },
+    {
+      "epoch": 2.9597423510466987,
+      "grad_norm": 1.0871576070785522,
+      "learning_rate": 1.1755568941579424e-05,
+      "loss": 0.1736,
+      "step": 920
+    },
+    {
+      "epoch": 2.975845410628019,
+      "grad_norm": 1.102060317993164,
+      "learning_rate": 1.160000595946068e-05,
+      "loss": 0.1767,
+      "step": 925
+    },
+    {
+      "epoch": 2.9919484702093397,
+      "grad_norm": 0.9936461448669434,
+      "learning_rate": 1.1444827527350954e-05,
+      "loss": 0.1695,
+      "step": 930
+    },
+    {
+      "epoch": 3.006441223832528,
+      "grad_norm": 0.9595783948898315,
+      "learning_rate": 1.1290051196418494e-05,
+      "loss": 0.1457,
+      "step": 935
+    },
+    {
+      "epoch": 3.0225442834138487,
+      "grad_norm": 1.0653377771377563,
+      "learning_rate": 1.1135694472352652e-05,
+      "loss": 0.1399,
+      "step": 940
+    },
+    {
+      "epoch": 3.0386473429951693,
+      "grad_norm": 1.34004807472229,
+      "learning_rate": 1.0981774813383916e-05,
+      "loss": 0.1538,
+      "step": 945
+    },
+    {
+      "epoch": 3.0547504025764893,
+      "grad_norm": 1.1128737926483154,
+      "learning_rate": 1.0828309628309347e-05,
+      "loss": 0.1252,
+      "step": 950
+    },
+    {
+      "epoch": 3.07085346215781,
+      "grad_norm": 1.0216444730758667,
+      "learning_rate": 1.0675316274523571e-05,
+      "loss": 0.1356,
+      "step": 955
+    },
+    {
+      "epoch": 3.0869565217391304,
+      "grad_norm": 1.013954997062683,
+      "learning_rate": 1.0522812056055617e-05,
+      "loss": 0.1334,
+      "step": 960
+    },
+    {
+      "epoch": 3.103059581320451,
+      "grad_norm": 1.003402590751648,
+      "learning_rate": 1.037081422161178e-05,
+      "loss": 0.1263,
+      "step": 965
+    },
+    {
+      "epoch": 3.1191626409017714,
+      "grad_norm": 1.1828265190124512,
+      "learning_rate": 1.0219339962624717e-05,
+      "loss": 0.1254,
+      "step": 970
+    },
+    {
+      "epoch": 3.135265700483092,
+      "grad_norm": 1.0109069347381592,
+      "learning_rate": 1.0068406411309068e-05,
+      "loss": 0.1307,
+      "step": 975
+    },
+    {
+      "epoch": 3.151368760064412,
+      "grad_norm": 1.0399396419525146,
+      "learning_rate": 9.918030638723725e-06,
+      "loss": 0.1195,
+      "step": 980
+    },
+    {
+      "epoch": 3.1674718196457325,
+      "grad_norm": 0.9721539616584778,
+      "learning_rate": 9.768229652841066e-06,
+      "loss": 0.1347,
+      "step": 985
+    },
+    {
+      "epoch": 3.183574879227053,
+      "grad_norm": 0.9482682943344116,
+      "learning_rate": 9.61902039662328e-06,
+      "loss": 0.147,
+      "step": 990
+    },
+    {
+      "epoch": 3.1996779388083736,
+      "grad_norm": 0.9318935871124268,
+      "learning_rate": 9.47041974610606e-06,
+      "loss": 0.1341,
+      "step": 995
+    },
+    {
+      "epoch": 3.215780998389694,
+      "grad_norm": 1.1986130475997925,
+      "learning_rate": 9.322444508489905e-06,
+      "loss": 0.1155,
+      "step": 1000
+    },
+    {
+      "epoch": 3.2318840579710146,
+      "grad_norm": 1.0879420042037964,
+      "learning_rate": 9.175111420239133e-06,
+      "loss": 0.1399,
+      "step": 1005
+    },
+    {
+      "epoch": 3.247987117552335,
+      "grad_norm": 0.9248384833335876,
+      "learning_rate": 9.028437145188963e-06,
+      "loss": 0.1218,
+      "step": 1010
+    },
+    {
+      "epoch": 3.2640901771336552,
+      "grad_norm": 1.1074427366256714,
+      "learning_rate": 8.882438272660755e-06,
+      "loss": 0.154,
+      "step": 1015
+    },
+    {
+      "epoch": 3.2801932367149758,
+      "grad_norm": 1.03411865234375,
+      "learning_rate": 8.737131315585736e-06,
+      "loss": 0.1329,
+      "step": 1020
+    },
+    {
+      "epoch": 3.2962962962962963,
+      "grad_norm": 0.847226619720459,
+      "learning_rate": 8.592532708637319e-06,
+      "loss": 0.1385,
+      "step": 1025
+    },
+    {
+      "epoch": 3.312399355877617,
+      "grad_norm": 0.9666579961776733,
+      "learning_rate": 8.448658806372289e-06,
+      "loss": 0.1359,
+      "step": 1030
+    },
+    {
+      "epoch": 3.3285024154589373,
+      "grad_norm": 1.055481195449829,
+      "learning_rate": 8.305525881381045e-06,
+      "loss": 0.1445,
+      "step": 1035
+    },
+    {
+      "epoch": 3.3446054750402574,
+      "grad_norm": 1.088436484336853,
+      "learning_rate": 8.16315012244715e-06,
+      "loss": 0.1073,
+      "step": 1040
+    },
+    {
+      "epoch": 3.360708534621578,
+      "grad_norm": 0.939185380935669,
+      "learning_rate": 8.021547632716291e-06,
+      "loss": 0.1106,
+      "step": 1045
+    },
+    {
+      "epoch": 3.3768115942028984,
+      "grad_norm": 1.077433705329895,
+      "learning_rate": 7.88073442787498e-06,
+      "loss": 0.1247,
+      "step": 1050
+    },
+    {
+      "epoch": 3.392914653784219,
+      "grad_norm": 1.1379334926605225,
+      "learning_rate": 7.740726434339125e-06,
+      "loss": 0.1339,
+      "step": 1055
+    },
+    {
+      "epoch": 3.4090177133655395,
+      "grad_norm": 0.8844690322875977,
+      "learning_rate": 7.60153948745272e-06,
+      "loss": 0.1169,
+      "step": 1060
+    },
+    {
+      "epoch": 3.42512077294686,
+      "grad_norm": 1.0570695400238037,
+      "learning_rate": 7.463189329696789e-06,
+      "loss": 0.1228,
+      "step": 1065
+    },
+    {
+      "epoch": 3.4412238325281805,
+      "grad_norm": 1.0314443111419678,
+      "learning_rate": 7.3256916089088825e-06,
+      "loss": 0.1339,
+      "step": 1070
+    },
+    {
+      "epoch": 3.4573268921095006,
+      "grad_norm": 0.9575313329696655,
+      "learning_rate": 7.189061876513241e-06,
+      "loss": 0.1418,
+      "step": 1075
+    },
+    {
+      "epoch": 3.473429951690821,
+      "grad_norm": 0.9746488332748413,
+      "learning_rate": 7.053315585761911e-06,
+      "loss": 0.1268,
+      "step": 1080
+    },
+    {
+      "epoch": 3.4895330112721417,
+      "grad_norm": 1.0186628103256226,
+      "learning_rate": 6.9184680899868955e-06,
+      "loss": 0.1346,
+      "step": 1085
+    },
+    {
+      "epoch": 3.505636070853462,
+      "grad_norm": 1.020210862159729,
+      "learning_rate": 6.784534640863668e-06,
+      "loss": 0.1264,
+      "step": 1090
+    },
+    {
+      "epoch": 3.5217391304347827,
+      "grad_norm": 0.9518364071846008,
+      "learning_rate": 6.651530386686148e-06,
+      "loss": 0.1332,
+      "step": 1095
+    },
+    {
+      "epoch": 3.537842190016103,
+      "grad_norm": 1.0923478603363037,
+      "learning_rate": 6.519470370653402e-06,
+      "loss": 0.1183,
+      "step": 1100
+    },
+    {
+      "epoch": 3.5539452495974233,
+      "grad_norm": 1.100982666015625,
+      "learning_rate": 6.388369529168181e-06,
+      "loss": 0.1185,
+      "step": 1105
+    },
+    {
+      "epoch": 3.570048309178744,
+      "grad_norm": 0.9394665956497192,
+      "learning_rate": 6.258242690147579e-06,
+      "loss": 0.1018,
+      "step": 1110
+    },
+    {
+      "epoch": 3.5861513687600644,
+      "grad_norm": 0.7709954380989075,
+      "learning_rate": 6.129104571345969e-06,
+      "loss": 0.1156,
+      "step": 1115
+    },
+    {
+      "epoch": 3.602254428341385,
+      "grad_norm": 0.9713561534881592,
+      "learning_rate": 6.000969778690342e-06,
+      "loss": 0.1164,
+      "step": 1120
+    },
+    {
+      "epoch": 3.6183574879227054,
+      "grad_norm": 0.8961610198020935,
+      "learning_rate": 5.873852804628358e-06,
+      "loss": 0.1068,
+      "step": 1125
+    },
+    {
+      "epoch": 3.634460547504026,
+      "grad_norm": 0.9026066064834595,
+      "learning_rate": 5.747768026489188e-06,
+      "loss": 0.1151,
+      "step": 1130
+    },
+    {
+      "epoch": 3.6505636070853464,
+      "grad_norm": 1.1281657218933105,
+      "learning_rate": 5.622729704857419e-06,
+      "loss": 0.1415,
+      "step": 1135
+    },
+    {
+      "epoch": 3.6666666666666665,
+      "grad_norm": 0.8988010883331299,
+      "learning_rate": 5.4987519819601e-06,
+      "loss": 0.0987,
+      "step": 1140
+    },
+    {
+      "epoch": 3.682769726247987,
+      "grad_norm": 1.0092145204544067,
+      "learning_rate": 5.37584888006724e-06,
+      "loss": 0.1018,
+      "step": 1145
+    },
+    {
+      "epoch": 3.6988727858293076,
+      "grad_norm": 0.853235125541687,
+      "learning_rate": 5.254034299905825e-06,
+      "loss": 0.1116,
+      "step": 1150
+    },
+    {
+      "epoch": 3.714975845410628,
+      "grad_norm": 1.030652642250061,
+      "learning_rate": 5.1333220190876366e-06,
+      "loss": 0.1178,
+      "step": 1155
+    },
+    {
+      "epoch": 3.7310789049919486,
+      "grad_norm": 1.0030436515808105,
+      "learning_rate": 5.0137256905509125e-06,
+      "loss": 0.0959,
+      "step": 1160
+    },
+    {
+      "epoch": 3.7471819645732687,
+      "grad_norm": 1.0387457609176636,
+      "learning_rate": 4.895258841016193e-06,
+      "loss": 0.1171,
+      "step": 1165
+    },
+    {
+      "epoch": 3.763285024154589,
+      "grad_norm": 0.9441067576408386,
+      "learning_rate": 4.777934869456392e-06,
+      "loss": 0.0964,
+      "step": 1170
+    },
+    {
+      "epoch": 3.7793880837359097,
+      "grad_norm": 1.0259944200515747,
+      "learning_rate": 4.661767045581333e-06,
+      "loss": 0.1132,
+      "step": 1175
+    },
+    {
+      "epoch": 3.7954911433172303,
+      "grad_norm": 0.8966212272644043,
+      "learning_rate": 4.546768508336897e-06,
+      "loss": 0.1011,
+      "step": 1180
+    },
+    {
+      "epoch": 3.8115942028985508,
+      "grad_norm": 0.9555193185806274,
+      "learning_rate": 4.432952264418995e-06,
+      "loss": 0.1013,
+      "step": 1185
+    },
+    {
+      "epoch": 3.8276972624798713,
+      "grad_norm": 1.0005531311035156,
+      "learning_rate": 4.320331186802423e-06,
+      "loss": 0.1013,
+      "step": 1190
+    },
+    {
+      "epoch": 3.843800322061192,
+      "grad_norm": 0.860589325428009,
+      "learning_rate": 4.208918013284927e-06,
+      "loss": 0.1059,
+      "step": 1195
+    },
+    {
+      "epoch": 3.8599033816425123,
+      "grad_norm": 1.0399781465530396,
+      "learning_rate": 4.098725345046495e-06,
+      "loss": 0.0976,
+      "step": 1200
+    },
+    {
+      "epoch": 3.8760064412238324,
+      "grad_norm": 0.7752050161361694,
+      "learning_rate": 3.9897656452241445e-06,
+      "loss": 0.1039,
+      "step": 1205
+    },
+    {
+      "epoch": 3.892109500805153,
+      "grad_norm": 0.708637535572052,
+      "learning_rate": 3.8820512375022834e-06,
+      "loss": 0.1156,
+      "step": 1210
+    },
+    {
+      "epoch": 3.9082125603864735,
+      "grad_norm": 0.9214046597480774,
+      "learning_rate": 3.7755943047188697e-06,
+      "loss": 0.1057,
+      "step": 1215
+    },
+    {
+      "epoch": 3.924315619967794,
+      "grad_norm": 0.935538649559021,
+      "learning_rate": 3.6704068874874917e-06,
+      "loss": 0.1086,
+      "step": 1220
+    },
+    {
+      "epoch": 3.940418679549114,
+      "grad_norm": 0.9492272138595581,
+      "learning_rate": 3.5665008828355566e-06,
+      "loss": 0.0996,
+      "step": 1225
+    },
+    {
+      "epoch": 3.9565217391304346,
+      "grad_norm": 0.919436514377594,
+      "learning_rate": 3.463888042858664e-06,
+      "loss": 0.1099,
+      "step": 1230
+    },
+    {
+      "epoch": 3.972624798711755,
+      "grad_norm": 0.9560986161231995,
+      "learning_rate": 3.3625799733914286e-06,
+      "loss": 0.0967,
+      "step": 1235
+    },
+    {
+      "epoch": 3.9887278582930756,
+      "grad_norm": 0.873535692691803,
+      "learning_rate": 3.2625881326948326e-06,
+      "loss": 0.1053,
+      "step": 1240
+    },
+    {
+      "epoch": 4.003220611916264,
+      "grad_norm": 0.7944340705871582,
+      "learning_rate": 3.1639238301602376e-06,
+      "loss": 0.0835,
+      "step": 1245
+    },
+    {
+      "epoch": 4.019323671497585,
+      "grad_norm": 0.7604539394378662,
+      "learning_rate": 3.06659822503027e-06,
+      "loss": 0.0741,
+      "step": 1250
+    },
+    {
+      "epoch": 4.035426731078905,
+      "grad_norm": 0.7764983177185059,
+      "learning_rate": 2.970622325136669e-06,
+      "loss": 0.1002,
+      "step": 1255
+    },
+    {
+      "epoch": 4.051529790660226,
+      "grad_norm": 0.7339351177215576,
+      "learning_rate": 2.8760069856552875e-06,
+      "loss": 0.073,
+      "step": 1260
+    },
+    {
+      "epoch": 4.067632850241546,
+      "grad_norm": 0.7452949285507202,
+      "learning_rate": 2.7827629078782996e-06,
+      "loss": 0.0806,
+      "step": 1265
+    },
+    {
+      "epoch": 4.083735909822867,
+      "grad_norm": 0.6853229999542236,
+      "learning_rate": 2.6909006380038874e-06,
+      "loss": 0.0835,
+      "step": 1270
+    },
+    {
+      "epoch": 4.099838969404187,
+      "grad_norm": 0.7205654382705688,
+      "learning_rate": 2.6004305659434024e-06,
+      "loss": 0.0807,
+      "step": 1275
+    },
+    {
+      "epoch": 4.115942028985507,
+      "grad_norm": 0.7608060240745544,
+      "learning_rate": 2.5113629241462703e-06,
+      "loss": 0.0802,
+      "step": 1280
+    },
+    {
+      "epoch": 4.132045088566827,
+      "grad_norm": 0.8705418109893799,
+      "learning_rate": 2.42370778644263e-06,
+      "loss": 0.0823,
+      "step": 1285
+    },
+    {
+      "epoch": 4.148148148148148,
+      "grad_norm": 0.7755470275878906,
+      "learning_rate": 2.337475066903973e-06,
+      "loss": 0.0783,
+      "step": 1290
+    },
+    {
+      "epoch": 4.164251207729468,
+      "grad_norm": 0.8142275214195251,
+      "learning_rate": 2.252674518721829e-06,
+      "loss": 0.0873,
+      "step": 1295
+    },
+    {
+      "epoch": 4.180354267310789,
+      "grad_norm": 0.9993813037872314,
+      "learning_rate": 2.1693157331046504e-06,
+      "loss": 0.0829,
+      "step": 1300
+    },
+    {
+      "epoch": 4.1964573268921095,
+      "grad_norm": 0.6725679039955139,
+      "learning_rate": 2.0874081381930073e-06,
+      "loss": 0.0668,
+      "step": 1305
+    },
+    {
+      "epoch": 4.21256038647343,
+      "grad_norm": 0.778102695941925,
+      "learning_rate": 2.0069609979932396e-06,
+      "loss": 0.0922,
+      "step": 1310
+    },
+    {
+      "epoch": 4.2286634460547505,
+      "grad_norm": 0.8950245380401611,
+      "learning_rate": 1.927983411329685e-06,
+      "loss": 0.0946,
+      "step": 1315
+    },
+    {
+      "epoch": 4.244766505636071,
+      "grad_norm": 0.7721431255340576,
+      "learning_rate": 1.850484310815541e-06,
+      "loss": 0.0651,
+      "step": 1320
+    },
+    {
+      "epoch": 4.260869565217392,
+      "grad_norm": 0.8024240732192993,
+      "learning_rate": 1.7744724618425867e-06,
+      "loss": 0.0876,
+      "step": 1325
+    },
+    {
+      "epoch": 4.276972624798712,
+      "grad_norm": 0.7815277576446533,
+      "learning_rate": 1.6999564615897706e-06,
+      "loss": 0.0946,
+      "step": 1330
+    },
+    {
+      "epoch": 4.293075684380033,
+      "grad_norm": 0.7834950089454651,
+      "learning_rate": 1.6269447380508657e-06,
+      "loss": 0.0769,
+      "step": 1335
+    },
+    {
+      "epoch": 4.309178743961352,
+      "grad_norm": 0.9737600088119507,
+      "learning_rate": 1.5554455490812048e-06,
+      "loss": 0.0797,
+      "step": 1340
+    },
+    {
+      "epoch": 4.325281803542673,
+      "grad_norm": 0.6915079951286316,
+      "learning_rate": 1.4854669814637145e-06,
+      "loss": 0.0817,
+      "step": 1345
+    },
+    {
+      "epoch": 4.341384863123993,
+      "grad_norm": 0.7295588850975037,
+      "learning_rate": 1.4170169499942753e-06,
+      "loss": 0.074,
+      "step": 1350
+    },
+    {
+      "epoch": 4.357487922705314,
+      "grad_norm": 0.7292076945304871,
+      "learning_rate": 1.3501031965865207e-06,
+      "loss": 0.0889,
+      "step": 1355
+    },
+    {
+      "epoch": 4.373590982286634,
+      "grad_norm": 0.831916332244873,
+      "learning_rate": 1.2847332893962076e-06,
+      "loss": 0.0746,
+      "step": 1360
+    },
+    {
+      "epoch": 4.389694041867955,
+      "grad_norm": 0.6844192743301392,
+      "learning_rate": 1.2209146219652335e-06,
+      "loss": 0.0839,
+      "step": 1365
+    },
+    {
+      "epoch": 4.405797101449275,
+      "grad_norm": 0.6453650593757629,
+      "learning_rate": 1.1586544123854138e-06,
+      "loss": 0.08,
+      "step": 1370
+    },
+    {
+      "epoch": 4.421900161030596,
+      "grad_norm": 0.8373687267303467,
+      "learning_rate": 1.0979597024820749e-06,
+      "loss": 0.0844,
+      "step": 1375
+    },
+    {
+      "epoch": 4.438003220611916,
+      "grad_norm": 0.7590744495391846,
+      "learning_rate": 1.038837357017612e-06,
+      "loss": 0.0838,
+      "step": 1380
+    },
+    {
+      "epoch": 4.454106280193237,
+      "grad_norm": 0.6972682476043701,
+      "learning_rate": 9.812940629150684e-07,
+      "loss": 0.0839,
+      "step": 1385
+    },
+    {
+      "epoch": 4.4702093397745575,
+      "grad_norm": 0.683718740940094,
+      "learning_rate": 9.253363285018118e-07,
+      "loss": 0.0758,
+      "step": 1390
+    },
+    {
+      "epoch": 4.486312399355878,
+      "grad_norm": 0.7787452340126038,
+      "learning_rate": 8.709704827734244e-07,
+      "loss": 0.0919,
+      "step": 1395
+    },
+    {
+      "epoch": 4.5024154589371985,
+      "grad_norm": 0.756025493144989,
+      "learning_rate": 8.182026746778787e-07,
+      "loss": 0.0727,
+      "step": 1400
+    },
+    {
+      "epoch": 4.518518518518518,
+      "grad_norm": 0.7247276902198792,
+      "learning_rate": 7.670388724200767e-07,
+      "loss": 0.0828,
+      "step": 1405
+    },
+    {
+      "epoch": 4.534621578099839,
+      "grad_norm": 0.7496559023857117,
+      "learning_rate": 7.174848627868097e-07,
+      "loss": 0.0802,
+      "step": 1410
+    },
+    {
+      "epoch": 4.550724637681159,
+      "grad_norm": 0.7700359225273132,
+      "learning_rate": 6.695462504922706e-07,
+      "loss": 0.0823,
+      "step": 1415
+    },
+    {
+      "epoch": 4.56682769726248,
+      "grad_norm": 0.627795398235321,
+      "learning_rate": 6.232284575441338e-07,
+      "loss": 0.0717,
+      "step": 1420
+    },
+    {
+      "epoch": 4.5829307568438,
+      "grad_norm": 0.7721079587936401,
+      "learning_rate": 5.785367226303212e-07,
+      "loss": 0.0764,
+      "step": 1425
+    },
+    {
+      "epoch": 4.599033816425121,
+      "grad_norm": 0.7804126739501953,
+      "learning_rate": 5.354761005264725e-07,
+      "loss": 0.0734,
+      "step": 1430
+    },
+    {
+      "epoch": 4.615136876006441,
+      "grad_norm": 0.7084224224090576,
+      "learning_rate": 4.940514615242475e-07,
+      "loss": 0.0673,
+      "step": 1435
+    },
+    {
+      "epoch": 4.631239935587762,
+      "grad_norm": 0.656895637512207,
+      "learning_rate": 4.5426749088047725e-07,
+      "loss": 0.0799,
+      "step": 1440
+    },
+    {
+      "epoch": 4.647342995169082,
+      "grad_norm": 0.9349786639213562,
+      "learning_rate": 4.161286882872456e-07,
+      "loss": 0.0932,
+      "step": 1445
+    },
+    {
+      "epoch": 4.663446054750403,
+      "grad_norm": 0.6597044467926025,
+      "learning_rate": 3.796393673629617e-07,
+      "loss": 0.0786,
+      "step": 1450
+    },
+    {
+      "epoch": 4.679549114331723,
+      "grad_norm": 0.7497778534889221,
+      "learning_rate": 3.4480365516447176e-07,
+      "loss": 0.0794,
+      "step": 1455
+    },
+    {
+      "epoch": 4.695652173913043,
+      "grad_norm": 0.7813300490379333,
+      "learning_rate": 3.116254917202843e-07,
+      "loss": 0.0896,
+      "step": 1460
+    },
+    {
+      "epoch": 4.7117552334943635,
+      "grad_norm": 0.7202969789505005,
+      "learning_rate": 2.8010862958493145e-07,
+      "loss": 0.0709,
+      "step": 1465
+    },
+    {
+      "epoch": 4.727858293075684,
+      "grad_norm": 0.6622485518455505,
+      "learning_rate": 2.502566334145501e-07,
+      "loss": 0.0811,
+      "step": 1470
+    },
+    {
+      "epoch": 4.743961352657005,
+      "grad_norm": 0.7711755633354187,
+      "learning_rate": 2.2207287956370283e-07,
+      "loss": 0.0755,
+      "step": 1475
+    },
+    {
+      "epoch": 4.760064412238325,
+      "grad_norm": 0.7787219882011414,
+      "learning_rate": 1.955605557035084e-07,
+      "loss": 0.0807,
+      "step": 1480
+    },
+    {
+      "epoch": 4.776167471819646,
+      "grad_norm": 0.7959200143814087,
+      "learning_rate": 1.7072266046109675e-07,
+      "loss": 0.0959,
+      "step": 1485
+    },
+    {
+      "epoch": 4.792270531400966,
+      "grad_norm": 0.6802248954772949,
+      "learning_rate": 1.4756200308046198e-07,
+      "loss": 0.0766,
+      "step": 1490
+    },
+    {
+      "epoch": 4.808373590982287,
+      "grad_norm": 0.7470690608024597,
+      "learning_rate": 1.260812031047265e-07,
+      "loss": 0.0793,
+      "step": 1495
+    },
+    {
+      "epoch": 4.824476650563607,
+      "grad_norm": 0.8416296243667603,
+      "learning_rate": 1.0628269007986469e-07,
+      "loss": 0.0797,
+      "step": 1500
+    },
+    {
+      "epoch": 4.840579710144928,
+      "grad_norm": 0.8034248948097229,
+      "learning_rate": 8.816870327990445e-08,
+      "loss": 0.0786,
+      "step": 1505
+    },
+    {
+      "epoch": 4.856682769726248,
+      "grad_norm": 0.9114562273025513,
+      "learning_rate": 7.17412914536697e-08,
+      "loss": 0.0912,
+      "step": 1510
+    },
+    {
+      "epoch": 4.872785829307569,
+      "grad_norm": 0.7382351756095886,
+      "learning_rate": 5.7002312593054105e-08,
+      "loss": 0.0719,
+      "step": 1515
+    },
+    {
+      "epoch": 4.888888888888889,
+      "grad_norm": 0.717256486415863,
+      "learning_rate": 4.3953433722877524e-08,
+      "loss": 0.0859,
+      "step": 1520
+    },
+    {
+      "epoch": 4.90499194847021,
+      "grad_norm": 0.7053928971290588,
+      "learning_rate": 3.2596130712340135e-08,
+      "loss": 0.0843,
+      "step": 1525
+    },
+    {
+      "epoch": 4.921095008051529,
+      "grad_norm": 0.7305659055709839,
+      "learning_rate": 2.2931688108099314e-08,
+      "loss": 0.0726,
+      "step": 1530
+    },
+    {
+      "epoch": 4.93719806763285,
+      "grad_norm": 0.6849284768104553,
+      "learning_rate": 1.4961198988979186e-08,
+      "loss": 0.0818,
+      "step": 1535
+    },
+    {
+      "epoch": 4.9533011272141705,
+      "grad_norm": 0.7415360808372498,
+      "learning_rate": 8.68556484234284e-09,
+      "loss": 0.0761,
+      "step": 1540
+    },
+    {
+      "epoch": 4.969404186795491,
+      "grad_norm": 0.7831457257270813,
+      "learning_rate": 4.105495462132236e-09,
+      "loss": 0.0845,
+      "step": 1545
+    },
+    {
+      "epoch": 4.9855072463768115,
+      "grad_norm": 0.6487026214599609,
+      "learning_rate": 1.2215088685840758e-09,
+      "loss": 0.0696,
+      "step": 1550
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 0.9624710083007812,
+      "learning_rate": 3.3931249641683257e-11,
+      "loss": 0.0819,
+      "step": 1555
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1555,
+  "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": 2.2083740614514442e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

28_128_e5_3e-5/checkpoint-1555/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3839c69e7206b6c68ac17087343cb6c4560ba7b2b5157d7f3e1c366218d509a5
+size 7736

28_128_e5_3e-5/checkpoint-1555/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)

28_128_e5_3e-5/checkpoint-311/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

28_128_e5_3e-5/checkpoint-311/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",
+    "q_proj",
+    "v_proj",
+    "o_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

28_128_e5_3e-5/checkpoint-311/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24f6dd44e7d68033097fea63575908a36ddc2fbcc657a83fc6a079a25b0fba76
+size 791751704

28_128_e5_3e-5/checkpoint-311/latest

@@ -0,0 +1 @@
+global_step311

28_128_e5_3e-5/checkpoint-311/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e9b945521cf32c3f0c626f62465922b364b7e3d71bfc119ac67e33efb9dbdb2
+size 15920

28_128_e5_3e-5/checkpoint-311/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6f5e371e3f2f0a57a7905ea24dddc24b2d99b54eed6fb4738ac750d29558ba7
+size 15920

28_128_e5_3e-5/checkpoint-311/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a7fa597bb47c3f9e0a2d0d1e03a4fd79a02bdab8604048869847c24fb5b667c
+size 15920