Upload folder using huggingface_hub

10_128_e5_3e-5/checkpoint-1023/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

10_128_e5_3e-5/checkpoint-1023/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": [
+    "k_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

10_128_e5_3e-5/checkpoint-1023/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab68248b99bf6430c1fd9dca74b6bc6528880537284abc36ce7427cffa37ff64
+size 791751704

10_128_e5_3e-5/checkpoint-1023/latest

@@ -0,0 +1 @@
+global_step1023

10_128_e5_3e-5/checkpoint-1023/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa7028fefa1406c1905563f3878d134bf03541f15bb34f485d01a44d1545fff4
+size 15920

10_128_e5_3e-5/checkpoint-1023/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c44f11e30b92a9d81f71a0ef970a7643683e85c44492a74e17175d56eb5f562
+size 15920

10_128_e5_3e-5/checkpoint-1023/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6fd94c9470bf051eb0e23dc99fccd592f5c9372734f2c7a4b6386aaea75863bf
+size 15920

10_128_e5_3e-5/checkpoint-1023/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1a668db750b4935c6d78aca05e29ff06a3aea67fb6ca130e4a9968849770fb8
+size 15920

10_128_e5_3e-5/checkpoint-1023/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6650d1b958d18eb79d233b2dfa0d01a1d2df49d54ee0109ad9b3b25609a9c149
+size 15920

10_128_e5_3e-5/checkpoint-1023/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d06cb247502097ab9928d0176b4212c87b68ad1c757de28770fb7d55853c4f9d
+size 15920

10_128_e5_3e-5/checkpoint-1023/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4dcdc1f95be877602fe8332dabd1f9f12196b73577b6804cbb1de6951937fd56
+size 15920

10_128_e5_3e-5/checkpoint-1023/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a935c0f25b77276d7676b6b25b5c93cc2b563f6dc71dd85aa5666a970d9139d
+size 15920

10_128_e5_3e-5/checkpoint-1023/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a17e315b920445562d76e9215d9eb43ca22f89f2aa46eb7a58bd567807d0a51a
+size 1064

10_128_e5_3e-5/checkpoint-1023/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
+  }
+}

10_128_e5_3e-5/checkpoint-1023/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
+}

10_128_e5_3e-5/checkpoint-1023/trainer_state.json

@@ -0,0 +1,1462 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1023,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01466275659824047,
+      "grad_norm": 1.4255214929580688,
+      "learning_rate": 1.3953488372093023e-06,
+      "loss": 1.29,
+      "step": 5
+    },
+    {
+      "epoch": 0.02932551319648094,
+      "grad_norm": 1.1072325706481934,
+      "learning_rate": 3.1395348837209307e-06,
+      "loss": 1.3093,
+      "step": 10
+    },
+    {
+      "epoch": 0.04398826979472141,
+      "grad_norm": 0.672154426574707,
+      "learning_rate": 4.883720930232559e-06,
+      "loss": 1.3215,
+      "step": 15
+    },
+    {
+      "epoch": 0.05865102639296188,
+      "grad_norm": 0.6654118895530701,
+      "learning_rate": 6.627906976744187e-06,
+      "loss": 1.3156,
+      "step": 20
+    },
+    {
+      "epoch": 0.07331378299120235,
+      "grad_norm": 0.5303452610969543,
+      "learning_rate": 8.372093023255815e-06,
+      "loss": 1.2544,
+      "step": 25
+    },
+    {
+      "epoch": 0.08797653958944282,
+      "grad_norm": 0.5655878186225891,
+      "learning_rate": 1.0116279069767442e-05,
+      "loss": 1.2429,
+      "step": 30
+    },
+    {
+      "epoch": 0.10263929618768329,
+      "grad_norm": 0.6195840835571289,
+      "learning_rate": 1.1860465116279069e-05,
+      "loss": 1.2536,
+      "step": 35
+    },
+    {
+      "epoch": 0.11730205278592376,
+      "grad_norm": 0.5693106055259705,
+      "learning_rate": 1.3604651162790698e-05,
+      "loss": 1.2467,
+      "step": 40
+    },
+    {
+      "epoch": 0.13196480938416422,
+      "grad_norm": 0.532855212688446,
+      "learning_rate": 1.5348837209302328e-05,
+      "loss": 1.2043,
+      "step": 45
+    },
+    {
+      "epoch": 0.1466275659824047,
+      "grad_norm": 0.5181292295455933,
+      "learning_rate": 1.7093023255813955e-05,
+      "loss": 1.1559,
+      "step": 50
+    },
+    {
+      "epoch": 0.16129032258064516,
+      "grad_norm": 0.42991209030151367,
+      "learning_rate": 1.8837209302325582e-05,
+      "loss": 1.2215,
+      "step": 55
+    },
+    {
+      "epoch": 0.17595307917888564,
+      "grad_norm": 0.4203603267669678,
+      "learning_rate": 2.058139534883721e-05,
+      "loss": 1.1913,
+      "step": 60
+    },
+    {
+      "epoch": 0.1906158357771261,
+      "grad_norm": 0.4520230293273926,
+      "learning_rate": 2.2325581395348837e-05,
+      "loss": 1.1667,
+      "step": 65
+    },
+    {
+      "epoch": 0.20527859237536658,
+      "grad_norm": 0.4934653043746948,
+      "learning_rate": 2.4069767441860464e-05,
+      "loss": 1.1766,
+      "step": 70
+    },
+    {
+      "epoch": 0.21994134897360704,
+      "grad_norm": 0.4670737087726593,
+      "learning_rate": 2.5813953488372094e-05,
+      "loss": 1.1401,
+      "step": 75
+    },
+    {
+      "epoch": 0.23460410557184752,
+      "grad_norm": 0.4692193269729614,
+      "learning_rate": 2.755813953488372e-05,
+      "loss": 1.1505,
+      "step": 80
+    },
+    {
+      "epoch": 0.24926686217008798,
+      "grad_norm": 0.6386131644248962,
+      "learning_rate": 2.930232558139535e-05,
+      "loss": 1.0667,
+      "step": 85
+    },
+    {
+      "epoch": 0.26392961876832843,
+      "grad_norm": 0.6102390289306641,
+      "learning_rate": 2.999974583918511e-05,
+      "loss": 1.0966,
+      "step": 90
+    },
+    {
+      "epoch": 0.2785923753665689,
+      "grad_norm": 0.5375617146492004,
+      "learning_rate": 2.999819266539633e-05,
+      "loss": 1.0831,
+      "step": 95
+    },
+    {
+      "epoch": 0.2932551319648094,
+      "grad_norm": 0.715238630771637,
+      "learning_rate": 2.999522766429942e-05,
+      "loss": 1.0329,
+      "step": 100
+    },
+    {
+      "epoch": 0.30791788856304986,
+      "grad_norm": 0.5940918922424316,
+      "learning_rate": 2.9990851114999312e-05,
+      "loss": 1.0805,
+      "step": 105
+    },
+    {
+      "epoch": 0.3225806451612903,
+      "grad_norm": 0.5568494200706482,
+      "learning_rate": 2.9985063429474443e-05,
+      "loss": 1.0535,
+      "step": 110
+    },
+    {
+      "epoch": 0.33724340175953077,
+      "grad_norm": 0.5415995121002197,
+      "learning_rate": 2.9977865152537936e-05,
+      "loss": 1.0271,
+      "step": 115
+    },
+    {
+      "epoch": 0.3519061583577713,
+      "grad_norm": 0.5505713224411011,
+      "learning_rate": 2.996925696178638e-05,
+      "loss": 1.0306,
+      "step": 120
+    },
+    {
+      "epoch": 0.36656891495601174,
+      "grad_norm": 0.5080408453941345,
+      "learning_rate": 2.9959239667535986e-05,
+      "loss": 1.0539,
+      "step": 125
+    },
+    {
+      "epoch": 0.3812316715542522,
+      "grad_norm": 0.6116588711738586,
+      "learning_rate": 2.9947814212746345e-05,
+      "loss": 1.0046,
+      "step": 130
+    },
+    {
+      "epoch": 0.39589442815249265,
+      "grad_norm": 0.6235987544059753,
+      "learning_rate": 2.993498167293165e-05,
+      "loss": 1.0165,
+      "step": 135
+    },
+    {
+      "epoch": 0.41055718475073316,
+      "grad_norm": 0.604079008102417,
+      "learning_rate": 2.9920743256059457e-05,
+      "loss": 0.9569,
+      "step": 140
+    },
+    {
+      "epoch": 0.4252199413489736,
+      "grad_norm": 0.6320441961288452,
+      "learning_rate": 2.9905100302436976e-05,
+      "loss": 0.9959,
+      "step": 145
+    },
+    {
+      "epoch": 0.4398826979472141,
+      "grad_norm": 0.6032307147979736,
+      "learning_rate": 2.9888054284584897e-05,
+      "loss": 0.9596,
+      "step": 150
+    },
+    {
+      "epoch": 0.45454545454545453,
+      "grad_norm": 0.6766473054885864,
+      "learning_rate": 2.9869606807098795e-05,
+      "loss": 0.972,
+      "step": 155
+    },
+    {
+      "epoch": 0.46920821114369504,
+      "grad_norm": 0.7806677222251892,
+      "learning_rate": 2.9849759606498072e-05,
+      "loss": 0.9601,
+      "step": 160
+    },
+    {
+      "epoch": 0.4838709677419355,
+      "grad_norm": 0.6947822570800781,
+      "learning_rate": 2.9828514551062484e-05,
+      "loss": 0.9369,
+      "step": 165
+    },
+    {
+      "epoch": 0.49853372434017595,
+      "grad_norm": 0.6939064860343933,
+      "learning_rate": 2.9805873640656285e-05,
+      "loss": 0.9276,
+      "step": 170
+    },
+    {
+      "epoch": 0.5131964809384164,
+      "grad_norm": 0.6706434488296509,
+      "learning_rate": 2.9781839006539975e-05,
+      "loss": 0.9151,
+      "step": 175
+    },
+    {
+      "epoch": 0.5278592375366569,
+      "grad_norm": 0.7009207606315613,
+      "learning_rate": 2.9756412911169677e-05,
+      "loss": 0.9136,
+      "step": 180
+    },
+    {
+      "epoch": 0.5425219941348973,
+      "grad_norm": 0.6738207936286926,
+      "learning_rate": 2.9729597747984153e-05,
+      "loss": 0.9049,
+      "step": 185
+    },
+    {
+      "epoch": 0.5571847507331378,
+      "grad_norm": 0.8600137829780579,
+      "learning_rate": 2.9701396041179532e-05,
+      "loss": 0.8703,
+      "step": 190
+    },
+    {
+      "epoch": 0.5718475073313783,
+      "grad_norm": 0.6539983153343201,
+      "learning_rate": 2.967181044547165e-05,
+      "loss": 0.8734,
+      "step": 195
+    },
+    {
+      "epoch": 0.5865102639296188,
+      "grad_norm": 0.8027879595756531,
+      "learning_rate": 2.9640843745846202e-05,
+      "loss": 0.8866,
+      "step": 200
+    },
+    {
+      "epoch": 0.6011730205278593,
+      "grad_norm": 0.7741113901138306,
+      "learning_rate": 2.9608498857296558e-05,
+      "loss": 0.8554,
+      "step": 205
+    },
+    {
+      "epoch": 0.6158357771260997,
+      "grad_norm": 0.7552165389060974,
+      "learning_rate": 2.9574778824549372e-05,
+      "loss": 0.8581,
+      "step": 210
+    },
+    {
+      "epoch": 0.6304985337243402,
+      "grad_norm": 0.9202144145965576,
+      "learning_rate": 2.953968682177796e-05,
+      "loss": 0.8359,
+      "step": 215
+    },
+    {
+      "epoch": 0.6451612903225806,
+      "grad_norm": 0.8514307141304016,
+      "learning_rate": 2.9503226152303538e-05,
+      "loss": 0.9121,
+      "step": 220
+    },
+    {
+      "epoch": 0.6598240469208211,
+      "grad_norm": 0.8031731247901917,
+      "learning_rate": 2.9465400248284216e-05,
+      "loss": 0.8577,
+      "step": 225
+    },
+    {
+      "epoch": 0.6744868035190615,
+      "grad_norm": 1.084228515625,
+      "learning_rate": 2.9426212670391974e-05,
+      "loss": 0.8434,
+      "step": 230
+    },
+    {
+      "epoch": 0.6891495601173021,
+      "grad_norm": 0.9767665863037109,
+      "learning_rate": 2.9385667107477448e-05,
+      "loss": 0.8034,
+      "step": 235
+    },
+    {
+      "epoch": 0.7038123167155426,
+      "grad_norm": 1.003465175628662,
+      "learning_rate": 2.9343767376222706e-05,
+      "loss": 0.8492,
+      "step": 240
+    },
+    {
+      "epoch": 0.718475073313783,
+      "grad_norm": 0.8748902678489685,
+      "learning_rate": 2.930051742078196e-05,
+      "loss": 0.8514,
+      "step": 245
+    },
+    {
+      "epoch": 0.7331378299120235,
+      "grad_norm": 0.8922749161720276,
+      "learning_rate": 2.9255921312410298e-05,
+      "loss": 0.7944,
+      "step": 250
+    },
+    {
+      "epoch": 0.7478005865102639,
+      "grad_norm": 0.8216958045959473,
+      "learning_rate": 2.920998324908043e-05,
+      "loss": 0.8304,
+      "step": 255
+    },
+    {
+      "epoch": 0.7624633431085044,
+      "grad_norm": 0.794455349445343,
+      "learning_rate": 2.916270755508755e-05,
+      "loss": 0.7566,
+      "step": 260
+    },
+    {
+      "epoch": 0.7771260997067448,
+      "grad_norm": 0.9645755887031555,
+      "learning_rate": 2.9114098680642236e-05,
+      "loss": 0.7874,
+      "step": 265
+    },
+    {
+      "epoch": 0.7917888563049853,
+      "grad_norm": 0.8249243497848511,
+      "learning_rate": 2.9064161201451572e-05,
+      "loss": 0.7494,
+      "step": 270
+    },
+    {
+      "epoch": 0.8064516129032258,
+      "grad_norm": 0.8988327383995056,
+      "learning_rate": 2.90128998182884e-05,
+      "loss": 0.7533,
+      "step": 275
+    },
+    {
+      "epoch": 0.8211143695014663,
+      "grad_norm": 0.8654042482376099,
+      "learning_rate": 2.896031935654885e-05,
+      "loss": 0.7925,
+      "step": 280
+    },
+    {
+      "epoch": 0.8357771260997068,
+      "grad_norm": 1.00701904296875,
+      "learning_rate": 2.890642476579805e-05,
+      "loss": 0.7175,
+      "step": 285
+    },
+    {
+      "epoch": 0.8504398826979472,
+      "grad_norm": 0.8109433054924011,
+      "learning_rate": 2.8851221119304282e-05,
+      "loss": 0.7541,
+      "step": 290
+    },
+    {
+      "epoch": 0.8651026392961877,
+      "grad_norm": 0.8286988735198975,
+      "learning_rate": 2.8794713613561376e-05,
+      "loss": 0.6744,
+      "step": 295
+    },
+    {
+      "epoch": 0.8797653958944281,
+      "grad_norm": 0.8909016847610474,
+      "learning_rate": 2.873690756779956e-05,
+      "loss": 0.7355,
+      "step": 300
+    },
+    {
+      "epoch": 0.8944281524926686,
+      "grad_norm": 1.0289263725280762,
+      "learning_rate": 2.8677808423484727e-05,
+      "loss": 0.7108,
+      "step": 305
+    },
+    {
+      "epoch": 0.9090909090909091,
+      "grad_norm": 0.9855186343193054,
+      "learning_rate": 2.8617421743806237e-05,
+      "loss": 0.6774,
+      "step": 310
+    },
+    {
+      "epoch": 0.9237536656891495,
+      "grad_norm": 0.8788015246391296,
+      "learning_rate": 2.8555753213153232e-05,
+      "loss": 0.7565,
+      "step": 315
+    },
+    {
+      "epoch": 0.9384164222873901,
+      "grad_norm": 0.9619746804237366,
+      "learning_rate": 2.8492808636579525e-05,
+      "loss": 0.7123,
+      "step": 320
+    },
+    {
+      "epoch": 0.9530791788856305,
+      "grad_norm": 1.0104520320892334,
+      "learning_rate": 2.8428593939257175e-05,
+      "loss": 0.713,
+      "step": 325
+    },
+    {
+      "epoch": 0.967741935483871,
+      "grad_norm": 0.9837501645088196,
+      "learning_rate": 2.8363115165918722e-05,
+      "loss": 0.6476,
+      "step": 330
+    },
+    {
+      "epoch": 0.9824046920821115,
+      "grad_norm": 0.9956397414207458,
+      "learning_rate": 2.8296378480288185e-05,
+      "loss": 0.6983,
+      "step": 335
+    },
+    {
+      "epoch": 0.9970674486803519,
+      "grad_norm": 1.2373675107955933,
+      "learning_rate": 2.8228390164500834e-05,
+      "loss": 0.7202,
+      "step": 340
+    },
+    {
+      "epoch": 1.0117302052785924,
+      "grad_norm": 0.9639426469802856,
+      "learning_rate": 2.8159156618511857e-05,
+      "loss": 0.6296,
+      "step": 345
+    },
+    {
+      "epoch": 1.0263929618768328,
+      "grad_norm": 0.9819639325141907,
+      "learning_rate": 2.8088684359493885e-05,
+      "loss": 0.5852,
+      "step": 350
+    },
+    {
+      "epoch": 1.0410557184750733,
+      "grad_norm": 1.0521001815795898,
+      "learning_rate": 2.801698002122354e-05,
+      "loss": 0.587,
+      "step": 355
+    },
+    {
+      "epoch": 1.0557184750733137,
+      "grad_norm": 1.1284842491149902,
+      "learning_rate": 2.794405035345695e-05,
+      "loss": 0.5954,
+      "step": 360
+    },
+    {
+      "epoch": 1.0703812316715542,
+      "grad_norm": 1.0381896495819092,
+      "learning_rate": 2.786990222129438e-05,
+      "loss": 0.6026,
+      "step": 365
+    },
+    {
+      "epoch": 1.0850439882697946,
+      "grad_norm": 1.0300935506820679,
+      "learning_rate": 2.7794542604534012e-05,
+      "loss": 0.5956,
+      "step": 370
+    },
+    {
+      "epoch": 1.099706744868035,
+      "grad_norm": 1.0741995573043823,
+      "learning_rate": 2.77179785970149e-05,
+      "loss": 0.5512,
+      "step": 375
+    },
+    {
+      "epoch": 1.1143695014662756,
+      "grad_norm": 1.0927764177322388,
+      "learning_rate": 2.7640217405949207e-05,
+      "loss": 0.5778,
+      "step": 380
+    },
+    {
+      "epoch": 1.129032258064516,
+      "grad_norm": 1.1936726570129395,
+      "learning_rate": 2.7561266351243777e-05,
+      "loss": 0.6097,
+      "step": 385
+    },
+    {
+      "epoch": 1.1436950146627567,
+      "grad_norm": 0.9053242206573486,
+      "learning_rate": 2.748113286481107e-05,
+      "loss": 0.5764,
+      "step": 390
+    },
+    {
+      "epoch": 1.1583577712609971,
+      "grad_norm": 1.0905089378356934,
+      "learning_rate": 2.7399824489869593e-05,
+      "loss": 0.5367,
+      "step": 395
+    },
+    {
+      "epoch": 1.1730205278592376,
+      "grad_norm": 1.0911023616790771,
+      "learning_rate": 2.7317348880233815e-05,
+      "loss": 0.5846,
+      "step": 400
+    },
+    {
+      "epoch": 1.187683284457478,
+      "grad_norm": 1.1443125009536743,
+      "learning_rate": 2.7233713799593714e-05,
+      "loss": 0.5371,
+      "step": 405
+    },
+    {
+      "epoch": 1.2023460410557185,
+      "grad_norm": 1.220241904258728,
+      "learning_rate": 2.7148927120783928e-05,
+      "loss": 0.5616,
+      "step": 410
+    },
+    {
+      "epoch": 1.217008797653959,
+      "grad_norm": 1.0743826627731323,
+      "learning_rate": 2.7062996825042675e-05,
+      "loss": 0.5559,
+      "step": 415
+    },
+    {
+      "epoch": 1.2316715542521994,
+      "grad_norm": 1.0664162635803223,
+      "learning_rate": 2.6975931001260444e-05,
+      "loss": 0.5379,
+      "step": 420
+    },
+    {
+      "epoch": 1.2463343108504399,
+      "grad_norm": 1.186277151107788,
+      "learning_rate": 2.688773784521859e-05,
+      "loss": 0.5162,
+      "step": 425
+    },
+    {
+      "epoch": 1.2609970674486803,
+      "grad_norm": 1.0952428579330444,
+      "learning_rate": 2.67984256588178e-05,
+      "loss": 0.5071,
+      "step": 430
+    },
+    {
+      "epoch": 1.2756598240469208,
+      "grad_norm": 1.0587871074676514,
+      "learning_rate": 2.6708002849296634e-05,
+      "loss": 0.5135,
+      "step": 435
+    },
+    {
+      "epoch": 1.2903225806451613,
+      "grad_norm": 0.9939508438110352,
+      "learning_rate": 2.6616477928440124e-05,
+      "loss": 0.5545,
+      "step": 440
+    },
+    {
+      "epoch": 1.3049853372434017,
+      "grad_norm": 1.0896538496017456,
+      "learning_rate": 2.6523859511778513e-05,
+      "loss": 0.5094,
+      "step": 445
+    },
+    {
+      "epoch": 1.3196480938416422,
+      "grad_norm": 1.031437873840332,
+      "learning_rate": 2.643015631777628e-05,
+      "loss": 0.5654,
+      "step": 450
+    },
+    {
+      "epoch": 1.3343108504398826,
+      "grad_norm": 1.1387972831726074,
+      "learning_rate": 2.6335377167011413e-05,
+      "loss": 0.5062,
+      "step": 455
+    },
+    {
+      "epoch": 1.3489736070381233,
+      "grad_norm": 1.1082990169525146,
+      "learning_rate": 2.6239530981345122e-05,
+      "loss": 0.5128,
+      "step": 460
+    },
+    {
+      "epoch": 1.3636363636363638,
+      "grad_norm": 1.121334195137024,
+      "learning_rate": 2.6142626783081974e-05,
+      "loss": 0.501,
+      "step": 465
+    },
+    {
+      "epoch": 1.3782991202346042,
+      "grad_norm": 1.0630511045455933,
+      "learning_rate": 2.6044673694120628e-05,
+      "loss": 0.5054,
+      "step": 470
+    },
+    {
+      "epoch": 1.3929618768328447,
+      "grad_norm": 0.9710410237312317,
+      "learning_rate": 2.5945680935095117e-05,
+      "loss": 0.4749,
+      "step": 475
+    },
+    {
+      "epoch": 1.4076246334310851,
+      "grad_norm": 1.0230484008789062,
+      "learning_rate": 2.5845657824506924e-05,
+      "loss": 0.5306,
+      "step": 480
+    },
+    {
+      "epoch": 1.4222873900293256,
+      "grad_norm": 1.1004149913787842,
+      "learning_rate": 2.5744613777847777e-05,
+      "loss": 0.4675,
+      "step": 485
+    },
+    {
+      "epoch": 1.436950146627566,
+      "grad_norm": 1.1122344732284546,
+      "learning_rate": 2.5642558306713343e-05,
+      "loss": 0.4768,
+      "step": 490
+    },
+    {
+      "epoch": 1.4516129032258065,
+      "grad_norm": 1.0336195230484009,
+      "learning_rate": 2.5539501017907873e-05,
+      "loss": 0.5324,
+      "step": 495
+    },
+    {
+      "epoch": 1.466275659824047,
+      "grad_norm": 1.364105463027954,
+      "learning_rate": 2.5435451612539895e-05,
+      "loss": 0.4729,
+      "step": 500
+    },
+    {
+      "epoch": 1.4809384164222874,
+      "grad_norm": 1.0328553915023804,
+      "learning_rate": 2.5330419885108992e-05,
+      "loss": 0.456,
+      "step": 505
+    },
+    {
+      "epoch": 1.4956011730205279,
+      "grad_norm": 1.1582674980163574,
+      "learning_rate": 2.522441572258383e-05,
+      "loss": 0.4445,
+      "step": 510
+    },
+    {
+      "epoch": 1.5102639296187683,
+      "grad_norm": 0.9491491913795471,
+      "learning_rate": 2.5117449103471477e-05,
+      "loss": 0.4651,
+      "step": 515
+    },
+    {
+      "epoch": 1.5249266862170088,
+      "grad_norm": 1.1487443447113037,
+      "learning_rate": 2.5009530096878076e-05,
+      "loss": 0.4637,
+      "step": 520
+    },
+    {
+      "epoch": 1.5395894428152492,
+      "grad_norm": 1.208156943321228,
+      "learning_rate": 2.4900668861561015e-05,
+      "loss": 0.478,
+      "step": 525
+    },
+    {
+      "epoch": 1.5542521994134897,
+      "grad_norm": 1.0690094232559204,
+      "learning_rate": 2.479087564497266e-05,
+      "loss": 0.4422,
+      "step": 530
+    },
+    {
+      "epoch": 1.5689149560117301,
+      "grad_norm": 0.9839731454849243,
+      "learning_rate": 2.4680160782295712e-05,
+      "loss": 0.4509,
+      "step": 535
+    },
+    {
+      "epoch": 1.5835777126099706,
+      "grad_norm": 1.233916997909546,
+      "learning_rate": 2.4568534695470333e-05,
+      "loss": 0.4431,
+      "step": 540
+    },
+    {
+      "epoch": 1.598240469208211,
+      "grad_norm": 1.1664586067199707,
+      "learning_rate": 2.4456007892213095e-05,
+      "loss": 0.4338,
+      "step": 545
+    },
+    {
+      "epoch": 1.6129032258064515,
+      "grad_norm": 1.0460413694381714,
+      "learning_rate": 2.434259096502787e-05,
+      "loss": 0.4385,
+      "step": 550
+    },
+    {
+      "epoch": 1.627565982404692,
+      "grad_norm": 1.2513885498046875,
+      "learning_rate": 2.42282945902087e-05,
+      "loss": 0.3922,
+      "step": 555
+    },
+    {
+      "epoch": 1.6422287390029324,
+      "grad_norm": 1.156153917312622,
+      "learning_rate": 2.411312952683481e-05,
+      "loss": 0.4229,
+      "step": 560
+    },
+    {
+      "epoch": 1.6568914956011729,
+      "grad_norm": 1.0419753789901733,
+      "learning_rate": 2.3997106615757825e-05,
+      "loss": 0.4139,
+      "step": 565
+    },
+    {
+      "epoch": 1.6715542521994133,
+      "grad_norm": 1.1947492361068726,
+      "learning_rate": 2.3880236778581305e-05,
+      "loss": 0.4207,
+      "step": 570
+    },
+    {
+      "epoch": 1.6862170087976538,
+      "grad_norm": 1.100502610206604,
+      "learning_rate": 2.376253101663262e-05,
+      "loss": 0.4272,
+      "step": 575
+    },
+    {
+      "epoch": 1.7008797653958945,
+      "grad_norm": 1.186316967010498,
+      "learning_rate": 2.3644000409927408e-05,
+      "loss": 0.4051,
+      "step": 580
+    },
+    {
+      "epoch": 1.715542521994135,
+      "grad_norm": 1.0992082357406616,
+      "learning_rate": 2.3524656116126518e-05,
+      "loss": 0.4302,
+      "step": 585
+    },
+    {
+      "epoch": 1.7302052785923754,
+      "grad_norm": 1.2629891633987427,
+      "learning_rate": 2.3404509369485767e-05,
+      "loss": 0.4393,
+      "step": 590
+    },
+    {
+      "epoch": 1.7448680351906158,
+      "grad_norm": 1.0620739459991455,
+      "learning_rate": 2.328357147979839e-05,
+      "loss": 0.41,
+      "step": 595
+    },
+    {
+      "epoch": 1.7595307917888563,
+      "grad_norm": 1.171355962753296,
+      "learning_rate": 2.3161853831330407e-05,
+      "loss": 0.4126,
+      "step": 600
+    },
+    {
+      "epoch": 1.7741935483870968,
+      "grad_norm": 1.1063891649246216,
+      "learning_rate": 2.303936788174901e-05,
+      "loss": 0.4131,
+      "step": 605
+    },
+    {
+      "epoch": 1.7888563049853372,
+      "grad_norm": 1.2573152780532837,
+      "learning_rate": 2.2916125161044003e-05,
+      "loss": 0.3722,
+      "step": 610
+    },
+    {
+      "epoch": 1.8035190615835777,
+      "grad_norm": 1.0304930210113525,
+      "learning_rate": 2.2792137270442447e-05,
+      "loss": 0.3695,
+      "step": 615
+    },
+    {
+      "epoch": 1.8181818181818183,
+      "grad_norm": 1.1280535459518433,
+      "learning_rate": 2.2667415881316603e-05,
+      "loss": 0.3849,
+      "step": 620
+    },
+    {
+      "epoch": 1.8328445747800588,
+      "grad_norm": 1.0633505582809448,
+      "learning_rate": 2.2541972734085263e-05,
+      "loss": 0.3865,
+      "step": 625
+    },
+    {
+      "epoch": 1.8475073313782993,
+      "grad_norm": 1.0254065990447998,
+      "learning_rate": 2.2415819637108596e-05,
+      "loss": 0.3978,
+      "step": 630
+    },
+    {
+      "epoch": 1.8621700879765397,
+      "grad_norm": 1.1003201007843018,
+      "learning_rate": 2.228896846557659e-05,
+      "loss": 0.3435,
+      "step": 635
+    },
+    {
+      "epoch": 1.8768328445747802,
+      "grad_norm": 1.1027531623840332,
+      "learning_rate": 2.2161431160391193e-05,
+      "loss": 0.3782,
+      "step": 640
+    },
+    {
+      "epoch": 1.8914956011730206,
+      "grad_norm": 0.9939024448394775,
+      "learning_rate": 2.2033219727042287e-05,
+      "loss": 0.3723,
+      "step": 645
+    },
+    {
+      "epoch": 1.906158357771261,
+      "grad_norm": 1.0867937803268433,
+      "learning_rate": 2.190434623447756e-05,
+      "loss": 0.4106,
+      "step": 650
+    },
+    {
+      "epoch": 1.9208211143695015,
+      "grad_norm": 1.262559175491333,
+      "learning_rate": 2.177482281396645e-05,
+      "loss": 0.3522,
+      "step": 655
+    },
+    {
+      "epoch": 1.935483870967742,
+      "grad_norm": 1.1688114404678345,
+      "learning_rate": 2.1644661657958137e-05,
+      "loss": 0.3592,
+      "step": 660
+    },
+    {
+      "epoch": 1.9501466275659824,
+      "grad_norm": 1.3537970781326294,
+      "learning_rate": 2.151387501893389e-05,
+      "loss": 0.3756,
+      "step": 665
+    },
+    {
+      "epoch": 1.964809384164223,
+      "grad_norm": 1.0283355712890625,
+      "learning_rate": 2.1382475208253654e-05,
+      "loss": 0.3409,
+      "step": 670
+    },
+    {
+      "epoch": 1.9794721407624634,
+      "grad_norm": 1.1959054470062256,
+      "learning_rate": 2.1250474594997174e-05,
+      "loss": 0.3548,
+      "step": 675
+    },
+    {
+      "epoch": 1.9941348973607038,
+      "grad_norm": 1.1744413375854492,
+      "learning_rate": 2.1117885604799623e-05,
+      "loss": 0.3535,
+      "step": 680
+    },
+    {
+      "epoch": 2.0087976539589443,
+      "grad_norm": 1.0488088130950928,
+      "learning_rate": 2.0984720718681974e-05,
+      "loss": 0.3104,
+      "step": 685
+    },
+    {
+      "epoch": 2.0234604105571847,
+      "grad_norm": 1.3794373273849487,
+      "learning_rate": 2.08509924718761e-05,
+      "loss": 0.2576,
+      "step": 690
+    },
+    {
+      "epoch": 2.038123167155425,
+      "grad_norm": 1.1910767555236816,
+      "learning_rate": 2.071671345264479e-05,
+      "loss": 0.2794,
+      "step": 695
+    },
+    {
+      "epoch": 2.0527859237536656,
+      "grad_norm": 1.3358043432235718,
+      "learning_rate": 2.0581896301096793e-05,
+      "loss": 0.2809,
+      "step": 700
+    },
+    {
+      "epoch": 2.067448680351906,
+      "grad_norm": 1.064995288848877,
+      "learning_rate": 2.044655370799695e-05,
+      "loss": 0.2718,
+      "step": 705
+    },
+    {
+      "epoch": 2.0821114369501466,
+      "grad_norm": 1.117784023284912,
+      "learning_rate": 2.031069841357157e-05,
+      "loss": 0.2761,
+      "step": 710
+    },
+    {
+      "epoch": 2.096774193548387,
+      "grad_norm": 1.070440411567688,
+      "learning_rate": 2.0174343206309178e-05,
+      "loss": 0.2655,
+      "step": 715
+    },
+    {
+      "epoch": 2.1114369501466275,
+      "grad_norm": 1.1525242328643799,
+      "learning_rate": 2.003750092175665e-05,
+      "loss": 0.3002,
+      "step": 720
+    },
+    {
+      "epoch": 2.126099706744868,
+      "grad_norm": 1.3735837936401367,
+      "learning_rate": 1.9900184441311012e-05,
+      "loss": 0.2979,
+      "step": 725
+    },
+    {
+      "epoch": 2.1407624633431084,
+      "grad_norm": 1.0924019813537598,
+      "learning_rate": 1.9762406691006838e-05,
+      "loss": 0.2743,
+      "step": 730
+    },
+    {
+      "epoch": 2.155425219941349,
+      "grad_norm": 1.26235830783844,
+      "learning_rate": 1.9624180640299486e-05,
+      "loss": 0.286,
+      "step": 735
+    },
+    {
+      "epoch": 2.1700879765395893,
+      "grad_norm": 1.2357171773910522,
+      "learning_rate": 1.9485519300844263e-05,
+      "loss": 0.2598,
+      "step": 740
+    },
+    {
+      "epoch": 2.1847507331378297,
+      "grad_norm": 1.2705949544906616,
+      "learning_rate": 1.9346435725271572e-05,
+      "loss": 0.265,
+      "step": 745
+    },
+    {
+      "epoch": 2.19941348973607,
+      "grad_norm": 1.1281880140304565,
+      "learning_rate": 1.9206943005958245e-05,
+      "loss": 0.2522,
+      "step": 750
+    },
+    {
+      "epoch": 2.2140762463343107,
+      "grad_norm": 1.297865390777588,
+      "learning_rate": 1.906705427379509e-05,
+      "loss": 0.3025,
+      "step": 755
+    },
+    {
+      "epoch": 2.228739002932551,
+      "grad_norm": 1.176975131034851,
+      "learning_rate": 1.892678269695089e-05,
+      "loss": 0.2605,
+      "step": 760
+    },
+    {
+      "epoch": 2.2434017595307916,
+      "grad_norm": 1.0944550037384033,
+      "learning_rate": 1.8786141479632774e-05,
+      "loss": 0.2394,
+      "step": 765
+    },
+    {
+      "epoch": 2.258064516129032,
+      "grad_norm": 1.336591362953186,
+      "learning_rate": 1.8645143860843325e-05,
+      "loss": 0.2568,
+      "step": 770
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 1.0364762544631958,
+      "learning_rate": 1.85038031131343e-05,
+      "loss": 0.2615,
+      "step": 775
+    },
+    {
+      "epoch": 2.2873900293255134,
+      "grad_norm": 1.3140968084335327,
+      "learning_rate": 1.8362132541357285e-05,
+      "loss": 0.2515,
+      "step": 780
+    },
+    {
+      "epoch": 2.302052785923754,
+      "grad_norm": 1.115621566772461,
+      "learning_rate": 1.8220145481411238e-05,
+      "loss": 0.2368,
+      "step": 785
+    },
+    {
+      "epoch": 2.3167155425219943,
+      "grad_norm": 1.102271318435669,
+      "learning_rate": 1.8077855298987155e-05,
+      "loss": 0.2486,
+      "step": 790
+    },
+    {
+      "epoch": 2.3313782991202348,
+      "grad_norm": 1.1389175653457642,
+      "learning_rate": 1.793527538830991e-05,
+      "loss": 0.2926,
+      "step": 795
+    },
+    {
+      "epoch": 2.346041055718475,
+      "grad_norm": 1.1549748182296753,
+      "learning_rate": 1.779241917087741e-05,
+      "loss": 0.2563,
+      "step": 800
+    },
+    {
+      "epoch": 2.3607038123167157,
+      "grad_norm": 1.1911712884902954,
+      "learning_rate": 1.7649300094197186e-05,
+      "loss": 0.234,
+      "step": 805
+    },
+    {
+      "epoch": 2.375366568914956,
+      "grad_norm": 1.3891445398330688,
+      "learning_rate": 1.7505931630520554e-05,
+      "loss": 0.252,
+      "step": 810
+    },
+    {
+      "epoch": 2.3900293255131966,
+      "grad_norm": 1.16497004032135,
+      "learning_rate": 1.7362327275574406e-05,
+      "loss": 0.2477,
+      "step": 815
+    },
+    {
+      "epoch": 2.404692082111437,
+      "grad_norm": 1.1270742416381836,
+      "learning_rate": 1.7218500547290824e-05,
+      "loss": 0.2573,
+      "step": 820
+    },
+    {
+      "epoch": 2.4193548387096775,
+      "grad_norm": 1.08439040184021,
+      "learning_rate": 1.7074464984534616e-05,
+      "loss": 0.2499,
+      "step": 825
+    },
+    {
+      "epoch": 2.434017595307918,
+      "grad_norm": 1.1671042442321777,
+      "learning_rate": 1.6930234145828813e-05,
+      "loss": 0.252,
+      "step": 830
+    },
+    {
+      "epoch": 2.4486803519061584,
+      "grad_norm": 1.2057565450668335,
+      "learning_rate": 1.6785821608078402e-05,
+      "loss": 0.2527,
+      "step": 835
+    },
+    {
+      "epoch": 2.463343108504399,
+      "grad_norm": 1.105715274810791,
+      "learning_rate": 1.664124096529227e-05,
+      "loss": 0.2513,
+      "step": 840
+    },
+    {
+      "epoch": 2.4780058651026393,
+      "grad_norm": 1.1525838375091553,
+      "learning_rate": 1.6496505827303565e-05,
+      "loss": 0.2174,
+      "step": 845
+    },
+    {
+      "epoch": 2.4926686217008798,
+      "grad_norm": 1.2670106887817383,
+      "learning_rate": 1.6351629818488568e-05,
+      "loss": 0.2305,
+      "step": 850
+    },
+    {
+      "epoch": 2.5073313782991202,
+      "grad_norm": 1.1649614572525024,
+      "learning_rate": 1.620662657648416e-05,
+      "loss": 0.2205,
+      "step": 855
+    },
+    {
+      "epoch": 2.5219941348973607,
+      "grad_norm": 1.0540505647659302,
+      "learning_rate": 1.6061509750904086e-05,
+      "loss": 0.2216,
+      "step": 860
+    },
+    {
+      "epoch": 2.536656891495601,
+      "grad_norm": 1.0484446287155151,
+      "learning_rate": 1.591629300205408e-05,
+      "loss": 0.2295,
+      "step": 865
+    },
+    {
+      "epoch": 2.5513196480938416,
+      "grad_norm": 1.1153535842895508,
+      "learning_rate": 1.577098999964598e-05,
+      "loss": 0.2317,
+      "step": 870
+    },
+    {
+      "epoch": 2.565982404692082,
+      "grad_norm": 1.1432286500930786,
+      "learning_rate": 1.5625614421510903e-05,
+      "loss": 0.2507,
+      "step": 875
+    },
+    {
+      "epoch": 2.5806451612903225,
+      "grad_norm": 1.3040791749954224,
+      "learning_rate": 1.54801799523118e-05,
+      "loss": 0.22,
+      "step": 880
+    },
+    {
+      "epoch": 2.595307917888563,
+      "grad_norm": 1.1595948934555054,
+      "learning_rate": 1.5334700282255197e-05,
+      "loss": 0.2076,
+      "step": 885
+    },
+    {
+      "epoch": 2.6099706744868034,
+      "grad_norm": 1.0780363082885742,
+      "learning_rate": 1.5189189105802516e-05,
+      "loss": 0.2223,
+      "step": 890
+    },
+    {
+      "epoch": 2.624633431085044,
+      "grad_norm": 1.2541507482528687,
+      "learning_rate": 1.5043660120380983e-05,
+      "loss": 0.2168,
+      "step": 895
+    },
+    {
+      "epoch": 2.6392961876832843,
+      "grad_norm": 1.1171681880950928,
+      "learning_rate": 1.4898127025094247e-05,
+      "loss": 0.217,
+      "step": 900
+    },
+    {
+      "epoch": 2.653958944281525,
+      "grad_norm": 1.2759246826171875,
+      "learning_rate": 1.475260351943281e-05,
+      "loss": 0.2078,
+      "step": 905
+    },
+    {
+      "epoch": 2.6686217008797652,
+      "grad_norm": 1.1182351112365723,
+      "learning_rate": 1.46071033019845e-05,
+      "loss": 0.2258,
+      "step": 910
+    },
+    {
+      "epoch": 2.6832844574780057,
+      "grad_norm": 1.0474587678909302,
+      "learning_rate": 1.4461640069144924e-05,
+      "loss": 0.221,
+      "step": 915
+    },
+    {
+      "epoch": 2.6979472140762466,
+      "grad_norm": 1.3205687999725342,
+      "learning_rate": 1.431622751382823e-05,
+      "loss": 0.2355,
+      "step": 920
+    },
+    {
+      "epoch": 2.712609970674487,
+      "grad_norm": 1.08077871799469,
+      "learning_rate": 1.4170879324178126e-05,
+      "loss": 0.2122,
+      "step": 925
+    },
+    {
+      "epoch": 2.7272727272727275,
+      "grad_norm": 1.1960148811340332,
+      "learning_rate": 1.402560918227937e-05,
+      "loss": 0.1938,
+      "step": 930
+    },
+    {
+      "epoch": 2.741935483870968,
+      "grad_norm": 1.1344069242477417,
+      "learning_rate": 1.3880430762869852e-05,
+      "loss": 0.2025,
+      "step": 935
+    },
+    {
+      "epoch": 2.7565982404692084,
+      "grad_norm": 1.1796761751174927,
+      "learning_rate": 1.3735357732053312e-05,
+      "loss": 0.2213,
+      "step": 940
+    },
+    {
+      "epoch": 2.771260997067449,
+      "grad_norm": 1.029355525970459,
+      "learning_rate": 1.3590403746012951e-05,
+      "loss": 0.2116,
+      "step": 945
+    },
+    {
+      "epoch": 2.7859237536656893,
+      "grad_norm": 1.4283525943756104,
+      "learning_rate": 1.3445582449725894e-05,
+      "loss": 0.2321,
+      "step": 950
+    },
+    {
+      "epoch": 2.80058651026393,
+      "grad_norm": 1.0230028629302979,
+      "learning_rate": 1.3300907475678766e-05,
+      "loss": 0.2056,
+      "step": 955
+    },
+    {
+      "epoch": 2.8152492668621703,
+      "grad_norm": 1.176578402519226,
+      "learning_rate": 1.3156392442584427e-05,
+      "loss": 0.2139,
+      "step": 960
+    },
+    {
+      "epoch": 2.8299120234604107,
+      "grad_norm": 1.1015251874923706,
+      "learning_rate": 1.3012050954099958e-05,
+      "loss": 0.1994,
+      "step": 965
+    },
+    {
+      "epoch": 2.844574780058651,
+      "grad_norm": 1.2235914468765259,
+      "learning_rate": 1.2867896597546172e-05,
+      "loss": 0.2097,
+      "step": 970
+    },
+    {
+      "epoch": 2.8592375366568916,
+      "grad_norm": 1.1241238117218018,
+      "learning_rate": 1.2723942942628542e-05,
+      "loss": 0.2102,
+      "step": 975
+    },
+    {
+      "epoch": 2.873900293255132,
+      "grad_norm": 1.2867295742034912,
+      "learning_rate": 1.2580203540159863e-05,
+      "loss": 0.2003,
+      "step": 980
+    },
+    {
+      "epoch": 2.8885630498533725,
+      "grad_norm": 1.1582754850387573,
+      "learning_rate": 1.2436691920784685e-05,
+      "loss": 0.1884,
+      "step": 985
+    },
+    {
+      "epoch": 2.903225806451613,
+      "grad_norm": 1.234466791152954,
+      "learning_rate": 1.2293421593705585e-05,
+      "loss": 0.1931,
+      "step": 990
+    },
+    {
+      "epoch": 2.9178885630498534,
+      "grad_norm": 1.1238077878952026,
+      "learning_rate": 1.2150406045411564e-05,
+      "loss": 0.1982,
+      "step": 995
+    },
+    {
+      "epoch": 2.932551319648094,
+      "grad_norm": 1.229047417640686,
+      "learning_rate": 1.2007658738408467e-05,
+      "loss": 0.1866,
+      "step": 1000
+    },
+    {
+      "epoch": 2.9472140762463344,
+      "grad_norm": 1.3575365543365479,
+      "learning_rate": 1.1865193109951758e-05,
+      "loss": 0.1826,
+      "step": 1005
+    },
+    {
+      "epoch": 2.961876832844575,
+      "grad_norm": 1.1395304203033447,
+      "learning_rate": 1.172302257078159e-05,
+      "loss": 0.1811,
+      "step": 1010
+    },
+    {
+      "epoch": 2.9765395894428153,
+      "grad_norm": 1.1130917072296143,
+      "learning_rate": 1.1581160503860433e-05,
+      "loss": 0.1948,
+      "step": 1015
+    },
+    {
+      "epoch": 2.9912023460410557,
+      "grad_norm": 1.1696594953536987,
+      "learning_rate": 1.1439620263113305e-05,
+      "loss": 0.1733,
+      "step": 1020
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1705,
+  "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": 1.4978820660778762e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

10_128_e5_3e-5/checkpoint-1023/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb5751790d9aa12da2be1b50c556f1abeb2ac2b9c9885833cb0d870294e55895
+size 7736

10_128_e5_3e-5/checkpoint-1023/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)

10_128_e5_3e-5/checkpoint-1364/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

10_128_e5_3e-5/checkpoint-1364/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": [
+    "k_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

10_128_e5_3e-5/checkpoint-1364/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb308f358b28e7f528f5c478885b9cf960769046adfbdb2ff25f24941ab1bcac
+size 791751704

10_128_e5_3e-5/checkpoint-1364/latest

@@ -0,0 +1 @@
+global_step1364

10_128_e5_3e-5/checkpoint-1364/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:086fabcb102f92c6e2e4fae52d0500a09fadd178f4894e9f9690ba04163b3ca9
+size 15920

10_128_e5_3e-5/checkpoint-1364/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:888becb354967f595a954bd00a4a04c0e75566325a11476d4326a545903fd631
+size 15920

10_128_e5_3e-5/checkpoint-1364/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e0efdd73479432ce6b781ab59e1a50093e4b9950005ba79d1819ccc8f39beea
+size 15920

10_128_e5_3e-5/checkpoint-1364/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daabc773ac8e66ee56f26b0f814eab54929c56f67a44d5c2cd1892c169896d30
+size 15920

10_128_e5_3e-5/checkpoint-1364/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b17cb2cc89d184edf3f9f245e2540b03d30b05c765a0bf355b715e061d4ee47
+size 15920

10_128_e5_3e-5/checkpoint-1364/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8490ad4608e7ba7c3366faba590a3d557c8826797f6b9f3ccae353d8f8d6d07
+size 15920

10_128_e5_3e-5/checkpoint-1364/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93b9948453e9d26190dd5a82b3e0ad36375ab6a0ea9df4d19bd173e66e497eb8
+size 15920

10_128_e5_3e-5/checkpoint-1364/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:929bee9eb4840d413f3bcc7eb57a93ea6b490e6d6653c4bfcc86415b366953c0
+size 15920

10_128_e5_3e-5/checkpoint-1364/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2de9babe7d9426ff78334d4842392fce6c25a0ad557f11082fbf52365763bd06
+size 1064

10_128_e5_3e-5/checkpoint-1364/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
+  }
+}

10_128_e5_3e-5/checkpoint-1364/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
+}

10_128_e5_3e-5/checkpoint-1364/trainer_state.json

@@ -0,0 +1,1938 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 1364,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01466275659824047,
+      "grad_norm": 1.4255214929580688,
+      "learning_rate": 1.3953488372093023e-06,
+      "loss": 1.29,
+      "step": 5
+    },
+    {
+      "epoch": 0.02932551319648094,
+      "grad_norm": 1.1072325706481934,
+      "learning_rate": 3.1395348837209307e-06,
+      "loss": 1.3093,
+      "step": 10
+    },
+    {
+      "epoch": 0.04398826979472141,
+      "grad_norm": 0.672154426574707,
+      "learning_rate": 4.883720930232559e-06,
+      "loss": 1.3215,
+      "step": 15
+    },
+    {
+      "epoch": 0.05865102639296188,
+      "grad_norm": 0.6654118895530701,
+      "learning_rate": 6.627906976744187e-06,
+      "loss": 1.3156,
+      "step": 20
+    },
+    {
+      "epoch": 0.07331378299120235,
+      "grad_norm": 0.5303452610969543,
+      "learning_rate": 8.372093023255815e-06,
+      "loss": 1.2544,
+      "step": 25
+    },
+    {
+      "epoch": 0.08797653958944282,
+      "grad_norm": 0.5655878186225891,
+      "learning_rate": 1.0116279069767442e-05,
+      "loss": 1.2429,
+      "step": 30
+    },
+    {
+      "epoch": 0.10263929618768329,
+      "grad_norm": 0.6195840835571289,
+      "learning_rate": 1.1860465116279069e-05,
+      "loss": 1.2536,
+      "step": 35
+    },
+    {
+      "epoch": 0.11730205278592376,
+      "grad_norm": 0.5693106055259705,
+      "learning_rate": 1.3604651162790698e-05,
+      "loss": 1.2467,
+      "step": 40
+    },
+    {
+      "epoch": 0.13196480938416422,
+      "grad_norm": 0.532855212688446,
+      "learning_rate": 1.5348837209302328e-05,
+      "loss": 1.2043,
+      "step": 45
+    },
+    {
+      "epoch": 0.1466275659824047,
+      "grad_norm": 0.5181292295455933,
+      "learning_rate": 1.7093023255813955e-05,
+      "loss": 1.1559,
+      "step": 50
+    },
+    {
+      "epoch": 0.16129032258064516,
+      "grad_norm": 0.42991209030151367,
+      "learning_rate": 1.8837209302325582e-05,
+      "loss": 1.2215,
+      "step": 55
+    },
+    {
+      "epoch": 0.17595307917888564,
+      "grad_norm": 0.4203603267669678,
+      "learning_rate": 2.058139534883721e-05,
+      "loss": 1.1913,
+      "step": 60
+    },
+    {
+      "epoch": 0.1906158357771261,
+      "grad_norm": 0.4520230293273926,
+      "learning_rate": 2.2325581395348837e-05,
+      "loss": 1.1667,
+      "step": 65
+    },
+    {
+      "epoch": 0.20527859237536658,
+      "grad_norm": 0.4934653043746948,
+      "learning_rate": 2.4069767441860464e-05,
+      "loss": 1.1766,
+      "step": 70
+    },
+    {
+      "epoch": 0.21994134897360704,
+      "grad_norm": 0.4670737087726593,
+      "learning_rate": 2.5813953488372094e-05,
+      "loss": 1.1401,
+      "step": 75
+    },
+    {
+      "epoch": 0.23460410557184752,
+      "grad_norm": 0.4692193269729614,
+      "learning_rate": 2.755813953488372e-05,
+      "loss": 1.1505,
+      "step": 80
+    },
+    {
+      "epoch": 0.24926686217008798,
+      "grad_norm": 0.6386131644248962,
+      "learning_rate": 2.930232558139535e-05,
+      "loss": 1.0667,
+      "step": 85
+    },
+    {
+      "epoch": 0.26392961876832843,
+      "grad_norm": 0.6102390289306641,
+      "learning_rate": 2.999974583918511e-05,
+      "loss": 1.0966,
+      "step": 90
+    },
+    {
+      "epoch": 0.2785923753665689,
+      "grad_norm": 0.5375617146492004,
+      "learning_rate": 2.999819266539633e-05,
+      "loss": 1.0831,
+      "step": 95
+    },
+    {
+      "epoch": 0.2932551319648094,
+      "grad_norm": 0.715238630771637,
+      "learning_rate": 2.999522766429942e-05,
+      "loss": 1.0329,
+      "step": 100
+    },
+    {
+      "epoch": 0.30791788856304986,
+      "grad_norm": 0.5940918922424316,
+      "learning_rate": 2.9990851114999312e-05,
+      "loss": 1.0805,
+      "step": 105
+    },
+    {
+      "epoch": 0.3225806451612903,
+      "grad_norm": 0.5568494200706482,
+      "learning_rate": 2.9985063429474443e-05,
+      "loss": 1.0535,
+      "step": 110
+    },
+    {
+      "epoch": 0.33724340175953077,
+      "grad_norm": 0.5415995121002197,
+      "learning_rate": 2.9977865152537936e-05,
+      "loss": 1.0271,
+      "step": 115
+    },
+    {
+      "epoch": 0.3519061583577713,
+      "grad_norm": 0.5505713224411011,
+      "learning_rate": 2.996925696178638e-05,
+      "loss": 1.0306,
+      "step": 120
+    },
+    {
+      "epoch": 0.36656891495601174,
+      "grad_norm": 0.5080408453941345,
+      "learning_rate": 2.9959239667535986e-05,
+      "loss": 1.0539,
+      "step": 125
+    },
+    {
+      "epoch": 0.3812316715542522,
+      "grad_norm": 0.6116588711738586,
+      "learning_rate": 2.9947814212746345e-05,
+      "loss": 1.0046,
+      "step": 130
+    },
+    {
+      "epoch": 0.39589442815249265,
+      "grad_norm": 0.6235987544059753,
+      "learning_rate": 2.993498167293165e-05,
+      "loss": 1.0165,
+      "step": 135
+    },
+    {
+      "epoch": 0.41055718475073316,
+      "grad_norm": 0.604079008102417,
+      "learning_rate": 2.9920743256059457e-05,
+      "loss": 0.9569,
+      "step": 140
+    },
+    {
+      "epoch": 0.4252199413489736,
+      "grad_norm": 0.6320441961288452,
+      "learning_rate": 2.9905100302436976e-05,
+      "loss": 0.9959,
+      "step": 145
+    },
+    {
+      "epoch": 0.4398826979472141,
+      "grad_norm": 0.6032307147979736,
+      "learning_rate": 2.9888054284584897e-05,
+      "loss": 0.9596,
+      "step": 150
+    },
+    {
+      "epoch": 0.45454545454545453,
+      "grad_norm": 0.6766473054885864,
+      "learning_rate": 2.9869606807098795e-05,
+      "loss": 0.972,
+      "step": 155
+    },
+    {
+      "epoch": 0.46920821114369504,
+      "grad_norm": 0.7806677222251892,
+      "learning_rate": 2.9849759606498072e-05,
+      "loss": 0.9601,
+      "step": 160
+    },
+    {
+      "epoch": 0.4838709677419355,
+      "grad_norm": 0.6947822570800781,
+      "learning_rate": 2.9828514551062484e-05,
+      "loss": 0.9369,
+      "step": 165
+    },
+    {
+      "epoch": 0.49853372434017595,
+      "grad_norm": 0.6939064860343933,
+      "learning_rate": 2.9805873640656285e-05,
+      "loss": 0.9276,
+      "step": 170
+    },
+    {
+      "epoch": 0.5131964809384164,
+      "grad_norm": 0.6706434488296509,
+      "learning_rate": 2.9781839006539975e-05,
+      "loss": 0.9151,
+      "step": 175
+    },
+    {
+      "epoch": 0.5278592375366569,
+      "grad_norm": 0.7009207606315613,
+      "learning_rate": 2.9756412911169677e-05,
+      "loss": 0.9136,
+      "step": 180
+    },
+    {
+      "epoch": 0.5425219941348973,
+      "grad_norm": 0.6738207936286926,
+      "learning_rate": 2.9729597747984153e-05,
+      "loss": 0.9049,
+      "step": 185
+    },
+    {
+      "epoch": 0.5571847507331378,
+      "grad_norm": 0.8600137829780579,
+      "learning_rate": 2.9701396041179532e-05,
+      "loss": 0.8703,
+      "step": 190
+    },
+    {
+      "epoch": 0.5718475073313783,
+      "grad_norm": 0.6539983153343201,
+      "learning_rate": 2.967181044547165e-05,
+      "loss": 0.8734,
+      "step": 195
+    },
+    {
+      "epoch": 0.5865102639296188,
+      "grad_norm": 0.8027879595756531,
+      "learning_rate": 2.9640843745846202e-05,
+      "loss": 0.8866,
+      "step": 200
+    },
+    {
+      "epoch": 0.6011730205278593,
+      "grad_norm": 0.7741113901138306,
+      "learning_rate": 2.9608498857296558e-05,
+      "loss": 0.8554,
+      "step": 205
+    },
+    {
+      "epoch": 0.6158357771260997,
+      "grad_norm": 0.7552165389060974,
+      "learning_rate": 2.9574778824549372e-05,
+      "loss": 0.8581,
+      "step": 210
+    },
+    {
+      "epoch": 0.6304985337243402,
+      "grad_norm": 0.9202144145965576,
+      "learning_rate": 2.953968682177796e-05,
+      "loss": 0.8359,
+      "step": 215
+    },
+    {
+      "epoch": 0.6451612903225806,
+      "grad_norm": 0.8514307141304016,
+      "learning_rate": 2.9503226152303538e-05,
+      "loss": 0.9121,
+      "step": 220
+    },
+    {
+      "epoch": 0.6598240469208211,
+      "grad_norm": 0.8031731247901917,
+      "learning_rate": 2.9465400248284216e-05,
+      "loss": 0.8577,
+      "step": 225
+    },
+    {
+      "epoch": 0.6744868035190615,
+      "grad_norm": 1.084228515625,
+      "learning_rate": 2.9426212670391974e-05,
+      "loss": 0.8434,
+      "step": 230
+    },
+    {
+      "epoch": 0.6891495601173021,
+      "grad_norm": 0.9767665863037109,
+      "learning_rate": 2.9385667107477448e-05,
+      "loss": 0.8034,
+      "step": 235
+    },
+    {
+      "epoch": 0.7038123167155426,
+      "grad_norm": 1.003465175628662,
+      "learning_rate": 2.9343767376222706e-05,
+      "loss": 0.8492,
+      "step": 240
+    },
+    {
+      "epoch": 0.718475073313783,
+      "grad_norm": 0.8748902678489685,
+      "learning_rate": 2.930051742078196e-05,
+      "loss": 0.8514,
+      "step": 245
+    },
+    {
+      "epoch": 0.7331378299120235,
+      "grad_norm": 0.8922749161720276,
+      "learning_rate": 2.9255921312410298e-05,
+      "loss": 0.7944,
+      "step": 250
+    },
+    {
+      "epoch": 0.7478005865102639,
+      "grad_norm": 0.8216958045959473,
+      "learning_rate": 2.920998324908043e-05,
+      "loss": 0.8304,
+      "step": 255
+    },
+    {
+      "epoch": 0.7624633431085044,
+      "grad_norm": 0.794455349445343,
+      "learning_rate": 2.916270755508755e-05,
+      "loss": 0.7566,
+      "step": 260
+    },
+    {
+      "epoch": 0.7771260997067448,
+      "grad_norm": 0.9645755887031555,
+      "learning_rate": 2.9114098680642236e-05,
+      "loss": 0.7874,
+      "step": 265
+    },
+    {
+      "epoch": 0.7917888563049853,
+      "grad_norm": 0.8249243497848511,
+      "learning_rate": 2.9064161201451572e-05,
+      "loss": 0.7494,
+      "step": 270
+    },
+    {
+      "epoch": 0.8064516129032258,
+      "grad_norm": 0.8988327383995056,
+      "learning_rate": 2.90128998182884e-05,
+      "loss": 0.7533,
+      "step": 275
+    },
+    {
+      "epoch": 0.8211143695014663,
+      "grad_norm": 0.8654042482376099,
+      "learning_rate": 2.896031935654885e-05,
+      "loss": 0.7925,
+      "step": 280
+    },
+    {
+      "epoch": 0.8357771260997068,
+      "grad_norm": 1.00701904296875,
+      "learning_rate": 2.890642476579805e-05,
+      "loss": 0.7175,
+      "step": 285
+    },
+    {
+      "epoch": 0.8504398826979472,
+      "grad_norm": 0.8109433054924011,
+      "learning_rate": 2.8851221119304282e-05,
+      "loss": 0.7541,
+      "step": 290
+    },
+    {
+      "epoch": 0.8651026392961877,
+      "grad_norm": 0.8286988735198975,
+      "learning_rate": 2.8794713613561376e-05,
+      "loss": 0.6744,
+      "step": 295
+    },
+    {
+      "epoch": 0.8797653958944281,
+      "grad_norm": 0.8909016847610474,
+      "learning_rate": 2.873690756779956e-05,
+      "loss": 0.7355,
+      "step": 300
+    },
+    {
+      "epoch": 0.8944281524926686,
+      "grad_norm": 1.0289263725280762,
+      "learning_rate": 2.8677808423484727e-05,
+      "loss": 0.7108,
+      "step": 305
+    },
+    {
+      "epoch": 0.9090909090909091,
+      "grad_norm": 0.9855186343193054,
+      "learning_rate": 2.8617421743806237e-05,
+      "loss": 0.6774,
+      "step": 310
+    },
+    {
+      "epoch": 0.9237536656891495,
+      "grad_norm": 0.8788015246391296,
+      "learning_rate": 2.8555753213153232e-05,
+      "loss": 0.7565,
+      "step": 315
+    },
+    {
+      "epoch": 0.9384164222873901,
+      "grad_norm": 0.9619746804237366,
+      "learning_rate": 2.8492808636579525e-05,
+      "loss": 0.7123,
+      "step": 320
+    },
+    {
+      "epoch": 0.9530791788856305,
+      "grad_norm": 1.0104520320892334,
+      "learning_rate": 2.8428593939257175e-05,
+      "loss": 0.713,
+      "step": 325
+    },
+    {
+      "epoch": 0.967741935483871,
+      "grad_norm": 0.9837501645088196,
+      "learning_rate": 2.8363115165918722e-05,
+      "loss": 0.6476,
+      "step": 330
+    },
+    {
+      "epoch": 0.9824046920821115,
+      "grad_norm": 0.9956397414207458,
+      "learning_rate": 2.8296378480288185e-05,
+      "loss": 0.6983,
+      "step": 335
+    },
+    {
+      "epoch": 0.9970674486803519,
+      "grad_norm": 1.2373675107955933,
+      "learning_rate": 2.8228390164500834e-05,
+      "loss": 0.7202,
+      "step": 340
+    },
+    {
+      "epoch": 1.0117302052785924,
+      "grad_norm": 0.9639426469802856,
+      "learning_rate": 2.8159156618511857e-05,
+      "loss": 0.6296,
+      "step": 345
+    },
+    {
+      "epoch": 1.0263929618768328,
+      "grad_norm": 0.9819639325141907,
+      "learning_rate": 2.8088684359493885e-05,
+      "loss": 0.5852,
+      "step": 350
+    },
+    {
+      "epoch": 1.0410557184750733,
+      "grad_norm": 1.0521001815795898,
+      "learning_rate": 2.801698002122354e-05,
+      "loss": 0.587,
+      "step": 355
+    },
+    {
+      "epoch": 1.0557184750733137,
+      "grad_norm": 1.1284842491149902,
+      "learning_rate": 2.794405035345695e-05,
+      "loss": 0.5954,
+      "step": 360
+    },
+    {
+      "epoch": 1.0703812316715542,
+      "grad_norm": 1.0381896495819092,
+      "learning_rate": 2.786990222129438e-05,
+      "loss": 0.6026,
+      "step": 365
+    },
+    {
+      "epoch": 1.0850439882697946,
+      "grad_norm": 1.0300935506820679,
+      "learning_rate": 2.7794542604534012e-05,
+      "loss": 0.5956,
+      "step": 370
+    },
+    {
+      "epoch": 1.099706744868035,
+      "grad_norm": 1.0741995573043823,
+      "learning_rate": 2.77179785970149e-05,
+      "loss": 0.5512,
+      "step": 375
+    },
+    {
+      "epoch": 1.1143695014662756,
+      "grad_norm": 1.0927764177322388,
+      "learning_rate": 2.7640217405949207e-05,
+      "loss": 0.5778,
+      "step": 380
+    },
+    {
+      "epoch": 1.129032258064516,
+      "grad_norm": 1.1936726570129395,
+      "learning_rate": 2.7561266351243777e-05,
+      "loss": 0.6097,
+      "step": 385
+    },
+    {
+      "epoch": 1.1436950146627567,
+      "grad_norm": 0.9053242206573486,
+      "learning_rate": 2.748113286481107e-05,
+      "loss": 0.5764,
+      "step": 390
+    },
+    {
+      "epoch": 1.1583577712609971,
+      "grad_norm": 1.0905089378356934,
+      "learning_rate": 2.7399824489869593e-05,
+      "loss": 0.5367,
+      "step": 395
+    },
+    {
+      "epoch": 1.1730205278592376,
+      "grad_norm": 1.0911023616790771,
+      "learning_rate": 2.7317348880233815e-05,
+      "loss": 0.5846,
+      "step": 400
+    },
+    {
+      "epoch": 1.187683284457478,
+      "grad_norm": 1.1443125009536743,
+      "learning_rate": 2.7233713799593714e-05,
+      "loss": 0.5371,
+      "step": 405
+    },
+    {
+      "epoch": 1.2023460410557185,
+      "grad_norm": 1.220241904258728,
+      "learning_rate": 2.7148927120783928e-05,
+      "loss": 0.5616,
+      "step": 410
+    },
+    {
+      "epoch": 1.217008797653959,
+      "grad_norm": 1.0743826627731323,
+      "learning_rate": 2.7062996825042675e-05,
+      "loss": 0.5559,
+      "step": 415
+    },
+    {
+      "epoch": 1.2316715542521994,
+      "grad_norm": 1.0664162635803223,
+      "learning_rate": 2.6975931001260444e-05,
+      "loss": 0.5379,
+      "step": 420
+    },
+    {
+      "epoch": 1.2463343108504399,
+      "grad_norm": 1.186277151107788,
+      "learning_rate": 2.688773784521859e-05,
+      "loss": 0.5162,
+      "step": 425
+    },
+    {
+      "epoch": 1.2609970674486803,
+      "grad_norm": 1.0952428579330444,
+      "learning_rate": 2.67984256588178e-05,
+      "loss": 0.5071,
+      "step": 430
+    },
+    {
+      "epoch": 1.2756598240469208,
+      "grad_norm": 1.0587871074676514,
+      "learning_rate": 2.6708002849296634e-05,
+      "loss": 0.5135,
+      "step": 435
+    },
+    {
+      "epoch": 1.2903225806451613,
+      "grad_norm": 0.9939508438110352,
+      "learning_rate": 2.6616477928440124e-05,
+      "loss": 0.5545,
+      "step": 440
+    },
+    {
+      "epoch": 1.3049853372434017,
+      "grad_norm": 1.0896538496017456,
+      "learning_rate": 2.6523859511778513e-05,
+      "loss": 0.5094,
+      "step": 445
+    },
+    {
+      "epoch": 1.3196480938416422,
+      "grad_norm": 1.031437873840332,
+      "learning_rate": 2.643015631777628e-05,
+      "loss": 0.5654,
+      "step": 450
+    },
+    {
+      "epoch": 1.3343108504398826,
+      "grad_norm": 1.1387972831726074,
+      "learning_rate": 2.6335377167011413e-05,
+      "loss": 0.5062,
+      "step": 455
+    },
+    {
+      "epoch": 1.3489736070381233,
+      "grad_norm": 1.1082990169525146,
+      "learning_rate": 2.6239530981345122e-05,
+      "loss": 0.5128,
+      "step": 460
+    },
+    {
+      "epoch": 1.3636363636363638,
+      "grad_norm": 1.121334195137024,
+      "learning_rate": 2.6142626783081974e-05,
+      "loss": 0.501,
+      "step": 465
+    },
+    {
+      "epoch": 1.3782991202346042,
+      "grad_norm": 1.0630511045455933,
+      "learning_rate": 2.6044673694120628e-05,
+      "loss": 0.5054,
+      "step": 470
+    },
+    {
+      "epoch": 1.3929618768328447,
+      "grad_norm": 0.9710410237312317,
+      "learning_rate": 2.5945680935095117e-05,
+      "loss": 0.4749,
+      "step": 475
+    },
+    {
+      "epoch": 1.4076246334310851,
+      "grad_norm": 1.0230484008789062,
+      "learning_rate": 2.5845657824506924e-05,
+      "loss": 0.5306,
+      "step": 480
+    },
+    {
+      "epoch": 1.4222873900293256,
+      "grad_norm": 1.1004149913787842,
+      "learning_rate": 2.5744613777847777e-05,
+      "loss": 0.4675,
+      "step": 485
+    },
+    {
+      "epoch": 1.436950146627566,
+      "grad_norm": 1.1122344732284546,
+      "learning_rate": 2.5642558306713343e-05,
+      "loss": 0.4768,
+      "step": 490
+    },
+    {
+      "epoch": 1.4516129032258065,
+      "grad_norm": 1.0336195230484009,
+      "learning_rate": 2.5539501017907873e-05,
+      "loss": 0.5324,
+      "step": 495
+    },
+    {
+      "epoch": 1.466275659824047,
+      "grad_norm": 1.364105463027954,
+      "learning_rate": 2.5435451612539895e-05,
+      "loss": 0.4729,
+      "step": 500
+    },
+    {
+      "epoch": 1.4809384164222874,
+      "grad_norm": 1.0328553915023804,
+      "learning_rate": 2.5330419885108992e-05,
+      "loss": 0.456,
+      "step": 505
+    },
+    {
+      "epoch": 1.4956011730205279,
+      "grad_norm": 1.1582674980163574,
+      "learning_rate": 2.522441572258383e-05,
+      "loss": 0.4445,
+      "step": 510
+    },
+    {
+      "epoch": 1.5102639296187683,
+      "grad_norm": 0.9491491913795471,
+      "learning_rate": 2.5117449103471477e-05,
+      "loss": 0.4651,
+      "step": 515
+    },
+    {
+      "epoch": 1.5249266862170088,
+      "grad_norm": 1.1487443447113037,
+      "learning_rate": 2.5009530096878076e-05,
+      "loss": 0.4637,
+      "step": 520
+    },
+    {
+      "epoch": 1.5395894428152492,
+      "grad_norm": 1.208156943321228,
+      "learning_rate": 2.4900668861561015e-05,
+      "loss": 0.478,
+      "step": 525
+    },
+    {
+      "epoch": 1.5542521994134897,
+      "grad_norm": 1.0690094232559204,
+      "learning_rate": 2.479087564497266e-05,
+      "loss": 0.4422,
+      "step": 530
+    },
+    {
+      "epoch": 1.5689149560117301,
+      "grad_norm": 0.9839731454849243,
+      "learning_rate": 2.4680160782295712e-05,
+      "loss": 0.4509,
+      "step": 535
+    },
+    {
+      "epoch": 1.5835777126099706,
+      "grad_norm": 1.233916997909546,
+      "learning_rate": 2.4568534695470333e-05,
+      "loss": 0.4431,
+      "step": 540
+    },
+    {
+      "epoch": 1.598240469208211,
+      "grad_norm": 1.1664586067199707,
+      "learning_rate": 2.4456007892213095e-05,
+      "loss": 0.4338,
+      "step": 545
+    },
+    {
+      "epoch": 1.6129032258064515,
+      "grad_norm": 1.0460413694381714,
+      "learning_rate": 2.434259096502787e-05,
+      "loss": 0.4385,
+      "step": 550
+    },
+    {
+      "epoch": 1.627565982404692,
+      "grad_norm": 1.2513885498046875,
+      "learning_rate": 2.42282945902087e-05,
+      "loss": 0.3922,
+      "step": 555
+    },
+    {
+      "epoch": 1.6422287390029324,
+      "grad_norm": 1.156153917312622,
+      "learning_rate": 2.411312952683481e-05,
+      "loss": 0.4229,
+      "step": 560
+    },
+    {
+      "epoch": 1.6568914956011729,
+      "grad_norm": 1.0419753789901733,
+      "learning_rate": 2.3997106615757825e-05,
+      "loss": 0.4139,
+      "step": 565
+    },
+    {
+      "epoch": 1.6715542521994133,
+      "grad_norm": 1.1947492361068726,
+      "learning_rate": 2.3880236778581305e-05,
+      "loss": 0.4207,
+      "step": 570
+    },
+    {
+      "epoch": 1.6862170087976538,
+      "grad_norm": 1.100502610206604,
+      "learning_rate": 2.376253101663262e-05,
+      "loss": 0.4272,
+      "step": 575
+    },
+    {
+      "epoch": 1.7008797653958945,
+      "grad_norm": 1.186316967010498,
+      "learning_rate": 2.3644000409927408e-05,
+      "loss": 0.4051,
+      "step": 580
+    },
+    {
+      "epoch": 1.715542521994135,
+      "grad_norm": 1.0992082357406616,
+      "learning_rate": 2.3524656116126518e-05,
+      "loss": 0.4302,
+      "step": 585
+    },
+    {
+      "epoch": 1.7302052785923754,
+      "grad_norm": 1.2629891633987427,
+      "learning_rate": 2.3404509369485767e-05,
+      "loss": 0.4393,
+      "step": 590
+    },
+    {
+      "epoch": 1.7448680351906158,
+      "grad_norm": 1.0620739459991455,
+      "learning_rate": 2.328357147979839e-05,
+      "loss": 0.41,
+      "step": 595
+    },
+    {
+      "epoch": 1.7595307917888563,
+      "grad_norm": 1.171355962753296,
+      "learning_rate": 2.3161853831330407e-05,
+      "loss": 0.4126,
+      "step": 600
+    },
+    {
+      "epoch": 1.7741935483870968,
+      "grad_norm": 1.1063891649246216,
+      "learning_rate": 2.303936788174901e-05,
+      "loss": 0.4131,
+      "step": 605
+    },
+    {
+      "epoch": 1.7888563049853372,
+      "grad_norm": 1.2573152780532837,
+      "learning_rate": 2.2916125161044003e-05,
+      "loss": 0.3722,
+      "step": 610
+    },
+    {
+      "epoch": 1.8035190615835777,
+      "grad_norm": 1.0304930210113525,
+      "learning_rate": 2.2792137270442447e-05,
+      "loss": 0.3695,
+      "step": 615
+    },
+    {
+      "epoch": 1.8181818181818183,
+      "grad_norm": 1.1280535459518433,
+      "learning_rate": 2.2667415881316603e-05,
+      "loss": 0.3849,
+      "step": 620
+    },
+    {
+      "epoch": 1.8328445747800588,
+      "grad_norm": 1.0633505582809448,
+      "learning_rate": 2.2541972734085263e-05,
+      "loss": 0.3865,
+      "step": 625
+    },
+    {
+      "epoch": 1.8475073313782993,
+      "grad_norm": 1.0254065990447998,
+      "learning_rate": 2.2415819637108596e-05,
+      "loss": 0.3978,
+      "step": 630
+    },
+    {
+      "epoch": 1.8621700879765397,
+      "grad_norm": 1.1003201007843018,
+      "learning_rate": 2.228896846557659e-05,
+      "loss": 0.3435,
+      "step": 635
+    },
+    {
+      "epoch": 1.8768328445747802,
+      "grad_norm": 1.1027531623840332,
+      "learning_rate": 2.2161431160391193e-05,
+      "loss": 0.3782,
+      "step": 640
+    },
+    {
+      "epoch": 1.8914956011730206,
+      "grad_norm": 0.9939024448394775,
+      "learning_rate": 2.2033219727042287e-05,
+      "loss": 0.3723,
+      "step": 645
+    },
+    {
+      "epoch": 1.906158357771261,
+      "grad_norm": 1.0867937803268433,
+      "learning_rate": 2.190434623447756e-05,
+      "loss": 0.4106,
+      "step": 650
+    },
+    {
+      "epoch": 1.9208211143695015,
+      "grad_norm": 1.262559175491333,
+      "learning_rate": 2.177482281396645e-05,
+      "loss": 0.3522,
+      "step": 655
+    },
+    {
+      "epoch": 1.935483870967742,
+      "grad_norm": 1.1688114404678345,
+      "learning_rate": 2.1644661657958137e-05,
+      "loss": 0.3592,
+      "step": 660
+    },
+    {
+      "epoch": 1.9501466275659824,
+      "grad_norm": 1.3537970781326294,
+      "learning_rate": 2.151387501893389e-05,
+      "loss": 0.3756,
+      "step": 665
+    },
+    {
+      "epoch": 1.964809384164223,
+      "grad_norm": 1.0283355712890625,
+      "learning_rate": 2.1382475208253654e-05,
+      "loss": 0.3409,
+      "step": 670
+    },
+    {
+      "epoch": 1.9794721407624634,
+      "grad_norm": 1.1959054470062256,
+      "learning_rate": 2.1250474594997174e-05,
+      "loss": 0.3548,
+      "step": 675
+    },
+    {
+      "epoch": 1.9941348973607038,
+      "grad_norm": 1.1744413375854492,
+      "learning_rate": 2.1117885604799623e-05,
+      "loss": 0.3535,
+      "step": 680
+    },
+    {
+      "epoch": 2.0087976539589443,
+      "grad_norm": 1.0488088130950928,
+      "learning_rate": 2.0984720718681974e-05,
+      "loss": 0.3104,
+      "step": 685
+    },
+    {
+      "epoch": 2.0234604105571847,
+      "grad_norm": 1.3794373273849487,
+      "learning_rate": 2.08509924718761e-05,
+      "loss": 0.2576,
+      "step": 690
+    },
+    {
+      "epoch": 2.038123167155425,
+      "grad_norm": 1.1910767555236816,
+      "learning_rate": 2.071671345264479e-05,
+      "loss": 0.2794,
+      "step": 695
+    },
+    {
+      "epoch": 2.0527859237536656,
+      "grad_norm": 1.3358043432235718,
+      "learning_rate": 2.0581896301096793e-05,
+      "loss": 0.2809,
+      "step": 700
+    },
+    {
+      "epoch": 2.067448680351906,
+      "grad_norm": 1.064995288848877,
+      "learning_rate": 2.044655370799695e-05,
+      "loss": 0.2718,
+      "step": 705
+    },
+    {
+      "epoch": 2.0821114369501466,
+      "grad_norm": 1.117784023284912,
+      "learning_rate": 2.031069841357157e-05,
+      "loss": 0.2761,
+      "step": 710
+    },
+    {
+      "epoch": 2.096774193548387,
+      "grad_norm": 1.070440411567688,
+      "learning_rate": 2.0174343206309178e-05,
+      "loss": 0.2655,
+      "step": 715
+    },
+    {
+      "epoch": 2.1114369501466275,
+      "grad_norm": 1.1525242328643799,
+      "learning_rate": 2.003750092175665e-05,
+      "loss": 0.3002,
+      "step": 720
+    },
+    {
+      "epoch": 2.126099706744868,
+      "grad_norm": 1.3735837936401367,
+      "learning_rate": 1.9900184441311012e-05,
+      "loss": 0.2979,
+      "step": 725
+    },
+    {
+      "epoch": 2.1407624633431084,
+      "grad_norm": 1.0924019813537598,
+      "learning_rate": 1.9762406691006838e-05,
+      "loss": 0.2743,
+      "step": 730
+    },
+    {
+      "epoch": 2.155425219941349,
+      "grad_norm": 1.26235830783844,
+      "learning_rate": 1.9624180640299486e-05,
+      "loss": 0.286,
+      "step": 735
+    },
+    {
+      "epoch": 2.1700879765395893,
+      "grad_norm": 1.2357171773910522,
+      "learning_rate": 1.9485519300844263e-05,
+      "loss": 0.2598,
+      "step": 740
+    },
+    {
+      "epoch": 2.1847507331378297,
+      "grad_norm": 1.2705949544906616,
+      "learning_rate": 1.9346435725271572e-05,
+      "loss": 0.265,
+      "step": 745
+    },
+    {
+      "epoch": 2.19941348973607,
+      "grad_norm": 1.1281880140304565,
+      "learning_rate": 1.9206943005958245e-05,
+      "loss": 0.2522,
+      "step": 750
+    },
+    {
+      "epoch": 2.2140762463343107,
+      "grad_norm": 1.297865390777588,
+      "learning_rate": 1.906705427379509e-05,
+      "loss": 0.3025,
+      "step": 755
+    },
+    {
+      "epoch": 2.228739002932551,
+      "grad_norm": 1.176975131034851,
+      "learning_rate": 1.892678269695089e-05,
+      "loss": 0.2605,
+      "step": 760
+    },
+    {
+      "epoch": 2.2434017595307916,
+      "grad_norm": 1.0944550037384033,
+      "learning_rate": 1.8786141479632774e-05,
+      "loss": 0.2394,
+      "step": 765
+    },
+    {
+      "epoch": 2.258064516129032,
+      "grad_norm": 1.336591362953186,
+      "learning_rate": 1.8645143860843325e-05,
+      "loss": 0.2568,
+      "step": 770
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 1.0364762544631958,
+      "learning_rate": 1.85038031131343e-05,
+      "loss": 0.2615,
+      "step": 775
+    },
+    {
+      "epoch": 2.2873900293255134,
+      "grad_norm": 1.3140968084335327,
+      "learning_rate": 1.8362132541357285e-05,
+      "loss": 0.2515,
+      "step": 780
+    },
+    {
+      "epoch": 2.302052785923754,
+      "grad_norm": 1.115621566772461,
+      "learning_rate": 1.8220145481411238e-05,
+      "loss": 0.2368,
+      "step": 785
+    },
+    {
+      "epoch": 2.3167155425219943,
+      "grad_norm": 1.102271318435669,
+      "learning_rate": 1.8077855298987155e-05,
+      "loss": 0.2486,
+      "step": 790
+    },
+    {
+      "epoch": 2.3313782991202348,
+      "grad_norm": 1.1389175653457642,
+      "learning_rate": 1.793527538830991e-05,
+      "loss": 0.2926,
+      "step": 795
+    },
+    {
+      "epoch": 2.346041055718475,
+      "grad_norm": 1.1549748182296753,
+      "learning_rate": 1.779241917087741e-05,
+      "loss": 0.2563,
+      "step": 800
+    },
+    {
+      "epoch": 2.3607038123167157,
+      "grad_norm": 1.1911712884902954,
+      "learning_rate": 1.7649300094197186e-05,
+      "loss": 0.234,
+      "step": 805
+    },
+    {
+      "epoch": 2.375366568914956,
+      "grad_norm": 1.3891445398330688,
+      "learning_rate": 1.7505931630520554e-05,
+      "loss": 0.252,
+      "step": 810
+    },
+    {
+      "epoch": 2.3900293255131966,
+      "grad_norm": 1.16497004032135,
+      "learning_rate": 1.7362327275574406e-05,
+      "loss": 0.2477,
+      "step": 815
+    },
+    {
+      "epoch": 2.404692082111437,
+      "grad_norm": 1.1270742416381836,
+      "learning_rate": 1.7218500547290824e-05,
+      "loss": 0.2573,
+      "step": 820
+    },
+    {
+      "epoch": 2.4193548387096775,
+      "grad_norm": 1.08439040184021,
+      "learning_rate": 1.7074464984534616e-05,
+      "loss": 0.2499,
+      "step": 825
+    },
+    {
+      "epoch": 2.434017595307918,
+      "grad_norm": 1.1671042442321777,
+      "learning_rate": 1.6930234145828813e-05,
+      "loss": 0.252,
+      "step": 830
+    },
+    {
+      "epoch": 2.4486803519061584,
+      "grad_norm": 1.2057565450668335,
+      "learning_rate": 1.6785821608078402e-05,
+      "loss": 0.2527,
+      "step": 835
+    },
+    {
+      "epoch": 2.463343108504399,
+      "grad_norm": 1.105715274810791,
+      "learning_rate": 1.664124096529227e-05,
+      "loss": 0.2513,
+      "step": 840
+    },
+    {
+      "epoch": 2.4780058651026393,
+      "grad_norm": 1.1525838375091553,
+      "learning_rate": 1.6496505827303565e-05,
+      "loss": 0.2174,
+      "step": 845
+    },
+    {
+      "epoch": 2.4926686217008798,
+      "grad_norm": 1.2670106887817383,
+      "learning_rate": 1.6351629818488568e-05,
+      "loss": 0.2305,
+      "step": 850
+    },
+    {
+      "epoch": 2.5073313782991202,
+      "grad_norm": 1.1649614572525024,
+      "learning_rate": 1.620662657648416e-05,
+      "loss": 0.2205,
+      "step": 855
+    },
+    {
+      "epoch": 2.5219941348973607,
+      "grad_norm": 1.0540505647659302,
+      "learning_rate": 1.6061509750904086e-05,
+      "loss": 0.2216,
+      "step": 860
+    },
+    {
+      "epoch": 2.536656891495601,
+      "grad_norm": 1.0484446287155151,
+      "learning_rate": 1.591629300205408e-05,
+      "loss": 0.2295,
+      "step": 865
+    },
+    {
+      "epoch": 2.5513196480938416,
+      "grad_norm": 1.1153535842895508,
+      "learning_rate": 1.577098999964598e-05,
+      "loss": 0.2317,
+      "step": 870
+    },
+    {
+      "epoch": 2.565982404692082,
+      "grad_norm": 1.1432286500930786,
+      "learning_rate": 1.5625614421510903e-05,
+      "loss": 0.2507,
+      "step": 875
+    },
+    {
+      "epoch": 2.5806451612903225,
+      "grad_norm": 1.3040791749954224,
+      "learning_rate": 1.54801799523118e-05,
+      "loss": 0.22,
+      "step": 880
+    },
+    {
+      "epoch": 2.595307917888563,
+      "grad_norm": 1.1595948934555054,
+      "learning_rate": 1.5334700282255197e-05,
+      "loss": 0.2076,
+      "step": 885
+    },
+    {
+      "epoch": 2.6099706744868034,
+      "grad_norm": 1.0780363082885742,
+      "learning_rate": 1.5189189105802516e-05,
+      "loss": 0.2223,
+      "step": 890
+    },
+    {
+      "epoch": 2.624633431085044,
+      "grad_norm": 1.2541507482528687,
+      "learning_rate": 1.5043660120380983e-05,
+      "loss": 0.2168,
+      "step": 895
+    },
+    {
+      "epoch": 2.6392961876832843,
+      "grad_norm": 1.1171681880950928,
+      "learning_rate": 1.4898127025094247e-05,
+      "loss": 0.217,
+      "step": 900
+    },
+    {
+      "epoch": 2.653958944281525,
+      "grad_norm": 1.2759246826171875,
+      "learning_rate": 1.475260351943281e-05,
+      "loss": 0.2078,
+      "step": 905
+    },
+    {
+      "epoch": 2.6686217008797652,
+      "grad_norm": 1.1182351112365723,
+      "learning_rate": 1.46071033019845e-05,
+      "loss": 0.2258,
+      "step": 910
+    },
+    {
+      "epoch": 2.6832844574780057,
+      "grad_norm": 1.0474587678909302,
+      "learning_rate": 1.4461640069144924e-05,
+      "loss": 0.221,
+      "step": 915
+    },
+    {
+      "epoch": 2.6979472140762466,
+      "grad_norm": 1.3205687999725342,
+      "learning_rate": 1.431622751382823e-05,
+      "loss": 0.2355,
+      "step": 920
+    },
+    {
+      "epoch": 2.712609970674487,
+      "grad_norm": 1.08077871799469,
+      "learning_rate": 1.4170879324178126e-05,
+      "loss": 0.2122,
+      "step": 925
+    },
+    {
+      "epoch": 2.7272727272727275,
+      "grad_norm": 1.1960148811340332,
+      "learning_rate": 1.402560918227937e-05,
+      "loss": 0.1938,
+      "step": 930
+    },
+    {
+      "epoch": 2.741935483870968,
+      "grad_norm": 1.1344069242477417,
+      "learning_rate": 1.3880430762869852e-05,
+      "loss": 0.2025,
+      "step": 935
+    },
+    {
+      "epoch": 2.7565982404692084,
+      "grad_norm": 1.1796761751174927,
+      "learning_rate": 1.3735357732053312e-05,
+      "loss": 0.2213,
+      "step": 940
+    },
+    {
+      "epoch": 2.771260997067449,
+      "grad_norm": 1.029355525970459,
+      "learning_rate": 1.3590403746012951e-05,
+      "loss": 0.2116,
+      "step": 945
+    },
+    {
+      "epoch": 2.7859237536656893,
+      "grad_norm": 1.4283525943756104,
+      "learning_rate": 1.3445582449725894e-05,
+      "loss": 0.2321,
+      "step": 950
+    },
+    {
+      "epoch": 2.80058651026393,
+      "grad_norm": 1.0230028629302979,
+      "learning_rate": 1.3300907475678766e-05,
+      "loss": 0.2056,
+      "step": 955
+    },
+    {
+      "epoch": 2.8152492668621703,
+      "grad_norm": 1.176578402519226,
+      "learning_rate": 1.3156392442584427e-05,
+      "loss": 0.2139,
+      "step": 960
+    },
+    {
+      "epoch": 2.8299120234604107,
+      "grad_norm": 1.1015251874923706,
+      "learning_rate": 1.3012050954099958e-05,
+      "loss": 0.1994,
+      "step": 965
+    },
+    {
+      "epoch": 2.844574780058651,
+      "grad_norm": 1.2235914468765259,
+      "learning_rate": 1.2867896597546172e-05,
+      "loss": 0.2097,
+      "step": 970
+    },
+    {
+      "epoch": 2.8592375366568916,
+      "grad_norm": 1.1241238117218018,
+      "learning_rate": 1.2723942942628542e-05,
+      "loss": 0.2102,
+      "step": 975
+    },
+    {
+      "epoch": 2.873900293255132,
+      "grad_norm": 1.2867295742034912,
+      "learning_rate": 1.2580203540159863e-05,
+      "loss": 0.2003,
+      "step": 980
+    },
+    {
+      "epoch": 2.8885630498533725,
+      "grad_norm": 1.1582754850387573,
+      "learning_rate": 1.2436691920784685e-05,
+      "loss": 0.1884,
+      "step": 985
+    },
+    {
+      "epoch": 2.903225806451613,
+      "grad_norm": 1.234466791152954,
+      "learning_rate": 1.2293421593705585e-05,
+      "loss": 0.1931,
+      "step": 990
+    },
+    {
+      "epoch": 2.9178885630498534,
+      "grad_norm": 1.1238077878952026,
+      "learning_rate": 1.2150406045411564e-05,
+      "loss": 0.1982,
+      "step": 995
+    },
+    {
+      "epoch": 2.932551319648094,
+      "grad_norm": 1.229047417640686,
+      "learning_rate": 1.2007658738408467e-05,
+      "loss": 0.1866,
+      "step": 1000
+    },
+    {
+      "epoch": 2.9472140762463344,
+      "grad_norm": 1.3575365543365479,
+      "learning_rate": 1.1865193109951758e-05,
+      "loss": 0.1826,
+      "step": 1005
+    },
+    {
+      "epoch": 2.961876832844575,
+      "grad_norm": 1.1395304203033447,
+      "learning_rate": 1.172302257078159e-05,
+      "loss": 0.1811,
+      "step": 1010
+    },
+    {
+      "epoch": 2.9765395894428153,
+      "grad_norm": 1.1130917072296143,
+      "learning_rate": 1.1581160503860433e-05,
+      "loss": 0.1948,
+      "step": 1015
+    },
+    {
+      "epoch": 2.9912023460410557,
+      "grad_norm": 1.1696594953536987,
+      "learning_rate": 1.1439620263113305e-05,
+      "loss": 0.1733,
+      "step": 1020
+    },
+    {
+      "epoch": 3.005865102639296,
+      "grad_norm": 0.9741109013557434,
+      "learning_rate": 1.1298415172170686e-05,
+      "loss": 0.1637,
+      "step": 1025
+    },
+    {
+      "epoch": 3.0205278592375366,
+      "grad_norm": 1.0648201704025269,
+      "learning_rate": 1.1157558523114361e-05,
+      "loss": 0.1536,
+      "step": 1030
+    },
+    {
+      "epoch": 3.035190615835777,
+      "grad_norm": 1.122677206993103,
+      "learning_rate": 1.1017063575226167e-05,
+      "loss": 0.1379,
+      "step": 1035
+    },
+    {
+      "epoch": 3.0498533724340176,
+      "grad_norm": 0.9806665182113647,
+      "learning_rate": 1.0876943553739868e-05,
+      "loss": 0.1334,
+      "step": 1040
+    },
+    {
+      "epoch": 3.064516129032258,
+      "grad_norm": 1.0356829166412354,
+      "learning_rate": 1.0737211648596236e-05,
+      "loss": 0.1409,
+      "step": 1045
+    },
+    {
+      "epoch": 3.0791788856304985,
+      "grad_norm": 1.0322378873825073,
+      "learning_rate": 1.0597881013201395e-05,
+      "loss": 0.1447,
+      "step": 1050
+    },
+    {
+      "epoch": 3.093841642228739,
+      "grad_norm": 1.1586244106292725,
+      "learning_rate": 1.0458964763188706e-05,
+      "loss": 0.155,
+      "step": 1055
+    },
+    {
+      "epoch": 3.1085043988269794,
+      "grad_norm": 1.136451244354248,
+      "learning_rate": 1.0320475975184105e-05,
+      "loss": 0.1279,
+      "step": 1060
+    },
+    {
+      "epoch": 3.12316715542522,
+      "grad_norm": 1.1477034091949463,
+      "learning_rate": 1.0182427685575193e-05,
+      "loss": 0.1372,
+      "step": 1065
+    },
+    {
+      "epoch": 3.1378299120234603,
+      "grad_norm": 1.0353872776031494,
+      "learning_rate": 1.0044832889284048e-05,
+      "loss": 0.1695,
+      "step": 1070
+    },
+    {
+      "epoch": 3.1524926686217007,
+      "grad_norm": 1.013253092765808,
+      "learning_rate": 9.907704538543996e-06,
+      "loss": 0.1348,
+      "step": 1075
+    },
+    {
+      "epoch": 3.167155425219941,
+      "grad_norm": 1.047311544418335,
+      "learning_rate": 9.771055541680374e-06,
+      "loss": 0.1303,
+      "step": 1080
+    },
+    {
+      "epoch": 3.1818181818181817,
+      "grad_norm": 1.0538437366485596,
+      "learning_rate": 9.634898761895409e-06,
+      "loss": 0.1375,
+      "step": 1085
+    },
+    {
+      "epoch": 3.196480938416422,
+      "grad_norm": 1.062599778175354,
+      "learning_rate": 9.499247016057392e-06,
+      "loss": 0.1505,
+      "step": 1090
+    },
+    {
+      "epoch": 3.2111436950146626,
+      "grad_norm": 0.9383370876312256,
+      "learning_rate": 9.364113073494153e-06,
+      "loss": 0.1299,
+      "step": 1095
+    },
+    {
+      "epoch": 3.225806451612903,
+      "grad_norm": 0.9133987426757812,
+      "learning_rate": 9.22950965479108e-06,
+      "loss": 0.1539,
+      "step": 1100
+    },
+    {
+      "epoch": 3.2404692082111435,
+      "grad_norm": 1.1534616947174072,
+      "learning_rate": 9.09544943059367e-06,
+      "loss": 0.1568,
+      "step": 1105
+    },
+    {
+      "epoch": 3.255131964809384,
+      "grad_norm": 0.9719287753105164,
+      "learning_rate": 8.96194502041479e-06,
+      "loss": 0.1428,
+      "step": 1110
+    },
+    {
+      "epoch": 3.2697947214076244,
+      "grad_norm": 0.8950256705284119,
+      "learning_rate": 8.829008991446794e-06,
+      "loss": 0.1411,
+      "step": 1115
+    },
+    {
+      "epoch": 3.2844574780058653,
+      "grad_norm": 1.1876524686813354,
+      "learning_rate": 8.696653857378497e-06,
+      "loss": 0.1416,
+      "step": 1120
+    },
+    {
+      "epoch": 3.2991202346041058,
+      "grad_norm": 0.8630915880203247,
+      "learning_rate": 8.564892077217253e-06,
+      "loss": 0.1196,
+      "step": 1125
+    },
+    {
+      "epoch": 3.313782991202346,
+      "grad_norm": 1.1253145933151245,
+      "learning_rate": 8.43373605411613e-06,
+      "loss": 0.1265,
+      "step": 1130
+    },
+    {
+      "epoch": 3.3284457478005867,
+      "grad_norm": 1.1297721862792969,
+      "learning_rate": 8.303198134206367e-06,
+      "loss": 0.1178,
+      "step": 1135
+    },
+    {
+      "epoch": 3.343108504398827,
+      "grad_norm": 1.2541459798812866,
+      "learning_rate": 8.173290605435205e-06,
+      "loss": 0.1207,
+      "step": 1140
+    },
+    {
+      "epoch": 3.3577712609970676,
+      "grad_norm": 1.001312255859375,
+      "learning_rate": 8.044025696409164e-06,
+      "loss": 0.1276,
+      "step": 1145
+    },
+    {
+      "epoch": 3.372434017595308,
+      "grad_norm": 1.1012049913406372,
+      "learning_rate": 7.915415575242936e-06,
+      "loss": 0.138,
+      "step": 1150
+    },
+    {
+      "epoch": 3.3870967741935485,
+      "grad_norm": 1.0312981605529785,
+      "learning_rate": 7.787472348413962e-06,
+      "loss": 0.1232,
+      "step": 1155
+    },
+    {
+      "epoch": 3.401759530791789,
+      "grad_norm": 0.9043558239936829,
+      "learning_rate": 7.660208059622805e-06,
+      "loss": 0.1194,
+      "step": 1160
+    },
+    {
+      "epoch": 3.4164222873900294,
+      "grad_norm": 1.3074191808700562,
+      "learning_rate": 7.533634688659459e-06,
+      "loss": 0.1356,
+      "step": 1165
+    },
+    {
+      "epoch": 3.43108504398827,
+      "grad_norm": 1.0586715936660767,
+      "learning_rate": 7.407764150275625e-06,
+      "loss": 0.1212,
+      "step": 1170
+    },
+    {
+      "epoch": 3.4457478005865103,
+      "grad_norm": 0.9470232129096985,
+      "learning_rate": 7.2826082930631496e-06,
+      "loss": 0.1129,
+      "step": 1175
+    },
+    {
+      "epoch": 3.4604105571847508,
+      "grad_norm": 1.0220160484313965,
+      "learning_rate": 7.158178898338684e-06,
+      "loss": 0.1381,
+      "step": 1180
+    },
+    {
+      "epoch": 3.4750733137829912,
+      "grad_norm": 1.060505747795105,
+      "learning_rate": 7.034487679034668e-06,
+      "loss": 0.1166,
+      "step": 1185
+    },
+    {
+      "epoch": 3.4897360703812317,
+      "grad_norm": 0.8480039834976196,
+      "learning_rate": 6.911546278596756e-06,
+      "loss": 0.1152,
+      "step": 1190
+    },
+    {
+      "epoch": 3.504398826979472,
+      "grad_norm": 1.098563313484192,
+      "learning_rate": 6.789366269887773e-06,
+      "loss": 0.1348,
+      "step": 1195
+    },
+    {
+      "epoch": 3.5190615835777126,
+      "grad_norm": 0.9839580655097961,
+      "learning_rate": 6.667959154098351e-06,
+      "loss": 0.1223,
+      "step": 1200
+    },
+    {
+      "epoch": 3.533724340175953,
+      "grad_norm": 1.0041512250900269,
+      "learning_rate": 6.547336359664249e-06,
+      "loss": 0.1085,
+      "step": 1205
+    },
+    {
+      "epoch": 3.5483870967741935,
+      "grad_norm": 1.0116904973983765,
+      "learning_rate": 6.42750924119058e-06,
+      "loss": 0.126,
+      "step": 1210
+    },
+    {
+      "epoch": 3.563049853372434,
+      "grad_norm": 0.9084716439247131,
+      "learning_rate": 6.308489078382969e-06,
+      "loss": 0.1163,
+      "step": 1215
+    },
+    {
+      "epoch": 3.5777126099706744,
+      "grad_norm": 1.0130903720855713,
+      "learning_rate": 6.190287074985742e-06,
+      "loss": 0.1255,
+      "step": 1220
+    },
+    {
+      "epoch": 3.592375366568915,
+      "grad_norm": 0.8881044387817383,
+      "learning_rate": 6.07291435772731e-06,
+      "loss": 0.1297,
+      "step": 1225
+    },
+    {
+      "epoch": 3.6070381231671553,
+      "grad_norm": 0.9945613741874695,
+      "learning_rate": 5.9563819752727385e-06,
+      "loss": 0.1165,
+      "step": 1230
+    },
+    {
+      "epoch": 3.621700879765396,
+      "grad_norm": 0.9453412294387817,
+      "learning_rate": 5.840700897183728e-06,
+      "loss": 0.1272,
+      "step": 1235
+    },
+    {
+      "epoch": 3.6363636363636362,
+      "grad_norm": 0.9602278470993042,
+      "learning_rate": 5.725882012886007e-06,
+      "loss": 0.1109,
+      "step": 1240
+    },
+    {
+      "epoch": 3.6510263929618767,
+      "grad_norm": 0.827931821346283,
+      "learning_rate": 5.6119361306442744e-06,
+      "loss": 0.1139,
+      "step": 1245
+    },
+    {
+      "epoch": 3.665689149560117,
+      "grad_norm": 1.1056544780731201,
+      "learning_rate": 5.498873976544777e-06,
+      "loss": 0.1246,
+      "step": 1250
+    },
+    {
+      "epoch": 3.6803519061583576,
+      "grad_norm": 0.9266889095306396,
+      "learning_rate": 5.386706193485662e-06,
+      "loss": 0.1269,
+      "step": 1255
+    },
+    {
+      "epoch": 3.6950146627565985,
+      "grad_norm": 0.8673803806304932,
+      "learning_rate": 5.275443340175074e-06,
+      "loss": 0.1101,
+      "step": 1260
+    },
+    {
+      "epoch": 3.709677419354839,
+      "grad_norm": 0.8577314615249634,
+      "learning_rate": 5.165095890137271e-06,
+      "loss": 0.1127,
+      "step": 1265
+    },
+    {
+      "epoch": 3.7243401759530794,
+      "grad_norm": 0.9616947174072266,
+      "learning_rate": 5.055674230726706e-06,
+      "loss": 0.128,
+      "step": 1270
+    },
+    {
+      "epoch": 3.73900293255132,
+      "grad_norm": 0.9855098724365234,
+      "learning_rate": 4.947188662150223e-06,
+      "loss": 0.1175,
+      "step": 1275
+    },
+    {
+      "epoch": 3.7536656891495603,
+      "grad_norm": 1.0042691230773926,
+      "learning_rate": 4.839649396497495e-06,
+      "loss": 0.1293,
+      "step": 1280
+    },
+    {
+      "epoch": 3.768328445747801,
+      "grad_norm": 0.9086982011795044,
+      "learning_rate": 4.733066556779691e-06,
+      "loss": 0.1231,
+      "step": 1285
+    },
+    {
+      "epoch": 3.7829912023460412,
+      "grad_norm": 0.8819266557693481,
+      "learning_rate": 4.627450175976588e-06,
+      "loss": 0.1074,
+      "step": 1290
+    },
+    {
+      "epoch": 3.7976539589442817,
+      "grad_norm": 0.9197854399681091,
+      "learning_rate": 4.522810196092126e-06,
+      "loss": 0.1139,
+      "step": 1295
+    },
+    {
+      "epoch": 3.812316715542522,
+      "grad_norm": 1.0163944959640503,
+      "learning_rate": 4.4191564672185384e-06,
+      "loss": 0.106,
+      "step": 1300
+    },
+    {
+      "epoch": 3.8269794721407626,
+      "grad_norm": 0.9286377429962158,
+      "learning_rate": 4.316498746609145e-06,
+      "loss": 0.1134,
+      "step": 1305
+    },
+    {
+      "epoch": 3.841642228739003,
+      "grad_norm": 0.9762731790542603,
+      "learning_rate": 4.214846697759832e-06,
+      "loss": 0.1097,
+      "step": 1310
+    },
+    {
+      "epoch": 3.8563049853372435,
+      "grad_norm": 0.9714452028274536,
+      "learning_rate": 4.114209889499447e-06,
+      "loss": 0.1304,
+      "step": 1315
+    },
+    {
+      "epoch": 3.870967741935484,
+      "grad_norm": 0.9020538926124573,
+      "learning_rate": 4.014597795089019e-06,
+      "loss": 0.1189,
+      "step": 1320
+    },
+    {
+      "epoch": 3.8856304985337244,
+      "grad_norm": 0.9676201343536377,
+      "learning_rate": 3.91601979133002e-06,
+      "loss": 0.1176,
+      "step": 1325
+    },
+    {
+      "epoch": 3.900293255131965,
+      "grad_norm": 1.0289324522018433,
+      "learning_rate": 3.818485157681712e-06,
+      "loss": 0.1103,
+      "step": 1330
+    },
+    {
+      "epoch": 3.9149560117302054,
+      "grad_norm": 0.8782296776771545,
+      "learning_rate": 3.7220030753875992e-06,
+      "loss": 0.1088,
+      "step": 1335
+    },
+    {
+      "epoch": 3.929618768328446,
+      "grad_norm": 0.8449089527130127,
+      "learning_rate": 3.626582626611235e-06,
+      "loss": 0.1111,
+      "step": 1340
+    },
+    {
+      "epoch": 3.9442815249266863,
+      "grad_norm": 0.8565375804901123,
+      "learning_rate": 3.5322327935812294e-06,
+      "loss": 0.0938,
+      "step": 1345
+    },
+    {
+      "epoch": 3.9589442815249267,
+      "grad_norm": 0.8982692360877991,
+      "learning_rate": 3.438962457745749e-06,
+      "loss": 0.106,
+      "step": 1350
+    },
+    {
+      "epoch": 3.973607038123167,
+      "grad_norm": 0.8681496977806091,
+      "learning_rate": 3.3467803989364793e-06,
+      "loss": 0.1062,
+      "step": 1355
+    },
+    {
+      "epoch": 3.9882697947214076,
+      "grad_norm": 0.8504953384399414,
+      "learning_rate": 3.2556952945421427e-06,
+      "loss": 0.1151,
+      "step": 1360
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1705,
+  "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": 1.996675708755116e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

10_128_e5_3e-5/checkpoint-1364/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb5751790d9aa12da2be1b50c556f1abeb2ac2b9c9885833cb0d870294e55895
+size 7736

10_128_e5_3e-5/checkpoint-1364/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)

10_128_e5_3e-5/checkpoint-1705/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

10_128_e5_3e-5/checkpoint-1705/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": [
+    "k_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

10_128_e5_3e-5/checkpoint-1705/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3929204f17f386933ec44c7fa992e117f3b3af4aa5cbada53780378c0cd00e5
+size 791751704

10_128_e5_3e-5/checkpoint-1705/latest

@@ -0,0 +1 @@
+global_step1705

10_128_e5_3e-5/checkpoint-1705/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4add6bbeebd70f53da1b914dd918fcf520ba6957bb8cba9535bb5e48a2fae32
+size 15920

10_128_e5_3e-5/checkpoint-1705/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ca22067e2d59c3d0d5c5ca47395f9e50a09d8329bd5025c9844d0f19603b39b
+size 15920

10_128_e5_3e-5/checkpoint-1705/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:970ca53d743c73b50036f2bb779eb7b4d17d9131bbba673172553cc5ff830c62
+size 15920