Upload folder using huggingface_hub

base/23_128_e3_3e-5/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

base/23_128_e3_3e-5/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",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "up_proj",
+    "v_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

base/23_128_e3_3e-5/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df3df34af52463ca41f56d2a33aeb4ec651a86afd17bde1e11829b752c9efd2b
+size 791751704

base/23_128_e3_3e-5/latest

@@ -0,0 +1 @@
+global_step1182

base/23_128_e3_3e-5/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b6e6a796e951b9357afa1252fe3a47b50167f69eca90b848c5f6ecf4c5c00c1
+size 16389

base/23_128_e3_3e-5/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9b7ab48c76cdbacb2b3a36509ecf263746e8bf08ac460c01bac7ef60f8f7bed
+size 16389

base/23_128_e3_3e-5/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:259de05b364a4c3b69fdd3a8ad5111dc0094aeb1fcc566955a287935125190ee
+size 16389

base/23_128_e3_3e-5/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b88c23791ef6f8ac2b399db38e8d6333c0bcd86cded0740bdb04e22ac3216341
+size 16389

base/23_128_e3_3e-5/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f3fc00718583da106f5850f8557c38c5dde498f6ca3ce8684f8f076a774b301
+size 16389

base/23_128_e3_3e-5/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:486077d00e408b23ab472c37925e02e739542479af15d0b8e283af5893062f28
+size 16389

base/23_128_e3_3e-5/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a86e484104e8f92f837eba98c2adbfee62733e51def9a96181e54769dce58f40
+size 16389

base/23_128_e3_3e-5/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c01f6cc9be53a7c6c82dc694afc00b29d8b695d89153e993f2a182f3ac6fa93
+size 16389

base/23_128_e3_3e-5/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ae31a98fd6d102b433d467e2dc282061bf33dd19afc095df32aa7d0c7a29274
+size 1401

base/23_128_e3_3e-5/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
+  }
+}

base/23_128_e3_3e-5/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
+}

base/23_128_e3_3e-5/trainer_state.json

@@ -0,0 +1,1686 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1182,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.012690355329949238,
+      "grad_norm": 0.9447579383850098,
+      "learning_rate": 2e-06,
+      "loss": 1.2873,
+      "step": 5
+    },
+    {
+      "epoch": 0.025380710659898477,
+      "grad_norm": 0.672385036945343,
+      "learning_rate": 4.5e-06,
+      "loss": 1.2804,
+      "step": 10
+    },
+    {
+      "epoch": 0.03807106598984772,
+      "grad_norm": 0.5162912607192993,
+      "learning_rate": 7e-06,
+      "loss": 1.2845,
+      "step": 15
+    },
+    {
+      "epoch": 0.050761421319796954,
+      "grad_norm": 0.4344571828842163,
+      "learning_rate": 9.5e-06,
+      "loss": 1.2721,
+      "step": 20
+    },
+    {
+      "epoch": 0.06345177664974619,
+      "grad_norm": 0.5180246829986572,
+      "learning_rate": 1.2e-05,
+      "loss": 1.276,
+      "step": 25
+    },
+    {
+      "epoch": 0.07614213197969544,
+      "grad_norm": 0.4248524010181427,
+      "learning_rate": 1.45e-05,
+      "loss": 1.1997,
+      "step": 30
+    },
+    {
+      "epoch": 0.08883248730964467,
+      "grad_norm": 0.43036776781082153,
+      "learning_rate": 1.7e-05,
+      "loss": 1.2517,
+      "step": 35
+    },
+    {
+      "epoch": 0.10152284263959391,
+      "grad_norm": 0.43224117159843445,
+      "learning_rate": 1.95e-05,
+      "loss": 1.2309,
+      "step": 40
+    },
+    {
+      "epoch": 0.11421319796954314,
+      "grad_norm": 0.41242945194244385,
+      "learning_rate": 2.2e-05,
+      "loss": 1.1913,
+      "step": 45
+    },
+    {
+      "epoch": 0.12690355329949238,
+      "grad_norm": 0.4566176235675812,
+      "learning_rate": 2.45e-05,
+      "loss": 1.1915,
+      "step": 50
+    },
+    {
+      "epoch": 0.13959390862944163,
+      "grad_norm": 0.4279966652393341,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 1.1724,
+      "step": 55
+    },
+    {
+      "epoch": 0.15228426395939088,
+      "grad_norm": 0.4518648684024811,
+      "learning_rate": 2.95e-05,
+      "loss": 1.096,
+      "step": 60
+    },
+    {
+      "epoch": 0.1649746192893401,
+      "grad_norm": 0.5022394061088562,
+      "learning_rate": 2.9999059214234348e-05,
+      "loss": 1.1295,
+      "step": 65
+    },
+    {
+      "epoch": 0.17766497461928935,
+      "grad_norm": 0.48533013463020325,
+      "learning_rate": 2.99952374743146e-05,
+      "loss": 1.1065,
+      "step": 70
+    },
+    {
+      "epoch": 0.19035532994923857,
+      "grad_norm": 0.519172728061676,
+      "learning_rate": 2.9988476729599466e-05,
+      "loss": 1.0699,
+      "step": 75
+    },
+    {
+      "epoch": 0.20304568527918782,
+      "grad_norm": 0.5410915017127991,
+      "learning_rate": 2.9978778305167067e-05,
+      "loss": 1.1086,
+      "step": 80
+    },
+    {
+      "epoch": 0.21573604060913706,
+      "grad_norm": 0.5492595434188843,
+      "learning_rate": 2.9966144101868642e-05,
+      "loss": 1.0151,
+      "step": 85
+    },
+    {
+      "epoch": 0.22842639593908629,
+      "grad_norm": 0.583027720451355,
+      "learning_rate": 2.995057659595603e-05,
+      "loss": 1.037,
+      "step": 90
+    },
+    {
+      "epoch": 0.24111675126903553,
+      "grad_norm": 0.6004746556282043,
+      "learning_rate": 2.993207883859627e-05,
+      "loss": 1.0161,
+      "step": 95
+    },
+    {
+      "epoch": 0.25380710659898476,
+      "grad_norm": 0.6045013666152954,
+      "learning_rate": 2.991065445527364e-05,
+      "loss": 0.9897,
+      "step": 100
+    },
+    {
+      "epoch": 0.26649746192893403,
+      "grad_norm": 0.589195191860199,
+      "learning_rate": 2.9886307645079037e-05,
+      "loss": 0.9608,
+      "step": 105
+    },
+    {
+      "epoch": 0.27918781725888325,
+      "grad_norm": 0.693678617477417,
+      "learning_rate": 2.985904317988701e-05,
+      "loss": 0.9702,
+      "step": 110
+    },
+    {
+      "epoch": 0.2918781725888325,
+      "grad_norm": 0.6283678412437439,
+      "learning_rate": 2.982886640342046e-05,
+      "loss": 0.9344,
+      "step": 115
+    },
+    {
+      "epoch": 0.30456852791878175,
+      "grad_norm": 0.950693666934967,
+      "learning_rate": 2.9795783230203305e-05,
+      "loss": 0.9218,
+      "step": 120
+    },
+    {
+      "epoch": 0.31725888324873097,
+      "grad_norm": 0.913914680480957,
+      "learning_rate": 2.9759800144401265e-05,
+      "loss": 0.8794,
+      "step": 125
+    },
+    {
+      "epoch": 0.3299492385786802,
+      "grad_norm": 0.7479637861251831,
+      "learning_rate": 2.9720924198550978e-05,
+      "loss": 0.866,
+      "step": 130
+    },
+    {
+      "epoch": 0.3426395939086294,
+      "grad_norm": 0.7791286110877991,
+      "learning_rate": 2.967916301217774e-05,
+      "loss": 0.8314,
+      "step": 135
+    },
+    {
+      "epoch": 0.3553299492385787,
+      "grad_norm": 0.8646045327186584,
+      "learning_rate": 2.9634524770302118e-05,
+      "loss": 0.8242,
+      "step": 140
+    },
+    {
+      "epoch": 0.3680203045685279,
+      "grad_norm": 0.8183127641677856,
+      "learning_rate": 2.9587018221835692e-05,
+      "loss": 0.8425,
+      "step": 145
+    },
+    {
+      "epoch": 0.38071065989847713,
+      "grad_norm": 0.7740488648414612,
+      "learning_rate": 2.953665267786634e-05,
+      "loss": 0.8075,
+      "step": 150
+    },
+    {
+      "epoch": 0.3934010152284264,
+      "grad_norm": 0.8975210189819336,
+      "learning_rate": 2.9483438009833267e-05,
+      "loss": 0.7802,
+      "step": 155
+    },
+    {
+      "epoch": 0.40609137055837563,
+      "grad_norm": 1.21845281124115,
+      "learning_rate": 2.942738464759229e-05,
+      "loss": 0.8045,
+      "step": 160
+    },
+    {
+      "epoch": 0.41878172588832485,
+      "grad_norm": 0.9910733699798584,
+      "learning_rate": 2.9368503577371562e-05,
+      "loss": 0.7205,
+      "step": 165
+    },
+    {
+      "epoch": 0.43147208121827413,
+      "grad_norm": 0.8701192736625671,
+      "learning_rate": 2.9306806339618394e-05,
+      "loss": 0.7298,
+      "step": 170
+    },
+    {
+      "epoch": 0.44416243654822335,
+      "grad_norm": 0.9356279969215393,
+      "learning_rate": 2.9242305026737313e-05,
+      "loss": 0.726,
+      "step": 175
+    },
+    {
+      "epoch": 0.45685279187817257,
+      "grad_norm": 1.0532931089401245,
+      "learning_rate": 2.9175012280720024e-05,
+      "loss": 0.7903,
+      "step": 180
+    },
+    {
+      "epoch": 0.46954314720812185,
+      "grad_norm": 0.927617073059082,
+      "learning_rate": 2.910494129066766e-05,
+      "loss": 0.7741,
+      "step": 185
+    },
+    {
+      "epoch": 0.48223350253807107,
+      "grad_norm": 0.9918464422225952,
+      "learning_rate": 2.9032105790205696e-05,
+      "loss": 0.6901,
+      "step": 190
+    },
+    {
+      "epoch": 0.4949238578680203,
+      "grad_norm": 0.9750902056694031,
+      "learning_rate": 2.8956520054792306e-05,
+      "loss": 0.7065,
+      "step": 195
+    },
+    {
+      "epoch": 0.5076142131979695,
+      "grad_norm": 0.9423856735229492,
+      "learning_rate": 2.887819889892037e-05,
+      "loss": 0.6643,
+      "step": 200
+    },
+    {
+      "epoch": 0.5203045685279187,
+      "grad_norm": 0.9457867741584778,
+      "learning_rate": 2.8797157673213918e-05,
+      "loss": 0.6407,
+      "step": 205
+    },
+    {
+      "epoch": 0.5329949238578681,
+      "grad_norm": 1.43992280960083,
+      "learning_rate": 2.871341226141946e-05,
+      "loss": 0.6492,
+      "step": 210
+    },
+    {
+      "epoch": 0.5456852791878173,
+      "grad_norm": 1.1639478206634521,
+      "learning_rate": 2.8626979077292856e-05,
+      "loss": 0.6793,
+      "step": 215
+    },
+    {
+      "epoch": 0.5583756345177665,
+      "grad_norm": 0.98136305809021,
+      "learning_rate": 2.853787506138226e-05,
+      "loss": 0.6475,
+      "step": 220
+    },
+    {
+      "epoch": 0.5710659898477157,
+      "grad_norm": 1.4634478092193604,
+      "learning_rate": 2.844611767770787e-05,
+      "loss": 0.597,
+      "step": 225
+    },
+    {
+      "epoch": 0.583756345177665,
+      "grad_norm": 1.1080286502838135,
+      "learning_rate": 2.835172491033903e-05,
+      "loss": 0.604,
+      "step": 230
+    },
+    {
+      "epoch": 0.5964467005076142,
+      "grad_norm": 1.0195153951644897,
+      "learning_rate": 2.825471525986945e-05,
+      "loss": 0.6248,
+      "step": 235
+    },
+    {
+      "epoch": 0.6091370558375635,
+      "grad_norm": 0.9799978733062744,
+      "learning_rate": 2.815510773979113e-05,
+      "loss": 0.5685,
+      "step": 240
+    },
+    {
+      "epoch": 0.6218274111675127,
+      "grad_norm": 1.087212324142456,
+      "learning_rate": 2.8052921872767832e-05,
+      "loss": 0.5918,
+      "step": 245
+    },
+    {
+      "epoch": 0.6345177664974619,
+      "grad_norm": 1.1969616413116455,
+      "learning_rate": 2.7948177686808695e-05,
+      "loss": 0.5964,
+      "step": 250
+    },
+    {
+      "epoch": 0.6472081218274112,
+      "grad_norm": 1.1130094528198242,
+      "learning_rate": 2.7840895711342836e-05,
+      "loss": 0.5766,
+      "step": 255
+    },
+    {
+      "epoch": 0.6598984771573604,
+      "grad_norm": 1.1325479745864868,
+      "learning_rate": 2.7731096973195664e-05,
+      "loss": 0.5451,
+      "step": 260
+    },
+    {
+      "epoch": 0.6725888324873096,
+      "grad_norm": 1.190353512763977,
+      "learning_rate": 2.7618802992467718e-05,
+      "loss": 0.5484,
+      "step": 265
+    },
+    {
+      "epoch": 0.6852791878172588,
+      "grad_norm": 1.0955568552017212,
+      "learning_rate": 2.750403577831679e-05,
+      "loss": 0.5522,
+      "step": 270
+    },
+    {
+      "epoch": 0.6979695431472082,
+      "grad_norm": 1.0358052253723145,
+      "learning_rate": 2.738681782464426e-05,
+      "loss": 0.5561,
+      "step": 275
+    },
+    {
+      "epoch": 0.7106598984771574,
+      "grad_norm": 1.1491012573242188,
+      "learning_rate": 2.726717210568634e-05,
+      "loss": 0.5652,
+      "step": 280
+    },
+    {
+      "epoch": 0.7233502538071066,
+      "grad_norm": 1.3060334920883179,
+      "learning_rate": 2.7145122071511253e-05,
+      "loss": 0.4775,
+      "step": 285
+    },
+    {
+      "epoch": 0.7360406091370558,
+      "grad_norm": 0.9798694849014282,
+      "learning_rate": 2.7020691643423105e-05,
+      "loss": 0.5621,
+      "step": 290
+    },
+    {
+      "epoch": 0.748730964467005,
+      "grad_norm": 1.2010518312454224,
+      "learning_rate": 2.6893905209273405e-05,
+      "loss": 0.5151,
+      "step": 295
+    },
+    {
+      "epoch": 0.7614213197969543,
+      "grad_norm": 1.2016711235046387,
+      "learning_rate": 2.676478761868114e-05,
+      "loss": 0.4906,
+      "step": 300
+    },
+    {
+      "epoch": 0.7741116751269036,
+      "grad_norm": 1.219832181930542,
+      "learning_rate": 2.663336417816238e-05,
+      "loss": 0.5293,
+      "step": 305
+    },
+    {
+      "epoch": 0.7868020304568528,
+      "grad_norm": 1.1408915519714355,
+      "learning_rate": 2.649966064617025e-05,
+      "loss": 0.5006,
+      "step": 310
+    },
+    {
+      "epoch": 0.799492385786802,
+      "grad_norm": 1.169193148612976,
+      "learning_rate": 2.6363703228046455e-05,
+      "loss": 0.532,
+      "step": 315
+    },
+    {
+      "epoch": 0.8121827411167513,
+      "grad_norm": 1.0420085191726685,
+      "learning_rate": 2.6225518570885083e-05,
+      "loss": 0.534,
+      "step": 320
+    },
+    {
+      "epoch": 0.8248730964467005,
+      "grad_norm": 1.147485613822937,
+      "learning_rate": 2.6085133758309887e-05,
+      "loss": 0.4802,
+      "step": 325
+    },
+    {
+      "epoch": 0.8375634517766497,
+      "grad_norm": 1.11233389377594,
+      "learning_rate": 2.5942576305166044e-05,
+      "loss": 0.4842,
+      "step": 330
+    },
+    {
+      "epoch": 0.850253807106599,
+      "grad_norm": 1.0376954078674316,
+      "learning_rate": 2.5797874152127325e-05,
+      "loss": 0.4971,
+      "step": 335
+    },
+    {
+      "epoch": 0.8629441624365483,
+      "grad_norm": 1.7565768957138062,
+      "learning_rate": 2.5651055660219847e-05,
+      "loss": 0.4627,
+      "step": 340
+    },
+    {
+      "epoch": 0.8756345177664975,
+      "grad_norm": 1.2290433645248413,
+      "learning_rate": 2.550214960526344e-05,
+      "loss": 0.4032,
+      "step": 345
+    },
+    {
+      "epoch": 0.8883248730964467,
+      "grad_norm": 1.3079357147216797,
+      "learning_rate": 2.5351185172231683e-05,
+      "loss": 0.4487,
+      "step": 350
+    },
+    {
+      "epoch": 0.9010152284263959,
+      "grad_norm": 1.1647276878356934,
+      "learning_rate": 2.5198191949531785e-05,
+      "loss": 0.4466,
+      "step": 355
+    },
+    {
+      "epoch": 0.9137055837563451,
+      "grad_norm": 1.0520508289337158,
+      "learning_rate": 2.5043199923205382e-05,
+      "loss": 0.4308,
+      "step": 360
+    },
+    {
+      "epoch": 0.9263959390862944,
+      "grad_norm": 1.2066047191619873,
+      "learning_rate": 2.488623947105138e-05,
+      "loss": 0.4161,
+      "step": 365
+    },
+    {
+      "epoch": 0.9390862944162437,
+      "grad_norm": 1.2423921823501587,
+      "learning_rate": 2.4727341356672057e-05,
+      "loss": 0.4169,
+      "step": 370
+    },
+    {
+      "epoch": 0.9517766497461929,
+      "grad_norm": 1.1672241687774658,
+      "learning_rate": 2.4566536723443486e-05,
+      "loss": 0.4397,
+      "step": 375
+    },
+    {
+      "epoch": 0.9644670050761421,
+      "grad_norm": 1.381141185760498,
+      "learning_rate": 2.44038570884116e-05,
+      "loss": 0.4128,
+      "step": 380
+    },
+    {
+      "epoch": 0.9771573604060914,
+      "grad_norm": 1.2625432014465332,
+      "learning_rate": 2.4239334336114955e-05,
+      "loss": 0.3926,
+      "step": 385
+    },
+    {
+      "epoch": 0.9898477157360406,
+      "grad_norm": 1.259828805923462,
+      "learning_rate": 2.4073000712335516e-05,
+      "loss": 0.4054,
+      "step": 390
+    },
+    {
+      "epoch": 1.00253807106599,
+      "grad_norm": 1.1191716194152832,
+      "learning_rate": 2.3904888817778583e-05,
+      "loss": 0.398,
+      "step": 395
+    },
+    {
+      "epoch": 1.015228426395939,
+      "grad_norm": 1.604065179824829,
+      "learning_rate": 2.373503160168321e-05,
+      "loss": 0.3259,
+      "step": 400
+    },
+    {
+      "epoch": 1.0279187817258884,
+      "grad_norm": 1.2008863687515259,
+      "learning_rate": 2.35634623553643e-05,
+      "loss": 0.3231,
+      "step": 405
+    },
+    {
+      "epoch": 1.0406091370558375,
+      "grad_norm": 1.2370153665542603,
+      "learning_rate": 2.3390214705687607e-05,
+      "loss": 0.3305,
+      "step": 410
+    },
+    {
+      "epoch": 1.0532994923857868,
+      "grad_norm": 1.3695017099380493,
+      "learning_rate": 2.321532260847905e-05,
+      "loss": 0.3386,
+      "step": 415
+    },
+    {
+      "epoch": 1.0659898477157361,
+      "grad_norm": 1.1977289915084839,
+      "learning_rate": 2.3038820341869486e-05,
+      "loss": 0.3545,
+      "step": 420
+    },
+    {
+      "epoch": 1.0786802030456852,
+      "grad_norm": 1.303430199623108,
+      "learning_rate": 2.286074249957634e-05,
+      "loss": 0.3152,
+      "step": 425
+    },
+    {
+      "epoch": 1.0913705583756346,
+      "grad_norm": 1.0870370864868164,
+      "learning_rate": 2.2681123984123388e-05,
+      "loss": 0.324,
+      "step": 430
+    },
+    {
+      "epoch": 1.1040609137055837,
+      "grad_norm": 1.257110595703125,
+      "learning_rate": 2.25e-05,
+      "loss": 0.3039,
+      "step": 435
+    },
+    {
+      "epoch": 1.116751269035533,
+      "grad_norm": 1.1624414920806885,
+      "learning_rate": 2.231740604676121e-05,
+      "loss": 0.2589,
+      "step": 440
+    },
+    {
+      "epoch": 1.1294416243654823,
+      "grad_norm": 1.2051514387130737,
+      "learning_rate": 2.2133377912069934e-05,
+      "loss": 0.3056,
+      "step": 445
+    },
+    {
+      "epoch": 1.1421319796954315,
+      "grad_norm": 1.2107893228530884,
+      "learning_rate": 2.194795166468275e-05,
+      "loss": 0.285,
+      "step": 450
+    },
+    {
+      "epoch": 1.1548223350253808,
+      "grad_norm": 1.126853108406067,
+      "learning_rate": 2.1761163647380583e-05,
+      "loss": 0.2793,
+      "step": 455
+    },
+    {
+      "epoch": 1.16751269035533,
+      "grad_norm": 1.0835230350494385,
+      "learning_rate": 2.157305046984563e-05,
+      "loss": 0.3078,
+      "step": 460
+    },
+    {
+      "epoch": 1.1802030456852792,
+      "grad_norm": 1.191496729850769,
+      "learning_rate": 2.1383649001486055e-05,
+      "loss": 0.309,
+      "step": 465
+    },
+    {
+      "epoch": 1.1928934010152283,
+      "grad_norm": 1.18590247631073,
+      "learning_rate": 2.1192996364209708e-05,
+      "loss": 0.2919,
+      "step": 470
+    },
+    {
+      "epoch": 1.2055837563451777,
+      "grad_norm": 1.7720372676849365,
+      "learning_rate": 2.10011299251484e-05,
+      "loss": 0.2851,
+      "step": 475
+    },
+    {
+      "epoch": 1.218274111675127,
+      "grad_norm": 1.2300515174865723,
+      "learning_rate": 2.080808728933409e-05,
+      "loss": 0.2903,
+      "step": 480
+    },
+    {
+      "epoch": 1.2309644670050761,
+      "grad_norm": 1.2837207317352295,
+      "learning_rate": 2.061390629232846e-05,
+      "loss": 0.2786,
+      "step": 485
+    },
+    {
+      "epoch": 1.2436548223350254,
+      "grad_norm": 1.1217468976974487,
+      "learning_rate": 2.0418624992807297e-05,
+      "loss": 0.3108,
+      "step": 490
+    },
+    {
+      "epoch": 1.2563451776649746,
+      "grad_norm": 1.1909114122390747,
+      "learning_rate": 2.0222281665101145e-05,
+      "loss": 0.2447,
+      "step": 495
+    },
+    {
+      "epoch": 1.2690355329949239,
+      "grad_norm": 1.1922444105148315,
+      "learning_rate": 2.0024914791693707e-05,
+      "loss": 0.2923,
+      "step": 500
+    },
+    {
+      "epoch": 1.281725888324873,
+      "grad_norm": 1.3241158723831177,
+      "learning_rate": 1.982656305567942e-05,
+      "loss": 0.2589,
+      "step": 505
+    },
+    {
+      "epoch": 1.2944162436548223,
+      "grad_norm": 1.0729427337646484,
+      "learning_rate": 1.962726533318175e-05,
+      "loss": 0.2403,
+      "step": 510
+    },
+    {
+      "epoch": 1.3071065989847717,
+      "grad_norm": 1.2227731943130493,
+      "learning_rate": 1.94270606857336e-05,
+      "loss": 0.2701,
+      "step": 515
+    },
+    {
+      "epoch": 1.3197969543147208,
+      "grad_norm": 1.3616466522216797,
+      "learning_rate": 1.9225988352621445e-05,
+      "loss": 0.2722,
+      "step": 520
+    },
+    {
+      "epoch": 1.33248730964467,
+      "grad_norm": 1.4032851457595825,
+      "learning_rate": 1.9024087743194566e-05,
+      "loss": 0.2386,
+      "step": 525
+    },
+    {
+      "epoch": 1.3451776649746192,
+      "grad_norm": 1.1879099607467651,
+      "learning_rate": 1.8821398429140996e-05,
+      "loss": 0.2697,
+      "step": 530
+    },
+    {
+      "epoch": 1.3578680203045685,
+      "grad_norm": 1.3883295059204102,
+      "learning_rate": 1.8617960136731627e-05,
+      "loss": 0.2331,
+      "step": 535
+    },
+    {
+      "epoch": 1.3705583756345177,
+      "grad_norm": 1.365457534790039,
+      "learning_rate": 1.841381273903402e-05,
+      "loss": 0.2389,
+      "step": 540
+    },
+    {
+      "epoch": 1.383248730964467,
+      "grad_norm": 1.2199925184249878,
+      "learning_rate": 1.8208996248097462e-05,
+      "loss": 0.2607,
+      "step": 545
+    },
+    {
+      "epoch": 1.3959390862944163,
+      "grad_norm": 1.192286491394043,
+      "learning_rate": 1.8003550807110744e-05,
+      "loss": 0.2404,
+      "step": 550
+    },
+    {
+      "epoch": 1.4086294416243654,
+      "grad_norm": 1.077062726020813,
+      "learning_rate": 1.7797516682534294e-05,
+      "loss": 0.2297,
+      "step": 555
+    },
+    {
+      "epoch": 1.4213197969543148,
+      "grad_norm": 1.2271517515182495,
+      "learning_rate": 1.759093425620809e-05,
+      "loss": 0.2417,
+      "step": 560
+    },
+    {
+      "epoch": 1.434010152284264,
+      "grad_norm": 1.3175173997879028,
+      "learning_rate": 1.7383844017436998e-05,
+      "loss": 0.26,
+      "step": 565
+    },
+    {
+      "epoch": 1.4467005076142132,
+      "grad_norm": 1.0850142240524292,
+      "learning_rate": 1.717628655505503e-05,
+      "loss": 0.2267,
+      "step": 570
+    },
+    {
+      "epoch": 1.4593908629441623,
+      "grad_norm": 1.144986629486084,
+      "learning_rate": 1.6968302549470097e-05,
+      "loss": 0.2395,
+      "step": 575
+    },
+    {
+      "epoch": 1.4720812182741116,
+      "grad_norm": 1.3780667781829834,
+      "learning_rate": 1.6759932764690834e-05,
+      "loss": 0.2416,
+      "step": 580
+    },
+    {
+      "epoch": 1.484771573604061,
+      "grad_norm": 1.169228196144104,
+      "learning_rate": 1.6551218040336994e-05,
+      "loss": 0.2008,
+      "step": 585
+    },
+    {
+      "epoch": 1.49746192893401,
+      "grad_norm": 1.2531604766845703,
+      "learning_rate": 1.6342199283635086e-05,
+      "loss": 0.2008,
+      "step": 590
+    },
+    {
+      "epoch": 1.5101522842639594,
+      "grad_norm": 1.0524369478225708,
+      "learning_rate": 1.6132917461400685e-05,
+      "loss": 0.2017,
+      "step": 595
+    },
+    {
+      "epoch": 1.5228426395939088,
+      "grad_norm": 1.2968204021453857,
+      "learning_rate": 1.5923413592009145e-05,
+      "loss": 0.1984,
+      "step": 600
+    },
+    {
+      "epoch": 1.5355329949238579,
+      "grad_norm": 1.1446210145950317,
+      "learning_rate": 1.5713728737356138e-05,
+      "loss": 0.1874,
+      "step": 605
+    },
+    {
+      "epoch": 1.548223350253807,
+      "grad_norm": 1.0703312158584595,
+      "learning_rate": 1.5503903994809704e-05,
+      "loss": 0.2144,
+      "step": 610
+    },
+    {
+      "epoch": 1.5609137055837563,
+      "grad_norm": 1.1338443756103516,
+      "learning_rate": 1.5293980489155335e-05,
+      "loss": 0.1974,
+      "step": 615
+    },
+    {
+      "epoch": 1.5736040609137056,
+      "grad_norm": 1.1431093215942383,
+      "learning_rate": 1.508399936453569e-05,
+      "loss": 0.1993,
+      "step": 620
+    },
+    {
+      "epoch": 1.5862944162436547,
+      "grad_norm": 1.4238122701644897,
+      "learning_rate": 1.4874001776386501e-05,
+      "loss": 0.2072,
+      "step": 625
+    },
+    {
+      "epoch": 1.598984771573604,
+      "grad_norm": 1.2045807838439941,
+      "learning_rate": 1.4664028883370285e-05,
+      "loss": 0.213,
+      "step": 630
+    },
+    {
+      "epoch": 1.6116751269035534,
+      "grad_norm": 1.327996850013733,
+      "learning_rate": 1.4454121839309416e-05,
+      "loss": 0.1877,
+      "step": 635
+    },
+    {
+      "epoch": 1.6243654822335025,
+      "grad_norm": 1.2039096355438232,
+      "learning_rate": 1.4244321785120149e-05,
+      "loss": 0.19,
+      "step": 640
+    },
+    {
+      "epoch": 1.6370558375634516,
+      "grad_norm": 1.051845669746399,
+      "learning_rate": 1.4034669840749155e-05,
+      "loss": 0.1963,
+      "step": 645
+    },
+    {
+      "epoch": 1.649746192893401,
+      "grad_norm": 1.1482218503952026,
+      "learning_rate": 1.3825207097114201e-05,
+      "loss": 0.2067,
+      "step": 650
+    },
+    {
+      "epoch": 1.6624365482233503,
+      "grad_norm": 1.2002408504486084,
+      "learning_rate": 1.3615974608050472e-05,
+      "loss": 0.1777,
+      "step": 655
+    },
+    {
+      "epoch": 1.6751269035532994,
+      "grad_norm": 1.2326092720031738,
+      "learning_rate": 1.3407013382264229e-05,
+      "loss": 0.1899,
+      "step": 660
+    },
+    {
+      "epoch": 1.6878172588832487,
+      "grad_norm": 1.0630570650100708,
+      "learning_rate": 1.3198364375295225e-05,
+      "loss": 0.1791,
+      "step": 665
+    },
+    {
+      "epoch": 1.700507614213198,
+      "grad_norm": 1.138698697090149,
+      "learning_rate": 1.2990068481489631e-05,
+      "loss": 0.1874,
+      "step": 670
+    },
+    {
+      "epoch": 1.7131979695431472,
+      "grad_norm": 1.212302565574646,
+      "learning_rate": 1.278216652598487e-05,
+      "loss": 0.1819,
+      "step": 675
+    },
+    {
+      "epoch": 1.7258883248730963,
+      "grad_norm": 1.1460216045379639,
+      "learning_rate": 1.2574699256708104e-05,
+      "loss": 0.17,
+      "step": 680
+    },
+    {
+      "epoch": 1.7385786802030458,
+      "grad_norm": 1.0779529809951782,
+      "learning_rate": 1.2367707336389761e-05,
+      "loss": 0.151,
+      "step": 685
+    },
+    {
+      "epoch": 1.751269035532995,
+      "grad_norm": 1.1173380613327026,
+      "learning_rate": 1.2161231334593853e-05,
+      "loss": 0.1848,
+      "step": 690
+    },
+    {
+      "epoch": 1.763959390862944,
+      "grad_norm": 1.185747504234314,
+      "learning_rate": 1.1955311719766463e-05,
+      "loss": 0.1571,
+      "step": 695
+    },
+    {
+      "epoch": 1.7766497461928934,
+      "grad_norm": 1.02059006690979,
+      "learning_rate": 1.1749988851304141e-05,
+      "loss": 0.131,
+      "step": 700
+    },
+    {
+      "epoch": 1.7893401015228427,
+      "grad_norm": 1.3492285013198853,
+      "learning_rate": 1.1545302971643591e-05,
+      "loss": 0.164,
+      "step": 705
+    },
+    {
+      "epoch": 1.8020304568527918,
+      "grad_norm": 1.014811396598816,
+      "learning_rate": 1.1341294198374342e-05,
+      "loss": 0.1438,
+      "step": 710
+    },
+    {
+      "epoch": 1.8147208121827412,
+      "grad_norm": 1.2959096431732178,
+      "learning_rate": 1.1138002516375865e-05,
+      "loss": 0.1498,
+      "step": 715
+    },
+    {
+      "epoch": 1.8274111675126905,
+      "grad_norm": 1.130133032798767,
+      "learning_rate": 1.0935467769980678e-05,
+      "loss": 0.1459,
+      "step": 720
+    },
+    {
+      "epoch": 1.8401015228426396,
+      "grad_norm": 1.1135902404785156,
+      "learning_rate": 1.0733729655165056e-05,
+      "loss": 0.1423,
+      "step": 725
+    },
+    {
+      "epoch": 1.8527918781725887,
+      "grad_norm": 1.4973711967468262,
+      "learning_rate": 1.0532827711768747e-05,
+      "loss": 0.1704,
+      "step": 730
+    },
+    {
+      "epoch": 1.865482233502538,
+      "grad_norm": 1.267324686050415,
+      "learning_rate": 1.0332801315745363e-05,
+      "loss": 0.1745,
+      "step": 735
+    },
+    {
+      "epoch": 1.8781725888324874,
+      "grad_norm": 1.2987362146377563,
+      "learning_rate": 1.0133689671444817e-05,
+      "loss": 0.1504,
+      "step": 740
+    },
+    {
+      "epoch": 1.8908629441624365,
+      "grad_norm": 1.2410354614257812,
+      "learning_rate": 9.93553180392947e-06,
+      "loss": 0.1465,
+      "step": 745
+    },
+    {
+      "epoch": 1.9035532994923858,
+      "grad_norm": 1.0641611814498901,
+      "learning_rate": 9.738366551325343e-06,
+      "loss": 0.1415,
+      "step": 750
+    },
+    {
+      "epoch": 1.9162436548223352,
+      "grad_norm": 1.072670578956604,
+      "learning_rate": 9.542232557210039e-06,
+      "loss": 0.1461,
+      "step": 755
+    },
+    {
+      "epoch": 1.9289340101522843,
+      "grad_norm": 0.8728611469268799,
+      "learning_rate": 9.347168263038728e-06,
+      "loss": 0.1237,
+      "step": 760
+    },
+    {
+      "epoch": 1.9416243654822334,
+      "grad_norm": 1.1794564723968506,
+      "learning_rate": 9.15321190060981e-06,
+      "loss": 0.1293,
+      "step": 765
+    },
+    {
+      "epoch": 1.9543147208121827,
+      "grad_norm": 0.9986603260040283,
+      "learning_rate": 8.960401484571612e-06,
+      "loss": 0.1578,
+      "step": 770
+    },
+    {
+      "epoch": 1.967005076142132,
+      "grad_norm": 1.0394880771636963,
+      "learning_rate": 8.768774804971705e-06,
+      "loss": 0.1223,
+      "step": 775
+    },
+    {
+      "epoch": 1.9796954314720812,
+      "grad_norm": 1.184678554534912,
+      "learning_rate": 8.57836941985017e-06,
+      "loss": 0.1413,
+      "step": 780
+    },
+    {
+      "epoch": 1.9923857868020305,
+      "grad_norm": 1.1253178119659424,
+      "learning_rate": 8.389222647878426e-06,
+      "loss": 0.1323,
+      "step": 785
+    },
+    {
+      "epoch": 2.00507614213198,
+      "grad_norm": 1.0249272584915161,
+      "learning_rate": 8.201371561044864e-06,
+      "loss": 0.1182,
+      "step": 790
+    },
+    {
+      "epoch": 2.017766497461929,
+      "grad_norm": 1.0803836584091187,
+      "learning_rate": 8.014852977388965e-06,
+      "loss": 0.123,
+      "step": 795
+    },
+    {
+      "epoch": 2.030456852791878,
+      "grad_norm": 0.941626250743866,
+      "learning_rate": 7.82970345378503e-06,
+      "loss": 0.1131,
+      "step": 800
+    },
+    {
+      "epoch": 2.0431472081218276,
+      "grad_norm": 0.7750261425971985,
+      "learning_rate": 7.645959278777271e-06,
+      "loss": 0.0939,
+      "step": 805
+    },
+    {
+      "epoch": 2.0558375634517767,
+      "grad_norm": 0.9910224676132202,
+      "learning_rate": 7.46365646546732e-06,
+      "loss": 0.0954,
+      "step": 810
+    },
+    {
+      "epoch": 2.068527918781726,
+      "grad_norm": 1.15106999874115,
+      "learning_rate": 7.282830744455896e-06,
+      "loss": 0.1042,
+      "step": 815
+    },
+    {
+      "epoch": 2.081218274111675,
+      "grad_norm": 1.4101669788360596,
+      "learning_rate": 7.103517556839661e-06,
+      "loss": 0.0901,
+      "step": 820
+    },
+    {
+      "epoch": 2.0939086294416245,
+      "grad_norm": 0.7868049740791321,
+      "learning_rate": 6.925752047265011e-06,
+      "loss": 0.0798,
+      "step": 825
+    },
+    {
+      "epoch": 2.1065989847715736,
+      "grad_norm": 0.832548975944519,
+      "learning_rate": 6.749569057039771e-06,
+      "loss": 0.1177,
+      "step": 830
+    },
+    {
+      "epoch": 2.1192893401015227,
+      "grad_norm": 0.9674516320228577,
+      "learning_rate": 6.575003117304535e-06,
+      "loss": 0.0829,
+      "step": 835
+    },
+    {
+      "epoch": 2.1319796954314723,
+      "grad_norm": 0.8840376734733582,
+      "learning_rate": 6.402088442264615e-06,
+      "loss": 0.0895,
+      "step": 840
+    },
+    {
+      "epoch": 2.1446700507614214,
+      "grad_norm": 0.9384085536003113,
+      "learning_rate": 6.230858922484289e-06,
+      "loss": 0.1074,
+      "step": 845
+    },
+    {
+      "epoch": 2.1573604060913705,
+      "grad_norm": 0.9792207479476929,
+      "learning_rate": 6.061348118244298e-06,
+      "loss": 0.0961,
+      "step": 850
+    },
+    {
+      "epoch": 2.1700507614213196,
+      "grad_norm": 0.8903322815895081,
+      "learning_rate": 5.893589252964258e-06,
+      "loss": 0.0787,
+      "step": 855
+    },
+    {
+      "epoch": 2.182741116751269,
+      "grad_norm": 0.8336995244026184,
+      "learning_rate": 5.727615206690921e-06,
+      "loss": 0.085,
+      "step": 860
+    },
+    {
+      "epoch": 2.1954314720812182,
+      "grad_norm": 0.8919497728347778,
+      "learning_rate": 5.563458509653905e-06,
+      "loss": 0.0961,
+      "step": 865
+    },
+    {
+      "epoch": 2.2081218274111674,
+      "grad_norm": 1.1389808654785156,
+      "learning_rate": 5.401151335889819e-06,
+      "loss": 0.0826,
+      "step": 870
+    },
+    {
+      "epoch": 2.220812182741117,
+      "grad_norm": 1.1736438274383545,
+      "learning_rate": 5.240725496936373e-06,
+      "loss": 0.0871,
+      "step": 875
+    },
+    {
+      "epoch": 2.233502538071066,
+      "grad_norm": 0.8246440887451172,
+      "learning_rate": 5.082212435597352e-06,
+      "loss": 0.0905,
+      "step": 880
+    },
+    {
+      "epoch": 2.246192893401015,
+      "grad_norm": 0.9815599322319031,
+      "learning_rate": 4.925643219780053e-06,
+      "loss": 0.0857,
+      "step": 885
+    },
+    {
+      "epoch": 2.2588832487309647,
+      "grad_norm": 0.923875629901886,
+      "learning_rate": 4.771048536406012e-06,
+      "loss": 0.085,
+      "step": 890
+    },
+    {
+      "epoch": 2.271573604060914,
+      "grad_norm": 1.0129884481430054,
+      "learning_rate": 4.618458685396579e-06,
+      "loss": 0.095,
+      "step": 895
+    },
+    {
+      "epoch": 2.284263959390863,
+      "grad_norm": 1.239827036857605,
+      "learning_rate": 4.467903573734174e-06,
+      "loss": 0.0822,
+      "step": 900
+    },
+    {
+      "epoch": 2.296954314720812,
+      "grad_norm": 0.8126388192176819,
+      "learning_rate": 4.319412709600723e-06,
+      "loss": 0.0693,
+      "step": 905
+    },
+    {
+      "epoch": 2.3096446700507616,
+      "grad_norm": 0.9403148889541626,
+      "learning_rate": 4.173015196594103e-06,
+      "loss": 0.0807,
+      "step": 910
+    },
+    {
+      "epoch": 2.3223350253807107,
+      "grad_norm": 0.8209843635559082,
+      "learning_rate": 4.028739728024023e-06,
+      "loss": 0.0961,
+      "step": 915
+    },
+    {
+      "epoch": 2.33502538071066,
+      "grad_norm": 0.9164144396781921,
+      "learning_rate": 3.886614581288187e-06,
+      "loss": 0.0839,
+      "step": 920
+    },
+    {
+      "epoch": 2.347715736040609,
+      "grad_norm": 0.758331298828125,
+      "learning_rate": 3.746667612330109e-06,
+      "loss": 0.0781,
+      "step": 925
+    },
+    {
+      "epoch": 2.3604060913705585,
+      "grad_norm": 0.7769868969917297,
+      "learning_rate": 3.608926250179392e-06,
+      "loss": 0.0785,
+      "step": 930
+    },
+    {
+      "epoch": 2.3730964467005076,
+      "grad_norm": 0.8135273456573486,
+      "learning_rate": 3.4734174915758245e-06,
+      "loss": 0.0756,
+      "step": 935
+    },
+    {
+      "epoch": 2.3857868020304567,
+      "grad_norm": 0.9612864255905151,
+      "learning_rate": 3.340167895678059e-06,
+      "loss": 0.0754,
+      "step": 940
+    },
+    {
+      "epoch": 2.3984771573604062,
+      "grad_norm": 0.8055295348167419,
+      "learning_rate": 3.2092035788581907e-06,
+      "loss": 0.083,
+      "step": 945
+    },
+    {
+      "epoch": 2.4111675126903553,
+      "grad_norm": 0.7819513082504272,
+      "learning_rate": 3.0805502095829987e-06,
+      "loss": 0.0671,
+      "step": 950
+    },
+    {
+      "epoch": 2.4238578680203045,
+      "grad_norm": 0.8910360932350159,
+      "learning_rate": 2.954233003383089e-06,
+      "loss": 0.0739,
+      "step": 955
+    },
+    {
+      "epoch": 2.436548223350254,
+      "grad_norm": 0.8987388610839844,
+      "learning_rate": 2.830276717910692e-06,
+      "loss": 0.0954,
+      "step": 960
+    },
+    {
+      "epoch": 2.449238578680203,
+      "grad_norm": 0.9984980225563049,
+      "learning_rate": 2.7087056480873322e-06,
+      "loss": 0.0835,
+      "step": 965
+    },
+    {
+      "epoch": 2.4619289340101522,
+      "grad_norm": 0.703774631023407,
+      "learning_rate": 2.5895436213420566e-06,
+      "loss": 0.0717,
+      "step": 970
+    },
+    {
+      "epoch": 2.4746192893401013,
+      "grad_norm": 0.9176628589630127,
+      "learning_rate": 2.4728139929414185e-06,
+      "loss": 0.0822,
+      "step": 975
+    },
+    {
+      "epoch": 2.487309644670051,
+      "grad_norm": 0.9267148375511169,
+      "learning_rate": 2.358539641411883e-06,
+      "loss": 0.0993,
+      "step": 980
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 1.0409682989120483,
+      "learning_rate": 2.2467429640557903e-06,
+      "loss": 0.076,
+      "step": 985
+    },
+    {
+      "epoch": 2.512690355329949,
+      "grad_norm": 0.8411082625389099,
+      "learning_rate": 2.1374458725615147e-06,
+      "loss": 0.0755,
+      "step": 990
+    },
+    {
+      "epoch": 2.525380710659898,
+      "grad_norm": 0.7068350911140442,
+      "learning_rate": 2.0306697887089238e-06,
+      "loss": 0.0791,
+      "step": 995
+    },
+    {
+      "epoch": 2.5380710659898478,
+      "grad_norm": 0.8504384756088257,
+      "learning_rate": 1.9264356401707402e-06,
+      "loss": 0.0744,
+      "step": 1000
+    },
+    {
+      "epoch": 2.550761421319797,
+      "grad_norm": 0.8340543508529663,
+      "learning_rate": 1.824763856410861e-06,
+      "loss": 0.0751,
+      "step": 1005
+    },
+    {
+      "epoch": 2.563451776649746,
+      "grad_norm": 0.8223659992218018,
+      "learning_rate": 1.7256743646802092e-06,
+      "loss": 0.089,
+      "step": 1010
+    },
+    {
+      "epoch": 2.5761421319796955,
+      "grad_norm": 0.6343728303909302,
+      "learning_rate": 1.6291865861111354e-06,
+      "loss": 0.0755,
+      "step": 1015
+    },
+    {
+      "epoch": 2.5888324873096447,
+      "grad_norm": 0.8592020273208618,
+      "learning_rate": 1.5353194319108916e-06,
+      "loss": 0.0708,
+      "step": 1020
+    },
+    {
+      "epoch": 2.6015228426395938,
+      "grad_norm": 0.8506338000297546,
+      "learning_rate": 1.4440912996551753e-06,
+      "loss": 0.0773,
+      "step": 1025
+    },
+    {
+      "epoch": 2.6142131979695433,
+      "grad_norm": 0.9186679720878601,
+      "learning_rate": 1.3555200696822235e-06,
+      "loss": 0.079,
+      "step": 1030
+    },
+    {
+      "epoch": 2.6269035532994924,
+      "grad_norm": 0.7725194096565247,
+      "learning_rate": 1.2696231015883914e-06,
+      "loss": 0.0674,
+      "step": 1035
+    },
+    {
+      "epoch": 2.6395939086294415,
+      "grad_norm": 0.7940559983253479,
+      "learning_rate": 1.186417230825695e-06,
+      "loss": 0.0888,
+      "step": 1040
+    },
+    {
+      "epoch": 2.652284263959391,
+      "grad_norm": 0.804581880569458,
+      "learning_rate": 1.1059187654021764e-06,
+      "loss": 0.0795,
+      "step": 1045
+    },
+    {
+      "epoch": 2.66497461928934,
+      "grad_norm": 0.674748957157135,
+      "learning_rate": 1.0281434826855647e-06,
+      "loss": 0.0796,
+      "step": 1050
+    },
+    {
+      "epoch": 2.6776649746192893,
+      "grad_norm": 0.779748797416687,
+      "learning_rate": 9.531066263109972e-07,
+      "loss": 0.0669,
+      "step": 1055
+    },
+    {
+      "epoch": 2.6903553299492384,
+      "grad_norm": 0.7440015077590942,
+      "learning_rate": 8.808229031933285e-07,
+      "loss": 0.0734,
+      "step": 1060
+    },
+    {
+      "epoch": 2.703045685279188,
+      "grad_norm": 0.6639752984046936,
+      "learning_rate": 8.113064806446286e-07,
+      "loss": 0.0755,
+      "step": 1065
+    },
+    {
+      "epoch": 2.715736040609137,
+      "grad_norm": 0.6232616305351257,
+      "learning_rate": 7.445709835974512e-07,
+      "loss": 0.0714,
+      "step": 1070
+    },
+    {
+      "epoch": 2.728426395939086,
+      "grad_norm": 0.7801710963249207,
+      "learning_rate": 6.806294919344053e-07,
+      "loss": 0.0758,
+      "step": 1075
+    },
+    {
+      "epoch": 2.7411167512690353,
+      "grad_norm": 0.7783412933349609,
+      "learning_rate": 6.194945379245498e-07,
+      "loss": 0.0847,
+      "step": 1080
+    },
+    {
+      "epoch": 2.753807106598985,
+      "grad_norm": 0.7201891541481018,
+      "learning_rate": 5.611781037671176e-07,
+      "loss": 0.0762,
+      "step": 1085
+    },
+    {
+      "epoch": 2.766497461928934,
+      "grad_norm": 0.693265974521637,
+      "learning_rate": 5.056916192430622e-07,
+      "loss": 0.0762,
+      "step": 1090
+    },
+    {
+      "epoch": 2.779187817258883,
+      "grad_norm": 0.8439576625823975,
+      "learning_rate": 4.5304595947485927e-07,
+      "loss": 0.0699,
+      "step": 1095
+    },
+    {
+      "epoch": 2.7918781725888326,
+      "grad_norm": 0.6810204982757568,
+      "learning_rate": 4.032514427950307e-07,
+      "loss": 0.0649,
+      "step": 1100
+    },
+    {
+      "epoch": 2.8045685279187818,
+      "grad_norm": 0.6787329912185669,
+      "learning_rate": 3.56317828723795e-07,
+      "loss": 0.0646,
+      "step": 1105
+    },
+    {
+      "epoch": 2.817258883248731,
+      "grad_norm": 0.8831884264945984,
+      "learning_rate": 3.1225431605624134e-07,
+      "loss": 0.0692,
+      "step": 1110
+    },
+    {
+      "epoch": 2.8299492385786804,
+      "grad_norm": 0.8658782243728638,
+      "learning_rate": 2.710695410593994e-07,
+      "loss": 0.0713,
+      "step": 1115
+    },
+    {
+      "epoch": 2.8426395939086295,
+      "grad_norm": 0.6840971112251282,
+      "learning_rate": 2.3277157577957398e-07,
+      "loss": 0.0673,
+      "step": 1120
+    },
+    {
+      "epoch": 2.8553299492385786,
+      "grad_norm": 0.7518882751464844,
+      "learning_rate": 1.9736792646024847e-07,
+      "loss": 0.0714,
+      "step": 1125
+    },
+    {
+      "epoch": 2.868020304568528,
+      "grad_norm": 0.6127795577049255,
+      "learning_rate": 1.6486553207090194e-07,
+      "loss": 0.0715,
+      "step": 1130
+    },
+    {
+      "epoch": 2.8807106598984773,
+      "grad_norm": 0.721886157989502,
+      "learning_rate": 1.3527076294698848e-07,
+      "loss": 0.0628,
+      "step": 1135
+    },
+    {
+      "epoch": 2.8934010152284264,
+      "grad_norm": 0.8697863221168518,
+      "learning_rate": 1.0858941954139112e-07,
+      "loss": 0.0626,
+      "step": 1140
+    },
+    {
+      "epoch": 2.9060913705583755,
+      "grad_norm": 0.7899842858314514,
+      "learning_rate": 8.482673128753948e-08,
+      "loss": 0.0608,
+      "step": 1145
+    },
+    {
+      "epoch": 2.9187817258883246,
+      "grad_norm": 0.6605833768844604,
+      "learning_rate": 6.398735557448299e-08,
+      "loss": 0.0616,
+      "step": 1150
+    },
+    {
+      "epoch": 2.931472081218274,
+      "grad_norm": 0.7344260215759277,
+      "learning_rate": 4.607537683404106e-08,
+      "loss": 0.0677,
+      "step": 1155
+    },
+    {
+      "epoch": 2.9441624365482233,
+      "grad_norm": 1.1058152914047241,
+      "learning_rate": 3.1094305740291816e-08,
+      "loss": 0.0852,
+      "step": 1160
+    },
+    {
+      "epoch": 2.9568527918781724,
+      "grad_norm": 0.7568081617355347,
+      "learning_rate": 1.9047078521474137e-08,
+      "loss": 0.0891,
+      "step": 1165
+    },
+    {
+      "epoch": 2.969543147208122,
+      "grad_norm": 0.8575447201728821,
+      "learning_rate": 9.93605638451467e-09,
+      "loss": 0.0715,
+      "step": 1170
+    },
+    {
+      "epoch": 2.982233502538071,
+      "grad_norm": 0.7651799321174622,
+      "learning_rate": 3.763025052231361e-09,
+      "loss": 0.0801,
+      "step": 1175
+    },
+    {
+      "epoch": 2.99492385786802,
+      "grad_norm": 0.6262444853782654,
+      "learning_rate": 5.29194413350087e-10,
+      "loss": 0.0682,
+      "step": 1180
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1182,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.0381572557271204e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

base/23_128_e3_3e-5/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bcf4f5a0d83158e3400e6c1e6375f6dc0fed49bfc4de5b391074c79b0049029
+size 8273

base/23_128_e3_3e-5/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)