Upload folder using huggingface_hub

29_128_e5_3e-5/checkpoint-1200/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

29_128_e5_3e-5/checkpoint-1200/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": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "v_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

29_128_e5_3e-5/checkpoint-1200/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98ae73188a9280340e8c4a7d7a75f8b63d130dad7c4553e29283b8a522fccef5
+size 791751704

29_128_e5_3e-5/checkpoint-1200/latest

@@ -0,0 +1 @@
+global_step1200

29_128_e5_3e-5/checkpoint-1200/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:395549f20ebf649866dabb0f42d70d4b377c30bbbabd0bd187402fa1173d5e8b
+size 15920

29_128_e5_3e-5/checkpoint-1200/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63a36a7c700a5c66ab37cf38f99c5e67a9dbbecb66e11284037bf8cf880f83df
+size 15920

29_128_e5_3e-5/checkpoint-1200/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20f4898f2e1eff06806b723d77a4a9b7a45eff92abcf456881a2d7d55d22cb4a
+size 15920

29_128_e5_3e-5/checkpoint-1200/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a6c733fef39192cb9a35a5ca0394aea1e728c17b6bdbb1f8588b4100346705c
+size 15920

29_128_e5_3e-5/checkpoint-1200/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61cd62e68c8bba98ce11d46e6b30c8b5b16e32f2a3cb6223636cfc0eda7e5bd2
+size 15920

29_128_e5_3e-5/checkpoint-1200/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64d4ae801cb390d8f664478ea3416e9ccc2f3c8a0bf32dd3bdb101e51337b840
+size 15920

29_128_e5_3e-5/checkpoint-1200/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6c376fab84bc4ea70d3ff327808968d969c489066fce2c451129f5a769f9114
+size 15920

29_128_e5_3e-5/checkpoint-1200/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42a0ea346ca498e12b2377b0c54493e98e49df78e82d06a6fe59f971eb64c13a
+size 15920

29_128_e5_3e-5/checkpoint-1200/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea5934dfa4faf58e9018aef3d9049a70a4693a4e30aa8146e066503639b9d21b
+size 1064

29_128_e5_3e-5/checkpoint-1200/special_tokens_map.json

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

29_128_e5_3e-5/checkpoint-1200/tokenizer_config.json

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

29_128_e5_3e-5/checkpoint-1200/trainer_state.json

@@ -0,0 +1,1714 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0125,
+      "grad_norm": 1.086734652519226,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 1.3059,
+      "step": 5
+    },
+    {
+      "epoch": 0.025,
+      "grad_norm": 0.9148593544960022,
+      "learning_rate": 2.7e-06,
+      "loss": 1.3276,
+      "step": 10
+    },
+    {
+      "epoch": 0.0375,
+      "grad_norm": 0.7434160113334656,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 1.3328,
+      "step": 15
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 0.6667717099189758,
+      "learning_rate": 5.7000000000000005e-06,
+      "loss": 1.3032,
+      "step": 20
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 0.5638660192489624,
+      "learning_rate": 7.2e-06,
+      "loss": 1.3357,
+      "step": 25
+    },
+    {
+      "epoch": 0.075,
+      "grad_norm": 0.5368859171867371,
+      "learning_rate": 8.7e-06,
+      "loss": 1.2642,
+      "step": 30
+    },
+    {
+      "epoch": 0.0875,
+      "grad_norm": 0.6104497313499451,
+      "learning_rate": 1.02e-05,
+      "loss": 1.3103,
+      "step": 35
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 0.5396920442581177,
+      "learning_rate": 1.1700000000000001e-05,
+      "loss": 1.2437,
+      "step": 40
+    },
+    {
+      "epoch": 0.1125,
+      "grad_norm": 0.5430600047111511,
+      "learning_rate": 1.32e-05,
+      "loss": 1.2582,
+      "step": 45
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.44843268394470215,
+      "learning_rate": 1.47e-05,
+      "loss": 1.2164,
+      "step": 50
+    },
+    {
+      "epoch": 0.1375,
+      "grad_norm": 0.5682841539382935,
+      "learning_rate": 1.62e-05,
+      "loss": 1.2429,
+      "step": 55
+    },
+    {
+      "epoch": 0.15,
+      "grad_norm": 0.5052217245101929,
+      "learning_rate": 1.77e-05,
+      "loss": 1.2212,
+      "step": 60
+    },
+    {
+      "epoch": 0.1625,
+      "grad_norm": 0.4474620223045349,
+      "learning_rate": 1.9200000000000003e-05,
+      "loss": 1.1847,
+      "step": 65
+    },
+    {
+      "epoch": 0.175,
+      "grad_norm": 0.5079185962677002,
+      "learning_rate": 2.07e-05,
+      "loss": 1.2262,
+      "step": 70
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.47006088495254517,
+      "learning_rate": 2.22e-05,
+      "loss": 1.159,
+      "step": 75
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 0.5115821361541748,
+      "learning_rate": 2.37e-05,
+      "loss": 1.1331,
+      "step": 80
+    },
+    {
+      "epoch": 0.2125,
+      "grad_norm": 0.5092158317565918,
+      "learning_rate": 2.52e-05,
+      "loss": 1.1752,
+      "step": 85
+    },
+    {
+      "epoch": 0.225,
+      "grad_norm": 0.7345715761184692,
+      "learning_rate": 2.6700000000000002e-05,
+      "loss": 1.172,
+      "step": 90
+    },
+    {
+      "epoch": 0.2375,
+      "grad_norm": 0.5289098024368286,
+      "learning_rate": 2.8199999999999998e-05,
+      "loss": 1.1589,
+      "step": 95
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.5456954836845398,
+      "learning_rate": 2.97e-05,
+      "loss": 1.0921,
+      "step": 100
+    },
+    {
+      "epoch": 0.2625,
+      "grad_norm": 0.5777515172958374,
+      "learning_rate": 2.9999671925703373e-05,
+      "loss": 1.1855,
+      "step": 105
+    },
+    {
+      "epoch": 0.275,
+      "grad_norm": 0.651767909526825,
+      "learning_rate": 2.9998339148469037e-05,
+      "loss": 1.1073,
+      "step": 110
+    },
+    {
+      "epoch": 0.2875,
+      "grad_norm": 0.4987607002258301,
+      "learning_rate": 2.9995981254676936e-05,
+      "loss": 1.1076,
+      "step": 115
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 0.5758209228515625,
+      "learning_rate": 2.9992598405485974e-05,
+      "loss": 1.0912,
+      "step": 120
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.8067079782485962,
+      "learning_rate": 2.998819083210941e-05,
+      "loss": 1.0431,
+      "step": 125
+    },
+    {
+      "epoch": 0.325,
+      "grad_norm": 0.537667453289032,
+      "learning_rate": 2.998275883579902e-05,
+      "loss": 1.0783,
+      "step": 130
+    },
+    {
+      "epoch": 0.3375,
+      "grad_norm": 0.6047284007072449,
+      "learning_rate": 2.997630278782454e-05,
+      "loss": 1.0636,
+      "step": 135
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 0.7209072113037109,
+      "learning_rate": 2.9968823129448274e-05,
+      "loss": 1.0524,
+      "step": 140
+    },
+    {
+      "epoch": 0.3625,
+      "grad_norm": 0.6621487140655518,
+      "learning_rate": 2.996032037189493e-05,
+      "loss": 1.0593,
+      "step": 145
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.6688674688339233,
+      "learning_rate": 2.9950795096316707e-05,
+      "loss": 0.9841,
+      "step": 150
+    },
+    {
+      "epoch": 0.3875,
+      "grad_norm": 0.6082017421722412,
+      "learning_rate": 2.9940247953753522e-05,
+      "loss": 1.0213,
+      "step": 155
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 0.7148690819740295,
+      "learning_rate": 2.9928679665088565e-05,
+      "loss": 1.0034,
+      "step": 160
+    },
+    {
+      "epoch": 0.4125,
+      "grad_norm": 0.6557154059410095,
+      "learning_rate": 2.9916091020999002e-05,
+      "loss": 0.9285,
+      "step": 165
+    },
+    {
+      "epoch": 0.425,
+      "grad_norm": 0.8312786817550659,
+      "learning_rate": 2.9902482881901934e-05,
+      "loss": 0.9875,
+      "step": 170
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.6497301459312439,
+      "learning_rate": 2.9887856177895586e-05,
+      "loss": 0.9698,
+      "step": 175
+    },
+    {
+      "epoch": 0.45,
+      "grad_norm": 0.6885787844657898,
+      "learning_rate": 2.987221190869575e-05,
+      "loss": 0.9614,
+      "step": 180
+    },
+    {
+      "epoch": 0.4625,
+      "grad_norm": 0.7662853002548218,
+      "learning_rate": 2.985555114356745e-05,
+      "loss": 0.9896,
+      "step": 185
+    },
+    {
+      "epoch": 0.475,
+      "grad_norm": 0.659321665763855,
+      "learning_rate": 2.983787502125185e-05,
+      "loss": 0.9846,
+      "step": 190
+    },
+    {
+      "epoch": 0.4875,
+      "grad_norm": 0.7467763423919678,
+      "learning_rate": 2.9819184749888423e-05,
+      "loss": 0.9759,
+      "step": 195
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.763093113899231,
+      "learning_rate": 2.979948160693239e-05,
+      "loss": 0.9271,
+      "step": 200
+    },
+    {
+      "epoch": 0.5125,
+      "grad_norm": 0.7503418326377869,
+      "learning_rate": 2.977876693906741e-05,
+      "loss": 0.9311,
+      "step": 205
+    },
+    {
+      "epoch": 0.525,
+      "grad_norm": 0.7104331254959106,
+      "learning_rate": 2.975704216211351e-05,
+      "loss": 0.9195,
+      "step": 210
+    },
+    {
+      "epoch": 0.5375,
+      "grad_norm": 0.838015615940094,
+      "learning_rate": 2.9734308760930333e-05,
+      "loss": 0.8963,
+      "step": 215
+    },
+    {
+      "epoch": 0.55,
+      "grad_norm": 0.8911941051483154,
+      "learning_rate": 2.9710568289315652e-05,
+      "loss": 0.8869,
+      "step": 220
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.7820862531661987,
+      "learning_rate": 2.968582236989917e-05,
+      "loss": 0.906,
+      "step": 225
+    },
+    {
+      "epoch": 0.575,
+      "grad_norm": 0.7517544627189636,
+      "learning_rate": 2.9660072694031605e-05,
+      "loss": 0.9284,
+      "step": 230
+    },
+    {
+      "epoch": 0.5875,
+      "grad_norm": 0.845384418964386,
+      "learning_rate": 2.9633321021669106e-05,
+      "loss": 0.8505,
+      "step": 235
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 0.7715607285499573,
+      "learning_rate": 2.9605569181252933e-05,
+      "loss": 0.8817,
+      "step": 240
+    },
+    {
+      "epoch": 0.6125,
+      "grad_norm": 3.4710471630096436,
+      "learning_rate": 2.957681906958452e-05,
+      "loss": 0.8737,
+      "step": 245
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.9051657319068909,
+      "learning_rate": 2.9547072651695803e-05,
+      "loss": 0.8843,
+      "step": 250
+    },
+    {
+      "epoch": 0.6375,
+      "grad_norm": 0.8575041890144348,
+      "learning_rate": 2.951633196071494e-05,
+      "loss": 0.832,
+      "step": 255
+    },
+    {
+      "epoch": 0.65,
+      "grad_norm": 0.995969831943512,
+      "learning_rate": 2.9484599097727317e-05,
+      "loss": 0.8513,
+      "step": 260
+    },
+    {
+      "epoch": 0.6625,
+      "grad_norm": 0.945318877696991,
+      "learning_rate": 2.9451876231631974e-05,
+      "loss": 0.8536,
+      "step": 265
+    },
+    {
+      "epoch": 0.675,
+      "grad_norm": 0.9573862552642822,
+      "learning_rate": 2.9418165598993335e-05,
+      "loss": 0.7936,
+      "step": 270
+    },
+    {
+      "epoch": 0.6875,
+      "grad_norm": 0.9231624603271484,
+      "learning_rate": 2.9383469503888366e-05,
+      "loss": 0.8143,
+      "step": 275
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 0.923611581325531,
+      "learning_rate": 2.934779031774908e-05,
+      "loss": 0.8392,
+      "step": 280
+    },
+    {
+      "epoch": 0.7125,
+      "grad_norm": 0.8286426067352295,
+      "learning_rate": 2.9311130479200447e-05,
+      "loss": 0.8319,
+      "step": 285
+    },
+    {
+      "epoch": 0.725,
+      "grad_norm": 0.9228174090385437,
+      "learning_rate": 2.9273492493893743e-05,
+      "loss": 0.8213,
+      "step": 290
+    },
+    {
+      "epoch": 0.7375,
+      "grad_norm": 0.9326543807983398,
+      "learning_rate": 2.9234878934335264e-05,
+      "loss": 0.8286,
+      "step": 295
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.9320464730262756,
+      "learning_rate": 2.9195292439710517e-05,
+      "loss": 0.8206,
+      "step": 300
+    },
+    {
+      "epoch": 0.7625,
+      "grad_norm": 0.9256623387336731,
+      "learning_rate": 2.9154735715703816e-05,
+      "loss": 0.7637,
+      "step": 305
+    },
+    {
+      "epoch": 0.775,
+      "grad_norm": 0.952567994594574,
+      "learning_rate": 2.9113211534313385e-05,
+      "loss": 0.8445,
+      "step": 310
+    },
+    {
+      "epoch": 0.7875,
+      "grad_norm": 0.9957364797592163,
+      "learning_rate": 2.9070722733661856e-05,
+      "loss": 0.7563,
+      "step": 315
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 0.9804989099502563,
+      "learning_rate": 2.902727221780231e-05,
+      "loss": 0.7775,
+      "step": 320
+    },
+    {
+      "epoch": 0.8125,
+      "grad_norm": 1.0001029968261719,
+      "learning_rate": 2.89828629565198e-05,
+      "loss": 0.7535,
+      "step": 325
+    },
+    {
+      "epoch": 0.825,
+      "grad_norm": 0.9708714485168457,
+      "learning_rate": 2.893749798512833e-05,
+      "loss": 0.7825,
+      "step": 330
+    },
+    {
+      "epoch": 0.8375,
+      "grad_norm": 1.0440568923950195,
+      "learning_rate": 2.889118040426345e-05,
+      "loss": 0.7497,
+      "step": 335
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 1.1043874025344849,
+      "learning_rate": 2.8843913379670276e-05,
+      "loss": 0.7402,
+      "step": 340
+    },
+    {
+      "epoch": 0.8625,
+      "grad_norm": 0.9350870847702026,
+      "learning_rate": 2.8795700141987153e-05,
+      "loss": 0.7163,
+      "step": 345
+    },
+    {
+      "epoch": 0.875,
+      "grad_norm": 0.9727185964584351,
+      "learning_rate": 2.874654398652483e-05,
+      "loss": 0.7286,
+      "step": 350
+    },
+    {
+      "epoch": 0.8875,
+      "grad_norm": 0.9859328269958496,
+      "learning_rate": 2.8696448273041243e-05,
+      "loss": 0.7221,
+      "step": 355
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 0.9797766804695129,
+      "learning_rate": 2.8645416425511855e-05,
+      "loss": 0.6722,
+      "step": 360
+    },
+    {
+      "epoch": 0.9125,
+      "grad_norm": 0.880566418170929,
+      "learning_rate": 2.859345193189567e-05,
+      "loss": 0.7379,
+      "step": 365
+    },
+    {
+      "epoch": 0.925,
+      "grad_norm": 0.9722506999969482,
+      "learning_rate": 2.8540558343896787e-05,
+      "loss": 0.6665,
+      "step": 370
+    },
+    {
+      "epoch": 0.9375,
+      "grad_norm": 1.0045404434204102,
+      "learning_rate": 2.848673927672169e-05,
+      "loss": 0.6749,
+      "step": 375
+    },
+    {
+      "epoch": 0.95,
+      "grad_norm": 0.9702157974243164,
+      "learning_rate": 2.843199840883213e-05,
+      "loss": 0.6908,
+      "step": 380
+    },
+    {
+      "epoch": 0.9625,
+      "grad_norm": 1.1106311082839966,
+      "learning_rate": 2.837633948169371e-05,
+      "loss": 0.7297,
+      "step": 385
+    },
+    {
+      "epoch": 0.975,
+      "grad_norm": 0.9999896883964539,
+      "learning_rate": 2.8319766299520164e-05,
+      "loss": 0.6577,
+      "step": 390
+    },
+    {
+      "epoch": 0.9875,
+      "grad_norm": 0.9784647822380066,
+      "learning_rate": 2.8262282729013353e-05,
+      "loss": 0.693,
+      "step": 395
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.9709898233413696,
+      "learning_rate": 2.8203892699098956e-05,
+      "loss": 0.6752,
+      "step": 400
+    },
+    {
+      "epoch": 1.0125,
+      "grad_norm": 0.9533920884132385,
+      "learning_rate": 2.8144600200657953e-05,
+      "loss": 0.6057,
+      "step": 405
+    },
+    {
+      "epoch": 1.025,
+      "grad_norm": 1.112910270690918,
+      "learning_rate": 2.808440928625386e-05,
+      "loss": 0.5958,
+      "step": 410
+    },
+    {
+      "epoch": 1.0375,
+      "grad_norm": 1.167204737663269,
+      "learning_rate": 2.8023324069855714e-05,
+      "loss": 0.6015,
+      "step": 415
+    },
+    {
+      "epoch": 1.05,
+      "grad_norm": 1.128758430480957,
+      "learning_rate": 2.7961348726556918e-05,
+      "loss": 0.6223,
+      "step": 420
+    },
+    {
+      "epoch": 1.0625,
+      "grad_norm": 1.3209761381149292,
+      "learning_rate": 2.7898487492289855e-05,
+      "loss": 0.5801,
+      "step": 425
+    },
+    {
+      "epoch": 1.075,
+      "grad_norm": 1.096300482749939,
+      "learning_rate": 2.7834744663536395e-05,
+      "loss": 0.59,
+      "step": 430
+    },
+    {
+      "epoch": 1.0875,
+      "grad_norm": 1.0881574153900146,
+      "learning_rate": 2.7770124597034205e-05,
+      "loss": 0.5658,
+      "step": 435
+    },
+    {
+      "epoch": 1.1,
+      "grad_norm": 1.2155952453613281,
+      "learning_rate": 2.770463170947899e-05,
+      "loss": 0.5828,
+      "step": 440
+    },
+    {
+      "epoch": 1.1125,
+      "grad_norm": 1.1294485330581665,
+      "learning_rate": 2.763827047722262e-05,
+      "loss": 0.5345,
+      "step": 445
+    },
+    {
+      "epoch": 1.125,
+      "grad_norm": 1.1405869722366333,
+      "learning_rate": 2.757104543596717e-05,
+      "loss": 0.5759,
+      "step": 450
+    },
+    {
+      "epoch": 1.1375,
+      "grad_norm": 1.0838080644607544,
+      "learning_rate": 2.7502961180454915e-05,
+      "loss": 0.5247,
+      "step": 455
+    },
+    {
+      "epoch": 1.15,
+      "grad_norm": 1.1288769245147705,
+      "learning_rate": 2.7434022364154272e-05,
+      "loss": 0.5777,
+      "step": 460
+    },
+    {
+      "epoch": 1.1625,
+      "grad_norm": 1.133094072341919,
+      "learning_rate": 2.7364233698941764e-05,
+      "loss": 0.5041,
+      "step": 465
+    },
+    {
+      "epoch": 1.175,
+      "grad_norm": 1.0078961849212646,
+      "learning_rate": 2.729359995477995e-05,
+      "loss": 0.5607,
+      "step": 470
+    },
+    {
+      "epoch": 1.1875,
+      "grad_norm": 1.1658203601837158,
+      "learning_rate": 2.7222125959391424e-05,
+      "loss": 0.588,
+      "step": 475
+    },
+    {
+      "epoch": 1.2,
+      "grad_norm": 0.9901076555252075,
+      "learning_rate": 2.7149816597928824e-05,
+      "loss": 0.5257,
+      "step": 480
+    },
+    {
+      "epoch": 1.2125,
+      "grad_norm": 1.1960549354553223,
+      "learning_rate": 2.7076676812640962e-05,
+      "loss": 0.5219,
+      "step": 485
+    },
+    {
+      "epoch": 1.225,
+      "grad_norm": 1.0868672132492065,
+      "learning_rate": 2.7002711602534995e-05,
+      "loss": 0.5205,
+      "step": 490
+    },
+    {
+      "epoch": 1.2375,
+      "grad_norm": 1.078203797340393,
+      "learning_rate": 2.6927926023034784e-05,
+      "loss": 0.5298,
+      "step": 495
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 1.0336695909500122,
+      "learning_rate": 2.6852325185635357e-05,
+      "loss": 0.5461,
+      "step": 500
+    },
+    {
+      "epoch": 1.2625,
+      "grad_norm": 0.9965822100639343,
+      "learning_rate": 2.6775914257553517e-05,
+      "loss": 0.5509,
+      "step": 505
+    },
+    {
+      "epoch": 1.275,
+      "grad_norm": 1.0736743211746216,
+      "learning_rate": 2.66986984613747e-05,
+      "loss": 0.5251,
+      "step": 510
+    },
+    {
+      "epoch": 1.2875,
+      "grad_norm": 0.9483274221420288,
+      "learning_rate": 2.6620683074696e-05,
+      "loss": 0.5098,
+      "step": 515
+    },
+    {
+      "epoch": 1.3,
+      "grad_norm": 1.1101971864700317,
+      "learning_rate": 2.654187342976548e-05,
+      "loss": 0.4986,
+      "step": 520
+    },
+    {
+      "epoch": 1.3125,
+      "grad_norm": 1.0614863634109497,
+      "learning_rate": 2.6462274913117666e-05,
+      "loss": 0.513,
+      "step": 525
+    },
+    {
+      "epoch": 1.325,
+      "grad_norm": 1.2074096202850342,
+      "learning_rate": 2.638189296520545e-05,
+      "loss": 0.5344,
+      "step": 530
+    },
+    {
+      "epoch": 1.3375,
+      "grad_norm": 1.2375788688659668,
+      "learning_rate": 2.630073308002819e-05,
+      "loss": 0.5142,
+      "step": 535
+    },
+    {
+      "epoch": 1.35,
+      "grad_norm": 1.0503743886947632,
+      "learning_rate": 2.6218800804756232e-05,
+      "loss": 0.4584,
+      "step": 540
+    },
+    {
+      "epoch": 1.3625,
+      "grad_norm": 0.9696715474128723,
+      "learning_rate": 2.613610173935175e-05,
+      "loss": 0.4718,
+      "step": 545
+    },
+    {
+      "epoch": 1.375,
+      "grad_norm": 1.0960568189620972,
+      "learning_rate": 2.605264153618602e-05,
+      "loss": 0.5352,
+      "step": 550
+    },
+    {
+      "epoch": 1.3875,
+      "grad_norm": 1.0279024839401245,
+      "learning_rate": 2.5968425899653055e-05,
+      "loss": 0.4806,
+      "step": 555
+    },
+    {
+      "epoch": 1.4,
+      "grad_norm": 1.0534144639968872,
+      "learning_rate": 2.5883460585779754e-05,
+      "loss": 0.4874,
+      "step": 560
+    },
+    {
+      "epoch": 1.4125,
+      "grad_norm": 1.0914931297302246,
+      "learning_rate": 2.579775140183245e-05,
+      "loss": 0.5225,
+      "step": 565
+    },
+    {
+      "epoch": 1.425,
+      "grad_norm": 1.0744789838790894,
+      "learning_rate": 2.5711304205920015e-05,
+      "loss": 0.5056,
+      "step": 570
+    },
+    {
+      "epoch": 1.4375,
+      "grad_norm": 1.2198079824447632,
+      "learning_rate": 2.5624124906593458e-05,
+      "loss": 0.4751,
+      "step": 575
+    },
+    {
+      "epoch": 1.45,
+      "grad_norm": 1.1285591125488281,
+      "learning_rate": 2.5536219462442086e-05,
+      "loss": 0.4315,
+      "step": 580
+    },
+    {
+      "epoch": 1.4625,
+      "grad_norm": 1.1472607851028442,
+      "learning_rate": 2.5447593881686235e-05,
+      "loss": 0.4825,
+      "step": 585
+    },
+    {
+      "epoch": 1.475,
+      "grad_norm": 1.0363106727600098,
+      "learning_rate": 2.5358254221766624e-05,
+      "loss": 0.4582,
+      "step": 590
+    },
+    {
+      "epoch": 1.4875,
+      "grad_norm": 1.2606053352355957,
+      "learning_rate": 2.5268206588930332e-05,
+      "loss": 0.4674,
+      "step": 595
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 1.1823961734771729,
+      "learning_rate": 2.517745713781345e-05,
+      "loss": 0.5222,
+      "step": 600
+    },
+    {
+      "epoch": 1.5125,
+      "grad_norm": 1.1270991563796997,
+      "learning_rate": 2.508601207102042e-05,
+      "loss": 0.438,
+      "step": 605
+    },
+    {
+      "epoch": 1.525,
+      "grad_norm": 1.0307538509368896,
+      "learning_rate": 2.4993877638700083e-05,
+      "loss": 0.4528,
+      "step": 610
+    },
+    {
+      "epoch": 1.5375,
+      "grad_norm": 1.1684446334838867,
+      "learning_rate": 2.4901060138118502e-05,
+      "loss": 0.4359,
+      "step": 615
+    },
+    {
+      "epoch": 1.55,
+      "grad_norm": 1.1035212278366089,
+      "learning_rate": 2.480756591322854e-05,
+      "loss": 0.4552,
+      "step": 620
+    },
+    {
+      "epoch": 1.5625,
+      "grad_norm": 1.2371423244476318,
+      "learning_rate": 2.4713401354236282e-05,
+      "loss": 0.4246,
+      "step": 625
+    },
+    {
+      "epoch": 1.575,
+      "grad_norm": 1.1511696577072144,
+      "learning_rate": 2.4618572897164255e-05,
+      "loss": 0.4018,
+      "step": 630
+    },
+    {
+      "epoch": 1.5875,
+      "grad_norm": 1.25965416431427,
+      "learning_rate": 2.452308702341153e-05,
+      "loss": 0.4366,
+      "step": 635
+    },
+    {
+      "epoch": 1.6,
+      "grad_norm": 1.0590370893478394,
+      "learning_rate": 2.442695025931074e-05,
+      "loss": 0.4153,
+      "step": 640
+    },
+    {
+      "epoch": 1.6125,
+      "grad_norm": 1.0804214477539062,
+      "learning_rate": 2.433016917568201e-05,
+      "loss": 0.4147,
+      "step": 645
+    },
+    {
+      "epoch": 1.625,
+      "grad_norm": 1.0489362478256226,
+      "learning_rate": 2.4232750387383847e-05,
+      "loss": 0.4363,
+      "step": 650
+    },
+    {
+      "epoch": 1.6375,
+      "grad_norm": 1.198287010192871,
+      "learning_rate": 2.4134700552861018e-05,
+      "loss": 0.4347,
+      "step": 655
+    },
+    {
+      "epoch": 1.65,
+      "grad_norm": 1.1126651763916016,
+      "learning_rate": 2.403602637368948e-05,
+      "loss": 0.4434,
+      "step": 660
+    },
+    {
+      "epoch": 1.6625,
+      "grad_norm": 1.137837290763855,
+      "learning_rate": 2.3936734594118295e-05,
+      "loss": 0.4437,
+      "step": 665
+    },
+    {
+      "epoch": 1.675,
+      "grad_norm": 1.1391346454620361,
+      "learning_rate": 2.3836832000608704e-05,
+      "loss": 0.4306,
+      "step": 670
+    },
+    {
+      "epoch": 1.6875,
+      "grad_norm": 0.9699942469596863,
+      "learning_rate": 2.3736325421370275e-05,
+      "loss": 0.4575,
+      "step": 675
+    },
+    {
+      "epoch": 1.7,
+      "grad_norm": 1.0101770162582397,
+      "learning_rate": 2.3635221725894186e-05,
+      "loss": 0.4321,
+      "step": 680
+    },
+    {
+      "epoch": 1.7125,
+      "grad_norm": 1.0216741561889648,
+      "learning_rate": 2.3533527824483722e-05,
+      "loss": 0.4074,
+      "step": 685
+    },
+    {
+      "epoch": 1.725,
+      "grad_norm": 1.1822564601898193,
+      "learning_rate": 2.3431250667781958e-05,
+      "loss": 0.451,
+      "step": 690
+    },
+    {
+      "epoch": 1.7375,
+      "grad_norm": 1.1009029150009155,
+      "learning_rate": 2.33283972462967e-05,
+      "loss": 0.4171,
+      "step": 695
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 1.347333550453186,
+      "learning_rate": 2.3224974589922673e-05,
+      "loss": 0.4304,
+      "step": 700
+    },
+    {
+      "epoch": 1.7625,
+      "grad_norm": 1.4213443994522095,
+      "learning_rate": 2.312098976746107e-05,
+      "loss": 0.3918,
+      "step": 705
+    },
+    {
+      "epoch": 1.775,
+      "grad_norm": 1.077433466911316,
+      "learning_rate": 2.3016449886136372e-05,
+      "loss": 0.4047,
+      "step": 710
+    },
+    {
+      "epoch": 1.7875,
+      "grad_norm": 1.1451928615570068,
+      "learning_rate": 2.2911362091110596e-05,
+      "loss": 0.3928,
+      "step": 715
+    },
+    {
+      "epoch": 1.8,
+      "grad_norm": 1.272578477859497,
+      "learning_rate": 2.2805733564994948e-05,
+      "loss": 0.3776,
+      "step": 720
+    },
+    {
+      "epoch": 1.8125,
+      "grad_norm": 1.2507826089859009,
+      "learning_rate": 2.269957152735887e-05,
+      "loss": 0.3815,
+      "step": 725
+    },
+    {
+      "epoch": 1.825,
+      "grad_norm": 1.2830368280410767,
+      "learning_rate": 2.259288323423662e-05,
+      "loss": 0.384,
+      "step": 730
+    },
+    {
+      "epoch": 1.8375,
+      "grad_norm": 1.164024829864502,
+      "learning_rate": 2.2485675977631308e-05,
+      "loss": 0.3858,
+      "step": 735
+    },
+    {
+      "epoch": 1.85,
+      "grad_norm": 1.3388005495071411,
+      "learning_rate": 2.237795708501652e-05,
+      "loss": 0.4192,
+      "step": 740
+    },
+    {
+      "epoch": 1.8625,
+      "grad_norm": 0.9997522830963135,
+      "learning_rate": 2.2269733918835478e-05,
+      "loss": 0.3802,
+      "step": 745
+    },
+    {
+      "epoch": 1.875,
+      "grad_norm": 1.1285563707351685,
+      "learning_rate": 2.216101387599782e-05,
+      "loss": 0.3711,
+      "step": 750
+    },
+    {
+      "epoch": 1.8875,
+      "grad_norm": 1.1473147869110107,
+      "learning_rate": 2.205180438737407e-05,
+      "loss": 0.3749,
+      "step": 755
+    },
+    {
+      "epoch": 1.9,
+      "grad_norm": 1.2977720499038696,
+      "learning_rate": 2.1942112917287704e-05,
+      "loss": 0.3863,
+      "step": 760
+    },
+    {
+      "epoch": 1.9125,
+      "grad_norm": 1.165961503982544,
+      "learning_rate": 2.1831946963004988e-05,
+      "loss": 0.3753,
+      "step": 765
+    },
+    {
+      "epoch": 1.925,
+      "grad_norm": 1.1040446758270264,
+      "learning_rate": 2.1721314054222563e-05,
+      "loss": 0.3467,
+      "step": 770
+    },
+    {
+      "epoch": 1.9375,
+      "grad_norm": 1.0992324352264404,
+      "learning_rate": 2.1610221752552784e-05,
+      "loss": 0.3624,
+      "step": 775
+    },
+    {
+      "epoch": 1.95,
+      "grad_norm": 1.210777759552002,
+      "learning_rate": 2.1498677651006906e-05,
+      "loss": 0.3796,
+      "step": 780
+    },
+    {
+      "epoch": 1.9625,
+      "grad_norm": 1.1452407836914062,
+      "learning_rate": 2.138668937347609e-05,
+      "loss": 0.3686,
+      "step": 785
+    },
+    {
+      "epoch": 1.975,
+      "grad_norm": 1.2745158672332764,
+      "learning_rate": 2.1274264574210352e-05,
+      "loss": 0.3553,
+      "step": 790
+    },
+    {
+      "epoch": 1.9875,
+      "grad_norm": 1.21009361743927,
+      "learning_rate": 2.1161410937295385e-05,
+      "loss": 0.3413,
+      "step": 795
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.2560657262802124,
+      "learning_rate": 2.1048136176127357e-05,
+      "loss": 0.3242,
+      "step": 800
+    },
+    {
+      "epoch": 2.0125,
+      "grad_norm": 1.1402417421340942,
+      "learning_rate": 2.0934448032885724e-05,
+      "loss": 0.2778,
+      "step": 805
+    },
+    {
+      "epoch": 2.025,
+      "grad_norm": 1.2693945169448853,
+      "learning_rate": 2.0820354278004066e-05,
+      "loss": 0.2728,
+      "step": 810
+    },
+    {
+      "epoch": 2.0375,
+      "grad_norm": 1.238933801651001,
+      "learning_rate": 2.0705862709638972e-05,
+      "loss": 0.303,
+      "step": 815
+    },
+    {
+      "epoch": 2.05,
+      "grad_norm": 1.1587833166122437,
+      "learning_rate": 2.059098115313705e-05,
+      "loss": 0.2636,
+      "step": 820
+    },
+    {
+      "epoch": 2.0625,
+      "grad_norm": 1.290188193321228,
+      "learning_rate": 2.0475717460500088e-05,
+      "loss": 0.2925,
+      "step": 825
+    },
+    {
+      "epoch": 2.075,
+      "grad_norm": 1.2627977132797241,
+      "learning_rate": 2.0360079509848366e-05,
+      "loss": 0.2911,
+      "step": 830
+    },
+    {
+      "epoch": 2.0875,
+      "grad_norm": 1.1960002183914185,
+      "learning_rate": 2.02440752048822e-05,
+      "loss": 0.2887,
+      "step": 835
+    },
+    {
+      "epoch": 2.1,
+      "grad_norm": 1.150658130645752,
+      "learning_rate": 2.0127712474341747e-05,
+      "loss": 0.2759,
+      "step": 840
+    },
+    {
+      "epoch": 2.1125,
+      "grad_norm": 1.3438881635665894,
+      "learning_rate": 2.001099927146507e-05,
+      "loss": 0.2718,
+      "step": 845
+    },
+    {
+      "epoch": 2.125,
+      "grad_norm": 1.0232552289962769,
+      "learning_rate": 1.989394357344454e-05,
+      "loss": 0.2865,
+      "step": 850
+    },
+    {
+      "epoch": 2.1375,
+      "grad_norm": 1.052276611328125,
+      "learning_rate": 1.9776553380881634e-05,
+      "loss": 0.2561,
+      "step": 855
+    },
+    {
+      "epoch": 2.15,
+      "grad_norm": 1.2404303550720215,
+      "learning_rate": 1.965883671724007e-05,
+      "loss": 0.2631,
+      "step": 860
+    },
+    {
+      "epoch": 2.1625,
+      "grad_norm": 1.1204473972320557,
+      "learning_rate": 1.954080162829745e-05,
+      "loss": 0.2489,
+      "step": 865
+    },
+    {
+      "epoch": 2.175,
+      "grad_norm": 1.102751612663269,
+      "learning_rate": 1.9422456181595304e-05,
+      "loss": 0.2536,
+      "step": 870
+    },
+    {
+      "epoch": 2.1875,
+      "grad_norm": 1.3035495281219482,
+      "learning_rate": 1.9303808465887713e-05,
+      "loss": 0.2617,
+      "step": 875
+    },
+    {
+      "epoch": 2.2,
+      "grad_norm": 1.1419627666473389,
+      "learning_rate": 1.918486659058844e-05,
+      "loss": 0.269,
+      "step": 880
+    },
+    {
+      "epoch": 2.2125,
+      "grad_norm": 1.2295540571212769,
+      "learning_rate": 1.9065638685216667e-05,
+      "loss": 0.2596,
+      "step": 885
+    },
+    {
+      "epoch": 2.225,
+      "grad_norm": 1.2283310890197754,
+      "learning_rate": 1.8946132898841344e-05,
+      "loss": 0.2427,
+      "step": 890
+    },
+    {
+      "epoch": 2.2375,
+      "grad_norm": 1.084255576133728,
+      "learning_rate": 1.8826357399524227e-05,
+      "loss": 0.2959,
+      "step": 895
+    },
+    {
+      "epoch": 2.25,
+      "grad_norm": 1.203917384147644,
+      "learning_rate": 1.8706320373761594e-05,
+      "loss": 0.2551,
+      "step": 900
+    },
+    {
+      "epoch": 2.2625,
+      "grad_norm": 1.0413402318954468,
+      "learning_rate": 1.8586030025924697e-05,
+      "loss": 0.2429,
+      "step": 905
+    },
+    {
+      "epoch": 2.275,
+      "grad_norm": 1.2268816232681274,
+      "learning_rate": 1.8465494577699033e-05,
+      "loss": 0.2476,
+      "step": 910
+    },
+    {
+      "epoch": 2.2875,
+      "grad_norm": 1.2103034257888794,
+      "learning_rate": 1.8344722267522377e-05,
+      "loss": 0.2472,
+      "step": 915
+    },
+    {
+      "epoch": 2.3,
+      "grad_norm": 1.125713586807251,
+      "learning_rate": 1.822372135002169e-05,
+      "loss": 0.2632,
+      "step": 920
+    },
+    {
+      "epoch": 2.3125,
+      "grad_norm": 1.2902660369873047,
+      "learning_rate": 1.810250009544895e-05,
+      "loss": 0.2424,
+      "step": 925
+    },
+    {
+      "epoch": 2.325,
+      "grad_norm": 1.1748820543289185,
+      "learning_rate": 1.7981066789115897e-05,
+      "loss": 0.2302,
+      "step": 930
+    },
+    {
+      "epoch": 2.3375,
+      "grad_norm": 1.058660626411438,
+      "learning_rate": 1.7859429730827704e-05,
+      "loss": 0.2237,
+      "step": 935
+    },
+    {
+      "epoch": 2.35,
+      "grad_norm": 1.1528538465499878,
+      "learning_rate": 1.7737597234315727e-05,
+      "loss": 0.2575,
+      "step": 940
+    },
+    {
+      "epoch": 2.3625,
+      "grad_norm": 1.188664197921753,
+      "learning_rate": 1.7615577626669274e-05,
+      "loss": 0.2512,
+      "step": 945
+    },
+    {
+      "epoch": 2.375,
+      "grad_norm": 1.1694797277450562,
+      "learning_rate": 1.7493379247766444e-05,
+      "loss": 0.2382,
+      "step": 950
+    },
+    {
+      "epoch": 2.3875,
+      "grad_norm": 1.1770079135894775,
+      "learning_rate": 1.7371010449704097e-05,
+      "loss": 0.2181,
+      "step": 955
+    },
+    {
+      "epoch": 2.4,
+      "grad_norm": 1.051956295967102,
+      "learning_rate": 1.724847959622705e-05,
+      "loss": 0.2397,
+      "step": 960
+    },
+    {
+      "epoch": 2.4125,
+      "grad_norm": 1.1248499155044556,
+      "learning_rate": 1.7125795062156363e-05,
+      "loss": 0.2478,
+      "step": 965
+    },
+    {
+      "epoch": 2.425,
+      "grad_norm": 1.0435619354248047,
+      "learning_rate": 1.7002965232816977e-05,
+      "loss": 0.274,
+      "step": 970
+    },
+    {
+      "epoch": 2.4375,
+      "grad_norm": 1.1685163974761963,
+      "learning_rate": 1.6879998503464565e-05,
+      "loss": 0.2389,
+      "step": 975
+    },
+    {
+      "epoch": 2.45,
+      "grad_norm": 1.136615514755249,
+      "learning_rate": 1.675690327871174e-05,
+      "loss": 0.2471,
+      "step": 980
+    },
+    {
+      "epoch": 2.4625,
+      "grad_norm": 1.2618733644485474,
+      "learning_rate": 1.663368797195359e-05,
+      "loss": 0.2371,
+      "step": 985
+    },
+    {
+      "epoch": 2.475,
+      "grad_norm": 1.3753706216812134,
+      "learning_rate": 1.6510361004792665e-05,
+      "loss": 0.2581,
+      "step": 990
+    },
+    {
+      "epoch": 2.4875,
+      "grad_norm": 1.141377329826355,
+      "learning_rate": 1.6386930806463355e-05,
+      "loss": 0.2336,
+      "step": 995
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 1.3642939329147339,
+      "learning_rate": 1.6263405813255748e-05,
+      "loss": 0.232,
+      "step": 1000
+    },
+    {
+      "epoch": 2.5125,
+      "grad_norm": 1.2523542642593384,
+      "learning_rate": 1.6139794467939047e-05,
+      "loss": 0.2128,
+      "step": 1005
+    },
+    {
+      "epoch": 2.525,
+      "grad_norm": 1.3189752101898193,
+      "learning_rate": 1.601610521918451e-05,
+      "loss": 0.2218,
+      "step": 1010
+    },
+    {
+      "epoch": 2.5375,
+      "grad_norm": 1.1008399724960327,
+      "learning_rate": 1.5892346520987986e-05,
+      "loss": 0.2463,
+      "step": 1015
+    },
+    {
+      "epoch": 2.55,
+      "grad_norm": 1.1347867250442505,
+      "learning_rate": 1.5768526832092097e-05,
+      "loss": 0.211,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5625,
+      "grad_norm": 1.186482310295105,
+      "learning_rate": 1.564465461540811e-05,
+      "loss": 0.2208,
+      "step": 1025
+    },
+    {
+      "epoch": 2.575,
+      "grad_norm": 1.1044975519180298,
+      "learning_rate": 1.5520738337437485e-05,
+      "loss": 0.2125,
+      "step": 1030
+    },
+    {
+      "epoch": 2.5875,
+      "grad_norm": 1.147139310836792,
+      "learning_rate": 1.5396786467693216e-05,
+      "loss": 0.2038,
+      "step": 1035
+    },
+    {
+      "epoch": 2.6,
+      "grad_norm": 1.3628555536270142,
+      "learning_rate": 1.5272807478120954e-05,
+      "loss": 0.2299,
+      "step": 1040
+    },
+    {
+      "epoch": 2.6125,
+      "grad_norm": 1.1701704263687134,
+      "learning_rate": 1.5148809842519945e-05,
+      "loss": 0.2116,
+      "step": 1045
+    },
+    {
+      "epoch": 2.625,
+      "grad_norm": 0.9667600393295288,
+      "learning_rate": 1.5024802035963873e-05,
+      "loss": 0.2344,
+      "step": 1050
+    },
+    {
+      "epoch": 2.6375,
+      "grad_norm": 1.1534322500228882,
+      "learning_rate": 1.4900792534221589e-05,
+      "loss": 0.2231,
+      "step": 1055
+    },
+    {
+      "epoch": 2.65,
+      "grad_norm": 1.2170480489730835,
+      "learning_rate": 1.4776789813177808e-05,
+      "loss": 0.2243,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6625,
+      "grad_norm": 1.1053290367126465,
+      "learning_rate": 1.4652802348253793e-05,
+      "loss": 0.2144,
+      "step": 1065
+    },
+    {
+      "epoch": 2.675,
+      "grad_norm": 1.067043423652649,
+      "learning_rate": 1.4528838613828076e-05,
+      "loss": 0.2341,
+      "step": 1070
+    },
+    {
+      "epoch": 2.6875,
+      "grad_norm": 1.0644820928573608,
+      "learning_rate": 1.4404907082657238e-05,
+      "loss": 0.199,
+      "step": 1075
+    },
+    {
+      "epoch": 2.7,
+      "grad_norm": 1.0699652433395386,
+      "learning_rate": 1.4281016225296808e-05,
+      "loss": 0.1893,
+      "step": 1080
+    },
+    {
+      "epoch": 2.7125,
+      "grad_norm": 1.2279590368270874,
+      "learning_rate": 1.4157174509522318e-05,
+      "loss": 0.2027,
+      "step": 1085
+    },
+    {
+      "epoch": 2.725,
+      "grad_norm": 1.0308759212493896,
+      "learning_rate": 1.4033390399750534e-05,
+      "loss": 0.197,
+      "step": 1090
+    },
+    {
+      "epoch": 2.7375,
+      "grad_norm": 1.281794786453247,
+      "learning_rate": 1.3909672356460934e-05,
+      "loss": 0.2161,
+      "step": 1095
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 1.0185565948486328,
+      "learning_rate": 1.378602883561744e-05,
+      "loss": 0.2212,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7625,
+      "grad_norm": 1.130090594291687,
+      "learning_rate": 1.3662468288090446e-05,
+      "loss": 0.1898,
+      "step": 1105
+    },
+    {
+      "epoch": 2.775,
+      "grad_norm": 1.1406450271606445,
+      "learning_rate": 1.3538999159079253e-05,
+      "loss": 0.1764,
+      "step": 1110
+    },
+    {
+      "epoch": 2.7875,
+      "grad_norm": 1.2482119798660278,
+      "learning_rate": 1.3415629887534817e-05,
+      "loss": 0.1768,
+      "step": 1115
+    },
+    {
+      "epoch": 2.8,
+      "grad_norm": 1.1973145008087158,
+      "learning_rate": 1.3292368905582972e-05,
+      "loss": 0.1947,
+      "step": 1120
+    },
+    {
+      "epoch": 2.8125,
+      "grad_norm": 1.1229867935180664,
+      "learning_rate": 1.3169224637948099e-05,
+      "loss": 0.2081,
+      "step": 1125
+    },
+    {
+      "epoch": 2.825,
+      "grad_norm": 1.129728078842163,
+      "learning_rate": 1.3046205501377313e-05,
+      "loss": 0.1761,
+      "step": 1130
+    },
+    {
+      "epoch": 2.8375,
+      "grad_norm": 1.0157361030578613,
+      "learning_rate": 1.2923319904065187e-05,
+      "loss": 0.206,
+      "step": 1135
+    },
+    {
+      "epoch": 2.85,
+      "grad_norm": 1.4206640720367432,
+      "learning_rate": 1.280057624507906e-05,
+      "loss": 0.1954,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8625,
+      "grad_norm": 1.1708303689956665,
+      "learning_rate": 1.267798291378498e-05,
+      "loss": 0.1919,
+      "step": 1145
+    },
+    {
+      "epoch": 2.875,
+      "grad_norm": 0.9822798371315002,
+      "learning_rate": 1.2555548289274293e-05,
+      "loss": 0.1725,
+      "step": 1150
+    },
+    {
+      "epoch": 2.8875,
+      "grad_norm": 1.2971844673156738,
+      "learning_rate": 1.243328073979095e-05,
+      "loss": 0.2,
+      "step": 1155
+    },
+    {
+      "epoch": 2.9,
+      "grad_norm": 1.1322238445281982,
+      "learning_rate": 1.231118862215954e-05,
+      "loss": 0.1632,
+      "step": 1160
+    },
+    {
+      "epoch": 2.9125,
+      "grad_norm": 1.1313915252685547,
+      "learning_rate": 1.2189280281214128e-05,
+      "loss": 0.1942,
+      "step": 1165
+    },
+    {
+      "epoch": 2.925,
+      "grad_norm": 1.2760839462280273,
+      "learning_rate": 1.2067564049227884e-05,
+      "loss": 0.1802,
+      "step": 1170
+    },
+    {
+      "epoch": 2.9375,
+      "grad_norm": 1.1183810234069824,
+      "learning_rate": 1.1946048245343568e-05,
+      "loss": 0.2114,
+      "step": 1175
+    },
+    {
+      "epoch": 2.95,
+      "grad_norm": 1.0835254192352295,
+      "learning_rate": 1.1824741175004967e-05,
+      "loss": 0.1931,
+      "step": 1180
+    },
+    {
+      "epoch": 2.9625,
+      "grad_norm": 1.1179535388946533,
+      "learning_rate": 1.1703651129389203e-05,
+      "loss": 0.1968,
+      "step": 1185
+    },
+    {
+      "epoch": 2.975,
+      "grad_norm": 1.0869146585464478,
+      "learning_rate": 1.1582786384840045e-05,
+      "loss": 0.1676,
+      "step": 1190
+    },
+    {
+      "epoch": 2.9875,
+      "grad_norm": 1.2248408794403076,
+      "learning_rate": 1.1462155202302227e-05,
+      "loss": 0.1867,
+      "step": 1195
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.1031678915023804,
+      "learning_rate": 1.134176582675684e-05,
+      "loss": 0.1632,
+      "step": 1200
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 2000,
+  "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": 1.64227002403704e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

29_128_e5_3e-5/checkpoint-1200/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:871812aa4697fc2cb08f60c4dde08e33ab05b528c39c3bc200d654ee41929393
+size 7736

29_128_e5_3e-5/checkpoint-1200/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)

29_128_e5_3e-5/checkpoint-1600/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

29_128_e5_3e-5/checkpoint-1600/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": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "v_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

29_128_e5_3e-5/checkpoint-1600/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8aa97a198fb9d657d1271fe3b8998c09daf3bf5a01af5bfe45a731a8c2276bdc
+size 791751704

29_128_e5_3e-5/checkpoint-1600/latest

@@ -0,0 +1 @@
+global_step1600

29_128_e5_3e-5/checkpoint-1600/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df77d3fd5983d714d62b5953af1a5d5a1ff10e087b906f75049c8f872c8ce4a9
+size 15920

29_128_e5_3e-5/checkpoint-1600/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3ac756345bdaa81846b51b4f4279937971435d5356b03c7fa2251fe89d6272c
+size 15920

29_128_e5_3e-5/checkpoint-1600/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c011fb1b61da1056f218f9173212ce8c62b44899d0bb682f9a650c17ad62c05
+size 15920

29_128_e5_3e-5/checkpoint-1600/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e67e39acc22b048a8bf7352b9ab4fa69c67fafb8a32af6048549eee36c99c51
+size 15920

29_128_e5_3e-5/checkpoint-1600/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0eb3718704783cdc27dd46fde86fad28811a667beaaa5675460aa19d9ddb7cb
+size 15920

29_128_e5_3e-5/checkpoint-1600/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:606561a590155b2fd8746fe4671bfa7387b5ee5a9a45ecaa0a5d6c1389f1ce45
+size 15920

29_128_e5_3e-5/checkpoint-1600/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76f853368408bf3c2b3d6e8134cedc2c610f74c80ec28dfddba618eb0024efd4
+size 15920

29_128_e5_3e-5/checkpoint-1600/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:120eae9bf5327834137c4a5c87c1d1eaeb90f2d5839850140d9c938f6c5c37bb
+size 15920

29_128_e5_3e-5/checkpoint-1600/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b988c0c72fc48eeb111ea8bd873c7d8ed9542882c86efb19d43f6dc9e95297e7
+size 1064

29_128_e5_3e-5/checkpoint-1600/special_tokens_map.json

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

29_128_e5_3e-5/checkpoint-1600/tokenizer_config.json

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

29_128_e5_3e-5/checkpoint-1600/trainer_state.json

@@ -0,0 +1,2274 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 1600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0125,
+      "grad_norm": 1.086734652519226,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 1.3059,
+      "step": 5
+    },
+    {
+      "epoch": 0.025,
+      "grad_norm": 0.9148593544960022,
+      "learning_rate": 2.7e-06,
+      "loss": 1.3276,
+      "step": 10
+    },
+    {
+      "epoch": 0.0375,
+      "grad_norm": 0.7434160113334656,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 1.3328,
+      "step": 15
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 0.6667717099189758,
+      "learning_rate": 5.7000000000000005e-06,
+      "loss": 1.3032,
+      "step": 20
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 0.5638660192489624,
+      "learning_rate": 7.2e-06,
+      "loss": 1.3357,
+      "step": 25
+    },
+    {
+      "epoch": 0.075,
+      "grad_norm": 0.5368859171867371,
+      "learning_rate": 8.7e-06,
+      "loss": 1.2642,
+      "step": 30
+    },
+    {
+      "epoch": 0.0875,
+      "grad_norm": 0.6104497313499451,
+      "learning_rate": 1.02e-05,
+      "loss": 1.3103,
+      "step": 35
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 0.5396920442581177,
+      "learning_rate": 1.1700000000000001e-05,
+      "loss": 1.2437,
+      "step": 40
+    },
+    {
+      "epoch": 0.1125,
+      "grad_norm": 0.5430600047111511,
+      "learning_rate": 1.32e-05,
+      "loss": 1.2582,
+      "step": 45
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.44843268394470215,
+      "learning_rate": 1.47e-05,
+      "loss": 1.2164,
+      "step": 50
+    },
+    {
+      "epoch": 0.1375,
+      "grad_norm": 0.5682841539382935,
+      "learning_rate": 1.62e-05,
+      "loss": 1.2429,
+      "step": 55
+    },
+    {
+      "epoch": 0.15,
+      "grad_norm": 0.5052217245101929,
+      "learning_rate": 1.77e-05,
+      "loss": 1.2212,
+      "step": 60
+    },
+    {
+      "epoch": 0.1625,
+      "grad_norm": 0.4474620223045349,
+      "learning_rate": 1.9200000000000003e-05,
+      "loss": 1.1847,
+      "step": 65
+    },
+    {
+      "epoch": 0.175,
+      "grad_norm": 0.5079185962677002,
+      "learning_rate": 2.07e-05,
+      "loss": 1.2262,
+      "step": 70
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.47006088495254517,
+      "learning_rate": 2.22e-05,
+      "loss": 1.159,
+      "step": 75
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 0.5115821361541748,
+      "learning_rate": 2.37e-05,
+      "loss": 1.1331,
+      "step": 80
+    },
+    {
+      "epoch": 0.2125,
+      "grad_norm": 0.5092158317565918,
+      "learning_rate": 2.52e-05,
+      "loss": 1.1752,
+      "step": 85
+    },
+    {
+      "epoch": 0.225,
+      "grad_norm": 0.7345715761184692,
+      "learning_rate": 2.6700000000000002e-05,
+      "loss": 1.172,
+      "step": 90
+    },
+    {
+      "epoch": 0.2375,
+      "grad_norm": 0.5289098024368286,
+      "learning_rate": 2.8199999999999998e-05,
+      "loss": 1.1589,
+      "step": 95
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.5456954836845398,
+      "learning_rate": 2.97e-05,
+      "loss": 1.0921,
+      "step": 100
+    },
+    {
+      "epoch": 0.2625,
+      "grad_norm": 0.5777515172958374,
+      "learning_rate": 2.9999671925703373e-05,
+      "loss": 1.1855,
+      "step": 105
+    },
+    {
+      "epoch": 0.275,
+      "grad_norm": 0.651767909526825,
+      "learning_rate": 2.9998339148469037e-05,
+      "loss": 1.1073,
+      "step": 110
+    },
+    {
+      "epoch": 0.2875,
+      "grad_norm": 0.4987607002258301,
+      "learning_rate": 2.9995981254676936e-05,
+      "loss": 1.1076,
+      "step": 115
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 0.5758209228515625,
+      "learning_rate": 2.9992598405485974e-05,
+      "loss": 1.0912,
+      "step": 120
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.8067079782485962,
+      "learning_rate": 2.998819083210941e-05,
+      "loss": 1.0431,
+      "step": 125
+    },
+    {
+      "epoch": 0.325,
+      "grad_norm": 0.537667453289032,
+      "learning_rate": 2.998275883579902e-05,
+      "loss": 1.0783,
+      "step": 130
+    },
+    {
+      "epoch": 0.3375,
+      "grad_norm": 0.6047284007072449,
+      "learning_rate": 2.997630278782454e-05,
+      "loss": 1.0636,
+      "step": 135
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 0.7209072113037109,
+      "learning_rate": 2.9968823129448274e-05,
+      "loss": 1.0524,
+      "step": 140
+    },
+    {
+      "epoch": 0.3625,
+      "grad_norm": 0.6621487140655518,
+      "learning_rate": 2.996032037189493e-05,
+      "loss": 1.0593,
+      "step": 145
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.6688674688339233,
+      "learning_rate": 2.9950795096316707e-05,
+      "loss": 0.9841,
+      "step": 150
+    },
+    {
+      "epoch": 0.3875,
+      "grad_norm": 0.6082017421722412,
+      "learning_rate": 2.9940247953753522e-05,
+      "loss": 1.0213,
+      "step": 155
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 0.7148690819740295,
+      "learning_rate": 2.9928679665088565e-05,
+      "loss": 1.0034,
+      "step": 160
+    },
+    {
+      "epoch": 0.4125,
+      "grad_norm": 0.6557154059410095,
+      "learning_rate": 2.9916091020999002e-05,
+      "loss": 0.9285,
+      "step": 165
+    },
+    {
+      "epoch": 0.425,
+      "grad_norm": 0.8312786817550659,
+      "learning_rate": 2.9902482881901934e-05,
+      "loss": 0.9875,
+      "step": 170
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.6497301459312439,
+      "learning_rate": 2.9887856177895586e-05,
+      "loss": 0.9698,
+      "step": 175
+    },
+    {
+      "epoch": 0.45,
+      "grad_norm": 0.6885787844657898,
+      "learning_rate": 2.987221190869575e-05,
+      "loss": 0.9614,
+      "step": 180
+    },
+    {
+      "epoch": 0.4625,
+      "grad_norm": 0.7662853002548218,
+      "learning_rate": 2.985555114356745e-05,
+      "loss": 0.9896,
+      "step": 185
+    },
+    {
+      "epoch": 0.475,
+      "grad_norm": 0.659321665763855,
+      "learning_rate": 2.983787502125185e-05,
+      "loss": 0.9846,
+      "step": 190
+    },
+    {
+      "epoch": 0.4875,
+      "grad_norm": 0.7467763423919678,
+      "learning_rate": 2.9819184749888423e-05,
+      "loss": 0.9759,
+      "step": 195
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.763093113899231,
+      "learning_rate": 2.979948160693239e-05,
+      "loss": 0.9271,
+      "step": 200
+    },
+    {
+      "epoch": 0.5125,
+      "grad_norm": 0.7503418326377869,
+      "learning_rate": 2.977876693906741e-05,
+      "loss": 0.9311,
+      "step": 205
+    },
+    {
+      "epoch": 0.525,
+      "grad_norm": 0.7104331254959106,
+      "learning_rate": 2.975704216211351e-05,
+      "loss": 0.9195,
+      "step": 210
+    },
+    {
+      "epoch": 0.5375,
+      "grad_norm": 0.838015615940094,
+      "learning_rate": 2.9734308760930333e-05,
+      "loss": 0.8963,
+      "step": 215
+    },
+    {
+      "epoch": 0.55,
+      "grad_norm": 0.8911941051483154,
+      "learning_rate": 2.9710568289315652e-05,
+      "loss": 0.8869,
+      "step": 220
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.7820862531661987,
+      "learning_rate": 2.968582236989917e-05,
+      "loss": 0.906,
+      "step": 225
+    },
+    {
+      "epoch": 0.575,
+      "grad_norm": 0.7517544627189636,
+      "learning_rate": 2.9660072694031605e-05,
+      "loss": 0.9284,
+      "step": 230
+    },
+    {
+      "epoch": 0.5875,
+      "grad_norm": 0.845384418964386,
+      "learning_rate": 2.9633321021669106e-05,
+      "loss": 0.8505,
+      "step": 235
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 0.7715607285499573,
+      "learning_rate": 2.9605569181252933e-05,
+      "loss": 0.8817,
+      "step": 240
+    },
+    {
+      "epoch": 0.6125,
+      "grad_norm": 3.4710471630096436,
+      "learning_rate": 2.957681906958452e-05,
+      "loss": 0.8737,
+      "step": 245
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.9051657319068909,
+      "learning_rate": 2.9547072651695803e-05,
+      "loss": 0.8843,
+      "step": 250
+    },
+    {
+      "epoch": 0.6375,
+      "grad_norm": 0.8575041890144348,
+      "learning_rate": 2.951633196071494e-05,
+      "loss": 0.832,
+      "step": 255
+    },
+    {
+      "epoch": 0.65,
+      "grad_norm": 0.995969831943512,
+      "learning_rate": 2.9484599097727317e-05,
+      "loss": 0.8513,
+      "step": 260
+    },
+    {
+      "epoch": 0.6625,
+      "grad_norm": 0.945318877696991,
+      "learning_rate": 2.9451876231631974e-05,
+      "loss": 0.8536,
+      "step": 265
+    },
+    {
+      "epoch": 0.675,
+      "grad_norm": 0.9573862552642822,
+      "learning_rate": 2.9418165598993335e-05,
+      "loss": 0.7936,
+      "step": 270
+    },
+    {
+      "epoch": 0.6875,
+      "grad_norm": 0.9231624603271484,
+      "learning_rate": 2.9383469503888366e-05,
+      "loss": 0.8143,
+      "step": 275
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 0.923611581325531,
+      "learning_rate": 2.934779031774908e-05,
+      "loss": 0.8392,
+      "step": 280
+    },
+    {
+      "epoch": 0.7125,
+      "grad_norm": 0.8286426067352295,
+      "learning_rate": 2.9311130479200447e-05,
+      "loss": 0.8319,
+      "step": 285
+    },
+    {
+      "epoch": 0.725,
+      "grad_norm": 0.9228174090385437,
+      "learning_rate": 2.9273492493893743e-05,
+      "loss": 0.8213,
+      "step": 290
+    },
+    {
+      "epoch": 0.7375,
+      "grad_norm": 0.9326543807983398,
+      "learning_rate": 2.9234878934335264e-05,
+      "loss": 0.8286,
+      "step": 295
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.9320464730262756,
+      "learning_rate": 2.9195292439710517e-05,
+      "loss": 0.8206,
+      "step": 300
+    },
+    {
+      "epoch": 0.7625,
+      "grad_norm": 0.9256623387336731,
+      "learning_rate": 2.9154735715703816e-05,
+      "loss": 0.7637,
+      "step": 305
+    },
+    {
+      "epoch": 0.775,
+      "grad_norm": 0.952567994594574,
+      "learning_rate": 2.9113211534313385e-05,
+      "loss": 0.8445,
+      "step": 310
+    },
+    {
+      "epoch": 0.7875,
+      "grad_norm": 0.9957364797592163,
+      "learning_rate": 2.9070722733661856e-05,
+      "loss": 0.7563,
+      "step": 315
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 0.9804989099502563,
+      "learning_rate": 2.902727221780231e-05,
+      "loss": 0.7775,
+      "step": 320
+    },
+    {
+      "epoch": 0.8125,
+      "grad_norm": 1.0001029968261719,
+      "learning_rate": 2.89828629565198e-05,
+      "loss": 0.7535,
+      "step": 325
+    },
+    {
+      "epoch": 0.825,
+      "grad_norm": 0.9708714485168457,
+      "learning_rate": 2.893749798512833e-05,
+      "loss": 0.7825,
+      "step": 330
+    },
+    {
+      "epoch": 0.8375,
+      "grad_norm": 1.0440568923950195,
+      "learning_rate": 2.889118040426345e-05,
+      "loss": 0.7497,
+      "step": 335
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 1.1043874025344849,
+      "learning_rate": 2.8843913379670276e-05,
+      "loss": 0.7402,
+      "step": 340
+    },
+    {
+      "epoch": 0.8625,
+      "grad_norm": 0.9350870847702026,
+      "learning_rate": 2.8795700141987153e-05,
+      "loss": 0.7163,
+      "step": 345
+    },
+    {
+      "epoch": 0.875,
+      "grad_norm": 0.9727185964584351,
+      "learning_rate": 2.874654398652483e-05,
+      "loss": 0.7286,
+      "step": 350
+    },
+    {
+      "epoch": 0.8875,
+      "grad_norm": 0.9859328269958496,
+      "learning_rate": 2.8696448273041243e-05,
+      "loss": 0.7221,
+      "step": 355
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 0.9797766804695129,
+      "learning_rate": 2.8645416425511855e-05,
+      "loss": 0.6722,
+      "step": 360
+    },
+    {
+      "epoch": 0.9125,
+      "grad_norm": 0.880566418170929,
+      "learning_rate": 2.859345193189567e-05,
+      "loss": 0.7379,
+      "step": 365
+    },
+    {
+      "epoch": 0.925,
+      "grad_norm": 0.9722506999969482,
+      "learning_rate": 2.8540558343896787e-05,
+      "loss": 0.6665,
+      "step": 370
+    },
+    {
+      "epoch": 0.9375,
+      "grad_norm": 1.0045404434204102,
+      "learning_rate": 2.848673927672169e-05,
+      "loss": 0.6749,
+      "step": 375
+    },
+    {
+      "epoch": 0.95,
+      "grad_norm": 0.9702157974243164,
+      "learning_rate": 2.843199840883213e-05,
+      "loss": 0.6908,
+      "step": 380
+    },
+    {
+      "epoch": 0.9625,
+      "grad_norm": 1.1106311082839966,
+      "learning_rate": 2.837633948169371e-05,
+      "loss": 0.7297,
+      "step": 385
+    },
+    {
+      "epoch": 0.975,
+      "grad_norm": 0.9999896883964539,
+      "learning_rate": 2.8319766299520164e-05,
+      "loss": 0.6577,
+      "step": 390
+    },
+    {
+      "epoch": 0.9875,
+      "grad_norm": 0.9784647822380066,
+      "learning_rate": 2.8262282729013353e-05,
+      "loss": 0.693,
+      "step": 395
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.9709898233413696,
+      "learning_rate": 2.8203892699098956e-05,
+      "loss": 0.6752,
+      "step": 400
+    },
+    {
+      "epoch": 1.0125,
+      "grad_norm": 0.9533920884132385,
+      "learning_rate": 2.8144600200657953e-05,
+      "loss": 0.6057,
+      "step": 405
+    },
+    {
+      "epoch": 1.025,
+      "grad_norm": 1.112910270690918,
+      "learning_rate": 2.808440928625386e-05,
+      "loss": 0.5958,
+      "step": 410
+    },
+    {
+      "epoch": 1.0375,
+      "grad_norm": 1.167204737663269,
+      "learning_rate": 2.8023324069855714e-05,
+      "loss": 0.6015,
+      "step": 415
+    },
+    {
+      "epoch": 1.05,
+      "grad_norm": 1.128758430480957,
+      "learning_rate": 2.7961348726556918e-05,
+      "loss": 0.6223,
+      "step": 420
+    },
+    {
+      "epoch": 1.0625,
+      "grad_norm": 1.3209761381149292,
+      "learning_rate": 2.7898487492289855e-05,
+      "loss": 0.5801,
+      "step": 425
+    },
+    {
+      "epoch": 1.075,
+      "grad_norm": 1.096300482749939,
+      "learning_rate": 2.7834744663536395e-05,
+      "loss": 0.59,
+      "step": 430
+    },
+    {
+      "epoch": 1.0875,
+      "grad_norm": 1.0881574153900146,
+      "learning_rate": 2.7770124597034205e-05,
+      "loss": 0.5658,
+      "step": 435
+    },
+    {
+      "epoch": 1.1,
+      "grad_norm": 1.2155952453613281,
+      "learning_rate": 2.770463170947899e-05,
+      "loss": 0.5828,
+      "step": 440
+    },
+    {
+      "epoch": 1.1125,
+      "grad_norm": 1.1294485330581665,
+      "learning_rate": 2.763827047722262e-05,
+      "loss": 0.5345,
+      "step": 445
+    },
+    {
+      "epoch": 1.125,
+      "grad_norm": 1.1405869722366333,
+      "learning_rate": 2.757104543596717e-05,
+      "loss": 0.5759,
+      "step": 450
+    },
+    {
+      "epoch": 1.1375,
+      "grad_norm": 1.0838080644607544,
+      "learning_rate": 2.7502961180454915e-05,
+      "loss": 0.5247,
+      "step": 455
+    },
+    {
+      "epoch": 1.15,
+      "grad_norm": 1.1288769245147705,
+      "learning_rate": 2.7434022364154272e-05,
+      "loss": 0.5777,
+      "step": 460
+    },
+    {
+      "epoch": 1.1625,
+      "grad_norm": 1.133094072341919,
+      "learning_rate": 2.7364233698941764e-05,
+      "loss": 0.5041,
+      "step": 465
+    },
+    {
+      "epoch": 1.175,
+      "grad_norm": 1.0078961849212646,
+      "learning_rate": 2.729359995477995e-05,
+      "loss": 0.5607,
+      "step": 470
+    },
+    {
+      "epoch": 1.1875,
+      "grad_norm": 1.1658203601837158,
+      "learning_rate": 2.7222125959391424e-05,
+      "loss": 0.588,
+      "step": 475
+    },
+    {
+      "epoch": 1.2,
+      "grad_norm": 0.9901076555252075,
+      "learning_rate": 2.7149816597928824e-05,
+      "loss": 0.5257,
+      "step": 480
+    },
+    {
+      "epoch": 1.2125,
+      "grad_norm": 1.1960549354553223,
+      "learning_rate": 2.7076676812640962e-05,
+      "loss": 0.5219,
+      "step": 485
+    },
+    {
+      "epoch": 1.225,
+      "grad_norm": 1.0868672132492065,
+      "learning_rate": 2.7002711602534995e-05,
+      "loss": 0.5205,
+      "step": 490
+    },
+    {
+      "epoch": 1.2375,
+      "grad_norm": 1.078203797340393,
+      "learning_rate": 2.6927926023034784e-05,
+      "loss": 0.5298,
+      "step": 495
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 1.0336695909500122,
+      "learning_rate": 2.6852325185635357e-05,
+      "loss": 0.5461,
+      "step": 500
+    },
+    {
+      "epoch": 1.2625,
+      "grad_norm": 0.9965822100639343,
+      "learning_rate": 2.6775914257553517e-05,
+      "loss": 0.5509,
+      "step": 505
+    },
+    {
+      "epoch": 1.275,
+      "grad_norm": 1.0736743211746216,
+      "learning_rate": 2.66986984613747e-05,
+      "loss": 0.5251,
+      "step": 510
+    },
+    {
+      "epoch": 1.2875,
+      "grad_norm": 0.9483274221420288,
+      "learning_rate": 2.6620683074696e-05,
+      "loss": 0.5098,
+      "step": 515
+    },
+    {
+      "epoch": 1.3,
+      "grad_norm": 1.1101971864700317,
+      "learning_rate": 2.654187342976548e-05,
+      "loss": 0.4986,
+      "step": 520
+    },
+    {
+      "epoch": 1.3125,
+      "grad_norm": 1.0614863634109497,
+      "learning_rate": 2.6462274913117666e-05,
+      "loss": 0.513,
+      "step": 525
+    },
+    {
+      "epoch": 1.325,
+      "grad_norm": 1.2074096202850342,
+      "learning_rate": 2.638189296520545e-05,
+      "loss": 0.5344,
+      "step": 530
+    },
+    {
+      "epoch": 1.3375,
+      "grad_norm": 1.2375788688659668,
+      "learning_rate": 2.630073308002819e-05,
+      "loss": 0.5142,
+      "step": 535
+    },
+    {
+      "epoch": 1.35,
+      "grad_norm": 1.0503743886947632,
+      "learning_rate": 2.6218800804756232e-05,
+      "loss": 0.4584,
+      "step": 540
+    },
+    {
+      "epoch": 1.3625,
+      "grad_norm": 0.9696715474128723,
+      "learning_rate": 2.613610173935175e-05,
+      "loss": 0.4718,
+      "step": 545
+    },
+    {
+      "epoch": 1.375,
+      "grad_norm": 1.0960568189620972,
+      "learning_rate": 2.605264153618602e-05,
+      "loss": 0.5352,
+      "step": 550
+    },
+    {
+      "epoch": 1.3875,
+      "grad_norm": 1.0279024839401245,
+      "learning_rate": 2.5968425899653055e-05,
+      "loss": 0.4806,
+      "step": 555
+    },
+    {
+      "epoch": 1.4,
+      "grad_norm": 1.0534144639968872,
+      "learning_rate": 2.5883460585779754e-05,
+      "loss": 0.4874,
+      "step": 560
+    },
+    {
+      "epoch": 1.4125,
+      "grad_norm": 1.0914931297302246,
+      "learning_rate": 2.579775140183245e-05,
+      "loss": 0.5225,
+      "step": 565
+    },
+    {
+      "epoch": 1.425,
+      "grad_norm": 1.0744789838790894,
+      "learning_rate": 2.5711304205920015e-05,
+      "loss": 0.5056,
+      "step": 570
+    },
+    {
+      "epoch": 1.4375,
+      "grad_norm": 1.2198079824447632,
+      "learning_rate": 2.5624124906593458e-05,
+      "loss": 0.4751,
+      "step": 575
+    },
+    {
+      "epoch": 1.45,
+      "grad_norm": 1.1285591125488281,
+      "learning_rate": 2.5536219462442086e-05,
+      "loss": 0.4315,
+      "step": 580
+    },
+    {
+      "epoch": 1.4625,
+      "grad_norm": 1.1472607851028442,
+      "learning_rate": 2.5447593881686235e-05,
+      "loss": 0.4825,
+      "step": 585
+    },
+    {
+      "epoch": 1.475,
+      "grad_norm": 1.0363106727600098,
+      "learning_rate": 2.5358254221766624e-05,
+      "loss": 0.4582,
+      "step": 590
+    },
+    {
+      "epoch": 1.4875,
+      "grad_norm": 1.2606053352355957,
+      "learning_rate": 2.5268206588930332e-05,
+      "loss": 0.4674,
+      "step": 595
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 1.1823961734771729,
+      "learning_rate": 2.517745713781345e-05,
+      "loss": 0.5222,
+      "step": 600
+    },
+    {
+      "epoch": 1.5125,
+      "grad_norm": 1.1270991563796997,
+      "learning_rate": 2.508601207102042e-05,
+      "loss": 0.438,
+      "step": 605
+    },
+    {
+      "epoch": 1.525,
+      "grad_norm": 1.0307538509368896,
+      "learning_rate": 2.4993877638700083e-05,
+      "loss": 0.4528,
+      "step": 610
+    },
+    {
+      "epoch": 1.5375,
+      "grad_norm": 1.1684446334838867,
+      "learning_rate": 2.4901060138118502e-05,
+      "loss": 0.4359,
+      "step": 615
+    },
+    {
+      "epoch": 1.55,
+      "grad_norm": 1.1035212278366089,
+      "learning_rate": 2.480756591322854e-05,
+      "loss": 0.4552,
+      "step": 620
+    },
+    {
+      "epoch": 1.5625,
+      "grad_norm": 1.2371423244476318,
+      "learning_rate": 2.4713401354236282e-05,
+      "loss": 0.4246,
+      "step": 625
+    },
+    {
+      "epoch": 1.575,
+      "grad_norm": 1.1511696577072144,
+      "learning_rate": 2.4618572897164255e-05,
+      "loss": 0.4018,
+      "step": 630
+    },
+    {
+      "epoch": 1.5875,
+      "grad_norm": 1.25965416431427,
+      "learning_rate": 2.452308702341153e-05,
+      "loss": 0.4366,
+      "step": 635
+    },
+    {
+      "epoch": 1.6,
+      "grad_norm": 1.0590370893478394,
+      "learning_rate": 2.442695025931074e-05,
+      "loss": 0.4153,
+      "step": 640
+    },
+    {
+      "epoch": 1.6125,
+      "grad_norm": 1.0804214477539062,
+      "learning_rate": 2.433016917568201e-05,
+      "loss": 0.4147,
+      "step": 645
+    },
+    {
+      "epoch": 1.625,
+      "grad_norm": 1.0489362478256226,
+      "learning_rate": 2.4232750387383847e-05,
+      "loss": 0.4363,
+      "step": 650
+    },
+    {
+      "epoch": 1.6375,
+      "grad_norm": 1.198287010192871,
+      "learning_rate": 2.4134700552861018e-05,
+      "loss": 0.4347,
+      "step": 655
+    },
+    {
+      "epoch": 1.65,
+      "grad_norm": 1.1126651763916016,
+      "learning_rate": 2.403602637368948e-05,
+      "loss": 0.4434,
+      "step": 660
+    },
+    {
+      "epoch": 1.6625,
+      "grad_norm": 1.137837290763855,
+      "learning_rate": 2.3936734594118295e-05,
+      "loss": 0.4437,
+      "step": 665
+    },
+    {
+      "epoch": 1.675,
+      "grad_norm": 1.1391346454620361,
+      "learning_rate": 2.3836832000608704e-05,
+      "loss": 0.4306,
+      "step": 670
+    },
+    {
+      "epoch": 1.6875,
+      "grad_norm": 0.9699942469596863,
+      "learning_rate": 2.3736325421370275e-05,
+      "loss": 0.4575,
+      "step": 675
+    },
+    {
+      "epoch": 1.7,
+      "grad_norm": 1.0101770162582397,
+      "learning_rate": 2.3635221725894186e-05,
+      "loss": 0.4321,
+      "step": 680
+    },
+    {
+      "epoch": 1.7125,
+      "grad_norm": 1.0216741561889648,
+      "learning_rate": 2.3533527824483722e-05,
+      "loss": 0.4074,
+      "step": 685
+    },
+    {
+      "epoch": 1.725,
+      "grad_norm": 1.1822564601898193,
+      "learning_rate": 2.3431250667781958e-05,
+      "loss": 0.451,
+      "step": 690
+    },
+    {
+      "epoch": 1.7375,
+      "grad_norm": 1.1009029150009155,
+      "learning_rate": 2.33283972462967e-05,
+      "loss": 0.4171,
+      "step": 695
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 1.347333550453186,
+      "learning_rate": 2.3224974589922673e-05,
+      "loss": 0.4304,
+      "step": 700
+    },
+    {
+      "epoch": 1.7625,
+      "grad_norm": 1.4213443994522095,
+      "learning_rate": 2.312098976746107e-05,
+      "loss": 0.3918,
+      "step": 705
+    },
+    {
+      "epoch": 1.775,
+      "grad_norm": 1.077433466911316,
+      "learning_rate": 2.3016449886136372e-05,
+      "loss": 0.4047,
+      "step": 710
+    },
+    {
+      "epoch": 1.7875,
+      "grad_norm": 1.1451928615570068,
+      "learning_rate": 2.2911362091110596e-05,
+      "loss": 0.3928,
+      "step": 715
+    },
+    {
+      "epoch": 1.8,
+      "grad_norm": 1.272578477859497,
+      "learning_rate": 2.2805733564994948e-05,
+      "loss": 0.3776,
+      "step": 720
+    },
+    {
+      "epoch": 1.8125,
+      "grad_norm": 1.2507826089859009,
+      "learning_rate": 2.269957152735887e-05,
+      "loss": 0.3815,
+      "step": 725
+    },
+    {
+      "epoch": 1.825,
+      "grad_norm": 1.2830368280410767,
+      "learning_rate": 2.259288323423662e-05,
+      "loss": 0.384,
+      "step": 730
+    },
+    {
+      "epoch": 1.8375,
+      "grad_norm": 1.164024829864502,
+      "learning_rate": 2.2485675977631308e-05,
+      "loss": 0.3858,
+      "step": 735
+    },
+    {
+      "epoch": 1.85,
+      "grad_norm": 1.3388005495071411,
+      "learning_rate": 2.237795708501652e-05,
+      "loss": 0.4192,
+      "step": 740
+    },
+    {
+      "epoch": 1.8625,
+      "grad_norm": 0.9997522830963135,
+      "learning_rate": 2.2269733918835478e-05,
+      "loss": 0.3802,
+      "step": 745
+    },
+    {
+      "epoch": 1.875,
+      "grad_norm": 1.1285563707351685,
+      "learning_rate": 2.216101387599782e-05,
+      "loss": 0.3711,
+      "step": 750
+    },
+    {
+      "epoch": 1.8875,
+      "grad_norm": 1.1473147869110107,
+      "learning_rate": 2.205180438737407e-05,
+      "loss": 0.3749,
+      "step": 755
+    },
+    {
+      "epoch": 1.9,
+      "grad_norm": 1.2977720499038696,
+      "learning_rate": 2.1942112917287704e-05,
+      "loss": 0.3863,
+      "step": 760
+    },
+    {
+      "epoch": 1.9125,
+      "grad_norm": 1.165961503982544,
+      "learning_rate": 2.1831946963004988e-05,
+      "loss": 0.3753,
+      "step": 765
+    },
+    {
+      "epoch": 1.925,
+      "grad_norm": 1.1040446758270264,
+      "learning_rate": 2.1721314054222563e-05,
+      "loss": 0.3467,
+      "step": 770
+    },
+    {
+      "epoch": 1.9375,
+      "grad_norm": 1.0992324352264404,
+      "learning_rate": 2.1610221752552784e-05,
+      "loss": 0.3624,
+      "step": 775
+    },
+    {
+      "epoch": 1.95,
+      "grad_norm": 1.210777759552002,
+      "learning_rate": 2.1498677651006906e-05,
+      "loss": 0.3796,
+      "step": 780
+    },
+    {
+      "epoch": 1.9625,
+      "grad_norm": 1.1452407836914062,
+      "learning_rate": 2.138668937347609e-05,
+      "loss": 0.3686,
+      "step": 785
+    },
+    {
+      "epoch": 1.975,
+      "grad_norm": 1.2745158672332764,
+      "learning_rate": 2.1274264574210352e-05,
+      "loss": 0.3553,
+      "step": 790
+    },
+    {
+      "epoch": 1.9875,
+      "grad_norm": 1.21009361743927,
+      "learning_rate": 2.1161410937295385e-05,
+      "loss": 0.3413,
+      "step": 795
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.2560657262802124,
+      "learning_rate": 2.1048136176127357e-05,
+      "loss": 0.3242,
+      "step": 800
+    },
+    {
+      "epoch": 2.0125,
+      "grad_norm": 1.1402417421340942,
+      "learning_rate": 2.0934448032885724e-05,
+      "loss": 0.2778,
+      "step": 805
+    },
+    {
+      "epoch": 2.025,
+      "grad_norm": 1.2693945169448853,
+      "learning_rate": 2.0820354278004066e-05,
+      "loss": 0.2728,
+      "step": 810
+    },
+    {
+      "epoch": 2.0375,
+      "grad_norm": 1.238933801651001,
+      "learning_rate": 2.0705862709638972e-05,
+      "loss": 0.303,
+      "step": 815
+    },
+    {
+      "epoch": 2.05,
+      "grad_norm": 1.1587833166122437,
+      "learning_rate": 2.059098115313705e-05,
+      "loss": 0.2636,
+      "step": 820
+    },
+    {
+      "epoch": 2.0625,
+      "grad_norm": 1.290188193321228,
+      "learning_rate": 2.0475717460500088e-05,
+      "loss": 0.2925,
+      "step": 825
+    },
+    {
+      "epoch": 2.075,
+      "grad_norm": 1.2627977132797241,
+      "learning_rate": 2.0360079509848366e-05,
+      "loss": 0.2911,
+      "step": 830
+    },
+    {
+      "epoch": 2.0875,
+      "grad_norm": 1.1960002183914185,
+      "learning_rate": 2.02440752048822e-05,
+      "loss": 0.2887,
+      "step": 835
+    },
+    {
+      "epoch": 2.1,
+      "grad_norm": 1.150658130645752,
+      "learning_rate": 2.0127712474341747e-05,
+      "loss": 0.2759,
+      "step": 840
+    },
+    {
+      "epoch": 2.1125,
+      "grad_norm": 1.3438881635665894,
+      "learning_rate": 2.001099927146507e-05,
+      "loss": 0.2718,
+      "step": 845
+    },
+    {
+      "epoch": 2.125,
+      "grad_norm": 1.0232552289962769,
+      "learning_rate": 1.989394357344454e-05,
+      "loss": 0.2865,
+      "step": 850
+    },
+    {
+      "epoch": 2.1375,
+      "grad_norm": 1.052276611328125,
+      "learning_rate": 1.9776553380881634e-05,
+      "loss": 0.2561,
+      "step": 855
+    },
+    {
+      "epoch": 2.15,
+      "grad_norm": 1.2404303550720215,
+      "learning_rate": 1.965883671724007e-05,
+      "loss": 0.2631,
+      "step": 860
+    },
+    {
+      "epoch": 2.1625,
+      "grad_norm": 1.1204473972320557,
+      "learning_rate": 1.954080162829745e-05,
+      "loss": 0.2489,
+      "step": 865
+    },
+    {
+      "epoch": 2.175,
+      "grad_norm": 1.102751612663269,
+      "learning_rate": 1.9422456181595304e-05,
+      "loss": 0.2536,
+      "step": 870
+    },
+    {
+      "epoch": 2.1875,
+      "grad_norm": 1.3035495281219482,
+      "learning_rate": 1.9303808465887713e-05,
+      "loss": 0.2617,
+      "step": 875
+    },
+    {
+      "epoch": 2.2,
+      "grad_norm": 1.1419627666473389,
+      "learning_rate": 1.918486659058844e-05,
+      "loss": 0.269,
+      "step": 880
+    },
+    {
+      "epoch": 2.2125,
+      "grad_norm": 1.2295540571212769,
+      "learning_rate": 1.9065638685216667e-05,
+      "loss": 0.2596,
+      "step": 885
+    },
+    {
+      "epoch": 2.225,
+      "grad_norm": 1.2283310890197754,
+      "learning_rate": 1.8946132898841344e-05,
+      "loss": 0.2427,
+      "step": 890
+    },
+    {
+      "epoch": 2.2375,
+      "grad_norm": 1.084255576133728,
+      "learning_rate": 1.8826357399524227e-05,
+      "loss": 0.2959,
+      "step": 895
+    },
+    {
+      "epoch": 2.25,
+      "grad_norm": 1.203917384147644,
+      "learning_rate": 1.8706320373761594e-05,
+      "loss": 0.2551,
+      "step": 900
+    },
+    {
+      "epoch": 2.2625,
+      "grad_norm": 1.0413402318954468,
+      "learning_rate": 1.8586030025924697e-05,
+      "loss": 0.2429,
+      "step": 905
+    },
+    {
+      "epoch": 2.275,
+      "grad_norm": 1.2268816232681274,
+      "learning_rate": 1.8465494577699033e-05,
+      "loss": 0.2476,
+      "step": 910
+    },
+    {
+      "epoch": 2.2875,
+      "grad_norm": 1.2103034257888794,
+      "learning_rate": 1.8344722267522377e-05,
+      "loss": 0.2472,
+      "step": 915
+    },
+    {
+      "epoch": 2.3,
+      "grad_norm": 1.125713586807251,
+      "learning_rate": 1.822372135002169e-05,
+      "loss": 0.2632,
+      "step": 920
+    },
+    {
+      "epoch": 2.3125,
+      "grad_norm": 1.2902660369873047,
+      "learning_rate": 1.810250009544895e-05,
+      "loss": 0.2424,
+      "step": 925
+    },
+    {
+      "epoch": 2.325,
+      "grad_norm": 1.1748820543289185,
+      "learning_rate": 1.7981066789115897e-05,
+      "loss": 0.2302,
+      "step": 930
+    },
+    {
+      "epoch": 2.3375,
+      "grad_norm": 1.058660626411438,
+      "learning_rate": 1.7859429730827704e-05,
+      "loss": 0.2237,
+      "step": 935
+    },
+    {
+      "epoch": 2.35,
+      "grad_norm": 1.1528538465499878,
+      "learning_rate": 1.7737597234315727e-05,
+      "loss": 0.2575,
+      "step": 940
+    },
+    {
+      "epoch": 2.3625,
+      "grad_norm": 1.188664197921753,
+      "learning_rate": 1.7615577626669274e-05,
+      "loss": 0.2512,
+      "step": 945
+    },
+    {
+      "epoch": 2.375,
+      "grad_norm": 1.1694797277450562,
+      "learning_rate": 1.7493379247766444e-05,
+      "loss": 0.2382,
+      "step": 950
+    },
+    {
+      "epoch": 2.3875,
+      "grad_norm": 1.1770079135894775,
+      "learning_rate": 1.7371010449704097e-05,
+      "loss": 0.2181,
+      "step": 955
+    },
+    {
+      "epoch": 2.4,
+      "grad_norm": 1.051956295967102,
+      "learning_rate": 1.724847959622705e-05,
+      "loss": 0.2397,
+      "step": 960
+    },
+    {
+      "epoch": 2.4125,
+      "grad_norm": 1.1248499155044556,
+      "learning_rate": 1.7125795062156363e-05,
+      "loss": 0.2478,
+      "step": 965
+    },
+    {
+      "epoch": 2.425,
+      "grad_norm": 1.0435619354248047,
+      "learning_rate": 1.7002965232816977e-05,
+      "loss": 0.274,
+      "step": 970
+    },
+    {
+      "epoch": 2.4375,
+      "grad_norm": 1.1685163974761963,
+      "learning_rate": 1.6879998503464565e-05,
+      "loss": 0.2389,
+      "step": 975
+    },
+    {
+      "epoch": 2.45,
+      "grad_norm": 1.136615514755249,
+      "learning_rate": 1.675690327871174e-05,
+      "loss": 0.2471,
+      "step": 980
+    },
+    {
+      "epoch": 2.4625,
+      "grad_norm": 1.2618733644485474,
+      "learning_rate": 1.663368797195359e-05,
+      "loss": 0.2371,
+      "step": 985
+    },
+    {
+      "epoch": 2.475,
+      "grad_norm": 1.3753706216812134,
+      "learning_rate": 1.6510361004792665e-05,
+      "loss": 0.2581,
+      "step": 990
+    },
+    {
+      "epoch": 2.4875,
+      "grad_norm": 1.141377329826355,
+      "learning_rate": 1.6386930806463355e-05,
+      "loss": 0.2336,
+      "step": 995
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 1.3642939329147339,
+      "learning_rate": 1.6263405813255748e-05,
+      "loss": 0.232,
+      "step": 1000
+    },
+    {
+      "epoch": 2.5125,
+      "grad_norm": 1.2523542642593384,
+      "learning_rate": 1.6139794467939047e-05,
+      "loss": 0.2128,
+      "step": 1005
+    },
+    {
+      "epoch": 2.525,
+      "grad_norm": 1.3189752101898193,
+      "learning_rate": 1.601610521918451e-05,
+      "loss": 0.2218,
+      "step": 1010
+    },
+    {
+      "epoch": 2.5375,
+      "grad_norm": 1.1008399724960327,
+      "learning_rate": 1.5892346520987986e-05,
+      "loss": 0.2463,
+      "step": 1015
+    },
+    {
+      "epoch": 2.55,
+      "grad_norm": 1.1347867250442505,
+      "learning_rate": 1.5768526832092097e-05,
+      "loss": 0.211,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5625,
+      "grad_norm": 1.186482310295105,
+      "learning_rate": 1.564465461540811e-05,
+      "loss": 0.2208,
+      "step": 1025
+    },
+    {
+      "epoch": 2.575,
+      "grad_norm": 1.1044975519180298,
+      "learning_rate": 1.5520738337437485e-05,
+      "loss": 0.2125,
+      "step": 1030
+    },
+    {
+      "epoch": 2.5875,
+      "grad_norm": 1.147139310836792,
+      "learning_rate": 1.5396786467693216e-05,
+      "loss": 0.2038,
+      "step": 1035
+    },
+    {
+      "epoch": 2.6,
+      "grad_norm": 1.3628555536270142,
+      "learning_rate": 1.5272807478120954e-05,
+      "loss": 0.2299,
+      "step": 1040
+    },
+    {
+      "epoch": 2.6125,
+      "grad_norm": 1.1701704263687134,
+      "learning_rate": 1.5148809842519945e-05,
+      "loss": 0.2116,
+      "step": 1045
+    },
+    {
+      "epoch": 2.625,
+      "grad_norm": 0.9667600393295288,
+      "learning_rate": 1.5024802035963873e-05,
+      "loss": 0.2344,
+      "step": 1050
+    },
+    {
+      "epoch": 2.6375,
+      "grad_norm": 1.1534322500228882,
+      "learning_rate": 1.4900792534221589e-05,
+      "loss": 0.2231,
+      "step": 1055
+    },
+    {
+      "epoch": 2.65,
+      "grad_norm": 1.2170480489730835,
+      "learning_rate": 1.4776789813177808e-05,
+      "loss": 0.2243,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6625,
+      "grad_norm": 1.1053290367126465,
+      "learning_rate": 1.4652802348253793e-05,
+      "loss": 0.2144,
+      "step": 1065
+    },
+    {
+      "epoch": 2.675,
+      "grad_norm": 1.067043423652649,
+      "learning_rate": 1.4528838613828076e-05,
+      "loss": 0.2341,
+      "step": 1070
+    },
+    {
+      "epoch": 2.6875,
+      "grad_norm": 1.0644820928573608,
+      "learning_rate": 1.4404907082657238e-05,
+      "loss": 0.199,
+      "step": 1075
+    },
+    {
+      "epoch": 2.7,
+      "grad_norm": 1.0699652433395386,
+      "learning_rate": 1.4281016225296808e-05,
+      "loss": 0.1893,
+      "step": 1080
+    },
+    {
+      "epoch": 2.7125,
+      "grad_norm": 1.2279590368270874,
+      "learning_rate": 1.4157174509522318e-05,
+      "loss": 0.2027,
+      "step": 1085
+    },
+    {
+      "epoch": 2.725,
+      "grad_norm": 1.0308759212493896,
+      "learning_rate": 1.4033390399750534e-05,
+      "loss": 0.197,
+      "step": 1090
+    },
+    {
+      "epoch": 2.7375,
+      "grad_norm": 1.281794786453247,
+      "learning_rate": 1.3909672356460934e-05,
+      "loss": 0.2161,
+      "step": 1095
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 1.0185565948486328,
+      "learning_rate": 1.378602883561744e-05,
+      "loss": 0.2212,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7625,
+      "grad_norm": 1.130090594291687,
+      "learning_rate": 1.3662468288090446e-05,
+      "loss": 0.1898,
+      "step": 1105
+    },
+    {
+      "epoch": 2.775,
+      "grad_norm": 1.1406450271606445,
+      "learning_rate": 1.3538999159079253e-05,
+      "loss": 0.1764,
+      "step": 1110
+    },
+    {
+      "epoch": 2.7875,
+      "grad_norm": 1.2482119798660278,
+      "learning_rate": 1.3415629887534817e-05,
+      "loss": 0.1768,
+      "step": 1115
+    },
+    {
+      "epoch": 2.8,
+      "grad_norm": 1.1973145008087158,
+      "learning_rate": 1.3292368905582972e-05,
+      "loss": 0.1947,
+      "step": 1120
+    },
+    {
+      "epoch": 2.8125,
+      "grad_norm": 1.1229867935180664,
+      "learning_rate": 1.3169224637948099e-05,
+      "loss": 0.2081,
+      "step": 1125
+    },
+    {
+      "epoch": 2.825,
+      "grad_norm": 1.129728078842163,
+      "learning_rate": 1.3046205501377313e-05,
+      "loss": 0.1761,
+      "step": 1130
+    },
+    {
+      "epoch": 2.8375,
+      "grad_norm": 1.0157361030578613,
+      "learning_rate": 1.2923319904065187e-05,
+      "loss": 0.206,
+      "step": 1135
+    },
+    {
+      "epoch": 2.85,
+      "grad_norm": 1.4206640720367432,
+      "learning_rate": 1.280057624507906e-05,
+      "loss": 0.1954,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8625,
+      "grad_norm": 1.1708303689956665,
+      "learning_rate": 1.267798291378498e-05,
+      "loss": 0.1919,
+      "step": 1145
+    },
+    {
+      "epoch": 2.875,
+      "grad_norm": 0.9822798371315002,
+      "learning_rate": 1.2555548289274293e-05,
+      "loss": 0.1725,
+      "step": 1150
+    },
+    {
+      "epoch": 2.8875,
+      "grad_norm": 1.2971844673156738,
+      "learning_rate": 1.243328073979095e-05,
+      "loss": 0.2,
+      "step": 1155
+    },
+    {
+      "epoch": 2.9,
+      "grad_norm": 1.1322238445281982,
+      "learning_rate": 1.231118862215954e-05,
+      "loss": 0.1632,
+      "step": 1160
+    },
+    {
+      "epoch": 2.9125,
+      "grad_norm": 1.1313915252685547,
+      "learning_rate": 1.2189280281214128e-05,
+      "loss": 0.1942,
+      "step": 1165
+    },
+    {
+      "epoch": 2.925,
+      "grad_norm": 1.2760839462280273,
+      "learning_rate": 1.2067564049227884e-05,
+      "loss": 0.1802,
+      "step": 1170
+    },
+    {
+      "epoch": 2.9375,
+      "grad_norm": 1.1183810234069824,
+      "learning_rate": 1.1946048245343568e-05,
+      "loss": 0.2114,
+      "step": 1175
+    },
+    {
+      "epoch": 2.95,
+      "grad_norm": 1.0835254192352295,
+      "learning_rate": 1.1824741175004967e-05,
+      "loss": 0.1931,
+      "step": 1180
+    },
+    {
+      "epoch": 2.9625,
+      "grad_norm": 1.1179535388946533,
+      "learning_rate": 1.1703651129389203e-05,
+      "loss": 0.1968,
+      "step": 1185
+    },
+    {
+      "epoch": 2.975,
+      "grad_norm": 1.0869146585464478,
+      "learning_rate": 1.1582786384840045e-05,
+      "loss": 0.1676,
+      "step": 1190
+    },
+    {
+      "epoch": 2.9875,
+      "grad_norm": 1.2248408794403076,
+      "learning_rate": 1.1462155202302227e-05,
+      "loss": 0.1867,
+      "step": 1195
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.1031678915023804,
+      "learning_rate": 1.134176582675684e-05,
+      "loss": 0.1632,
+      "step": 1200
+    },
+    {
+      "epoch": 3.0125,
+      "grad_norm": 1.1282730102539062,
+      "learning_rate": 1.1221626486657781e-05,
+      "loss": 0.1466,
+      "step": 1205
+    },
+    {
+      "epoch": 3.025,
+      "grad_norm": 1.097623348236084,
+      "learning_rate": 1.1101745393369359e-05,
+      "loss": 0.1429,
+      "step": 1210
+    },
+    {
+      "epoch": 3.0375,
+      "grad_norm": 0.9805450439453125,
+      "learning_rate": 1.0982130740605056e-05,
+      "loss": 0.1303,
+      "step": 1215
+    },
+    {
+      "epoch": 3.05,
+      "grad_norm": 1.0525195598602295,
+      "learning_rate": 1.0862790703867501e-05,
+      "loss": 0.1452,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0625,
+      "grad_norm": 1.0902045965194702,
+      "learning_rate": 1.0743733439889675e-05,
+      "loss": 0.1131,
+      "step": 1225
+    },
+    {
+      "epoch": 3.075,
+      "grad_norm": 1.029802918434143,
+      "learning_rate": 1.0624967086077427e-05,
+      "loss": 0.1472,
+      "step": 1230
+    },
+    {
+      "epoch": 3.0875,
+      "grad_norm": 1.017521858215332,
+      "learning_rate": 1.0506499759953277e-05,
+      "loss": 0.1425,
+      "step": 1235
+    },
+    {
+      "epoch": 3.1,
+      "grad_norm": 1.1488157510757446,
+      "learning_rate": 1.038833955860159e-05,
+      "loss": 0.145,
+      "step": 1240
+    },
+    {
+      "epoch": 3.1125,
+      "grad_norm": 1.2664741277694702,
+      "learning_rate": 1.0270494558115179e-05,
+      "loss": 0.1332,
+      "step": 1245
+    },
+    {
+      "epoch": 3.125,
+      "grad_norm": 1.004277229309082,
+      "learning_rate": 1.0152972813043285e-05,
+      "loss": 0.1262,
+      "step": 1250
+    },
+    {
+      "epoch": 3.1375,
+      "grad_norm": 1.0977530479431152,
+      "learning_rate": 1.0035782355841076e-05,
+      "loss": 0.1371,
+      "step": 1255
+    },
+    {
+      "epoch": 3.15,
+      "grad_norm": 1.162043809890747,
+      "learning_rate": 9.91893119632063e-06,
+      "loss": 0.1425,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1625,
+      "grad_norm": 0.9929468035697937,
+      "learning_rate": 9.802427321103482e-06,
+      "loss": 0.125,
+      "step": 1265
+    },
+    {
+      "epoch": 3.175,
+      "grad_norm": 1.1833415031433105,
+      "learning_rate": 9.686278693074743e-06,
+      "loss": 0.1304,
+      "step": 1270
+    },
+    {
+      "epoch": 3.1875,
+      "grad_norm": 1.1010042428970337,
+      "learning_rate": 9.570493250838855e-06,
+      "loss": 0.1295,
+      "step": 1275
+    },
+    {
+      "epoch": 3.2,
+      "grad_norm": 0.8267909288406372,
+      "learning_rate": 9.455078908176995e-06,
+      "loss": 0.1206,
+      "step": 1280
+    },
+    {
+      "epoch": 3.2125,
+      "grad_norm": 0.8930376172065735,
+      "learning_rate": 9.340043553506164e-06,
+      "loss": 0.1186,
+      "step": 1285
+    },
+    {
+      "epoch": 3.225,
+      "grad_norm": 1.034348964691162,
+      "learning_rate": 9.225395049340055e-06,
+      "loss": 0.1409,
+      "step": 1290
+    },
+    {
+      "epoch": 3.2375,
+      "grad_norm": 1.0826964378356934,
+      "learning_rate": 9.11114123175163e-06,
+      "loss": 0.1231,
+      "step": 1295
+    },
+    {
+      "epoch": 3.25,
+      "grad_norm": 1.0175362825393677,
+      "learning_rate": 8.997289909837553e-06,
+      "loss": 0.1456,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2625,
+      "grad_norm": 1.0682710409164429,
+      "learning_rate": 8.88384886518445e-06,
+      "loss": 0.1298,
+      "step": 1305
+    },
+    {
+      "epoch": 3.275,
+      "grad_norm": 1.1369993686676025,
+      "learning_rate": 8.770825851337022e-06,
+      "loss": 0.1267,
+      "step": 1310
+    },
+    {
+      "epoch": 3.2875,
+      "grad_norm": 0.9807403683662415,
+      "learning_rate": 8.65822859326813e-06,
+      "loss": 0.1368,
+      "step": 1315
+    },
+    {
+      "epoch": 3.3,
+      "grad_norm": 1.0629318952560425,
+      "learning_rate": 8.546064786850807e-06,
+      "loss": 0.129,
+      "step": 1320
+    },
+    {
+      "epoch": 3.3125,
+      "grad_norm": 0.9835811257362366,
+      "learning_rate": 8.434342098332222e-06,
+      "loss": 0.1299,
+      "step": 1325
+    },
+    {
+      "epoch": 3.325,
+      "grad_norm": 1.0027168989181519,
+      "learning_rate": 8.32306816380973e-06,
+      "loss": 0.1241,
+      "step": 1330
+    },
+    {
+      "epoch": 3.3375,
+      "grad_norm": 1.2881447076797485,
+      "learning_rate": 8.212250588708943e-06,
+      "loss": 0.1173,
+      "step": 1335
+    },
+    {
+      "epoch": 3.35,
+      "grad_norm": 0.9684204459190369,
+      "learning_rate": 8.101896947263916e-06,
+      "loss": 0.1309,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3625,
+      "grad_norm": 0.9748321175575256,
+      "learning_rate": 7.992014781999454e-06,
+      "loss": 0.1367,
+      "step": 1345
+    },
+    {
+      "epoch": 3.375,
+      "grad_norm": 1.0800721645355225,
+      "learning_rate": 7.88261160321559e-06,
+      "loss": 0.1235,
+      "step": 1350
+    },
+    {
+      "epoch": 3.3875,
+      "grad_norm": 1.133420705795288,
+      "learning_rate": 7.773694888474268e-06,
+      "loss": 0.1144,
+      "step": 1355
+    },
+    {
+      "epoch": 3.4,
+      "grad_norm": 0.8932015895843506,
+      "learning_rate": 7.665272082088263e-06,
+      "loss": 0.1085,
+      "step": 1360
+    },
+    {
+      "epoch": 3.4125,
+      "grad_norm": 1.0200635194778442,
+      "learning_rate": 7.557350594612366e-06,
+      "loss": 0.1243,
+      "step": 1365
+    },
+    {
+      "epoch": 3.425,
+      "grad_norm": 1.0771801471710205,
+      "learning_rate": 7.449937802336896e-06,
+      "loss": 0.1148,
+      "step": 1370
+    },
+    {
+      "epoch": 3.4375,
+      "grad_norm": 1.0405155420303345,
+      "learning_rate": 7.343041046783511e-06,
+      "loss": 0.1098,
+      "step": 1375
+    },
+    {
+      "epoch": 3.45,
+      "grad_norm": 1.036603331565857,
+      "learning_rate": 7.236667634203463e-06,
+      "loss": 0.126,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4625,
+      "grad_norm": 0.9559935331344604,
+      "learning_rate": 7.1308248350782e-06,
+      "loss": 0.1278,
+      "step": 1385
+    },
+    {
+      "epoch": 3.475,
+      "grad_norm": 1.1467918157577515,
+      "learning_rate": 7.025519883622446e-06,
+      "loss": 0.1156,
+      "step": 1390
+    },
+    {
+      "epoch": 3.4875,
+      "grad_norm": 0.8284739255905151,
+      "learning_rate": 6.920759977289752e-06,
+      "loss": 0.1173,
+      "step": 1395
+    },
+    {
+      "epoch": 3.5,
+      "grad_norm": 0.8840073347091675,
+      "learning_rate": 6.816552276280558e-06,
+      "loss": 0.12,
+      "step": 1400
+    },
+    {
+      "epoch": 3.5125,
+      "grad_norm": 0.9407573342323303,
+      "learning_rate": 6.712903903052802e-06,
+      "loss": 0.1234,
+      "step": 1405
+    },
+    {
+      "epoch": 3.525,
+      "grad_norm": 1.0713573694229126,
+      "learning_rate": 6.609821941835117e-06,
+      "loss": 0.1222,
+      "step": 1410
+    },
+    {
+      "epoch": 3.5375,
+      "grad_norm": 1.1578398942947388,
+      "learning_rate": 6.50731343814262e-06,
+      "loss": 0.1246,
+      "step": 1415
+    },
+    {
+      "epoch": 3.55,
+      "grad_norm": 0.851016104221344,
+      "learning_rate": 6.405385398295372e-06,
+      "loss": 0.1193,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5625,
+      "grad_norm": 0.8984037041664124,
+      "learning_rate": 6.304044788939499e-06,
+      "loss": 0.1047,
+      "step": 1425
+    },
+    {
+      "epoch": 3.575,
+      "grad_norm": 0.8706745505332947,
+      "learning_rate": 6.2032985365710384e-06,
+      "loss": 0.1219,
+      "step": 1430
+    },
+    {
+      "epoch": 3.5875,
+      "grad_norm": 0.8364382386207581,
+      "learning_rate": 6.10315352706251e-06,
+      "loss": 0.1187,
+      "step": 1435
+    },
+    {
+      "epoch": 3.6,
+      "grad_norm": 1.0387898683547974,
+      "learning_rate": 6.0036166051922955e-06,
+      "loss": 0.1045,
+      "step": 1440
+    },
+    {
+      "epoch": 3.6125,
+      "grad_norm": 1.0929104089736938,
+      "learning_rate": 5.9046945741767686e-06,
+      "loss": 0.1054,
+      "step": 1445
+    },
+    {
+      "epoch": 3.625,
+      "grad_norm": 1.0163252353668213,
+      "learning_rate": 5.806394195205357e-06,
+      "loss": 0.1111,
+      "step": 1450
+    },
+    {
+      "epoch": 3.6375,
+      "grad_norm": 0.9308605194091797,
+      "learning_rate": 5.708722186978381e-06,
+      "loss": 0.1109,
+      "step": 1455
+    },
+    {
+      "epoch": 3.65,
+      "grad_norm": 1.1388764381408691,
+      "learning_rate": 5.611685225247865e-06,
+      "loss": 0.117,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6625,
+      "grad_norm": 1.0151485204696655,
+      "learning_rate": 5.515289942361242e-06,
+      "loss": 0.1192,
+      "step": 1465
+    },
+    {
+      "epoch": 3.675,
+      "grad_norm": 0.878633975982666,
+      "learning_rate": 5.419542926808051e-06,
+      "loss": 0.1154,
+      "step": 1470
+    },
+    {
+      "epoch": 3.6875,
+      "grad_norm": 1.3095695972442627,
+      "learning_rate": 5.3244507227696186e-06,
+      "loss": 0.1123,
+      "step": 1475
+    },
+    {
+      "epoch": 3.7,
+      "grad_norm": 1.0636982917785645,
+      "learning_rate": 5.230019829671778e-06,
+      "loss": 0.106,
+      "step": 1480
+    },
+    {
+      "epoch": 3.7125,
+      "grad_norm": 0.9404662251472473,
+      "learning_rate": 5.136256701740633e-06,
+      "loss": 0.1299,
+      "step": 1485
+    },
+    {
+      "epoch": 3.725,
+      "grad_norm": 1.0156927108764648,
+      "learning_rate": 5.043167747561427e-06,
+      "loss": 0.1169,
+      "step": 1490
+    },
+    {
+      "epoch": 3.7375,
+      "grad_norm": 0.8949950337409973,
+      "learning_rate": 4.950759329640527e-06,
+      "loss": 0.0994,
+      "step": 1495
+    },
+    {
+      "epoch": 3.75,
+      "grad_norm": 1.032090425491333,
+      "learning_rate": 4.859037763970542e-06,
+      "loss": 0.1196,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7625,
+      "grad_norm": 0.9257729053497314,
+      "learning_rate": 4.768009319598653e-06,
+      "loss": 0.119,
+      "step": 1505
+    },
+    {
+      "epoch": 3.775,
+      "grad_norm": 0.8927642107009888,
+      "learning_rate": 4.677680218198114e-06,
+      "loss": 0.1141,
+      "step": 1510
+    },
+    {
+      "epoch": 3.7875,
+      "grad_norm": 0.8775412440299988,
+      "learning_rate": 4.588056633643011e-06,
+      "loss": 0.1134,
+      "step": 1515
+    },
+    {
+      "epoch": 3.8,
+      "grad_norm": 1.0772385597229004,
+      "learning_rate": 4.499144691586303e-06,
+      "loss": 0.1032,
+      "step": 1520
+    },
+    {
+      "epoch": 3.8125,
+      "grad_norm": 1.0644309520721436,
+      "learning_rate": 4.410950469041121e-06,
+      "loss": 0.1157,
+      "step": 1525
+    },
+    {
+      "epoch": 3.825,
+      "grad_norm": 0.9224784970283508,
+      "learning_rate": 4.323479993965423e-06,
+      "loss": 0.1077,
+      "step": 1530
+    },
+    {
+      "epoch": 3.8375,
+      "grad_norm": 0.8817914128303528,
+      "learning_rate": 4.236739244849991e-06,
+      "loss": 0.1008,
+      "step": 1535
+    },
+    {
+      "epoch": 3.85,
+      "grad_norm": 0.9970330595970154,
+      "learning_rate": 4.150734150309804e-06,
+      "loss": 0.1072,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8625,
+      "grad_norm": 0.9122308492660522,
+      "learning_rate": 4.06547058867883e-06,
+      "loss": 0.1006,
+      "step": 1545
+    },
+    {
+      "epoch": 3.875,
+      "grad_norm": 0.9633174538612366,
+      "learning_rate": 3.9809543876082445e-06,
+      "loss": 0.1096,
+      "step": 1550
+    },
+    {
+      "epoch": 3.8875,
+      "grad_norm": 0.9273152351379395,
+      "learning_rate": 3.897191323668126e-06,
+      "loss": 0.0932,
+      "step": 1555
+    },
+    {
+      "epoch": 3.9,
+      "grad_norm": 0.9404770135879517,
+      "learning_rate": 3.8141871219526275e-06,
+      "loss": 0.1058,
+      "step": 1560
+    },
+    {
+      "epoch": 3.9125,
+      "grad_norm": 1.0484914779663086,
+      "learning_rate": 3.731947455688677e-06,
+      "loss": 0.1038,
+      "step": 1565
+    },
+    {
+      "epoch": 3.925,
+      "grad_norm": 0.9468738436698914,
+      "learning_rate": 3.6504779458482196e-06,
+      "loss": 0.0908,
+      "step": 1570
+    },
+    {
+      "epoch": 3.9375,
+      "grad_norm": 0.9193241000175476,
+      "learning_rate": 3.569784160764036e-06,
+      "loss": 0.1057,
+      "step": 1575
+    },
+    {
+      "epoch": 3.95,
+      "grad_norm": 1.0435298681259155,
+      "learning_rate": 3.489871615749145e-06,
+      "loss": 0.1101,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9625,
+      "grad_norm": 0.9926328063011169,
+      "learning_rate": 3.4107457727198465e-06,
+      "loss": 0.099,
+      "step": 1585
+    },
+    {
+      "epoch": 3.975,
+      "grad_norm": 0.9997310042381287,
+      "learning_rate": 3.332412039822406e-06,
+      "loss": 0.1033,
+      "step": 1590
+    },
+    {
+      "epoch": 3.9875,
+      "grad_norm": 0.9943178296089172,
+      "learning_rate": 3.254875771063412e-06,
+      "loss": 0.0923,
+      "step": 1595
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 0.7756332159042358,
+      "learning_rate": 3.1781422659438387e-06,
+      "loss": 0.0988,
+      "step": 1600
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 2000,
+  "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": 2.1884385599065948e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

29_128_e5_3e-5/checkpoint-1600/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:871812aa4697fc2cb08f60c4dde08e33ab05b528c39c3bc200d654ee41929393
+size 7736

29_128_e5_3e-5/checkpoint-1600/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)

29_128_e5_3e-5/checkpoint-2000/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

29_128_e5_3e-5/checkpoint-2000/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": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "v_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

29_128_e5_3e-5/checkpoint-2000/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fa650af5b4c4bfcecdadc6ac1af9cd5bcfd973bda7537b3a6dbf91564da22b2
+size 791751704

29_128_e5_3e-5/checkpoint-2000/latest

@@ -0,0 +1 @@
+global_step2000

29_128_e5_3e-5/checkpoint-2000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f8798d1ff8bd3ca93dda9483812bc6d8ca45fc468301c3f8acb94b5b2080884
+size 15920

29_128_e5_3e-5/checkpoint-2000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:255b4a0b5969a0ed21ef292346060b5b16975d83725402cf56dabb7807633b64
+size 15920

29_128_e5_3e-5/checkpoint-2000/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c524d9c6a2acbd719e89fe7645859044ea12192fe4be0ea79b27edbd0bc0704
+size 15920