Upload folder using huggingface_hub

35_128_e5_3e-5/checkpoint-1038/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

35_128_e5_3e-5/checkpoint-1038/adapter_config.json

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

35_128_e5_3e-5/checkpoint-1038/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71324a967edebfdfd24c5860b2f6f8e6b849d59c48d962ea03cbc03a52335624
+size 791751704

35_128_e5_3e-5/checkpoint-1038/latest

@@ -0,0 +1 @@
+global_step1038

35_128_e5_3e-5/checkpoint-1038/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e45cae7db344365fc0bc7eb090c51ab144bd569c80d9fbba2bff867f9bbffff
+size 15920

35_128_e5_3e-5/checkpoint-1038/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3292e30e37764c4f26f4bd3feeea11a938dd841d13877684b726b5ad255bfb32
+size 15920

35_128_e5_3e-5/checkpoint-1038/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ada1068f7c77c220ce486c7b3d1bac25ecdea4d5ab02ff925280cf1ebd08b1a4
+size 15920

35_128_e5_3e-5/checkpoint-1038/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5838ba64be215eba5225eed6f53e0188d89441a5c8a94cbeca0cde83a87a6cf5
+size 15920

35_128_e5_3e-5/checkpoint-1038/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1292a4d6d8a14d23157c63906927a240d27c839b600bc0e59d3911b9b1be88c1
+size 15920

35_128_e5_3e-5/checkpoint-1038/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57214967512aeec3704ab20b5b4b1d6133589c22aa649ae2f50bec48db171155
+size 15920

35_128_e5_3e-5/checkpoint-1038/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2dae40b35c079cfb2b804bfce4d6c9632c58c3bd53032ef1c004743256ef7a9
+size 15920

35_128_e5_3e-5/checkpoint-1038/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:790631d11ae3818a75c88e07128a0ca3417ec6e00c8c51d49e26e356ad794b4f
+size 15920

35_128_e5_3e-5/checkpoint-1038/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44ee8a7751dcdc2b400b152514c9bd2ea9bedf7af14f1fafb7eb08ef67be9a74
+size 1064

35_128_e5_3e-5/checkpoint-1038/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
+  }
+}

35_128_e5_3e-5/checkpoint-1038/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
+}

35_128_e5_3e-5/checkpoint-1038/trainer_state.json

@@ -0,0 +1,1483 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1038,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01447178002894356,
+      "grad_norm": 1.2414709329605103,
+      "learning_rate": 1.3793103448275862e-06,
+      "loss": 1.2575,
+      "step": 5
+    },
+    {
+      "epoch": 0.02894356005788712,
+      "grad_norm": 0.995053231716156,
+      "learning_rate": 3.103448275862069e-06,
+      "loss": 1.2964,
+      "step": 10
+    },
+    {
+      "epoch": 0.04341534008683068,
+      "grad_norm": 0.6870602369308472,
+      "learning_rate": 4.827586206896552e-06,
+      "loss": 1.2136,
+      "step": 15
+    },
+    {
+      "epoch": 0.05788712011577424,
+      "grad_norm": 0.7477986216545105,
+      "learning_rate": 6.551724137931035e-06,
+      "loss": 1.208,
+      "step": 20
+    },
+    {
+      "epoch": 0.0723589001447178,
+      "grad_norm": 0.503740131855011,
+      "learning_rate": 8.275862068965517e-06,
+      "loss": 1.262,
+      "step": 25
+    },
+    {
+      "epoch": 0.08683068017366136,
+      "grad_norm": 0.5463724732398987,
+      "learning_rate": 9.999999999999999e-06,
+      "loss": 1.1591,
+      "step": 30
+    },
+    {
+      "epoch": 0.10130246020260492,
+      "grad_norm": 0.5826733112335205,
+      "learning_rate": 1.1724137931034483e-05,
+      "loss": 1.1859,
+      "step": 35
+    },
+    {
+      "epoch": 0.11577424023154848,
+      "grad_norm": 0.478679895401001,
+      "learning_rate": 1.3448275862068966e-05,
+      "loss": 1.1185,
+      "step": 40
+    },
+    {
+      "epoch": 0.13024602026049203,
+      "grad_norm": 0.45855218172073364,
+      "learning_rate": 1.517241379310345e-05,
+      "loss": 1.0916,
+      "step": 45
+    },
+    {
+      "epoch": 0.1447178002894356,
+      "grad_norm": 0.4334068298339844,
+      "learning_rate": 1.6896551724137932e-05,
+      "loss": 1.1103,
+      "step": 50
+    },
+    {
+      "epoch": 0.15918958031837915,
+      "grad_norm": 0.49706199765205383,
+      "learning_rate": 1.8620689655172415e-05,
+      "loss": 1.1351,
+      "step": 55
+    },
+    {
+      "epoch": 0.1736613603473227,
+      "grad_norm": 0.5013290643692017,
+      "learning_rate": 2.0344827586206894e-05,
+      "loss": 1.1267,
+      "step": 60
+    },
+    {
+      "epoch": 0.18813314037626627,
+      "grad_norm": 0.4919586181640625,
+      "learning_rate": 2.206896551724138e-05,
+      "loss": 1.0824,
+      "step": 65
+    },
+    {
+      "epoch": 0.20260492040520983,
+      "grad_norm": 0.4955993890762329,
+      "learning_rate": 2.3793103448275862e-05,
+      "loss": 1.088,
+      "step": 70
+    },
+    {
+      "epoch": 0.2170767004341534,
+      "grad_norm": 0.47736915946006775,
+      "learning_rate": 2.5517241379310345e-05,
+      "loss": 1.097,
+      "step": 75
+    },
+    {
+      "epoch": 0.23154848046309695,
+      "grad_norm": 0.4829305112361908,
+      "learning_rate": 2.7241379310344827e-05,
+      "loss": 1.0582,
+      "step": 80
+    },
+    {
+      "epoch": 0.2460202604920405,
+      "grad_norm": 0.5232416391372681,
+      "learning_rate": 2.8965517241379313e-05,
+      "loss": 1.0447,
+      "step": 85
+    },
+    {
+      "epoch": 0.26049204052098407,
+      "grad_norm": 0.5268356800079346,
+      "learning_rate": 2.999989031547876e-05,
+      "loss": 1.0425,
+      "step": 90
+    },
+    {
+      "epoch": 0.27496382054992763,
+      "grad_norm": 0.6044050455093384,
+      "learning_rate": 2.9998656383036702e-05,
+      "loss": 0.9967,
+      "step": 95
+    },
+    {
+      "epoch": 0.2894356005788712,
+      "grad_norm": 0.5722519755363464,
+      "learning_rate": 2.9996051525662343e-05,
+      "loss": 0.9925,
+      "step": 100
+    },
+    {
+      "epoch": 0.30390738060781475,
+      "grad_norm": 0.5845656991004944,
+      "learning_rate": 2.9992075981447947e-05,
+      "loss": 0.9589,
+      "step": 105
+    },
+    {
+      "epoch": 0.3183791606367583,
+      "grad_norm": 0.6687687635421753,
+      "learning_rate": 2.9986730113770898e-05,
+      "loss": 0.9607,
+      "step": 110
+    },
+    {
+      "epoch": 0.33285094066570187,
+      "grad_norm": 0.5653572678565979,
+      "learning_rate": 2.9980014411260523e-05,
+      "loss": 0.9765,
+      "step": 115
+    },
+    {
+      "epoch": 0.3473227206946454,
+      "grad_norm": 0.6110051274299622,
+      "learning_rate": 2.9971929487753402e-05,
+      "loss": 0.9347,
+      "step": 120
+    },
+    {
+      "epoch": 0.361794500723589,
+      "grad_norm": 0.6042832136154175,
+      "learning_rate": 2.9962476082237285e-05,
+      "loss": 0.8882,
+      "step": 125
+    },
+    {
+      "epoch": 0.37626628075253254,
+      "grad_norm": 0.6558710932731628,
+      "learning_rate": 2.9951655058783517e-05,
+      "loss": 0.9611,
+      "step": 130
+    },
+    {
+      "epoch": 0.3907380607814761,
+      "grad_norm": 0.6648174524307251,
+      "learning_rate": 2.993946740646809e-05,
+      "loss": 0.9514,
+      "step": 135
+    },
+    {
+      "epoch": 0.40520984081041966,
+      "grad_norm": 0.6010047793388367,
+      "learning_rate": 2.992591423928121e-05,
+      "loss": 0.8977,
+      "step": 140
+    },
+    {
+      "epoch": 0.4196816208393632,
+      "grad_norm": 0.594023585319519,
+      "learning_rate": 2.99109967960255e-05,
+      "loss": 0.9169,
+      "step": 145
+    },
+    {
+      "epoch": 0.4341534008683068,
+      "grad_norm": 0.6303651928901672,
+      "learning_rate": 2.9894716440202756e-05,
+      "loss": 0.8779,
+      "step": 150
+    },
+    {
+      "epoch": 0.44862518089725034,
+      "grad_norm": 0.6532771587371826,
+      "learning_rate": 2.9877074659889316e-05,
+      "loss": 0.8661,
+      "step": 155
+    },
+    {
+      "epoch": 0.4630969609261939,
+      "grad_norm": 0.7434287071228027,
+      "learning_rate": 2.9858073067600054e-05,
+      "loss": 0.8627,
+      "step": 160
+    },
+    {
+      "epoch": 0.47756874095513746,
+      "grad_norm": 0.7219240069389343,
+      "learning_rate": 2.983771340014098e-05,
+      "loss": 0.8855,
+      "step": 165
+    },
+    {
+      "epoch": 0.492040520984081,
+      "grad_norm": 0.7225666046142578,
+      "learning_rate": 2.981599751845051e-05,
+      "loss": 0.8531,
+      "step": 170
+    },
+    {
+      "epoch": 0.5065123010130246,
+      "grad_norm": 0.7195663452148438,
+      "learning_rate": 2.9792927407429344e-05,
+      "loss": 0.8174,
+      "step": 175
+    },
+    {
+      "epoch": 0.5209840810419681,
+      "grad_norm": 0.7346758842468262,
+      "learning_rate": 2.976850517575906e-05,
+      "loss": 0.8491,
+      "step": 180
+    },
+    {
+      "epoch": 0.5354558610709117,
+      "grad_norm": 0.7614021897315979,
+      "learning_rate": 2.9742733055709366e-05,
+      "loss": 0.8102,
+      "step": 185
+    },
+    {
+      "epoch": 0.5499276410998553,
+      "grad_norm": 0.7299443483352661,
+      "learning_rate": 2.9715613402934064e-05,
+      "loss": 0.7749,
+      "step": 190
+    },
+    {
+      "epoch": 0.5643994211287988,
+      "grad_norm": 0.7258509993553162,
+      "learning_rate": 2.9687148696255737e-05,
+      "loss": 0.8408,
+      "step": 195
+    },
+    {
+      "epoch": 0.5788712011577424,
+      "grad_norm": 0.8355885148048401,
+      "learning_rate": 2.9657341537439176e-05,
+      "loss": 0.7793,
+      "step": 200
+    },
+    {
+      "epoch": 0.5933429811866859,
+      "grad_norm": 0.7875196933746338,
+      "learning_rate": 2.9626194650953563e-05,
+      "loss": 0.8168,
+      "step": 205
+    },
+    {
+      "epoch": 0.6078147612156295,
+      "grad_norm": 1.2102893590927124,
+      "learning_rate": 2.9593710883723465e-05,
+      "loss": 0.7554,
+      "step": 210
+    },
+    {
+      "epoch": 0.622286541244573,
+      "grad_norm": 0.9075624346733093,
+      "learning_rate": 2.95598932048686e-05,
+      "loss": 0.7782,
+      "step": 215
+    },
+    {
+      "epoch": 0.6367583212735166,
+      "grad_norm": 0.9814792275428772,
+      "learning_rate": 2.9524744705432446e-05,
+      "loss": 0.7679,
+      "step": 220
+    },
+    {
+      "epoch": 0.6512301013024602,
+      "grad_norm": 0.8673927187919617,
+      "learning_rate": 2.9488268598099735e-05,
+      "loss": 0.7131,
+      "step": 225
+    },
+    {
+      "epoch": 0.6657018813314037,
+      "grad_norm": 1.2335319519042969,
+      "learning_rate": 2.945046821690277e-05,
+      "loss": 0.7368,
+      "step": 230
+    },
+    {
+      "epoch": 0.6801736613603473,
+      "grad_norm": 1.0025559663772583,
+      "learning_rate": 2.9411347016916704e-05,
+      "loss": 0.7447,
+      "step": 235
+    },
+    {
+      "epoch": 0.6946454413892909,
+      "grad_norm": 0.8002074360847473,
+      "learning_rate": 2.9370908573943737e-05,
+      "loss": 0.7396,
+      "step": 240
+    },
+    {
+      "epoch": 0.7091172214182344,
+      "grad_norm": 0.7924606800079346,
+      "learning_rate": 2.9329156584186267e-05,
+      "loss": 0.7821,
+      "step": 245
+    },
+    {
+      "epoch": 0.723589001447178,
+      "grad_norm": 0.8294292092323303,
+      "learning_rate": 2.9286094863909053e-05,
+      "loss": 0.7546,
+      "step": 250
+    },
+    {
+      "epoch": 0.7380607814761215,
+      "grad_norm": 0.8875476121902466,
+      "learning_rate": 2.924172734909038e-05,
+      "loss": 0.7541,
+      "step": 255
+    },
+    {
+      "epoch": 0.7525325615050651,
+      "grad_norm": 1.0516374111175537,
+      "learning_rate": 2.919605809506233e-05,
+      "loss": 0.7339,
+      "step": 260
+    },
+    {
+      "epoch": 0.7670043415340086,
+      "grad_norm": 0.9758723378181458,
+      "learning_rate": 2.9149091276140066e-05,
+      "loss": 0.7286,
+      "step": 265
+    },
+    {
+      "epoch": 0.7814761215629522,
+      "grad_norm": 0.8721650838851929,
+      "learning_rate": 2.910083118524034e-05,
+      "loss": 0.7224,
+      "step": 270
+    },
+    {
+      "epoch": 0.7959479015918958,
+      "grad_norm": 0.8282304406166077,
+      "learning_rate": 2.9051282233489065e-05,
+      "loss": 0.6857,
+      "step": 275
+    },
+    {
+      "epoch": 0.8104196816208393,
+      "grad_norm": 0.9779431223869324,
+      "learning_rate": 2.900044894981813e-05,
+      "loss": 0.6938,
+      "step": 280
+    },
+    {
+      "epoch": 0.8248914616497829,
+      "grad_norm": 0.842149019241333,
+      "learning_rate": 2.894833598055147e-05,
+      "loss": 0.6996,
+      "step": 285
+    },
+    {
+      "epoch": 0.8393632416787264,
+      "grad_norm": 0.8657100200653076,
+      "learning_rate": 2.8894948088980338e-05,
+      "loss": 0.6343,
+      "step": 290
+    },
+    {
+      "epoch": 0.85383502170767,
+      "grad_norm": 0.8805454969406128,
+      "learning_rate": 2.8840290154927955e-05,
+      "loss": 0.6506,
+      "step": 295
+    },
+    {
+      "epoch": 0.8683068017366136,
+      "grad_norm": 0.8671079874038696,
+      "learning_rate": 2.878436717430346e-05,
+      "loss": 0.7051,
+      "step": 300
+    },
+    {
+      "epoch": 0.8827785817655571,
+      "grad_norm": 0.9610555768013,
+      "learning_rate": 2.8727184258645276e-05,
+      "loss": 0.6888,
+      "step": 305
+    },
+    {
+      "epoch": 0.8972503617945007,
+      "grad_norm": 0.9625582695007324,
+      "learning_rate": 2.8668746634653908e-05,
+      "loss": 0.6669,
+      "step": 310
+    },
+    {
+      "epoch": 0.9117221418234442,
+      "grad_norm": 0.8685387372970581,
+      "learning_rate": 2.860905964371418e-05,
+      "loss": 0.645,
+      "step": 315
+    },
+    {
+      "epoch": 0.9261939218523878,
+      "grad_norm": 1.1596406698226929,
+      "learning_rate": 2.8548128741407043e-05,
+      "loss": 0.6432,
+      "step": 320
+    },
+    {
+      "epoch": 0.9406657018813314,
+      "grad_norm": 0.9292768239974976,
+      "learning_rate": 2.8485959497010906e-05,
+      "loss": 0.6599,
+      "step": 325
+    },
+    {
+      "epoch": 0.9551374819102749,
+      "grad_norm": 1.0150357484817505,
+      "learning_rate": 2.8422557592992584e-05,
+      "loss": 0.5829,
+      "step": 330
+    },
+    {
+      "epoch": 0.9696092619392185,
+      "grad_norm": 1.0730253458023071,
+      "learning_rate": 2.83579288244879e-05,
+      "loss": 0.615,
+      "step": 335
+    },
+    {
+      "epoch": 0.984081041968162,
+      "grad_norm": 1.0182080268859863,
+      "learning_rate": 2.829207909877201e-05,
+      "loss": 0.6109,
+      "step": 340
+    },
+    {
+      "epoch": 0.9985528219971056,
+      "grad_norm": 0.9661368131637573,
+      "learning_rate": 2.8225014434719423e-05,
+      "loss": 0.5798,
+      "step": 345
+    },
+    {
+      "epoch": 1.0115774240231548,
+      "grad_norm": 1.1847275495529175,
+      "learning_rate": 2.8156740962253904e-05,
+      "loss": 0.5366,
+      "step": 350
+    },
+    {
+      "epoch": 1.0260492040520983,
+      "grad_norm": 1.029125690460205,
+      "learning_rate": 2.808726492178813e-05,
+      "loss": 0.579,
+      "step": 355
+    },
+    {
+      "epoch": 1.0405209840810419,
+      "grad_norm": 1.106046199798584,
+      "learning_rate": 2.801659266365335e-05,
+      "loss": 0.5351,
+      "step": 360
+    },
+    {
+      "epoch": 1.0549927641099854,
+      "grad_norm": 0.9720465540885925,
+      "learning_rate": 2.794473064751889e-05,
+      "loss": 0.5171,
+      "step": 365
+    },
+    {
+      "epoch": 1.069464544138929,
+      "grad_norm": 1.0641460418701172,
+      "learning_rate": 2.787168544180174e-05,
+      "loss": 0.4973,
+      "step": 370
+    },
+    {
+      "epoch": 1.0839363241678726,
+      "grad_norm": 1.0609314441680908,
+      "learning_rate": 2.7797463723066224e-05,
+      "loss": 0.5228,
+      "step": 375
+    },
+    {
+      "epoch": 1.0984081041968161,
+      "grad_norm": 1.1373050212860107,
+      "learning_rate": 2.7722072275413647e-05,
+      "loss": 0.5296,
+      "step": 380
+    },
+    {
+      "epoch": 1.1128798842257597,
+      "grad_norm": 1.0449209213256836,
+      "learning_rate": 2.7645517989862304e-05,
+      "loss": 0.5076,
+      "step": 385
+    },
+    {
+      "epoch": 1.1273516642547032,
+      "grad_norm": 0.9028365612030029,
+      "learning_rate": 2.7567807863717544e-05,
+      "loss": 0.5403,
+      "step": 390
+    },
+    {
+      "epoch": 1.1418234442836468,
+      "grad_norm": 0.9618139863014221,
+      "learning_rate": 2.748894899993225e-05,
+      "loss": 0.5063,
+      "step": 395
+    },
+    {
+      "epoch": 1.1562952243125904,
+      "grad_norm": 0.9099822044372559,
+      "learning_rate": 2.740894860645756e-05,
+      "loss": 0.4521,
+      "step": 400
+    },
+    {
+      "epoch": 1.170767004341534,
+      "grad_norm": 0.9606701135635376,
+      "learning_rate": 2.732781399558408e-05,
+      "loss": 0.4984,
+      "step": 405
+    },
+    {
+      "epoch": 1.1852387843704775,
+      "grad_norm": 1.0110416412353516,
+      "learning_rate": 2.7245552583273475e-05,
+      "loss": 0.494,
+      "step": 410
+    },
+    {
+      "epoch": 1.199710564399421,
+      "grad_norm": 0.947682797908783,
+      "learning_rate": 2.716217188848066e-05,
+      "loss": 0.4795,
+      "step": 415
+    },
+    {
+      "epoch": 1.2141823444283646,
+      "grad_norm": 0.9584565162658691,
+      "learning_rate": 2.707767953246654e-05,
+      "loss": 0.5109,
+      "step": 420
+    },
+    {
+      "epoch": 1.2286541244573081,
+      "grad_norm": 1.0691497325897217,
+      "learning_rate": 2.699208323810137e-05,
+      "loss": 0.4361,
+      "step": 425
+    },
+    {
+      "epoch": 1.2431259044862517,
+      "grad_norm": 0.9888468384742737,
+      "learning_rate": 2.6905390829158903e-05,
+      "loss": 0.4539,
+      "step": 430
+    },
+    {
+      "epoch": 1.2575976845151953,
+      "grad_norm": 1.3001909255981445,
+      "learning_rate": 2.6817610229601253e-05,
+      "loss": 0.5022,
+      "step": 435
+    },
+    {
+      "epoch": 1.2720694645441388,
+      "grad_norm": 1.2359240055084229,
+      "learning_rate": 2.6728749462854606e-05,
+      "loss": 0.4434,
+      "step": 440
+    },
+    {
+      "epoch": 1.2865412445730824,
+      "grad_norm": 1.2030582427978516,
+      "learning_rate": 2.6638816651075877e-05,
+      "loss": 0.4496,
+      "step": 445
+    },
+    {
+      "epoch": 1.301013024602026,
+      "grad_norm": 1.0069599151611328,
+      "learning_rate": 2.6547820014410312e-05,
+      "loss": 0.412,
+      "step": 450
+    },
+    {
+      "epoch": 1.3154848046309695,
+      "grad_norm": 1.0469765663146973,
+      "learning_rate": 2.645576787024013e-05,
+      "loss": 0.4558,
+      "step": 455
+    },
+    {
+      "epoch": 1.329956584659913,
+      "grad_norm": 1.2109603881835938,
+      "learning_rate": 2.6362668632424302e-05,
+      "loss": 0.4665,
+      "step": 460
+    },
+    {
+      "epoch": 1.3444283646888566,
+      "grad_norm": 0.938054084777832,
+      "learning_rate": 2.6268530810529483e-05,
+      "loss": 0.3953,
+      "step": 465
+    },
+    {
+      "epoch": 1.3589001447178002,
+      "grad_norm": 0.8923037052154541,
+      "learning_rate": 2.6173363009052225e-05,
+      "loss": 0.4289,
+      "step": 470
+    },
+    {
+      "epoch": 1.3733719247467437,
+      "grad_norm": 1.0760612487792969,
+      "learning_rate": 2.60771739266325e-05,
+      "loss": 0.4314,
+      "step": 475
+    },
+    {
+      "epoch": 1.3878437047756873,
+      "grad_norm": 1.0694377422332764,
+      "learning_rate": 2.597997235525858e-05,
+      "loss": 0.44,
+      "step": 480
+    },
+    {
+      "epoch": 1.4023154848046309,
+      "grad_norm": 0.8559190630912781,
+      "learning_rate": 2.5881767179463493e-05,
+      "loss": 0.4688,
+      "step": 485
+    },
+    {
+      "epoch": 1.4167872648335744,
+      "grad_norm": 1.0568691492080688,
+      "learning_rate": 2.578256737551286e-05,
+      "loss": 0.4462,
+      "step": 490
+    },
+    {
+      "epoch": 1.431259044862518,
+      "grad_norm": 1.139095425605774,
+      "learning_rate": 2.568238201058451e-05,
+      "loss": 0.4281,
+      "step": 495
+    },
+    {
+      "epoch": 1.4457308248914615,
+      "grad_norm": 1.0132529735565186,
+      "learning_rate": 2.558122024193968e-05,
+      "loss": 0.3677,
+      "step": 500
+    },
+    {
+      "epoch": 1.460202604920405,
+      "grad_norm": 0.9675812721252441,
+      "learning_rate": 2.5479091316086008e-05,
+      "loss": 0.4631,
+      "step": 505
+    },
+    {
+      "epoch": 1.4746743849493487,
+      "grad_norm": 1.1200926303863525,
+      "learning_rate": 2.5376004567932376e-05,
+      "loss": 0.4088,
+      "step": 510
+    },
+    {
+      "epoch": 1.4891461649782922,
+      "grad_norm": 1.3555203676223755,
+      "learning_rate": 2.527196941993569e-05,
+      "loss": 0.3786,
+      "step": 515
+    },
+    {
+      "epoch": 1.5036179450072358,
+      "grad_norm": 1.0300952196121216,
+      "learning_rate": 2.5166995381239593e-05,
+      "loss": 0.3828,
+      "step": 520
+    },
+    {
+      "epoch": 1.5180897250361793,
+      "grad_norm": 1.009810209274292,
+      "learning_rate": 2.5061092046805348e-05,
+      "loss": 0.4285,
+      "step": 525
+    },
+    {
+      "epoch": 1.532561505065123,
+      "grad_norm": 0.9876683354377747,
+      "learning_rate": 2.4954269096534798e-05,
+      "loss": 0.4121,
+      "step": 530
+    },
+    {
+      "epoch": 1.5470332850940665,
+      "grad_norm": 1.0438982248306274,
+      "learning_rate": 2.48465362943856e-05,
+      "loss": 0.394,
+      "step": 535
+    },
+    {
+      "epoch": 1.56150506512301,
+      "grad_norm": 1.14638352394104,
+      "learning_rate": 2.4737903487478785e-05,
+      "loss": 0.4183,
+      "step": 540
+    },
+    {
+      "epoch": 1.5759768451519536,
+      "grad_norm": 1.0417720079421997,
+      "learning_rate": 2.462838060519865e-05,
+      "loss": 0.3917,
+      "step": 545
+    },
+    {
+      "epoch": 1.5904486251808971,
+      "grad_norm": 1.1312050819396973,
+      "learning_rate": 2.4517977658285247e-05,
+      "loss": 0.3921,
+      "step": 550
+    },
+    {
+      "epoch": 1.6049204052098407,
+      "grad_norm": 0.9650654196739197,
+      "learning_rate": 2.4406704737919317e-05,
+      "loss": 0.3452,
+      "step": 555
+    },
+    {
+      "epoch": 1.6193921852387843,
+      "grad_norm": 0.9657004475593567,
+      "learning_rate": 2.429457201479995e-05,
+      "loss": 0.4089,
+      "step": 560
+    },
+    {
+      "epoch": 1.6338639652677278,
+      "grad_norm": 1.0962070226669312,
+      "learning_rate": 2.4181589738214945e-05,
+      "loss": 0.3596,
+      "step": 565
+    },
+    {
+      "epoch": 1.6483357452966714,
+      "grad_norm": 1.0828574895858765,
+      "learning_rate": 2.4067768235103985e-05,
+      "loss": 0.3542,
+      "step": 570
+    },
+    {
+      "epoch": 1.662807525325615,
+      "grad_norm": 0.9871302843093872,
+      "learning_rate": 2.3953117909114746e-05,
+      "loss": 0.3368,
+      "step": 575
+    },
+    {
+      "epoch": 1.6772793053545585,
+      "grad_norm": 0.9643537402153015,
+      "learning_rate": 2.3837649239651934e-05,
+      "loss": 0.3755,
+      "step": 580
+    },
+    {
+      "epoch": 1.691751085383502,
+      "grad_norm": 1.0887973308563232,
+      "learning_rate": 2.3721372780919454e-05,
+      "loss": 0.3198,
+      "step": 585
+    },
+    {
+      "epoch": 1.7062228654124456,
+      "grad_norm": 1.0346442461013794,
+      "learning_rate": 2.360429916095574e-05,
+      "loss": 0.3677,
+      "step": 590
+    },
+    {
+      "epoch": 1.7206946454413892,
+      "grad_norm": 1.3537055253982544,
+      "learning_rate": 2.3486439080662283e-05,
+      "loss": 0.4023,
+      "step": 595
+    },
+    {
+      "epoch": 1.7351664254703327,
+      "grad_norm": 0.938713788986206,
+      "learning_rate": 2.3367803312825556e-05,
+      "loss": 0.352,
+      "step": 600
+    },
+    {
+      "epoch": 1.7496382054992763,
+      "grad_norm": 1.1569691896438599,
+      "learning_rate": 2.324840270113234e-05,
+      "loss": 0.3667,
+      "step": 605
+    },
+    {
+      "epoch": 1.7641099855282198,
+      "grad_norm": 1.043257713317871,
+      "learning_rate": 2.3128248159178583e-05,
+      "loss": 0.3612,
+      "step": 610
+    },
+    {
+      "epoch": 1.7785817655571634,
+      "grad_norm": 1.0356690883636475,
+      "learning_rate": 2.3007350669471866e-05,
+      "loss": 0.3484,
+      "step": 615
+    },
+    {
+      "epoch": 1.793053545586107,
+      "grad_norm": 1.0480122566223145,
+      "learning_rate": 2.2885721282427528e-05,
+      "loss": 0.3147,
+      "step": 620
+    },
+    {
+      "epoch": 1.8075253256150505,
+      "grad_norm": 1.1496782302856445,
+      "learning_rate": 2.2763371115358695e-05,
+      "loss": 0.3543,
+      "step": 625
+    },
+    {
+      "epoch": 1.821997105643994,
+      "grad_norm": 1.021693229675293,
+      "learning_rate": 2.264031135146006e-05,
+      "loss": 0.3454,
+      "step": 630
+    },
+    {
+      "epoch": 1.8364688856729376,
+      "grad_norm": 1.196956753730774,
+      "learning_rate": 2.2516553238785725e-05,
+      "loss": 0.3422,
+      "step": 635
+    },
+    {
+      "epoch": 1.8509406657018812,
+      "grad_norm": 1.0382981300354004,
+      "learning_rate": 2.239210808922109e-05,
+      "loss": 0.3425,
+      "step": 640
+    },
+    {
+      "epoch": 1.8654124457308248,
+      "grad_norm": 1.1502225399017334,
+      "learning_rate": 2.226698727744893e-05,
+      "loss": 0.3152,
+      "step": 645
+    },
+    {
+      "epoch": 1.8798842257597683,
+      "grad_norm": 1.0866633653640747,
+      "learning_rate": 2.2141202239909678e-05,
+      "loss": 0.3472,
+      "step": 650
+    },
+    {
+      "epoch": 1.8943560057887119,
+      "grad_norm": 1.0475261211395264,
+      "learning_rate": 2.2014764473756125e-05,
+      "loss": 0.3328,
+      "step": 655
+    },
+    {
+      "epoch": 1.9088277858176554,
+      "grad_norm": 1.0879319906234741,
+      "learning_rate": 2.1887685535802532e-05,
+      "loss": 0.3371,
+      "step": 660
+    },
+    {
+      "epoch": 1.923299565846599,
+      "grad_norm": 1.155945897102356,
+      "learning_rate": 2.1759977041468306e-05,
+      "loss": 0.3394,
+      "step": 665
+    },
+    {
+      "epoch": 1.9377713458755426,
+      "grad_norm": 1.0616649389266968,
+      "learning_rate": 2.1631650663716313e-05,
+      "loss": 0.2906,
+      "step": 670
+    },
+    {
+      "epoch": 1.9522431259044861,
+      "grad_norm": 1.1703593730926514,
+      "learning_rate": 2.1502718131985916e-05,
+      "loss": 0.3032,
+      "step": 675
+    },
+    {
+      "epoch": 1.9667149059334297,
+      "grad_norm": 1.0248326063156128,
+      "learning_rate": 2.1373191231120904e-05,
+      "loss": 0.2967,
+      "step": 680
+    },
+    {
+      "epoch": 1.9811866859623732,
+      "grad_norm": 1.1739745140075684,
+      "learning_rate": 2.1243081800292278e-05,
+      "loss": 0.3183,
+      "step": 685
+    },
+    {
+      "epoch": 1.9956584659913168,
+      "grad_norm": 0.9827852249145508,
+      "learning_rate": 2.1112401731916142e-05,
+      "loss": 0.3146,
+      "step": 690
+    },
+    {
+      "epoch": 2.008683068017366,
+      "grad_norm": 0.925191342830658,
+      "learning_rate": 2.0981162970566676e-05,
+      "loss": 0.2637,
+      "step": 695
+    },
+    {
+      "epoch": 2.0231548480463095,
+      "grad_norm": 0.987099289894104,
+      "learning_rate": 2.0849377511884396e-05,
+      "loss": 0.257,
+      "step": 700
+    },
+    {
+      "epoch": 2.037626628075253,
+      "grad_norm": 1.0751880407333374,
+      "learning_rate": 2.0717057401479672e-05,
+      "loss": 0.2402,
+      "step": 705
+    },
+    {
+      "epoch": 2.0520984081041966,
+      "grad_norm": 0.9453009366989136,
+      "learning_rate": 2.0584214733831765e-05,
+      "loss": 0.2241,
+      "step": 710
+    },
+    {
+      "epoch": 2.06657018813314,
+      "grad_norm": 0.9541075229644775,
+      "learning_rate": 2.0450861651183305e-05,
+      "loss": 0.2527,
+      "step": 715
+    },
+    {
+      "epoch": 2.0810419681620838,
+      "grad_norm": 1.1164172887802124,
+      "learning_rate": 2.0317010342430507e-05,
+      "loss": 0.2389,
+      "step": 720
+    },
+    {
+      "epoch": 2.0955137481910273,
+      "grad_norm": 1.2189607620239258,
+      "learning_rate": 2.0182673042009014e-05,
+      "loss": 0.2173,
+      "step": 725
+    },
+    {
+      "epoch": 2.109985528219971,
+      "grad_norm": 0.9571680426597595,
+      "learning_rate": 2.0047862028775665e-05,
+      "loss": 0.241,
+      "step": 730
+    },
+    {
+      "epoch": 2.1244573082489144,
+      "grad_norm": 0.9751089811325073,
+      "learning_rate": 1.9912589624886163e-05,
+      "loss": 0.224,
+      "step": 735
+    },
+    {
+      "epoch": 2.138929088277858,
+      "grad_norm": 1.1041796207427979,
+      "learning_rate": 1.9776868194668774e-05,
+      "loss": 0.2474,
+      "step": 740
+    },
+    {
+      "epoch": 2.1534008683068016,
+      "grad_norm": 1.0402464866638184,
+      "learning_rate": 1.9640710143494206e-05,
+      "loss": 0.2204,
+      "step": 745
+    },
+    {
+      "epoch": 2.167872648335745,
+      "grad_norm": 1.0665066242218018,
+      "learning_rate": 1.9504127916641704e-05,
+      "loss": 0.2359,
+      "step": 750
+    },
+    {
+      "epoch": 2.1823444283646887,
+      "grad_norm": 1.169087290763855,
+      "learning_rate": 1.936713399816153e-05,
+      "loss": 0.2123,
+      "step": 755
+    },
+    {
+      "epoch": 2.1968162083936322,
+      "grad_norm": 0.9681061506271362,
+      "learning_rate": 1.9229740909733883e-05,
+      "loss": 0.2253,
+      "step": 760
+    },
+    {
+      "epoch": 2.211287988422576,
+      "grad_norm": 1.0454251766204834,
+      "learning_rate": 1.9091961209524333e-05,
+      "loss": 0.2344,
+      "step": 765
+    },
+    {
+      "epoch": 2.2257597684515193,
+      "grad_norm": 1.1429376602172852,
+      "learning_rate": 1.8953807491036015e-05,
+      "loss": 0.2307,
+      "step": 770
+    },
+    {
+      "epoch": 2.240231548480463,
+      "grad_norm": 0.9582192897796631,
+      "learning_rate": 1.8815292381958534e-05,
+      "loss": 0.1943,
+      "step": 775
+    },
+    {
+      "epoch": 2.2547033285094065,
+      "grad_norm": 1.0701148509979248,
+      "learning_rate": 1.8676428543013733e-05,
+      "loss": 0.2386,
+      "step": 780
+    },
+    {
+      "epoch": 2.26917510853835,
+      "grad_norm": 0.897097110748291,
+      "learning_rate": 1.853722866679848e-05,
+      "loss": 0.2343,
+      "step": 785
+    },
+    {
+      "epoch": 2.2836468885672936,
+      "grad_norm": 1.074778437614441,
+      "learning_rate": 1.8397705476624513e-05,
+      "loss": 0.2285,
+      "step": 790
+    },
+    {
+      "epoch": 2.298118668596237,
+      "grad_norm": 1.1128685474395752,
+      "learning_rate": 1.8257871725355503e-05,
+      "loss": 0.2283,
+      "step": 795
+    },
+    {
+      "epoch": 2.3125904486251807,
+      "grad_norm": 1.0342803001403809,
+      "learning_rate": 1.8117740194241393e-05,
+      "loss": 0.2356,
+      "step": 800
+    },
+    {
+      "epoch": 2.3270622286541243,
+      "grad_norm": 1.1430531740188599,
+      "learning_rate": 1.7977323691750135e-05,
+      "loss": 0.2126,
+      "step": 805
+    },
+    {
+      "epoch": 2.341534008683068,
+      "grad_norm": 1.017957091331482,
+      "learning_rate": 1.7836635052396998e-05,
+      "loss": 0.2116,
+      "step": 810
+    },
+    {
+      "epoch": 2.3560057887120114,
+      "grad_norm": 1.1819818019866943,
+      "learning_rate": 1.769568713557141e-05,
+      "loss": 0.2317,
+      "step": 815
+    },
+    {
+      "epoch": 2.370477568740955,
+      "grad_norm": 1.094393014907837,
+      "learning_rate": 1.755449282436159e-05,
+      "loss": 0.208,
+      "step": 820
+    },
+    {
+      "epoch": 2.384949348769899,
+      "grad_norm": 1.2002899646759033,
+      "learning_rate": 1.7413065024376977e-05,
+      "loss": 0.2103,
+      "step": 825
+    },
+    {
+      "epoch": 2.399421128798842,
+      "grad_norm": 0.8949219584465027,
+      "learning_rate": 1.7271416662568652e-05,
+      "loss": 0.1954,
+      "step": 830
+    },
+    {
+      "epoch": 2.413892908827786,
+      "grad_norm": 1.070077657699585,
+      "learning_rate": 1.7129560686047725e-05,
+      "loss": 0.2238,
+      "step": 835
+    },
+    {
+      "epoch": 2.428364688856729,
+      "grad_norm": 1.0196245908737183,
+      "learning_rate": 1.6987510060901964e-05,
+      "loss": 0.2154,
+      "step": 840
+    },
+    {
+      "epoch": 2.442836468885673,
+      "grad_norm": 1.1058320999145508,
+      "learning_rate": 1.684527777101063e-05,
+      "loss": 0.2009,
+      "step": 845
+    },
+    {
+      "epoch": 2.4573082489146163,
+      "grad_norm": 0.9543981552124023,
+      "learning_rate": 1.6702876816857737e-05,
+      "loss": 0.1815,
+      "step": 850
+    },
+    {
+      "epoch": 2.4717800289435603,
+      "grad_norm": 1.2461034059524536,
+      "learning_rate": 1.6560320214343733e-05,
+      "loss": 0.1965,
+      "step": 855
+    },
+    {
+      "epoch": 2.4862518089725034,
+      "grad_norm": 0.9611772298812866,
+      "learning_rate": 1.641762099359583e-05,
+      "loss": 0.2176,
+      "step": 860
+    },
+    {
+      "epoch": 2.5007235890014474,
+      "grad_norm": 0.9620625972747803,
+      "learning_rate": 1.6274792197776996e-05,
+      "loss": 0.1894,
+      "step": 865
+    },
+    {
+      "epoch": 2.5151953690303905,
+      "grad_norm": 1.0272494554519653,
+      "learning_rate": 1.6131846881893777e-05,
+      "loss": 0.2037,
+      "step": 870
+    },
+    {
+      "epoch": 2.5296671490593345,
+      "grad_norm": 1.015389323234558,
+      "learning_rate": 1.5988798111603028e-05,
+      "loss": 0.2076,
+      "step": 875
+    },
+    {
+      "epoch": 2.5441389290882777,
+      "grad_norm": 1.0375982522964478,
+      "learning_rate": 1.584565896201765e-05,
+      "loss": 0.2029,
+      "step": 880
+    },
+    {
+      "epoch": 2.5586107091172217,
+      "grad_norm": 1.0840239524841309,
+      "learning_rate": 1.570244251651152e-05,
+      "loss": 0.1713,
+      "step": 885
+    },
+    {
+      "epoch": 2.5730824891461648,
+      "grad_norm": 0.8886203765869141,
+      "learning_rate": 1.5559161865523587e-05,
+      "loss": 0.1963,
+      "step": 890
+    },
+    {
+      "epoch": 2.587554269175109,
+      "grad_norm": 0.8905720710754395,
+      "learning_rate": 1.5415830105361403e-05,
+      "loss": 0.1914,
+      "step": 895
+    },
+    {
+      "epoch": 2.602026049204052,
+      "grad_norm": 1.0297545194625854,
+      "learning_rate": 1.527246033700405e-05,
+      "loss": 0.1928,
+      "step": 900
+    },
+    {
+      "epoch": 2.616497829232996,
+      "grad_norm": 0.8823828101158142,
+      "learning_rate": 1.512906566490469e-05,
+      "loss": 0.1775,
+      "step": 905
+    },
+    {
+      "epoch": 2.630969609261939,
+      "grad_norm": 1.3193711042404175,
+      "learning_rate": 1.4985659195792757e-05,
+      "loss": 0.1738,
+      "step": 910
+    },
+    {
+      "epoch": 2.645441389290883,
+      "grad_norm": 1.008651614189148,
+      "learning_rate": 1.4842254037475971e-05,
+      "loss": 0.2031,
+      "step": 915
+    },
+    {
+      "epoch": 2.659913169319826,
+      "grad_norm": 1.1403210163116455,
+      "learning_rate": 1.469886329764225e-05,
+      "loss": 0.1993,
+      "step": 920
+    },
+    {
+      "epoch": 2.67438494934877,
+      "grad_norm": 0.9636655449867249,
+      "learning_rate": 1.4555500082661603e-05,
+      "loss": 0.1639,
+      "step": 925
+    },
+    {
+      "epoch": 2.6888567293777133,
+      "grad_norm": 1.0645697116851807,
+      "learning_rate": 1.4412177496388196e-05,
+      "loss": 0.2048,
+      "step": 930
+    },
+    {
+      "epoch": 2.7033285094066573,
+      "grad_norm": 1.0411231517791748,
+      "learning_rate": 1.426890863896258e-05,
+      "loss": 0.1927,
+      "step": 935
+    },
+    {
+      "epoch": 2.7178002894356004,
+      "grad_norm": 1.0877488851547241,
+      "learning_rate": 1.4125706605614333e-05,
+      "loss": 0.1782,
+      "step": 940
+    },
+    {
+      "epoch": 2.7322720694645444,
+      "grad_norm": 1.1112526655197144,
+      "learning_rate": 1.3982584485465111e-05,
+      "loss": 0.1844,
+      "step": 945
+    },
+    {
+      "epoch": 2.7467438494934875,
+      "grad_norm": 0.9065836668014526,
+      "learning_rate": 1.3839555360332223e-05,
+      "loss": 0.1746,
+      "step": 950
+    },
+    {
+      "epoch": 2.7612156295224315,
+      "grad_norm": 1.3411282300949097,
+      "learning_rate": 1.3696632303532955e-05,
+      "loss": 0.1959,
+      "step": 955
+    },
+    {
+      "epoch": 2.7756874095513746,
+      "grad_norm": 0.9463226795196533,
+      "learning_rate": 1.3553828378689597e-05,
+      "loss": 0.165,
+      "step": 960
+    },
+    {
+      "epoch": 2.7901591895803186,
+      "grad_norm": 1.0703129768371582,
+      "learning_rate": 1.3411156638535422e-05,
+      "loss": 0.168,
+      "step": 965
+    },
+    {
+      "epoch": 2.8046309696092617,
+      "grad_norm": 1.1796904802322388,
+      "learning_rate": 1.3268630123721578e-05,
+      "loss": 0.1578,
+      "step": 970
+    },
+    {
+      "epoch": 2.8191027496382057,
+      "grad_norm": 1.1187974214553833,
+      "learning_rate": 1.3126261861625163e-05,
+      "loss": 0.1682,
+      "step": 975
+    },
+    {
+      "epoch": 2.833574529667149,
+      "grad_norm": 0.8996317982673645,
+      "learning_rate": 1.2984064865158473e-05,
+      "loss": 0.1711,
+      "step": 980
+    },
+    {
+      "epoch": 2.848046309696093,
+      "grad_norm": 0.8830468058586121,
+      "learning_rate": 1.2842052131579584e-05,
+      "loss": 0.1707,
+      "step": 985
+    },
+    {
+      "epoch": 2.862518089725036,
+      "grad_norm": 1.0437378883361816,
+      "learning_rate": 1.2700236641304356e-05,
+      "loss": 0.1688,
+      "step": 990
+    },
+    {
+      "epoch": 2.87698986975398,
+      "grad_norm": 1.0894595384597778,
+      "learning_rate": 1.2558631356719976e-05,
+      "loss": 0.1673,
+      "step": 995
+    },
+    {
+      "epoch": 2.891461649782923,
+      "grad_norm": 1.1124073266983032,
+      "learning_rate": 1.2417249221000169e-05,
+      "loss": 0.154,
+      "step": 1000
+    },
+    {
+      "epoch": 2.905933429811867,
+      "grad_norm": 0.9627780318260193,
+      "learning_rate": 1.2276103156922141e-05,
+      "loss": 0.1383,
+      "step": 1005
+    },
+    {
+      "epoch": 2.92040520984081,
+      "grad_norm": 0.9764670133590698,
+      "learning_rate": 1.2135206065685403e-05,
+      "loss": 0.1812,
+      "step": 1010
+    },
+    {
+      "epoch": 2.934876989869754,
+      "grad_norm": 0.8468527793884277,
+      "learning_rate": 1.199457082573255e-05,
+      "loss": 0.1721,
+      "step": 1015
+    },
+    {
+      "epoch": 2.9493487698986973,
+      "grad_norm": 1.1576621532440186,
+      "learning_rate": 1.185421029157214e-05,
+      "loss": 0.1804,
+      "step": 1020
+    },
+    {
+      "epoch": 2.9638205499276413,
+      "grad_norm": 1.082723617553711,
+      "learning_rate": 1.1714137292603738e-05,
+      "loss": 0.1693,
+      "step": 1025
+    },
+    {
+      "epoch": 2.9782923299565844,
+      "grad_norm": 1.0707679986953735,
+      "learning_rate": 1.1574364631945295e-05,
+      "loss": 0.156,
+      "step": 1030
+    },
+    {
+      "epoch": 2.9927641099855284,
+      "grad_norm": 1.0685093402862549,
+      "learning_rate": 1.1434905085262873e-05,
+      "loss": 0.1578,
+      "step": 1035
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1730,
+  "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": 1.5632361777696604e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

35_128_e5_3e-5/checkpoint-1038/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1380162c4a3d25f180c1ed514fe3a9a55a6f17c202b64210412030c30a27b748
+size 7736

35_128_e5_3e-5/checkpoint-1038/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)

35_128_e5_3e-5/checkpoint-1384/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

35_128_e5_3e-5/checkpoint-1384/adapter_config.json

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

35_128_e5_3e-5/checkpoint-1384/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4929da680dedef768459df3eabd1dbe3055bdffa2a22e2f0a6a766eb7f00f05
+size 791751704

35_128_e5_3e-5/checkpoint-1384/latest

@@ -0,0 +1 @@
+global_step1384

35_128_e5_3e-5/checkpoint-1384/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a396ba79e7c4f94f5052b2b19151821860810fe785eb6493fd8ff7fe1b8603a
+size 15920

35_128_e5_3e-5/checkpoint-1384/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:509833714747f5c009420c517380605b3036f6ca52964fa1a463ff2e3dd5b111
+size 15920

35_128_e5_3e-5/checkpoint-1384/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99a693f9dc6d5e60143d4d6d4ce040a487650300691109ec7d6c0863166c6c42
+size 15920

35_128_e5_3e-5/checkpoint-1384/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7fbd955e8e9ee727d97bf77a3d4f95b11e0cdc728d16931354cbc8d2b19ce8b
+size 15920

35_128_e5_3e-5/checkpoint-1384/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:747c6e09e6a5f57bccdbf8516523b9c2ca9d2c1dc64a0b6220980688b87e7d43
+size 15920

35_128_e5_3e-5/checkpoint-1384/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89d740293d3888b5c9267390ff1111ffb3be3df6865c92664714586fe5833294
+size 15920

35_128_e5_3e-5/checkpoint-1384/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:871cdd2e7eb530ffbebe493c83ec10cee418cb5c8c65e811b78a581a06970485
+size 15920

35_128_e5_3e-5/checkpoint-1384/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:43fcb0dbc2149b7029ae6e2c94316ec8d1994eba64496e8bc6cc7eb80ee762f9
+size 15920

35_128_e5_3e-5/checkpoint-1384/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb12f465e2e96cba298dfda8a1d2cfa055dfbdf3723c354114f75ad1ae97e122
+size 1064

35_128_e5_3e-5/checkpoint-1384/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
+  }
+}

35_128_e5_3e-5/checkpoint-1384/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
+}

35_128_e5_3e-5/checkpoint-1384/trainer_state.json

@@ -0,0 +1,1966 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 1384,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01447178002894356,
+      "grad_norm": 1.2414709329605103,
+      "learning_rate": 1.3793103448275862e-06,
+      "loss": 1.2575,
+      "step": 5
+    },
+    {
+      "epoch": 0.02894356005788712,
+      "grad_norm": 0.995053231716156,
+      "learning_rate": 3.103448275862069e-06,
+      "loss": 1.2964,
+      "step": 10
+    },
+    {
+      "epoch": 0.04341534008683068,
+      "grad_norm": 0.6870602369308472,
+      "learning_rate": 4.827586206896552e-06,
+      "loss": 1.2136,
+      "step": 15
+    },
+    {
+      "epoch": 0.05788712011577424,
+      "grad_norm": 0.7477986216545105,
+      "learning_rate": 6.551724137931035e-06,
+      "loss": 1.208,
+      "step": 20
+    },
+    {
+      "epoch": 0.0723589001447178,
+      "grad_norm": 0.503740131855011,
+      "learning_rate": 8.275862068965517e-06,
+      "loss": 1.262,
+      "step": 25
+    },
+    {
+      "epoch": 0.08683068017366136,
+      "grad_norm": 0.5463724732398987,
+      "learning_rate": 9.999999999999999e-06,
+      "loss": 1.1591,
+      "step": 30
+    },
+    {
+      "epoch": 0.10130246020260492,
+      "grad_norm": 0.5826733112335205,
+      "learning_rate": 1.1724137931034483e-05,
+      "loss": 1.1859,
+      "step": 35
+    },
+    {
+      "epoch": 0.11577424023154848,
+      "grad_norm": 0.478679895401001,
+      "learning_rate": 1.3448275862068966e-05,
+      "loss": 1.1185,
+      "step": 40
+    },
+    {
+      "epoch": 0.13024602026049203,
+      "grad_norm": 0.45855218172073364,
+      "learning_rate": 1.517241379310345e-05,
+      "loss": 1.0916,
+      "step": 45
+    },
+    {
+      "epoch": 0.1447178002894356,
+      "grad_norm": 0.4334068298339844,
+      "learning_rate": 1.6896551724137932e-05,
+      "loss": 1.1103,
+      "step": 50
+    },
+    {
+      "epoch": 0.15918958031837915,
+      "grad_norm": 0.49706199765205383,
+      "learning_rate": 1.8620689655172415e-05,
+      "loss": 1.1351,
+      "step": 55
+    },
+    {
+      "epoch": 0.1736613603473227,
+      "grad_norm": 0.5013290643692017,
+      "learning_rate": 2.0344827586206894e-05,
+      "loss": 1.1267,
+      "step": 60
+    },
+    {
+      "epoch": 0.18813314037626627,
+      "grad_norm": 0.4919586181640625,
+      "learning_rate": 2.206896551724138e-05,
+      "loss": 1.0824,
+      "step": 65
+    },
+    {
+      "epoch": 0.20260492040520983,
+      "grad_norm": 0.4955993890762329,
+      "learning_rate": 2.3793103448275862e-05,
+      "loss": 1.088,
+      "step": 70
+    },
+    {
+      "epoch": 0.2170767004341534,
+      "grad_norm": 0.47736915946006775,
+      "learning_rate": 2.5517241379310345e-05,
+      "loss": 1.097,
+      "step": 75
+    },
+    {
+      "epoch": 0.23154848046309695,
+      "grad_norm": 0.4829305112361908,
+      "learning_rate": 2.7241379310344827e-05,
+      "loss": 1.0582,
+      "step": 80
+    },
+    {
+      "epoch": 0.2460202604920405,
+      "grad_norm": 0.5232416391372681,
+      "learning_rate": 2.8965517241379313e-05,
+      "loss": 1.0447,
+      "step": 85
+    },
+    {
+      "epoch": 0.26049204052098407,
+      "grad_norm": 0.5268356800079346,
+      "learning_rate": 2.999989031547876e-05,
+      "loss": 1.0425,
+      "step": 90
+    },
+    {
+      "epoch": 0.27496382054992763,
+      "grad_norm": 0.6044050455093384,
+      "learning_rate": 2.9998656383036702e-05,
+      "loss": 0.9967,
+      "step": 95
+    },
+    {
+      "epoch": 0.2894356005788712,
+      "grad_norm": 0.5722519755363464,
+      "learning_rate": 2.9996051525662343e-05,
+      "loss": 0.9925,
+      "step": 100
+    },
+    {
+      "epoch": 0.30390738060781475,
+      "grad_norm": 0.5845656991004944,
+      "learning_rate": 2.9992075981447947e-05,
+      "loss": 0.9589,
+      "step": 105
+    },
+    {
+      "epoch": 0.3183791606367583,
+      "grad_norm": 0.6687687635421753,
+      "learning_rate": 2.9986730113770898e-05,
+      "loss": 0.9607,
+      "step": 110
+    },
+    {
+      "epoch": 0.33285094066570187,
+      "grad_norm": 0.5653572678565979,
+      "learning_rate": 2.9980014411260523e-05,
+      "loss": 0.9765,
+      "step": 115
+    },
+    {
+      "epoch": 0.3473227206946454,
+      "grad_norm": 0.6110051274299622,
+      "learning_rate": 2.9971929487753402e-05,
+      "loss": 0.9347,
+      "step": 120
+    },
+    {
+      "epoch": 0.361794500723589,
+      "grad_norm": 0.6042832136154175,
+      "learning_rate": 2.9962476082237285e-05,
+      "loss": 0.8882,
+      "step": 125
+    },
+    {
+      "epoch": 0.37626628075253254,
+      "grad_norm": 0.6558710932731628,
+      "learning_rate": 2.9951655058783517e-05,
+      "loss": 0.9611,
+      "step": 130
+    },
+    {
+      "epoch": 0.3907380607814761,
+      "grad_norm": 0.6648174524307251,
+      "learning_rate": 2.993946740646809e-05,
+      "loss": 0.9514,
+      "step": 135
+    },
+    {
+      "epoch": 0.40520984081041966,
+      "grad_norm": 0.6010047793388367,
+      "learning_rate": 2.992591423928121e-05,
+      "loss": 0.8977,
+      "step": 140
+    },
+    {
+      "epoch": 0.4196816208393632,
+      "grad_norm": 0.594023585319519,
+      "learning_rate": 2.99109967960255e-05,
+      "loss": 0.9169,
+      "step": 145
+    },
+    {
+      "epoch": 0.4341534008683068,
+      "grad_norm": 0.6303651928901672,
+      "learning_rate": 2.9894716440202756e-05,
+      "loss": 0.8779,
+      "step": 150
+    },
+    {
+      "epoch": 0.44862518089725034,
+      "grad_norm": 0.6532771587371826,
+      "learning_rate": 2.9877074659889316e-05,
+      "loss": 0.8661,
+      "step": 155
+    },
+    {
+      "epoch": 0.4630969609261939,
+      "grad_norm": 0.7434287071228027,
+      "learning_rate": 2.9858073067600054e-05,
+      "loss": 0.8627,
+      "step": 160
+    },
+    {
+      "epoch": 0.47756874095513746,
+      "grad_norm": 0.7219240069389343,
+      "learning_rate": 2.983771340014098e-05,
+      "loss": 0.8855,
+      "step": 165
+    },
+    {
+      "epoch": 0.492040520984081,
+      "grad_norm": 0.7225666046142578,
+      "learning_rate": 2.981599751845051e-05,
+      "loss": 0.8531,
+      "step": 170
+    },
+    {
+      "epoch": 0.5065123010130246,
+      "grad_norm": 0.7195663452148438,
+      "learning_rate": 2.9792927407429344e-05,
+      "loss": 0.8174,
+      "step": 175
+    },
+    {
+      "epoch": 0.5209840810419681,
+      "grad_norm": 0.7346758842468262,
+      "learning_rate": 2.976850517575906e-05,
+      "loss": 0.8491,
+      "step": 180
+    },
+    {
+      "epoch": 0.5354558610709117,
+      "grad_norm": 0.7614021897315979,
+      "learning_rate": 2.9742733055709366e-05,
+      "loss": 0.8102,
+      "step": 185
+    },
+    {
+      "epoch": 0.5499276410998553,
+      "grad_norm": 0.7299443483352661,
+      "learning_rate": 2.9715613402934064e-05,
+      "loss": 0.7749,
+      "step": 190
+    },
+    {
+      "epoch": 0.5643994211287988,
+      "grad_norm": 0.7258509993553162,
+      "learning_rate": 2.9687148696255737e-05,
+      "loss": 0.8408,
+      "step": 195
+    },
+    {
+      "epoch": 0.5788712011577424,
+      "grad_norm": 0.8355885148048401,
+      "learning_rate": 2.9657341537439176e-05,
+      "loss": 0.7793,
+      "step": 200
+    },
+    {
+      "epoch": 0.5933429811866859,
+      "grad_norm": 0.7875196933746338,
+      "learning_rate": 2.9626194650953563e-05,
+      "loss": 0.8168,
+      "step": 205
+    },
+    {
+      "epoch": 0.6078147612156295,
+      "grad_norm": 1.2102893590927124,
+      "learning_rate": 2.9593710883723465e-05,
+      "loss": 0.7554,
+      "step": 210
+    },
+    {
+      "epoch": 0.622286541244573,
+      "grad_norm": 0.9075624346733093,
+      "learning_rate": 2.95598932048686e-05,
+      "loss": 0.7782,
+      "step": 215
+    },
+    {
+      "epoch": 0.6367583212735166,
+      "grad_norm": 0.9814792275428772,
+      "learning_rate": 2.9524744705432446e-05,
+      "loss": 0.7679,
+      "step": 220
+    },
+    {
+      "epoch": 0.6512301013024602,
+      "grad_norm": 0.8673927187919617,
+      "learning_rate": 2.9488268598099735e-05,
+      "loss": 0.7131,
+      "step": 225
+    },
+    {
+      "epoch": 0.6657018813314037,
+      "grad_norm": 1.2335319519042969,
+      "learning_rate": 2.945046821690277e-05,
+      "loss": 0.7368,
+      "step": 230
+    },
+    {
+      "epoch": 0.6801736613603473,
+      "grad_norm": 1.0025559663772583,
+      "learning_rate": 2.9411347016916704e-05,
+      "loss": 0.7447,
+      "step": 235
+    },
+    {
+      "epoch": 0.6946454413892909,
+      "grad_norm": 0.8002074360847473,
+      "learning_rate": 2.9370908573943737e-05,
+      "loss": 0.7396,
+      "step": 240
+    },
+    {
+      "epoch": 0.7091172214182344,
+      "grad_norm": 0.7924606800079346,
+      "learning_rate": 2.9329156584186267e-05,
+      "loss": 0.7821,
+      "step": 245
+    },
+    {
+      "epoch": 0.723589001447178,
+      "grad_norm": 0.8294292092323303,
+      "learning_rate": 2.9286094863909053e-05,
+      "loss": 0.7546,
+      "step": 250
+    },
+    {
+      "epoch": 0.7380607814761215,
+      "grad_norm": 0.8875476121902466,
+      "learning_rate": 2.924172734909038e-05,
+      "loss": 0.7541,
+      "step": 255
+    },
+    {
+      "epoch": 0.7525325615050651,
+      "grad_norm": 1.0516374111175537,
+      "learning_rate": 2.919605809506233e-05,
+      "loss": 0.7339,
+      "step": 260
+    },
+    {
+      "epoch": 0.7670043415340086,
+      "grad_norm": 0.9758723378181458,
+      "learning_rate": 2.9149091276140066e-05,
+      "loss": 0.7286,
+      "step": 265
+    },
+    {
+      "epoch": 0.7814761215629522,
+      "grad_norm": 0.8721650838851929,
+      "learning_rate": 2.910083118524034e-05,
+      "loss": 0.7224,
+      "step": 270
+    },
+    {
+      "epoch": 0.7959479015918958,
+      "grad_norm": 0.8282304406166077,
+      "learning_rate": 2.9051282233489065e-05,
+      "loss": 0.6857,
+      "step": 275
+    },
+    {
+      "epoch": 0.8104196816208393,
+      "grad_norm": 0.9779431223869324,
+      "learning_rate": 2.900044894981813e-05,
+      "loss": 0.6938,
+      "step": 280
+    },
+    {
+      "epoch": 0.8248914616497829,
+      "grad_norm": 0.842149019241333,
+      "learning_rate": 2.894833598055147e-05,
+      "loss": 0.6996,
+      "step": 285
+    },
+    {
+      "epoch": 0.8393632416787264,
+      "grad_norm": 0.8657100200653076,
+      "learning_rate": 2.8894948088980338e-05,
+      "loss": 0.6343,
+      "step": 290
+    },
+    {
+      "epoch": 0.85383502170767,
+      "grad_norm": 0.8805454969406128,
+      "learning_rate": 2.8840290154927955e-05,
+      "loss": 0.6506,
+      "step": 295
+    },
+    {
+      "epoch": 0.8683068017366136,
+      "grad_norm": 0.8671079874038696,
+      "learning_rate": 2.878436717430346e-05,
+      "loss": 0.7051,
+      "step": 300
+    },
+    {
+      "epoch": 0.8827785817655571,
+      "grad_norm": 0.9610555768013,
+      "learning_rate": 2.8727184258645276e-05,
+      "loss": 0.6888,
+      "step": 305
+    },
+    {
+      "epoch": 0.8972503617945007,
+      "grad_norm": 0.9625582695007324,
+      "learning_rate": 2.8668746634653908e-05,
+      "loss": 0.6669,
+      "step": 310
+    },
+    {
+      "epoch": 0.9117221418234442,
+      "grad_norm": 0.8685387372970581,
+      "learning_rate": 2.860905964371418e-05,
+      "loss": 0.645,
+      "step": 315
+    },
+    {
+      "epoch": 0.9261939218523878,
+      "grad_norm": 1.1596406698226929,
+      "learning_rate": 2.8548128741407043e-05,
+      "loss": 0.6432,
+      "step": 320
+    },
+    {
+      "epoch": 0.9406657018813314,
+      "grad_norm": 0.9292768239974976,
+      "learning_rate": 2.8485959497010906e-05,
+      "loss": 0.6599,
+      "step": 325
+    },
+    {
+      "epoch": 0.9551374819102749,
+      "grad_norm": 1.0150357484817505,
+      "learning_rate": 2.8422557592992584e-05,
+      "loss": 0.5829,
+      "step": 330
+    },
+    {
+      "epoch": 0.9696092619392185,
+      "grad_norm": 1.0730253458023071,
+      "learning_rate": 2.83579288244879e-05,
+      "loss": 0.615,
+      "step": 335
+    },
+    {
+      "epoch": 0.984081041968162,
+      "grad_norm": 1.0182080268859863,
+      "learning_rate": 2.829207909877201e-05,
+      "loss": 0.6109,
+      "step": 340
+    },
+    {
+      "epoch": 0.9985528219971056,
+      "grad_norm": 0.9661368131637573,
+      "learning_rate": 2.8225014434719423e-05,
+      "loss": 0.5798,
+      "step": 345
+    },
+    {
+      "epoch": 1.0115774240231548,
+      "grad_norm": 1.1847275495529175,
+      "learning_rate": 2.8156740962253904e-05,
+      "loss": 0.5366,
+      "step": 350
+    },
+    {
+      "epoch": 1.0260492040520983,
+      "grad_norm": 1.029125690460205,
+      "learning_rate": 2.808726492178813e-05,
+      "loss": 0.579,
+      "step": 355
+    },
+    {
+      "epoch": 1.0405209840810419,
+      "grad_norm": 1.106046199798584,
+      "learning_rate": 2.801659266365335e-05,
+      "loss": 0.5351,
+      "step": 360
+    },
+    {
+      "epoch": 1.0549927641099854,
+      "grad_norm": 0.9720465540885925,
+      "learning_rate": 2.794473064751889e-05,
+      "loss": 0.5171,
+      "step": 365
+    },
+    {
+      "epoch": 1.069464544138929,
+      "grad_norm": 1.0641460418701172,
+      "learning_rate": 2.787168544180174e-05,
+      "loss": 0.4973,
+      "step": 370
+    },
+    {
+      "epoch": 1.0839363241678726,
+      "grad_norm": 1.0609314441680908,
+      "learning_rate": 2.7797463723066224e-05,
+      "loss": 0.5228,
+      "step": 375
+    },
+    {
+      "epoch": 1.0984081041968161,
+      "grad_norm": 1.1373050212860107,
+      "learning_rate": 2.7722072275413647e-05,
+      "loss": 0.5296,
+      "step": 380
+    },
+    {
+      "epoch": 1.1128798842257597,
+      "grad_norm": 1.0449209213256836,
+      "learning_rate": 2.7645517989862304e-05,
+      "loss": 0.5076,
+      "step": 385
+    },
+    {
+      "epoch": 1.1273516642547032,
+      "grad_norm": 0.9028365612030029,
+      "learning_rate": 2.7567807863717544e-05,
+      "loss": 0.5403,
+      "step": 390
+    },
+    {
+      "epoch": 1.1418234442836468,
+      "grad_norm": 0.9618139863014221,
+      "learning_rate": 2.748894899993225e-05,
+      "loss": 0.5063,
+      "step": 395
+    },
+    {
+      "epoch": 1.1562952243125904,
+      "grad_norm": 0.9099822044372559,
+      "learning_rate": 2.740894860645756e-05,
+      "loss": 0.4521,
+      "step": 400
+    },
+    {
+      "epoch": 1.170767004341534,
+      "grad_norm": 0.9606701135635376,
+      "learning_rate": 2.732781399558408e-05,
+      "loss": 0.4984,
+      "step": 405
+    },
+    {
+      "epoch": 1.1852387843704775,
+      "grad_norm": 1.0110416412353516,
+      "learning_rate": 2.7245552583273475e-05,
+      "loss": 0.494,
+      "step": 410
+    },
+    {
+      "epoch": 1.199710564399421,
+      "grad_norm": 0.947682797908783,
+      "learning_rate": 2.716217188848066e-05,
+      "loss": 0.4795,
+      "step": 415
+    },
+    {
+      "epoch": 1.2141823444283646,
+      "grad_norm": 0.9584565162658691,
+      "learning_rate": 2.707767953246654e-05,
+      "loss": 0.5109,
+      "step": 420
+    },
+    {
+      "epoch": 1.2286541244573081,
+      "grad_norm": 1.0691497325897217,
+      "learning_rate": 2.699208323810137e-05,
+      "loss": 0.4361,
+      "step": 425
+    },
+    {
+      "epoch": 1.2431259044862517,
+      "grad_norm": 0.9888468384742737,
+      "learning_rate": 2.6905390829158903e-05,
+      "loss": 0.4539,
+      "step": 430
+    },
+    {
+      "epoch": 1.2575976845151953,
+      "grad_norm": 1.3001909255981445,
+      "learning_rate": 2.6817610229601253e-05,
+      "loss": 0.5022,
+      "step": 435
+    },
+    {
+      "epoch": 1.2720694645441388,
+      "grad_norm": 1.2359240055084229,
+      "learning_rate": 2.6728749462854606e-05,
+      "loss": 0.4434,
+      "step": 440
+    },
+    {
+      "epoch": 1.2865412445730824,
+      "grad_norm": 1.2030582427978516,
+      "learning_rate": 2.6638816651075877e-05,
+      "loss": 0.4496,
+      "step": 445
+    },
+    {
+      "epoch": 1.301013024602026,
+      "grad_norm": 1.0069599151611328,
+      "learning_rate": 2.6547820014410312e-05,
+      "loss": 0.412,
+      "step": 450
+    },
+    {
+      "epoch": 1.3154848046309695,
+      "grad_norm": 1.0469765663146973,
+      "learning_rate": 2.645576787024013e-05,
+      "loss": 0.4558,
+      "step": 455
+    },
+    {
+      "epoch": 1.329956584659913,
+      "grad_norm": 1.2109603881835938,
+      "learning_rate": 2.6362668632424302e-05,
+      "loss": 0.4665,
+      "step": 460
+    },
+    {
+      "epoch": 1.3444283646888566,
+      "grad_norm": 0.938054084777832,
+      "learning_rate": 2.6268530810529483e-05,
+      "loss": 0.3953,
+      "step": 465
+    },
+    {
+      "epoch": 1.3589001447178002,
+      "grad_norm": 0.8923037052154541,
+      "learning_rate": 2.6173363009052225e-05,
+      "loss": 0.4289,
+      "step": 470
+    },
+    {
+      "epoch": 1.3733719247467437,
+      "grad_norm": 1.0760612487792969,
+      "learning_rate": 2.60771739266325e-05,
+      "loss": 0.4314,
+      "step": 475
+    },
+    {
+      "epoch": 1.3878437047756873,
+      "grad_norm": 1.0694377422332764,
+      "learning_rate": 2.597997235525858e-05,
+      "loss": 0.44,
+      "step": 480
+    },
+    {
+      "epoch": 1.4023154848046309,
+      "grad_norm": 0.8559190630912781,
+      "learning_rate": 2.5881767179463493e-05,
+      "loss": 0.4688,
+      "step": 485
+    },
+    {
+      "epoch": 1.4167872648335744,
+      "grad_norm": 1.0568691492080688,
+      "learning_rate": 2.578256737551286e-05,
+      "loss": 0.4462,
+      "step": 490
+    },
+    {
+      "epoch": 1.431259044862518,
+      "grad_norm": 1.139095425605774,
+      "learning_rate": 2.568238201058451e-05,
+      "loss": 0.4281,
+      "step": 495
+    },
+    {
+      "epoch": 1.4457308248914615,
+      "grad_norm": 1.0132529735565186,
+      "learning_rate": 2.558122024193968e-05,
+      "loss": 0.3677,
+      "step": 500
+    },
+    {
+      "epoch": 1.460202604920405,
+      "grad_norm": 0.9675812721252441,
+      "learning_rate": 2.5479091316086008e-05,
+      "loss": 0.4631,
+      "step": 505
+    },
+    {
+      "epoch": 1.4746743849493487,
+      "grad_norm": 1.1200926303863525,
+      "learning_rate": 2.5376004567932376e-05,
+      "loss": 0.4088,
+      "step": 510
+    },
+    {
+      "epoch": 1.4891461649782922,
+      "grad_norm": 1.3555203676223755,
+      "learning_rate": 2.527196941993569e-05,
+      "loss": 0.3786,
+      "step": 515
+    },
+    {
+      "epoch": 1.5036179450072358,
+      "grad_norm": 1.0300952196121216,
+      "learning_rate": 2.5166995381239593e-05,
+      "loss": 0.3828,
+      "step": 520
+    },
+    {
+      "epoch": 1.5180897250361793,
+      "grad_norm": 1.009810209274292,
+      "learning_rate": 2.5061092046805348e-05,
+      "loss": 0.4285,
+      "step": 525
+    },
+    {
+      "epoch": 1.532561505065123,
+      "grad_norm": 0.9876683354377747,
+      "learning_rate": 2.4954269096534798e-05,
+      "loss": 0.4121,
+      "step": 530
+    },
+    {
+      "epoch": 1.5470332850940665,
+      "grad_norm": 1.0438982248306274,
+      "learning_rate": 2.48465362943856e-05,
+      "loss": 0.394,
+      "step": 535
+    },
+    {
+      "epoch": 1.56150506512301,
+      "grad_norm": 1.14638352394104,
+      "learning_rate": 2.4737903487478785e-05,
+      "loss": 0.4183,
+      "step": 540
+    },
+    {
+      "epoch": 1.5759768451519536,
+      "grad_norm": 1.0417720079421997,
+      "learning_rate": 2.462838060519865e-05,
+      "loss": 0.3917,
+      "step": 545
+    },
+    {
+      "epoch": 1.5904486251808971,
+      "grad_norm": 1.1312050819396973,
+      "learning_rate": 2.4517977658285247e-05,
+      "loss": 0.3921,
+      "step": 550
+    },
+    {
+      "epoch": 1.6049204052098407,
+      "grad_norm": 0.9650654196739197,
+      "learning_rate": 2.4406704737919317e-05,
+      "loss": 0.3452,
+      "step": 555
+    },
+    {
+      "epoch": 1.6193921852387843,
+      "grad_norm": 0.9657004475593567,
+      "learning_rate": 2.429457201479995e-05,
+      "loss": 0.4089,
+      "step": 560
+    },
+    {
+      "epoch": 1.6338639652677278,
+      "grad_norm": 1.0962070226669312,
+      "learning_rate": 2.4181589738214945e-05,
+      "loss": 0.3596,
+      "step": 565
+    },
+    {
+      "epoch": 1.6483357452966714,
+      "grad_norm": 1.0828574895858765,
+      "learning_rate": 2.4067768235103985e-05,
+      "loss": 0.3542,
+      "step": 570
+    },
+    {
+      "epoch": 1.662807525325615,
+      "grad_norm": 0.9871302843093872,
+      "learning_rate": 2.3953117909114746e-05,
+      "loss": 0.3368,
+      "step": 575
+    },
+    {
+      "epoch": 1.6772793053545585,
+      "grad_norm": 0.9643537402153015,
+      "learning_rate": 2.3837649239651934e-05,
+      "loss": 0.3755,
+      "step": 580
+    },
+    {
+      "epoch": 1.691751085383502,
+      "grad_norm": 1.0887973308563232,
+      "learning_rate": 2.3721372780919454e-05,
+      "loss": 0.3198,
+      "step": 585
+    },
+    {
+      "epoch": 1.7062228654124456,
+      "grad_norm": 1.0346442461013794,
+      "learning_rate": 2.360429916095574e-05,
+      "loss": 0.3677,
+      "step": 590
+    },
+    {
+      "epoch": 1.7206946454413892,
+      "grad_norm": 1.3537055253982544,
+      "learning_rate": 2.3486439080662283e-05,
+      "loss": 0.4023,
+      "step": 595
+    },
+    {
+      "epoch": 1.7351664254703327,
+      "grad_norm": 0.938713788986206,
+      "learning_rate": 2.3367803312825556e-05,
+      "loss": 0.352,
+      "step": 600
+    },
+    {
+      "epoch": 1.7496382054992763,
+      "grad_norm": 1.1569691896438599,
+      "learning_rate": 2.324840270113234e-05,
+      "loss": 0.3667,
+      "step": 605
+    },
+    {
+      "epoch": 1.7641099855282198,
+      "grad_norm": 1.043257713317871,
+      "learning_rate": 2.3128248159178583e-05,
+      "loss": 0.3612,
+      "step": 610
+    },
+    {
+      "epoch": 1.7785817655571634,
+      "grad_norm": 1.0356690883636475,
+      "learning_rate": 2.3007350669471866e-05,
+      "loss": 0.3484,
+      "step": 615
+    },
+    {
+      "epoch": 1.793053545586107,
+      "grad_norm": 1.0480122566223145,
+      "learning_rate": 2.2885721282427528e-05,
+      "loss": 0.3147,
+      "step": 620
+    },
+    {
+      "epoch": 1.8075253256150505,
+      "grad_norm": 1.1496782302856445,
+      "learning_rate": 2.2763371115358695e-05,
+      "loss": 0.3543,
+      "step": 625
+    },
+    {
+      "epoch": 1.821997105643994,
+      "grad_norm": 1.021693229675293,
+      "learning_rate": 2.264031135146006e-05,
+      "loss": 0.3454,
+      "step": 630
+    },
+    {
+      "epoch": 1.8364688856729376,
+      "grad_norm": 1.196956753730774,
+      "learning_rate": 2.2516553238785725e-05,
+      "loss": 0.3422,
+      "step": 635
+    },
+    {
+      "epoch": 1.8509406657018812,
+      "grad_norm": 1.0382981300354004,
+      "learning_rate": 2.239210808922109e-05,
+      "loss": 0.3425,
+      "step": 640
+    },
+    {
+      "epoch": 1.8654124457308248,
+      "grad_norm": 1.1502225399017334,
+      "learning_rate": 2.226698727744893e-05,
+      "loss": 0.3152,
+      "step": 645
+    },
+    {
+      "epoch": 1.8798842257597683,
+      "grad_norm": 1.0866633653640747,
+      "learning_rate": 2.2141202239909678e-05,
+      "loss": 0.3472,
+      "step": 650
+    },
+    {
+      "epoch": 1.8943560057887119,
+      "grad_norm": 1.0475261211395264,
+      "learning_rate": 2.2014764473756125e-05,
+      "loss": 0.3328,
+      "step": 655
+    },
+    {
+      "epoch": 1.9088277858176554,
+      "grad_norm": 1.0879319906234741,
+      "learning_rate": 2.1887685535802532e-05,
+      "loss": 0.3371,
+      "step": 660
+    },
+    {
+      "epoch": 1.923299565846599,
+      "grad_norm": 1.155945897102356,
+      "learning_rate": 2.1759977041468306e-05,
+      "loss": 0.3394,
+      "step": 665
+    },
+    {
+      "epoch": 1.9377713458755426,
+      "grad_norm": 1.0616649389266968,
+      "learning_rate": 2.1631650663716313e-05,
+      "loss": 0.2906,
+      "step": 670
+    },
+    {
+      "epoch": 1.9522431259044861,
+      "grad_norm": 1.1703593730926514,
+      "learning_rate": 2.1502718131985916e-05,
+      "loss": 0.3032,
+      "step": 675
+    },
+    {
+      "epoch": 1.9667149059334297,
+      "grad_norm": 1.0248326063156128,
+      "learning_rate": 2.1373191231120904e-05,
+      "loss": 0.2967,
+      "step": 680
+    },
+    {
+      "epoch": 1.9811866859623732,
+      "grad_norm": 1.1739745140075684,
+      "learning_rate": 2.1243081800292278e-05,
+      "loss": 0.3183,
+      "step": 685
+    },
+    {
+      "epoch": 1.9956584659913168,
+      "grad_norm": 0.9827852249145508,
+      "learning_rate": 2.1112401731916142e-05,
+      "loss": 0.3146,
+      "step": 690
+    },
+    {
+      "epoch": 2.008683068017366,
+      "grad_norm": 0.925191342830658,
+      "learning_rate": 2.0981162970566676e-05,
+      "loss": 0.2637,
+      "step": 695
+    },
+    {
+      "epoch": 2.0231548480463095,
+      "grad_norm": 0.987099289894104,
+      "learning_rate": 2.0849377511884396e-05,
+      "loss": 0.257,
+      "step": 700
+    },
+    {
+      "epoch": 2.037626628075253,
+      "grad_norm": 1.0751880407333374,
+      "learning_rate": 2.0717057401479672e-05,
+      "loss": 0.2402,
+      "step": 705
+    },
+    {
+      "epoch": 2.0520984081041966,
+      "grad_norm": 0.9453009366989136,
+      "learning_rate": 2.0584214733831765e-05,
+      "loss": 0.2241,
+      "step": 710
+    },
+    {
+      "epoch": 2.06657018813314,
+      "grad_norm": 0.9541075229644775,
+      "learning_rate": 2.0450861651183305e-05,
+      "loss": 0.2527,
+      "step": 715
+    },
+    {
+      "epoch": 2.0810419681620838,
+      "grad_norm": 1.1164172887802124,
+      "learning_rate": 2.0317010342430507e-05,
+      "loss": 0.2389,
+      "step": 720
+    },
+    {
+      "epoch": 2.0955137481910273,
+      "grad_norm": 1.2189607620239258,
+      "learning_rate": 2.0182673042009014e-05,
+      "loss": 0.2173,
+      "step": 725
+    },
+    {
+      "epoch": 2.109985528219971,
+      "grad_norm": 0.9571680426597595,
+      "learning_rate": 2.0047862028775665e-05,
+      "loss": 0.241,
+      "step": 730
+    },
+    {
+      "epoch": 2.1244573082489144,
+      "grad_norm": 0.9751089811325073,
+      "learning_rate": 1.9912589624886163e-05,
+      "loss": 0.224,
+      "step": 735
+    },
+    {
+      "epoch": 2.138929088277858,
+      "grad_norm": 1.1041796207427979,
+      "learning_rate": 1.9776868194668774e-05,
+      "loss": 0.2474,
+      "step": 740
+    },
+    {
+      "epoch": 2.1534008683068016,
+      "grad_norm": 1.0402464866638184,
+      "learning_rate": 1.9640710143494206e-05,
+      "loss": 0.2204,
+      "step": 745
+    },
+    {
+      "epoch": 2.167872648335745,
+      "grad_norm": 1.0665066242218018,
+      "learning_rate": 1.9504127916641704e-05,
+      "loss": 0.2359,
+      "step": 750
+    },
+    {
+      "epoch": 2.1823444283646887,
+      "grad_norm": 1.169087290763855,
+      "learning_rate": 1.936713399816153e-05,
+      "loss": 0.2123,
+      "step": 755
+    },
+    {
+      "epoch": 2.1968162083936322,
+      "grad_norm": 0.9681061506271362,
+      "learning_rate": 1.9229740909733883e-05,
+      "loss": 0.2253,
+      "step": 760
+    },
+    {
+      "epoch": 2.211287988422576,
+      "grad_norm": 1.0454251766204834,
+      "learning_rate": 1.9091961209524333e-05,
+      "loss": 0.2344,
+      "step": 765
+    },
+    {
+      "epoch": 2.2257597684515193,
+      "grad_norm": 1.1429376602172852,
+      "learning_rate": 1.8953807491036015e-05,
+      "loss": 0.2307,
+      "step": 770
+    },
+    {
+      "epoch": 2.240231548480463,
+      "grad_norm": 0.9582192897796631,
+      "learning_rate": 1.8815292381958534e-05,
+      "loss": 0.1943,
+      "step": 775
+    },
+    {
+      "epoch": 2.2547033285094065,
+      "grad_norm": 1.0701148509979248,
+      "learning_rate": 1.8676428543013733e-05,
+      "loss": 0.2386,
+      "step": 780
+    },
+    {
+      "epoch": 2.26917510853835,
+      "grad_norm": 0.897097110748291,
+      "learning_rate": 1.853722866679848e-05,
+      "loss": 0.2343,
+      "step": 785
+    },
+    {
+      "epoch": 2.2836468885672936,
+      "grad_norm": 1.074778437614441,
+      "learning_rate": 1.8397705476624513e-05,
+      "loss": 0.2285,
+      "step": 790
+    },
+    {
+      "epoch": 2.298118668596237,
+      "grad_norm": 1.1128685474395752,
+      "learning_rate": 1.8257871725355503e-05,
+      "loss": 0.2283,
+      "step": 795
+    },
+    {
+      "epoch": 2.3125904486251807,
+      "grad_norm": 1.0342803001403809,
+      "learning_rate": 1.8117740194241393e-05,
+      "loss": 0.2356,
+      "step": 800
+    },
+    {
+      "epoch": 2.3270622286541243,
+      "grad_norm": 1.1430531740188599,
+      "learning_rate": 1.7977323691750135e-05,
+      "loss": 0.2126,
+      "step": 805
+    },
+    {
+      "epoch": 2.341534008683068,
+      "grad_norm": 1.017957091331482,
+      "learning_rate": 1.7836635052396998e-05,
+      "loss": 0.2116,
+      "step": 810
+    },
+    {
+      "epoch": 2.3560057887120114,
+      "grad_norm": 1.1819818019866943,
+      "learning_rate": 1.769568713557141e-05,
+      "loss": 0.2317,
+      "step": 815
+    },
+    {
+      "epoch": 2.370477568740955,
+      "grad_norm": 1.094393014907837,
+      "learning_rate": 1.755449282436159e-05,
+      "loss": 0.208,
+      "step": 820
+    },
+    {
+      "epoch": 2.384949348769899,
+      "grad_norm": 1.2002899646759033,
+      "learning_rate": 1.7413065024376977e-05,
+      "loss": 0.2103,
+      "step": 825
+    },
+    {
+      "epoch": 2.399421128798842,
+      "grad_norm": 0.8949219584465027,
+      "learning_rate": 1.7271416662568652e-05,
+      "loss": 0.1954,
+      "step": 830
+    },
+    {
+      "epoch": 2.413892908827786,
+      "grad_norm": 1.070077657699585,
+      "learning_rate": 1.7129560686047725e-05,
+      "loss": 0.2238,
+      "step": 835
+    },
+    {
+      "epoch": 2.428364688856729,
+      "grad_norm": 1.0196245908737183,
+      "learning_rate": 1.6987510060901964e-05,
+      "loss": 0.2154,
+      "step": 840
+    },
+    {
+      "epoch": 2.442836468885673,
+      "grad_norm": 1.1058320999145508,
+      "learning_rate": 1.684527777101063e-05,
+      "loss": 0.2009,
+      "step": 845
+    },
+    {
+      "epoch": 2.4573082489146163,
+      "grad_norm": 0.9543981552124023,
+      "learning_rate": 1.6702876816857737e-05,
+      "loss": 0.1815,
+      "step": 850
+    },
+    {
+      "epoch": 2.4717800289435603,
+      "grad_norm": 1.2461034059524536,
+      "learning_rate": 1.6560320214343733e-05,
+      "loss": 0.1965,
+      "step": 855
+    },
+    {
+      "epoch": 2.4862518089725034,
+      "grad_norm": 0.9611772298812866,
+      "learning_rate": 1.641762099359583e-05,
+      "loss": 0.2176,
+      "step": 860
+    },
+    {
+      "epoch": 2.5007235890014474,
+      "grad_norm": 0.9620625972747803,
+      "learning_rate": 1.6274792197776996e-05,
+      "loss": 0.1894,
+      "step": 865
+    },
+    {
+      "epoch": 2.5151953690303905,
+      "grad_norm": 1.0272494554519653,
+      "learning_rate": 1.6131846881893777e-05,
+      "loss": 0.2037,
+      "step": 870
+    },
+    {
+      "epoch": 2.5296671490593345,
+      "grad_norm": 1.015389323234558,
+      "learning_rate": 1.5988798111603028e-05,
+      "loss": 0.2076,
+      "step": 875
+    },
+    {
+      "epoch": 2.5441389290882777,
+      "grad_norm": 1.0375982522964478,
+      "learning_rate": 1.584565896201765e-05,
+      "loss": 0.2029,
+      "step": 880
+    },
+    {
+      "epoch": 2.5586107091172217,
+      "grad_norm": 1.0840239524841309,
+      "learning_rate": 1.570244251651152e-05,
+      "loss": 0.1713,
+      "step": 885
+    },
+    {
+      "epoch": 2.5730824891461648,
+      "grad_norm": 0.8886203765869141,
+      "learning_rate": 1.5559161865523587e-05,
+      "loss": 0.1963,
+      "step": 890
+    },
+    {
+      "epoch": 2.587554269175109,
+      "grad_norm": 0.8905720710754395,
+      "learning_rate": 1.5415830105361403e-05,
+      "loss": 0.1914,
+      "step": 895
+    },
+    {
+      "epoch": 2.602026049204052,
+      "grad_norm": 1.0297545194625854,
+      "learning_rate": 1.527246033700405e-05,
+      "loss": 0.1928,
+      "step": 900
+    },
+    {
+      "epoch": 2.616497829232996,
+      "grad_norm": 0.8823828101158142,
+      "learning_rate": 1.512906566490469e-05,
+      "loss": 0.1775,
+      "step": 905
+    },
+    {
+      "epoch": 2.630969609261939,
+      "grad_norm": 1.3193711042404175,
+      "learning_rate": 1.4985659195792757e-05,
+      "loss": 0.1738,
+      "step": 910
+    },
+    {
+      "epoch": 2.645441389290883,
+      "grad_norm": 1.008651614189148,
+      "learning_rate": 1.4842254037475971e-05,
+      "loss": 0.2031,
+      "step": 915
+    },
+    {
+      "epoch": 2.659913169319826,
+      "grad_norm": 1.1403210163116455,
+      "learning_rate": 1.469886329764225e-05,
+      "loss": 0.1993,
+      "step": 920
+    },
+    {
+      "epoch": 2.67438494934877,
+      "grad_norm": 0.9636655449867249,
+      "learning_rate": 1.4555500082661603e-05,
+      "loss": 0.1639,
+      "step": 925
+    },
+    {
+      "epoch": 2.6888567293777133,
+      "grad_norm": 1.0645697116851807,
+      "learning_rate": 1.4412177496388196e-05,
+      "loss": 0.2048,
+      "step": 930
+    },
+    {
+      "epoch": 2.7033285094066573,
+      "grad_norm": 1.0411231517791748,
+      "learning_rate": 1.426890863896258e-05,
+      "loss": 0.1927,
+      "step": 935
+    },
+    {
+      "epoch": 2.7178002894356004,
+      "grad_norm": 1.0877488851547241,
+      "learning_rate": 1.4125706605614333e-05,
+      "loss": 0.1782,
+      "step": 940
+    },
+    {
+      "epoch": 2.7322720694645444,
+      "grad_norm": 1.1112526655197144,
+      "learning_rate": 1.3982584485465111e-05,
+      "loss": 0.1844,
+      "step": 945
+    },
+    {
+      "epoch": 2.7467438494934875,
+      "grad_norm": 0.9065836668014526,
+      "learning_rate": 1.3839555360332223e-05,
+      "loss": 0.1746,
+      "step": 950
+    },
+    {
+      "epoch": 2.7612156295224315,
+      "grad_norm": 1.3411282300949097,
+      "learning_rate": 1.3696632303532955e-05,
+      "loss": 0.1959,
+      "step": 955
+    },
+    {
+      "epoch": 2.7756874095513746,
+      "grad_norm": 0.9463226795196533,
+      "learning_rate": 1.3553828378689597e-05,
+      "loss": 0.165,
+      "step": 960
+    },
+    {
+      "epoch": 2.7901591895803186,
+      "grad_norm": 1.0703129768371582,
+      "learning_rate": 1.3411156638535422e-05,
+      "loss": 0.168,
+      "step": 965
+    },
+    {
+      "epoch": 2.8046309696092617,
+      "grad_norm": 1.1796904802322388,
+      "learning_rate": 1.3268630123721578e-05,
+      "loss": 0.1578,
+      "step": 970
+    },
+    {
+      "epoch": 2.8191027496382057,
+      "grad_norm": 1.1187974214553833,
+      "learning_rate": 1.3126261861625163e-05,
+      "loss": 0.1682,
+      "step": 975
+    },
+    {
+      "epoch": 2.833574529667149,
+      "grad_norm": 0.8996317982673645,
+      "learning_rate": 1.2984064865158473e-05,
+      "loss": 0.1711,
+      "step": 980
+    },
+    {
+      "epoch": 2.848046309696093,
+      "grad_norm": 0.8830468058586121,
+      "learning_rate": 1.2842052131579584e-05,
+      "loss": 0.1707,
+      "step": 985
+    },
+    {
+      "epoch": 2.862518089725036,
+      "grad_norm": 1.0437378883361816,
+      "learning_rate": 1.2700236641304356e-05,
+      "loss": 0.1688,
+      "step": 990
+    },
+    {
+      "epoch": 2.87698986975398,
+      "grad_norm": 1.0894595384597778,
+      "learning_rate": 1.2558631356719976e-05,
+      "loss": 0.1673,
+      "step": 995
+    },
+    {
+      "epoch": 2.891461649782923,
+      "grad_norm": 1.1124073266983032,
+      "learning_rate": 1.2417249221000169e-05,
+      "loss": 0.154,
+      "step": 1000
+    },
+    {
+      "epoch": 2.905933429811867,
+      "grad_norm": 0.9627780318260193,
+      "learning_rate": 1.2276103156922141e-05,
+      "loss": 0.1383,
+      "step": 1005
+    },
+    {
+      "epoch": 2.92040520984081,
+      "grad_norm": 0.9764670133590698,
+      "learning_rate": 1.2135206065685403e-05,
+      "loss": 0.1812,
+      "step": 1010
+    },
+    {
+      "epoch": 2.934876989869754,
+      "grad_norm": 0.8468527793884277,
+      "learning_rate": 1.199457082573255e-05,
+      "loss": 0.1721,
+      "step": 1015
+    },
+    {
+      "epoch": 2.9493487698986973,
+      "grad_norm": 1.1576621532440186,
+      "learning_rate": 1.185421029157214e-05,
+      "loss": 0.1804,
+      "step": 1020
+    },
+    {
+      "epoch": 2.9638205499276413,
+      "grad_norm": 1.082723617553711,
+      "learning_rate": 1.1714137292603738e-05,
+      "loss": 0.1693,
+      "step": 1025
+    },
+    {
+      "epoch": 2.9782923299565844,
+      "grad_norm": 1.0707679986953735,
+      "learning_rate": 1.1574364631945295e-05,
+      "loss": 0.156,
+      "step": 1030
+    },
+    {
+      "epoch": 2.9927641099855284,
+      "grad_norm": 1.0685093402862549,
+      "learning_rate": 1.1434905085262873e-05,
+      "loss": 0.1578,
+      "step": 1035
+    },
+    {
+      "epoch": 3.0057887120115776,
+      "grad_norm": 0.9791037440299988,
+      "learning_rate": 1.1295771399602918e-05,
+      "loss": 0.1199,
+      "step": 1040
+    },
+    {
+      "epoch": 3.020260492040521,
+      "grad_norm": 0.968901515007019,
+      "learning_rate": 1.1156976292227142e-05,
+      "loss": 0.1273,
+      "step": 1045
+    },
+    {
+      "epoch": 3.0347322720694647,
+      "grad_norm": 0.8842726945877075,
+      "learning_rate": 1.1018532449450136e-05,
+      "loss": 0.1153,
+      "step": 1050
+    },
+    {
+      "epoch": 3.0492040520984083,
+      "grad_norm": 0.9021174907684326,
+      "learning_rate": 1.0880452525479773e-05,
+      "loss": 0.1316,
+      "step": 1055
+    },
+    {
+      "epoch": 3.063675832127352,
+      "grad_norm": 1.0647761821746826,
+      "learning_rate": 1.07427491412606e-05,
+      "loss": 0.1236,
+      "step": 1060
+    },
+    {
+      "epoch": 3.0781476121562954,
+      "grad_norm": 0.9823290705680847,
+      "learning_rate": 1.0605434883320215e-05,
+      "loss": 0.1131,
+      "step": 1065
+    },
+    {
+      "epoch": 3.092619392185239,
+      "grad_norm": 0.9973510503768921,
+      "learning_rate": 1.0468522302618865e-05,
+      "loss": 0.1067,
+      "step": 1070
+    },
+    {
+      "epoch": 3.1070911722141825,
+      "grad_norm": 0.9438299536705017,
+      "learning_rate": 1.0332023913402197e-05,
+      "loss": 0.1202,
+      "step": 1075
+    },
+    {
+      "epoch": 3.121562952243126,
+      "grad_norm": 0.9125979542732239,
+      "learning_rate": 1.0195952192057439e-05,
+      "loss": 0.1068,
+      "step": 1080
+    },
+    {
+      "epoch": 3.1360347322720696,
+      "grad_norm": 0.9468740224838257,
+      "learning_rate": 1.0060319575973022e-05,
+      "loss": 0.1309,
+      "step": 1085
+    },
+    {
+      "epoch": 3.150506512301013,
+      "grad_norm": 0.8789544105529785,
+      "learning_rate": 9.925138462401755e-06,
+      "loss": 0.11,
+      "step": 1090
+    },
+    {
+      "epoch": 3.1649782923299568,
+      "grad_norm": 1.0050874948501587,
+      "learning_rate": 9.790421207327698e-06,
+      "loss": 0.109,
+      "step": 1095
+    },
+    {
+      "epoch": 3.1794500723589003,
+      "grad_norm": 1.0932433605194092,
+      "learning_rate": 9.656180124336738e-06,
+      "loss": 0.1196,
+      "step": 1100
+    },
+    {
+      "epoch": 3.193921852387844,
+      "grad_norm": 0.943759024143219,
+      "learning_rate": 9.52242748349114e-06,
+      "loss": 0.0988,
+      "step": 1105
+    },
+    {
+      "epoch": 3.2083936324167874,
+      "grad_norm": 0.9907084107398987,
+      "learning_rate": 9.389175510207992e-06,
+      "loss": 0.0984,
+      "step": 1110
+    },
+    {
+      "epoch": 3.222865412445731,
+      "grad_norm": 1.0540509223937988,
+      "learning_rate": 9.256436384141781e-06,
+      "loss": 0.1145,
+      "step": 1115
+    },
+    {
+      "epoch": 3.2373371924746746,
+      "grad_norm": 1.028218388557434,
+      "learning_rate": 9.124222238071122e-06,
+      "loss": 0.1204,
+      "step": 1120
+    },
+    {
+      "epoch": 3.251808972503618,
+      "grad_norm": 0.8883622288703918,
+      "learning_rate": 8.992545156789783e-06,
+      "loss": 0.114,
+      "step": 1125
+    },
+    {
+      "epoch": 3.2662807525325617,
+      "grad_norm": 0.8343327045440674,
+      "learning_rate": 8.861417176002107e-06,
+      "loss": 0.1049,
+      "step": 1130
+    },
+    {
+      "epoch": 3.2807525325615052,
+      "grad_norm": 1.1146796941757202,
+      "learning_rate": 8.7308502812229e-06,
+      "loss": 0.1102,
+      "step": 1135
+    },
+    {
+      "epoch": 3.295224312590449,
+      "grad_norm": 0.9057011604309082,
+      "learning_rate": 8.600856406681937e-06,
+      "loss": 0.1261,
+      "step": 1140
+    },
+    {
+      "epoch": 3.3096960926193923,
+      "grad_norm": 0.964340329170227,
+      "learning_rate": 8.471447434233087e-06,
+      "loss": 0.0973,
+      "step": 1145
+    },
+    {
+      "epoch": 3.324167872648336,
+      "grad_norm": 0.7725487947463989,
+      "learning_rate": 8.342635192268333e-06,
+      "loss": 0.1112,
+      "step": 1150
+    },
+    {
+      "epoch": 3.3386396526772795,
+      "grad_norm": 0.9697093367576599,
+      "learning_rate": 8.21443145463658e-06,
+      "loss": 0.1119,
+      "step": 1155
+    },
+    {
+      "epoch": 3.353111432706223,
+      "grad_norm": 0.973771333694458,
+      "learning_rate": 8.086847939567511e-06,
+      "loss": 0.1184,
+      "step": 1160
+    },
+    {
+      "epoch": 3.3675832127351666,
+      "grad_norm": 0.7942235469818115,
+      "learning_rate": 7.959896308600485e-06,
+      "loss": 0.0994,
+      "step": 1165
+    },
+    {
+      "epoch": 3.38205499276411,
+      "grad_norm": 0.8822111487388611,
+      "learning_rate": 7.833588165518642e-06,
+      "loss": 0.1037,
+      "step": 1170
+    },
+    {
+      "epoch": 3.3965267727930537,
+      "grad_norm": 0.9486744403839111,
+      "learning_rate": 7.707935055288282e-06,
+      "loss": 0.1058,
+      "step": 1175
+    },
+    {
+      "epoch": 3.4109985528219973,
+      "grad_norm": 1.0627591609954834,
+      "learning_rate": 7.5829484630036185e-06,
+      "loss": 0.1263,
+      "step": 1180
+    },
+    {
+      "epoch": 3.425470332850941,
+      "grad_norm": 0.9766060709953308,
+      "learning_rate": 7.458639812837006e-06,
+      "loss": 0.1057,
+      "step": 1185
+    },
+    {
+      "epoch": 3.4399421128798844,
+      "grad_norm": 0.8696818947792053,
+      "learning_rate": 7.335020466994727e-06,
+      "loss": 0.119,
+      "step": 1190
+    },
+    {
+      "epoch": 3.454413892908828,
+      "grad_norm": 0.7666686177253723,
+      "learning_rate": 7.212101724678451e-06,
+      "loss": 0.1024,
+      "step": 1195
+    },
+    {
+      "epoch": 3.4688856729377715,
+      "grad_norm": 1.0751121044158936,
+      "learning_rate": 7.089894821052474e-06,
+      "loss": 0.101,
+      "step": 1200
+    },
+    {
+      "epoch": 3.483357452966715,
+      "grad_norm": 0.9100741147994995,
+      "learning_rate": 6.968410926216749e-06,
+      "loss": 0.1014,
+      "step": 1205
+    },
+    {
+      "epoch": 3.4978292329956586,
+      "grad_norm": 0.9838671088218689,
+      "learning_rate": 6.847661144185934e-06,
+      "loss": 0.1237,
+      "step": 1210
+    },
+    {
+      "epoch": 3.512301013024602,
+      "grad_norm": 0.9876192212104797,
+      "learning_rate": 6.727656511874437e-06,
+      "loss": 0.1035,
+      "step": 1215
+    },
+    {
+      "epoch": 3.5267727930535457,
+      "grad_norm": 0.8223061561584473,
+      "learning_rate": 6.608407998087615e-06,
+      "loss": 0.0946,
+      "step": 1220
+    },
+    {
+      "epoch": 3.5412445730824893,
+      "grad_norm": 0.8498020172119141,
+      "learning_rate": 6.489926502519177e-06,
+      "loss": 0.0941,
+      "step": 1225
+    },
+    {
+      "epoch": 3.555716353111433,
+      "grad_norm": 0.9287440776824951,
+      "learning_rate": 6.372222854754932e-06,
+      "loss": 0.1148,
+      "step": 1230
+    },
+    {
+      "epoch": 3.5701881331403764,
+      "grad_norm": 0.903258204460144,
+      "learning_rate": 6.255307813282922e-06,
+      "loss": 0.0966,
+      "step": 1235
+    },
+    {
+      "epoch": 3.58465991316932,
+      "grad_norm": 0.8735595345497131,
+      "learning_rate": 6.139192064510058e-06,
+      "loss": 0.0893,
+      "step": 1240
+    },
+    {
+      "epoch": 3.5991316931982635,
+      "grad_norm": 0.7219606637954712,
+      "learning_rate": 6.023886221785371e-06,
+      "loss": 0.0973,
+      "step": 1245
+    },
+    {
+      "epoch": 3.613603473227207,
+      "grad_norm": 0.8071395754814148,
+      "learning_rate": 5.9094008244298855e-06,
+      "loss": 0.1079,
+      "step": 1250
+    },
+    {
+      "epoch": 3.6280752532561507,
+      "grad_norm": 0.8985153436660767,
+      "learning_rate": 5.79574633677332e-06,
+      "loss": 0.0864,
+      "step": 1255
+    },
+    {
+      "epoch": 3.642547033285094,
+      "grad_norm": 0.9126312732696533,
+      "learning_rate": 5.682933147197573e-06,
+      "loss": 0.1192,
+      "step": 1260
+    },
+    {
+      "epoch": 3.6570188133140378,
+      "grad_norm": 0.9061828851699829,
+      "learning_rate": 5.570971567187263e-06,
+      "loss": 0.099,
+      "step": 1265
+    },
+    {
+      "epoch": 3.6714905933429813,
+      "grad_norm": 0.7799466252326965,
+      "learning_rate": 5.459871830387157e-06,
+      "loss": 0.0838,
+      "step": 1270
+    },
+    {
+      "epoch": 3.685962373371925,
+      "grad_norm": 0.8942389488220215,
+      "learning_rate": 5.349644091666815e-06,
+      "loss": 0.1015,
+      "step": 1275
+    },
+    {
+      "epoch": 3.7004341534008685,
+      "grad_norm": 0.6946765780448914,
+      "learning_rate": 5.240298426192397e-06,
+      "loss": 0.089,
+      "step": 1280
+    },
+    {
+      "epoch": 3.714905933429812,
+      "grad_norm": 0.986676812171936,
+      "learning_rate": 5.131844828505756e-06,
+      "loss": 0.1028,
+      "step": 1285
+    },
+    {
+      "epoch": 3.7293777134587556,
+      "grad_norm": 0.9233500361442566,
+      "learning_rate": 5.0242932116109305e-06,
+      "loss": 0.0916,
+      "step": 1290
+    },
+    {
+      "epoch": 3.743849493487699,
+      "grad_norm": 0.8703968524932861,
+      "learning_rate": 4.917653406068015e-06,
+      "loss": 0.0963,
+      "step": 1295
+    },
+    {
+      "epoch": 3.7583212735166427,
+      "grad_norm": 0.9567957520484924,
+      "learning_rate": 4.81193515909466e-06,
+      "loss": 0.1017,
+      "step": 1300
+    },
+    {
+      "epoch": 3.7727930535455863,
+      "grad_norm": 0.8062012195587158,
+      "learning_rate": 4.707148133675123e-06,
+      "loss": 0.1002,
+      "step": 1305
+    },
+    {
+      "epoch": 3.78726483357453,
+      "grad_norm": 0.7574873566627502,
+      "learning_rate": 4.603301907677061e-06,
+      "loss": 0.1105,
+      "step": 1310
+    },
+    {
+      "epoch": 3.8017366136034734,
+      "grad_norm": 0.9099677801132202,
+      "learning_rate": 4.5004059729760574e-06,
+      "loss": 0.0951,
+      "step": 1315
+    },
+    {
+      "epoch": 3.816208393632417,
+      "grad_norm": 1.002095103263855,
+      "learning_rate": 4.398469734588047e-06,
+      "loss": 0.1034,
+      "step": 1320
+    },
+    {
+      "epoch": 3.8306801736613605,
+      "grad_norm": 0.7883458733558655,
+      "learning_rate": 4.297502509809668e-06,
+      "loss": 0.1017,
+      "step": 1325
+    },
+    {
+      "epoch": 3.845151953690304,
+      "grad_norm": 0.7947147488594055,
+      "learning_rate": 4.1975135273666345e-06,
+      "loss": 0.0904,
+      "step": 1330
+    },
+    {
+      "epoch": 3.8596237337192476,
+      "grad_norm": 0.8625684380531311,
+      "learning_rate": 4.098511926570193e-06,
+      "loss": 0.0943,
+      "step": 1335
+    },
+    {
+      "epoch": 3.874095513748191,
+      "grad_norm": 0.7970901131629944,
+      "learning_rate": 4.0005067564817665e-06,
+      "loss": 0.0795,
+      "step": 1340
+    },
+    {
+      "epoch": 3.8885672937771347,
+      "grad_norm": 0.7038248181343079,
+      "learning_rate": 3.903506975085839e-06,
+      "loss": 0.1004,
+      "step": 1345
+    },
+    {
+      "epoch": 3.9030390738060783,
+      "grad_norm": 0.8135579228401184,
+      "learning_rate": 3.8075214484711645e-06,
+      "loss": 0.0937,
+      "step": 1350
+    },
+    {
+      "epoch": 3.917510853835022,
+      "grad_norm": 0.6718750596046448,
+      "learning_rate": 3.7125589500203942e-06,
+      "loss": 0.0898,
+      "step": 1355
+    },
+    {
+      "epoch": 3.9319826338639654,
+      "grad_norm": 0.7840124368667603,
+      "learning_rate": 3.618628159608137e-06,
+      "loss": 0.1043,
+      "step": 1360
+    },
+    {
+      "epoch": 3.946454413892909,
+      "grad_norm": 0.8345990180969238,
+      "learning_rate": 3.5257376628076053e-06,
+      "loss": 0.078,
+      "step": 1365
+    },
+    {
+      "epoch": 3.9609261939218525,
+      "grad_norm": 0.8957901000976562,
+      "learning_rate": 3.4338959501058675e-06,
+      "loss": 0.0942,
+      "step": 1370
+    },
+    {
+      "epoch": 3.975397973950796,
+      "grad_norm": 0.9078422784805298,
+      "learning_rate": 3.3431114161277835e-06,
+      "loss": 0.0993,
+      "step": 1375
+    },
+    {
+      "epoch": 3.9898697539797396,
+      "grad_norm": 0.845564603805542,
+      "learning_rate": 3.253392358868718e-06,
+      "loss": 0.0911,
+      "step": 1380
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1730,
+  "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": 2.085122293545042e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

35_128_e5_3e-5/checkpoint-1384/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1380162c4a3d25f180c1ed514fe3a9a55a6f17c202b64210412030c30a27b748
+size 7736

35_128_e5_3e-5/checkpoint-1384/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)

35_128_e5_3e-5/checkpoint-1730/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

35_128_e5_3e-5/checkpoint-1730/adapter_config.json

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

35_128_e5_3e-5/checkpoint-1730/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:179ae47134a644df6c5105b224789bc0e6367553e2fcf07106ab0a25850ba77b
+size 791751704

35_128_e5_3e-5/checkpoint-1730/latest

@@ -0,0 +1 @@
+global_step1730

35_128_e5_3e-5/checkpoint-1730/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76cd74fa4f298b1b554fe52aef03d80f07528320f1dd4438f2373e5ed29ee46e
+size 15920

35_128_e5_3e-5/checkpoint-1730/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4b2df04858d6d672645081348f52b42297d0dee138e5b02b935361c955e8981
+size 15920

35_128_e5_3e-5/checkpoint-1730/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1ba9ccecb667e9dc09d8965fd908d7ed3b4862313bd46f20ed26e4e1548f833
+size 15920