Upload folder using huggingface_hub

base/25_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/25_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": [
+    "down_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "up_proj",
+    "o_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

base/25_128_e3_3e-5/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd80ae2a720b764d6063c3922f8919a0877d9ba0bec6cfdb397c94b37e0e0571
+size 791751704

base/25_128_e3_3e-5/latest

@@ -0,0 +1 @@
+global_step1173

base/25_128_e3_3e-5/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed7786dd6dd52ab17035d1ff7eb47207b0eec8ee1f4087dcffeddbd4240ae610
+size 16389

base/25_128_e3_3e-5/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b7adf668d023d9684ba4ee3c3c692883b3837d6f8d447e67213459e1ae0e146
+size 16389

base/25_128_e3_3e-5/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ef9076894578e7e67ab966d3508f4fa53bf6586aa6346ffdb55d274e3b97fb1
+size 16389

base/25_128_e3_3e-5/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ddd7f6749f2330f3d39e682919dec256699f957d141bee4dd0a8e0d7106ff53a
+size 16389

base/25_128_e3_3e-5/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4b6fd0fe0bc7fa7cf4aff8bc234bb6ab41b87d7547c834e6b67d8d420b41ca6
+size 16389

base/25_128_e3_3e-5/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:646b94734becb45848cf90616e0573c3ed67d0cdd4ec0909cdcf0f41c14dc32a
+size 16389

base/25_128_e3_3e-5/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7af9afed4affda52e95456aa7c357721f88e3a719f8a06085b73e7c9dc3633a
+size 16389

base/25_128_e3_3e-5/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a529496496acf1686f1fcf65035ce966da60a691d019b1a0695e5d9f37b256b
+size 16389

base/25_128_e3_3e-5/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:718707ab1d3b08a0078e0c993f33e502910f1dc2662595975575cec09d0f1179
+size 1401

base/25_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/25_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/25_128_e3_3e-5/trainer_state.json

@@ -0,0 +1,1672 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1173,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01278772378516624,
+      "grad_norm": 0.7916995286941528,
+      "learning_rate": 2.033898305084746e-06,
+      "loss": 1.3113,
+      "step": 5
+    },
+    {
+      "epoch": 0.02557544757033248,
+      "grad_norm": 0.5998683571815491,
+      "learning_rate": 4.576271186440678e-06,
+      "loss": 1.274,
+      "step": 10
+    },
+    {
+      "epoch": 0.03836317135549872,
+      "grad_norm": 0.5380151867866516,
+      "learning_rate": 7.1186440677966106e-06,
+      "loss": 1.2552,
+      "step": 15
+    },
+    {
+      "epoch": 0.05115089514066496,
+      "grad_norm": 0.4696876108646393,
+      "learning_rate": 9.661016949152542e-06,
+      "loss": 1.2869,
+      "step": 20
+    },
+    {
+      "epoch": 0.0639386189258312,
+      "grad_norm": 0.4699367880821228,
+      "learning_rate": 1.2203389830508475e-05,
+      "loss": 1.2419,
+      "step": 25
+    },
+    {
+      "epoch": 0.07672634271099744,
+      "grad_norm": 0.43068233132362366,
+      "learning_rate": 1.4745762711864408e-05,
+      "loss": 1.1577,
+      "step": 30
+    },
+    {
+      "epoch": 0.08951406649616368,
+      "grad_norm": 0.39510229229927063,
+      "learning_rate": 1.728813559322034e-05,
+      "loss": 1.2293,
+      "step": 35
+    },
+    {
+      "epoch": 0.10230179028132992,
+      "grad_norm": 0.42271342873573303,
+      "learning_rate": 1.983050847457627e-05,
+      "loss": 1.1766,
+      "step": 40
+    },
+    {
+      "epoch": 0.11508951406649616,
+      "grad_norm": 0.41696181893348694,
+      "learning_rate": 2.2372881355932205e-05,
+      "loss": 1.1886,
+      "step": 45
+    },
+    {
+      "epoch": 0.1278772378516624,
+      "grad_norm": 0.4276786148548126,
+      "learning_rate": 2.4915254237288138e-05,
+      "loss": 1.1092,
+      "step": 50
+    },
+    {
+      "epoch": 0.14066496163682865,
+      "grad_norm": 0.4313717484474182,
+      "learning_rate": 2.7457627118644068e-05,
+      "loss": 1.1656,
+      "step": 55
+    },
+    {
+      "epoch": 0.1534526854219949,
+      "grad_norm": 0.43457067012786865,
+      "learning_rate": 3e-05,
+      "loss": 1.1431,
+      "step": 60
+    },
+    {
+      "epoch": 0.16624040920716113,
+      "grad_norm": 0.44670233130455017,
+      "learning_rate": 2.999850884276484e-05,
+      "loss": 1.1213,
+      "step": 65
+    },
+    {
+      "epoch": 0.17902813299232737,
+      "grad_norm": 0.4918992221355438,
+      "learning_rate": 2.999403566753267e-05,
+      "loss": 1.0153,
+      "step": 70
+    },
+    {
+      "epoch": 0.1918158567774936,
+      "grad_norm": 0.4775982201099396,
+      "learning_rate": 2.9986581363664512e-05,
+      "loss": 1.1052,
+      "step": 75
+    },
+    {
+      "epoch": 0.20460358056265984,
+      "grad_norm": 0.5761969089508057,
+      "learning_rate": 2.997614741323225e-05,
+      "loss": 1.0464,
+      "step": 80
+    },
+    {
+      "epoch": 0.21739130434782608,
+      "grad_norm": 0.6014876365661621,
+      "learning_rate": 2.9962735890723977e-05,
+      "loss": 0.9465,
+      "step": 85
+    },
+    {
+      "epoch": 0.23017902813299232,
+      "grad_norm": 0.6022469997406006,
+      "learning_rate": 2.994634946263153e-05,
+      "loss": 0.9536,
+      "step": 90
+    },
+    {
+      "epoch": 0.24296675191815856,
+      "grad_norm": 0.7171408534049988,
+      "learning_rate": 2.9926991386920353e-05,
+      "loss": 0.9643,
+      "step": 95
+    },
+    {
+      "epoch": 0.2557544757033248,
+      "grad_norm": 0.5690737366676331,
+      "learning_rate": 2.9904665512381735e-05,
+      "loss": 0.9731,
+      "step": 100
+    },
+    {
+      "epoch": 0.26854219948849106,
+      "grad_norm": 0.6398865580558777,
+      "learning_rate": 2.987937627786759e-05,
+      "loss": 0.919,
+      "step": 105
+    },
+    {
+      "epoch": 0.2813299232736573,
+      "grad_norm": 0.7070958018302917,
+      "learning_rate": 2.985112871140792e-05,
+      "loss": 0.9101,
+      "step": 110
+    },
+    {
+      "epoch": 0.29411764705882354,
+      "grad_norm": 0.7677425146102905,
+      "learning_rate": 2.9819928429211133e-05,
+      "loss": 0.8449,
+      "step": 115
+    },
+    {
+      "epoch": 0.3069053708439898,
+      "grad_norm": 0.7605948448181152,
+      "learning_rate": 2.9785781634547438e-05,
+      "loss": 0.91,
+      "step": 120
+    },
+    {
+      "epoch": 0.319693094629156,
+      "grad_norm": 0.7572303414344788,
+      "learning_rate": 2.9748695116515496e-05,
+      "loss": 0.9308,
+      "step": 125
+    },
+    {
+      "epoch": 0.33248081841432225,
+      "grad_norm": 0.9410399198532104,
+      "learning_rate": 2.970867624869259e-05,
+      "loss": 0.8835,
+      "step": 130
+    },
+    {
+      "epoch": 0.3452685421994885,
+      "grad_norm": 0.8266528248786926,
+      "learning_rate": 2.9665732987668633e-05,
+      "loss": 0.7929,
+      "step": 135
+    },
+    {
+      "epoch": 0.35805626598465473,
+      "grad_norm": 0.7973781228065491,
+      "learning_rate": 2.9619873871464203e-05,
+      "loss": 0.7902,
+      "step": 140
+    },
+    {
+      "epoch": 0.37084398976982097,
+      "grad_norm": 0.8212129473686218,
+      "learning_rate": 2.957110801783303e-05,
+      "loss": 0.7872,
+      "step": 145
+    },
+    {
+      "epoch": 0.3836317135549872,
+      "grad_norm": 0.8619541525840759,
+      "learning_rate": 2.9519445122449174e-05,
+      "loss": 0.7991,
+      "step": 150
+    },
+    {
+      "epoch": 0.39641943734015345,
+      "grad_norm": 0.8315053582191467,
+      "learning_rate": 2.946489545697933e-05,
+      "loss": 0.7884,
+      "step": 155
+    },
+    {
+      "epoch": 0.4092071611253197,
+      "grad_norm": 0.8699889183044434,
+      "learning_rate": 2.9407469867040615e-05,
+      "loss": 0.7542,
+      "step": 160
+    },
+    {
+      "epoch": 0.4219948849104859,
+      "grad_norm": 1.0452126264572144,
+      "learning_rate": 2.9347179770044217e-05,
+      "loss": 0.7435,
+      "step": 165
+    },
+    {
+      "epoch": 0.43478260869565216,
+      "grad_norm": 1.0428762435913086,
+      "learning_rate": 2.928403715292538e-05,
+      "loss": 0.6918,
+      "step": 170
+    },
+    {
+      "epoch": 0.4475703324808184,
+      "grad_norm": 1.3039144277572632,
+      "learning_rate": 2.921805456976016e-05,
+      "loss": 0.7402,
+      "step": 175
+    },
+    {
+      "epoch": 0.46035805626598464,
+      "grad_norm": 1.246919870376587,
+      "learning_rate": 2.914924513926938e-05,
+      "loss": 0.715,
+      "step": 180
+    },
+    {
+      "epoch": 0.4731457800511509,
+      "grad_norm": 1.0014393329620361,
+      "learning_rate": 2.9077622542210405e-05,
+      "loss": 0.7144,
+      "step": 185
+    },
+    {
+      "epoch": 0.4859335038363171,
+      "grad_norm": 1.0207639932632446,
+      "learning_rate": 2.9003201018657063e-05,
+      "loss": 0.6787,
+      "step": 190
+    },
+    {
+      "epoch": 0.49872122762148335,
+      "grad_norm": 0.9752850532531738,
+      "learning_rate": 2.8925995365168474e-05,
+      "loss": 0.6677,
+      "step": 195
+    },
+    {
+      "epoch": 0.5115089514066496,
+      "grad_norm": 1.052742838859558,
+      "learning_rate": 2.8846020931847138e-05,
+      "loss": 0.671,
+      "step": 200
+    },
+    {
+      "epoch": 0.5242966751918159,
+      "grad_norm": 1.0179582834243774,
+      "learning_rate": 2.8763293619287032e-05,
+      "loss": 0.6694,
+      "step": 205
+    },
+    {
+      "epoch": 0.5370843989769821,
+      "grad_norm": 1.1208581924438477,
+      "learning_rate": 2.867782987541225e-05,
+      "loss": 0.6965,
+      "step": 210
+    },
+    {
+      "epoch": 0.5498721227621484,
+      "grad_norm": 1.1096495389938354,
+      "learning_rate": 2.85896466922068e-05,
+      "loss": 0.6083,
+      "step": 215
+    },
+    {
+      "epoch": 0.5626598465473146,
+      "grad_norm": 0.9785555601119995,
+      "learning_rate": 2.849876160233623e-05,
+      "loss": 0.612,
+      "step": 220
+    },
+    {
+      "epoch": 0.5754475703324808,
+      "grad_norm": 1.0387229919433594,
+      "learning_rate": 2.8405192675661782e-05,
+      "loss": 0.6503,
+      "step": 225
+    },
+    {
+      "epoch": 0.5882352941176471,
+      "grad_norm": 0.9402236938476562,
+      "learning_rate": 2.830895851564773e-05,
+      "loss": 0.5921,
+      "step": 230
+    },
+    {
+      "epoch": 0.6010230179028133,
+      "grad_norm": 1.0591496229171753,
+      "learning_rate": 2.82100782556626e-05,
+      "loss": 0.6193,
+      "step": 235
+    },
+    {
+      "epoch": 0.6138107416879796,
+      "grad_norm": 0.9969694018363953,
+      "learning_rate": 2.810857155517507e-05,
+      "loss": 0.5764,
+      "step": 240
+    },
+    {
+      "epoch": 0.6265984654731458,
+      "grad_norm": 1.1126652956008911,
+      "learning_rate": 2.8004458595845253e-05,
+      "loss": 0.5547,
+      "step": 245
+    },
+    {
+      "epoch": 0.639386189258312,
+      "grad_norm": 0.9842098355293274,
+      "learning_rate": 2.789776007751216e-05,
+      "loss": 0.5815,
+      "step": 250
+    },
+    {
+      "epoch": 0.6521739130434783,
+      "grad_norm": 1.2084976434707642,
+      "learning_rate": 2.778849721407814e-05,
+      "loss": 0.5101,
+      "step": 255
+    },
+    {
+      "epoch": 0.6649616368286445,
+      "grad_norm": 1.322525978088379,
+      "learning_rate": 2.7676691729291103e-05,
+      "loss": 0.5149,
+      "step": 260
+    },
+    {
+      "epoch": 0.6777493606138107,
+      "grad_norm": 1.1127458810806274,
+      "learning_rate": 2.756236585242539e-05,
+      "loss": 0.5291,
+      "step": 265
+    },
+    {
+      "epoch": 0.690537084398977,
+      "grad_norm": 1.100934386253357,
+      "learning_rate": 2.744554231386213e-05,
+      "loss": 0.5787,
+      "step": 270
+    },
+    {
+      "epoch": 0.7033248081841432,
+      "grad_norm": 0.9788468480110168,
+      "learning_rate": 2.732624434056996e-05,
+      "loss": 0.5263,
+      "step": 275
+    },
+    {
+      "epoch": 0.7161125319693095,
+      "grad_norm": 1.1739388704299927,
+      "learning_rate": 2.720449565148701e-05,
+      "loss": 0.5414,
+      "step": 280
+    },
+    {
+      "epoch": 0.7289002557544757,
+      "grad_norm": 1.2678508758544922,
+      "learning_rate": 2.70803204528051e-05,
+      "loss": 0.5638,
+      "step": 285
+    },
+    {
+      "epoch": 0.7416879795396419,
+      "grad_norm": 1.1300591230392456,
+      "learning_rate": 2.695374343315702e-05,
+      "loss": 0.4322,
+      "step": 290
+    },
+    {
+      "epoch": 0.7544757033248082,
+      "grad_norm": 1.3103854656219482,
+      "learning_rate": 2.6824789758707913e-05,
+      "loss": 0.5019,
+      "step": 295
+    },
+    {
+      "epoch": 0.7672634271099744,
+      "grad_norm": 1.189692735671997,
+      "learning_rate": 2.6693485068151756e-05,
+      "loss": 0.5334,
+      "step": 300
+    },
+    {
+      "epoch": 0.7800511508951407,
+      "grad_norm": 1.183565616607666,
+      "learning_rate": 2.6559855467613774e-05,
+      "loss": 0.4694,
+      "step": 305
+    },
+    {
+      "epoch": 0.7928388746803069,
+      "grad_norm": 0.9612398743629456,
+      "learning_rate": 2.6423927525460067e-05,
+      "loss": 0.4971,
+      "step": 310
+    },
+    {
+      "epoch": 0.8056265984654731,
+      "grad_norm": 1.322892189025879,
+      "learning_rate": 2.6285728267015212e-05,
+      "loss": 0.4555,
+      "step": 315
+    },
+    {
+      "epoch": 0.8184143222506394,
+      "grad_norm": 1.0568349361419678,
+      "learning_rate": 2.6145285169189106e-05,
+      "loss": 0.424,
+      "step": 320
+    },
+    {
+      "epoch": 0.8312020460358056,
+      "grad_norm": 1.2171348333358765,
+      "learning_rate": 2.600262615501393e-05,
+      "loss": 0.4884,
+      "step": 325
+    },
+    {
+      "epoch": 0.8439897698209718,
+      "grad_norm": 1.1603668928146362,
+      "learning_rate": 2.5857779588092513e-05,
+      "loss": 0.4749,
+      "step": 330
+    },
+    {
+      "epoch": 0.8567774936061381,
+      "grad_norm": 1.1250090599060059,
+      "learning_rate": 2.5710774266959015e-05,
+      "loss": 0.4021,
+      "step": 335
+    },
+    {
+      "epoch": 0.8695652173913043,
+      "grad_norm": 1.1668307781219482,
+      "learning_rate": 2.55616394193532e-05,
+      "loss": 0.4352,
+      "step": 340
+    },
+    {
+      "epoch": 0.8823529411764706,
+      "grad_norm": 1.1451588869094849,
+      "learning_rate": 2.541040469640934e-05,
+      "loss": 0.4406,
+      "step": 345
+    },
+    {
+      "epoch": 0.8951406649616368,
+      "grad_norm": 1.2641648054122925,
+      "learning_rate": 2.5257100166760942e-05,
+      "loss": 0.42,
+      "step": 350
+    },
+    {
+      "epoch": 0.907928388746803,
+      "grad_norm": 1.0896860361099243,
+      "learning_rate": 2.5101756310562493e-05,
+      "loss": 0.3782,
+      "step": 355
+    },
+    {
+      "epoch": 0.9207161125319693,
+      "grad_norm": 1.0954163074493408,
+      "learning_rate": 2.4944404013429323e-05,
+      "loss": 0.4164,
+      "step": 360
+    },
+    {
+      "epoch": 0.9335038363171355,
+      "grad_norm": 1.0470174551010132,
+      "learning_rate": 2.4785074560296953e-05,
+      "loss": 0.3916,
+      "step": 365
+    },
+    {
+      "epoch": 0.9462915601023018,
+      "grad_norm": 1.2319450378417969,
+      "learning_rate": 2.462379962920096e-05,
+      "loss": 0.4137,
+      "step": 370
+    },
+    {
+      "epoch": 0.959079283887468,
+      "grad_norm": 1.2191568613052368,
+      "learning_rate": 2.446061128497872e-05,
+      "loss": 0.415,
+      "step": 375
+    },
+    {
+      "epoch": 0.9718670076726342,
+      "grad_norm": 1.1343295574188232,
+      "learning_rate": 2.429554197289426e-05,
+      "loss": 0.4196,
+      "step": 380
+    },
+    {
+      "epoch": 0.9846547314578005,
+      "grad_norm": 1.2405723333358765,
+      "learning_rate": 2.4128624512187444e-05,
+      "loss": 0.3605,
+      "step": 385
+    },
+    {
+      "epoch": 0.9974424552429667,
+      "grad_norm": 1.1308526992797852,
+      "learning_rate": 2.3959892089548844e-05,
+      "loss": 0.476,
+      "step": 390
+    },
+    {
+      "epoch": 1.010230179028133,
+      "grad_norm": 1.169920802116394,
+      "learning_rate": 2.3789378252521497e-05,
+      "loss": 0.367,
+      "step": 395
+    },
+    {
+      "epoch": 1.0230179028132993,
+      "grad_norm": 1.3428298234939575,
+      "learning_rate": 2.3617116902830967e-05,
+      "loss": 0.3173,
+      "step": 400
+    },
+    {
+      "epoch": 1.0358056265984654,
+      "grad_norm": 1.1855430603027344,
+      "learning_rate": 2.3443142289644987e-05,
+      "loss": 0.2885,
+      "step": 405
+    },
+    {
+      "epoch": 1.0485933503836318,
+      "grad_norm": 1.1801083087921143,
+      "learning_rate": 2.3267489002763977e-05,
+      "loss": 0.3569,
+      "step": 410
+    },
+    {
+      "epoch": 1.061381074168798,
+      "grad_norm": 1.283064365386963,
+      "learning_rate": 2.309019196574389e-05,
+      "loss": 0.3356,
+      "step": 415
+    },
+    {
+      "epoch": 1.0741687979539642,
+      "grad_norm": 1.0822687149047852,
+      "learning_rate": 2.2911286428952657e-05,
+      "loss": 0.3517,
+      "step": 420
+    },
+    {
+      "epoch": 1.0869565217391304,
+      "grad_norm": 1.1616040468215942,
+      "learning_rate": 2.2730807962561697e-05,
+      "loss": 0.2919,
+      "step": 425
+    },
+    {
+      "epoch": 1.0997442455242967,
+      "grad_norm": 1.1775623559951782,
+      "learning_rate": 2.25487924494738e-05,
+      "loss": 0.3241,
+      "step": 430
+    },
+    {
+      "epoch": 1.1125319693094629,
+      "grad_norm": 1.2192127704620361,
+      "learning_rate": 2.2365276078188864e-05,
+      "loss": 0.2889,
+      "step": 435
+    },
+    {
+      "epoch": 1.1253196930946292,
+      "grad_norm": 1.2917542457580566,
+      "learning_rate": 2.218029533560887e-05,
+      "loss": 0.2807,
+      "step": 440
+    },
+    {
+      "epoch": 1.1381074168797953,
+      "grad_norm": 1.1888532638549805,
+      "learning_rate": 2.19938869997835e-05,
+      "loss": 0.3101,
+      "step": 445
+    },
+    {
+      "epoch": 1.1508951406649617,
+      "grad_norm": 1.1672463417053223,
+      "learning_rate": 2.1806088132597914e-05,
+      "loss": 0.2699,
+      "step": 450
+    },
+    {
+      "epoch": 1.1636828644501278,
+      "grad_norm": 1.3224337100982666,
+      "learning_rate": 2.161693607240405e-05,
+      "loss": 0.3179,
+      "step": 455
+    },
+    {
+      "epoch": 1.1764705882352942,
+      "grad_norm": 1.224916696548462,
+      "learning_rate": 2.142646842659699e-05,
+      "loss": 0.3254,
+      "step": 460
+    },
+    {
+      "epoch": 1.1892583120204603,
+      "grad_norm": 1.1037440299987793,
+      "learning_rate": 2.1234723064137814e-05,
+      "loss": 0.2837,
+      "step": 465
+    },
+    {
+      "epoch": 1.2020460358056266,
+      "grad_norm": 1.1876534223556519,
+      "learning_rate": 2.1041738108024463e-05,
+      "loss": 0.2666,
+      "step": 470
+    },
+    {
+      "epoch": 1.2148337595907928,
+      "grad_norm": 1.047129511833191,
+      "learning_rate": 2.084755192771208e-05,
+      "loss": 0.2677,
+      "step": 475
+    },
+    {
+      "epoch": 1.227621483375959,
+      "grad_norm": 1.139037013053894,
+      "learning_rate": 2.0652203131484365e-05,
+      "loss": 0.2535,
+      "step": 480
+    },
+    {
+      "epoch": 1.2404092071611252,
+      "grad_norm": 1.1990001201629639,
+      "learning_rate": 2.0455730558777427e-05,
+      "loss": 0.252,
+      "step": 485
+    },
+    {
+      "epoch": 1.2531969309462916,
+      "grad_norm": 1.0957801342010498,
+      "learning_rate": 2.0258173272457724e-05,
+      "loss": 0.2677,
+      "step": 490
+    },
+    {
+      "epoch": 1.265984654731458,
+      "grad_norm": 1.1132164001464844,
+      "learning_rate": 2.005957055105548e-05,
+      "loss": 0.2479,
+      "step": 495
+    },
+    {
+      "epoch": 1.278772378516624,
+      "grad_norm": 1.2894468307495117,
+      "learning_rate": 1.9859961880955373e-05,
+      "loss": 0.2808,
+      "step": 500
+    },
+    {
+      "epoch": 1.2915601023017902,
+      "grad_norm": 1.2949382066726685,
+      "learning_rate": 1.965938694854575e-05,
+      "loss": 0.2615,
+      "step": 505
+    },
+    {
+      "epoch": 1.3043478260869565,
+      "grad_norm": 1.2229321002960205,
+      "learning_rate": 1.9457885632328155e-05,
+      "loss": 0.2665,
+      "step": 510
+    },
+    {
+      "epoch": 1.317135549872123,
+      "grad_norm": 1.1090989112854004,
+      "learning_rate": 1.9255497994988672e-05,
+      "loss": 0.2161,
+      "step": 515
+    },
+    {
+      "epoch": 1.329923273657289,
+      "grad_norm": 1.3068841695785522,
+      "learning_rate": 1.9052264275432602e-05,
+      "loss": 0.2204,
+      "step": 520
+    },
+    {
+      "epoch": 1.3427109974424551,
+      "grad_norm": 1.1525979042053223,
+      "learning_rate": 1.8848224880784106e-05,
+      "loss": 0.246,
+      "step": 525
+    },
+    {
+      "epoch": 1.3554987212276215,
+      "grad_norm": 1.1829029321670532,
+      "learning_rate": 1.8643420378352484e-05,
+      "loss": 0.2398,
+      "step": 530
+    },
+    {
+      "epoch": 1.3682864450127878,
+      "grad_norm": 1.3426706790924072,
+      "learning_rate": 1.843789148756647e-05,
+      "loss": 0.233,
+      "step": 535
+    },
+    {
+      "epoch": 1.381074168797954,
+      "grad_norm": 1.099814534187317,
+      "learning_rate": 1.8231679071878406e-05,
+      "loss": 0.203,
+      "step": 540
+    },
+    {
+      "epoch": 1.39386189258312,
+      "grad_norm": 1.2812471389770508,
+      "learning_rate": 1.8024824130639707e-05,
+      "loss": 0.2241,
+      "step": 545
+    },
+    {
+      "epoch": 1.4066496163682864,
+      "grad_norm": 1.1535215377807617,
+      "learning_rate": 1.7817367790949344e-05,
+      "loss": 0.1954,
+      "step": 550
+    },
+    {
+      "epoch": 1.4194373401534528,
+      "grad_norm": 1.127975583076477,
+      "learning_rate": 1.7609351299476898e-05,
+      "loss": 0.2411,
+      "step": 555
+    },
+    {
+      "epoch": 1.432225063938619,
+      "grad_norm": 1.216868281364441,
+      "learning_rate": 1.740081601426188e-05,
+      "loss": 0.2337,
+      "step": 560
+    },
+    {
+      "epoch": 1.445012787723785,
+      "grad_norm": 1.1923608779907227,
+      "learning_rate": 1.719180339649087e-05,
+      "loss": 0.2442,
+      "step": 565
+    },
+    {
+      "epoch": 1.4578005115089514,
+      "grad_norm": 1.1720415353775024,
+      "learning_rate": 1.698235500225416e-05,
+      "loss": 0.2193,
+      "step": 570
+    },
+    {
+      "epoch": 1.4705882352941178,
+      "grad_norm": 1.237507939338684,
+      "learning_rate": 1.6772512474283548e-05,
+      "loss": 0.2515,
+      "step": 575
+    },
+    {
+      "epoch": 1.4833759590792839,
+      "grad_norm": 1.4944435358047485,
+      "learning_rate": 1.6562317533672877e-05,
+      "loss": 0.192,
+      "step": 580
+    },
+    {
+      "epoch": 1.49616368286445,
+      "grad_norm": 1.1600385904312134,
+      "learning_rate": 1.6351811971583008e-05,
+      "loss": 0.2016,
+      "step": 585
+    },
+    {
+      "epoch": 1.5089514066496164,
+      "grad_norm": 1.0613391399383545,
+      "learning_rate": 1.6141037640932882e-05,
+      "loss": 0.229,
+      "step": 590
+    },
+    {
+      "epoch": 1.5217391304347827,
+      "grad_norm": 1.2351491451263428,
+      "learning_rate": 1.5930036448078234e-05,
+      "loss": 0.1518,
+      "step": 595
+    },
+    {
+      "epoch": 1.5345268542199488,
+      "grad_norm": 1.0561158657073975,
+      "learning_rate": 1.5718850344479778e-05,
+      "loss": 0.1951,
+      "step": 600
+    },
+    {
+      "epoch": 1.547314578005115,
+      "grad_norm": 1.1745907068252563,
+      "learning_rate": 1.5507521318362372e-05,
+      "loss": 0.206,
+      "step": 605
+    },
+    {
+      "epoch": 1.5601023017902813,
+      "grad_norm": 1.0715268850326538,
+      "learning_rate": 1.529609138636685e-05,
+      "loss": 0.1722,
+      "step": 610
+    },
+    {
+      "epoch": 1.5728900255754477,
+      "grad_norm": 1.1858606338500977,
+      "learning_rate": 1.5084602585196249e-05,
+      "loss": 0.1823,
+      "step": 615
+    },
+    {
+      "epoch": 1.5856777493606138,
+      "grad_norm": 1.1178280115127563,
+      "learning_rate": 1.4873096963258052e-05,
+      "loss": 0.1796,
+      "step": 620
+    },
+    {
+      "epoch": 1.59846547314578,
+      "grad_norm": 1.1538869142532349,
+      "learning_rate": 1.4661616572304036e-05,
+      "loss": 0.1923,
+      "step": 625
+    },
+    {
+      "epoch": 1.6112531969309463,
+      "grad_norm": 1.2692184448242188,
+      "learning_rate": 1.445020345906955e-05,
+      "loss": 0.1709,
+      "step": 630
+    },
+    {
+      "epoch": 1.6240409207161126,
+      "grad_norm": 1.2500762939453125,
+      "learning_rate": 1.423889965691372e-05,
+      "loss": 0.1826,
+      "step": 635
+    },
+    {
+      "epoch": 1.6368286445012787,
+      "grad_norm": 1.1280200481414795,
+      "learning_rate": 1.4027747177462318e-05,
+      "loss": 0.1738,
+      "step": 640
+    },
+    {
+      "epoch": 1.6496163682864449,
+      "grad_norm": 1.1322662830352783,
+      "learning_rate": 1.3816788002255019e-05,
+      "loss": 0.1828,
+      "step": 645
+    },
+    {
+      "epoch": 1.6624040920716112,
+      "grad_norm": 1.2282993793487549,
+      "learning_rate": 1.3606064074398544e-05,
+      "loss": 0.1832,
+      "step": 650
+    },
+    {
+      "epoch": 1.6751918158567776,
+      "grad_norm": 1.3571609258651733,
+      "learning_rate": 1.3395617290227505e-05,
+      "loss": 0.192,
+      "step": 655
+    },
+    {
+      "epoch": 1.6879795396419437,
+      "grad_norm": 1.020384669303894,
+      "learning_rate": 1.3185489490974556e-05,
+      "loss": 0.1691,
+      "step": 660
+    },
+    {
+      "epoch": 1.7007672634271098,
+      "grad_norm": 1.1239757537841797,
+      "learning_rate": 1.2975722454451454e-05,
+      "loss": 0.1983,
+      "step": 665
+    },
+    {
+      "epoch": 1.7135549872122762,
+      "grad_norm": 1.1790050268173218,
+      "learning_rate": 1.2766357886742744e-05,
+      "loss": 0.1839,
+      "step": 670
+    },
+    {
+      "epoch": 1.7263427109974425,
+      "grad_norm": 1.2545019388198853,
+      "learning_rate": 1.2557437413913767e-05,
+      "loss": 0.1925,
+      "step": 675
+    },
+    {
+      "epoch": 1.7391304347826086,
+      "grad_norm": 1.1638193130493164,
+      "learning_rate": 1.2349002573734469e-05,
+      "loss": 0.1564,
+      "step": 680
+    },
+    {
+      "epoch": 1.7519181585677748,
+      "grad_norm": 1.2566914558410645,
+      "learning_rate": 1.214109480742084e-05,
+      "loss": 0.1849,
+      "step": 685
+    },
+    {
+      "epoch": 1.7647058823529411,
+      "grad_norm": 1.0654963254928589,
+      "learning_rate": 1.1933755451395556e-05,
+      "loss": 0.1586,
+      "step": 690
+    },
+    {
+      "epoch": 1.7774936061381075,
+      "grad_norm": 1.2069470882415771,
+      "learning_rate": 1.17270257290694e-05,
+      "loss": 0.1597,
+      "step": 695
+    },
+    {
+      "epoch": 1.7902813299232738,
+      "grad_norm": 1.2823141813278198,
+      "learning_rate": 1.1520946742645184e-05,
+      "loss": 0.1843,
+      "step": 700
+    },
+    {
+      "epoch": 1.80306905370844,
+      "grad_norm": 1.0918316841125488,
+      "learning_rate": 1.13155594649458e-05,
+      "loss": 0.1746,
+      "step": 705
+    },
+    {
+      "epoch": 1.815856777493606,
+      "grad_norm": 1.194413423538208,
+      "learning_rate": 1.111090473126793e-05,
+      "loss": 0.1503,
+      "step": 710
+    },
+    {
+      "epoch": 1.8286445012787724,
+      "grad_norm": 0.9708653688430786,
+      "learning_rate": 1.0907023231263158e-05,
+      "loss": 0.1508,
+      "step": 715
+    },
+    {
+      "epoch": 1.8414322250639388,
+      "grad_norm": 0.997572660446167,
+      "learning_rate": 1.0703955500847993e-05,
+      "loss": 0.1365,
+      "step": 720
+    },
+    {
+      "epoch": 1.854219948849105,
+      "grad_norm": 1.2436491250991821,
+      "learning_rate": 1.050174191414449e-05,
+      "loss": 0.1287,
+      "step": 725
+    },
+    {
+      "epoch": 1.867007672634271,
+      "grad_norm": 1.132411241531372,
+      "learning_rate": 1.0300422675453038e-05,
+      "loss": 0.1624,
+      "step": 730
+    },
+    {
+      "epoch": 1.8797953964194374,
+      "grad_norm": 1.1288777589797974,
+      "learning_rate": 1.0100037811258878e-05,
+      "loss": 0.1433,
+      "step": 735
+    },
+    {
+      "epoch": 1.8925831202046037,
+      "grad_norm": 1.199167251586914,
+      "learning_rate": 9.900627162274017e-06,
+      "loss": 0.1433,
+      "step": 740
+    },
+    {
+      "epoch": 1.9053708439897699,
+      "grad_norm": 1.098994255065918,
+      "learning_rate": 9.702230375516064e-06,
+      "loss": 0.1363,
+      "step": 745
+    },
+    {
+      "epoch": 1.918158567774936,
+      "grad_norm": 1.3296716213226318,
+      "learning_rate": 9.504886896425545e-06,
+      "loss": 0.1305,
+      "step": 750
+    },
+    {
+      "epoch": 1.9309462915601023,
+      "grad_norm": 1.2086325883865356,
+      "learning_rate": 9.308635961023348e-06,
+      "loss": 0.1256,
+      "step": 755
+    },
+    {
+      "epoch": 1.9437340153452687,
+      "grad_norm": 1.1661462783813477,
+      "learning_rate": 9.113516588109773e-06,
+      "loss": 0.1423,
+      "step": 760
+    },
+    {
+      "epoch": 1.9565217391304348,
+      "grad_norm": 1.044481873512268,
+      "learning_rate": 8.919567571506777e-06,
+      "loss": 0.1474,
+      "step": 765
+    },
+    {
+      "epoch": 1.969309462915601,
+      "grad_norm": 0.998723566532135,
+      "learning_rate": 8.72682747234493e-06,
+      "loss": 0.1272,
+      "step": 770
+    },
+    {
+      "epoch": 1.9820971867007673,
+      "grad_norm": 1.1387728452682495,
+      "learning_rate": 8.53533461139669e-06,
+      "loss": 0.1155,
+      "step": 775
+    },
+    {
+      "epoch": 1.9948849104859336,
+      "grad_norm": 1.1012626886367798,
+      "learning_rate": 8.3451270614574e-06,
+      "loss": 0.1399,
+      "step": 780
+    },
+    {
+      "epoch": 2.0076726342710995,
+      "grad_norm": 0.9846829771995544,
+      "learning_rate": 8.15624263977563e-06,
+      "loss": 0.1053,
+      "step": 785
+    },
+    {
+      "epoch": 2.020460358056266,
+      "grad_norm": 0.922254204750061,
+      "learning_rate": 7.968718900534311e-06,
+      "loss": 0.1151,
+      "step": 790
+    },
+    {
+      "epoch": 2.0332480818414322,
+      "grad_norm": 1.0824726819992065,
+      "learning_rate": 7.782593127384184e-06,
+      "loss": 0.0946,
+      "step": 795
+    },
+    {
+      "epoch": 2.0460358056265986,
+      "grad_norm": 0.971753716468811,
+      "learning_rate": 7.597902326031018e-06,
+      "loss": 0.101,
+      "step": 800
+    },
+    {
+      "epoch": 2.0588235294117645,
+      "grad_norm": 1.745219111442566,
+      "learning_rate": 7.4146832168781085e-06,
+      "loss": 0.1016,
+      "step": 805
+    },
+    {
+      "epoch": 2.071611253196931,
+      "grad_norm": 1.0103071928024292,
+      "learning_rate": 7.232972227725485e-06,
+      "loss": 0.102,
+      "step": 810
+    },
+    {
+      "epoch": 2.084398976982097,
+      "grad_norm": 1.0176327228546143,
+      "learning_rate": 7.052805486527307e-06,
+      "loss": 0.0785,
+      "step": 815
+    },
+    {
+      "epoch": 2.0971867007672635,
+      "grad_norm": 0.9036539196968079,
+      "learning_rate": 6.874218814208863e-06,
+      "loss": 0.1105,
+      "step": 820
+    },
+    {
+      "epoch": 2.10997442455243,
+      "grad_norm": 0.9635645747184753,
+      "learning_rate": 6.6972477175446255e-06,
+      "loss": 0.0917,
+      "step": 825
+    },
+    {
+      "epoch": 2.122762148337596,
+      "grad_norm": 0.9675960540771484,
+      "learning_rate": 6.521927382098753e-06,
+      "loss": 0.0979,
+      "step": 830
+    },
+    {
+      "epoch": 2.135549872122762,
+      "grad_norm": 1.0831254720687866,
+      "learning_rate": 6.3482926652294695e-06,
+      "loss": 0.1051,
+      "step": 835
+    },
+    {
+      "epoch": 2.1483375959079285,
+      "grad_norm": 0.8053269386291504,
+      "learning_rate": 6.176378089158686e-06,
+      "loss": 0.1024,
+      "step": 840
+    },
+    {
+      "epoch": 2.1611253196930944,
+      "grad_norm": 0.9517676830291748,
+      "learning_rate": 6.006217834108261e-06,
+      "loss": 0.0857,
+      "step": 845
+    },
+    {
+      "epoch": 2.1739130434782608,
+      "grad_norm": 1.0134209394454956,
+      "learning_rate": 5.8378457315042576e-06,
+      "loss": 0.0805,
+      "step": 850
+    },
+    {
+      "epoch": 2.186700767263427,
+      "grad_norm": 1.147398829460144,
+      "learning_rate": 5.671295257250537e-06,
+      "loss": 0.089,
+      "step": 855
+    },
+    {
+      "epoch": 2.1994884910485935,
+      "grad_norm": 0.985681414604187,
+      "learning_rate": 5.506599525073064e-06,
+      "loss": 0.0863,
+      "step": 860
+    },
+    {
+      "epoch": 2.21227621483376,
+      "grad_norm": 1.0436224937438965,
+      "learning_rate": 5.343791279936189e-06,
+      "loss": 0.1137,
+      "step": 865
+    },
+    {
+      "epoch": 2.2250639386189257,
+      "grad_norm": 1.0183159112930298,
+      "learning_rate": 5.182902891532267e-06,
+      "loss": 0.104,
+      "step": 870
+    },
+    {
+      "epoch": 2.237851662404092,
+      "grad_norm": 1.01872980594635,
+      "learning_rate": 5.023966347845892e-06,
+      "loss": 0.0909,
+      "step": 875
+    },
+    {
+      "epoch": 2.2506393861892584,
+      "grad_norm": 0.889183521270752,
+      "learning_rate": 4.867013248794e-06,
+      "loss": 0.0798,
+      "step": 880
+    },
+    {
+      "epoch": 2.2634271099744243,
+      "grad_norm": 0.920059084892273,
+      "learning_rate": 4.712074799943158e-06,
+      "loss": 0.0713,
+      "step": 885
+    },
+    {
+      "epoch": 2.2762148337595907,
+      "grad_norm": 1.1006776094436646,
+      "learning_rate": 4.5591818063052315e-06,
+      "loss": 0.086,
+      "step": 890
+    },
+    {
+      "epoch": 2.289002557544757,
+      "grad_norm": 0.7678653597831726,
+      "learning_rate": 4.408364666212712e-06,
+      "loss": 0.0696,
+      "step": 895
+    },
+    {
+      "epoch": 2.3017902813299234,
+      "grad_norm": 0.9565536975860596,
+      "learning_rate": 4.2596533652748836e-06,
+      "loss": 0.0891,
+      "step": 900
+    },
+    {
+      "epoch": 2.3145780051150897,
+      "grad_norm": 1.0630682706832886,
+      "learning_rate": 4.113077470416057e-06,
+      "loss": 0.0886,
+      "step": 905
+    },
+    {
+      "epoch": 2.3273657289002556,
+      "grad_norm": 0.859937310218811,
+      "learning_rate": 3.9686661239970466e-06,
+      "loss": 0.1011,
+      "step": 910
+    },
+    {
+      "epoch": 2.340153452685422,
+      "grad_norm": 0.9706724286079407,
+      "learning_rate": 3.8264480380210686e-06,
+      "loss": 0.079,
+      "step": 915
+    },
+    {
+      "epoch": 2.3529411764705883,
+      "grad_norm": 0.7542669773101807,
+      "learning_rate": 3.6864514884251648e-06,
+      "loss": 0.062,
+      "step": 920
+    },
+    {
+      "epoch": 2.3657289002557547,
+      "grad_norm": 1.0022294521331787,
+      "learning_rate": 3.5487043094583756e-06,
+      "loss": 0.0743,
+      "step": 925
+    },
+    {
+      "epoch": 2.3785166240409206,
+      "grad_norm": 1.0501244068145752,
+      "learning_rate": 3.413233888147715e-06,
+      "loss": 0.0748,
+      "step": 930
+    },
+    {
+      "epoch": 2.391304347826087,
+      "grad_norm": 1.0656195878982544,
+      "learning_rate": 3.280067158853034e-06,
+      "loss": 0.0811,
+      "step": 935
+    },
+    {
+      "epoch": 2.4040920716112533,
+      "grad_norm": 1.061787486076355,
+      "learning_rate": 3.149230597911907e-06,
+      "loss": 0.0843,
+      "step": 940
+    },
+    {
+      "epoch": 2.4168797953964196,
+      "grad_norm": 0.8419595956802368,
+      "learning_rate": 3.020750218375605e-06,
+      "loss": 0.0658,
+      "step": 945
+    },
+    {
+      "epoch": 2.4296675191815855,
+      "grad_norm": 0.7697025537490845,
+      "learning_rate": 2.8946515648371303e-06,
+      "loss": 0.07,
+      "step": 950
+    },
+    {
+      "epoch": 2.442455242966752,
+      "grad_norm": 0.7373659014701843,
+      "learning_rate": 2.770959708352418e-06,
+      "loss": 0.0701,
+      "step": 955
+    },
+    {
+      "epoch": 2.455242966751918,
+      "grad_norm": 1.1907942295074463,
+      "learning_rate": 2.6496992414557053e-06,
+      "loss": 0.0842,
+      "step": 960
+    },
+    {
+      "epoch": 2.4680306905370846,
+      "grad_norm": 0.8502811789512634,
+      "learning_rate": 2.530894273270002e-06,
+      "loss": 0.0861,
+      "step": 965
+    },
+    {
+      "epoch": 2.4808184143222505,
+      "grad_norm": 0.8233861327171326,
+      "learning_rate": 2.4145684247136807e-06,
+      "loss": 0.0758,
+      "step": 970
+    },
+    {
+      "epoch": 2.493606138107417,
+      "grad_norm": 0.9293055534362793,
+      "learning_rate": 2.300744823804181e-06,
+      "loss": 0.0769,
+      "step": 975
+    },
+    {
+      "epoch": 2.506393861892583,
+      "grad_norm": 0.8738946318626404,
+      "learning_rate": 2.1894461010596396e-06,
+      "loss": 0.0719,
+      "step": 980
+    },
+    {
+      "epoch": 2.5191815856777495,
+      "grad_norm": 0.7666301131248474,
+      "learning_rate": 2.080694384999469e-06,
+      "loss": 0.0673,
+      "step": 985
+    },
+    {
+      "epoch": 2.531969309462916,
+      "grad_norm": 0.8043603301048279,
+      "learning_rate": 1.974511297744782e-06,
+      "loss": 0.0632,
+      "step": 990
+    },
+    {
+      "epoch": 2.544757033248082,
+      "grad_norm": 0.9332150816917419,
+      "learning_rate": 1.8709179507194158e-06,
+      "loss": 0.072,
+      "step": 995
+    },
+    {
+      "epoch": 2.557544757033248,
+      "grad_norm": 0.6791574954986572,
+      "learning_rate": 1.769934940452554e-06,
+      "loss": 0.0742,
+      "step": 1000
+    },
+    {
+      "epoch": 2.5703324808184145,
+      "grad_norm": 0.8469293117523193,
+      "learning_rate": 1.6715823444837241e-06,
+      "loss": 0.0623,
+      "step": 1005
+    },
+    {
+      "epoch": 2.5831202046035804,
+      "grad_norm": 0.7878550887107849,
+      "learning_rate": 1.5758797173709327e-06,
+      "loss": 0.0833,
+      "step": 1010
+    },
+    {
+      "epoch": 2.5959079283887467,
+      "grad_norm": 0.7265278100967407,
+      "learning_rate": 1.4828460868028277e-06,
+      "loss": 0.0685,
+      "step": 1015
+    },
+    {
+      "epoch": 2.608695652173913,
+      "grad_norm": 0.6198123693466187,
+      "learning_rate": 1.3924999498155832e-06,
+      "loss": 0.0557,
+      "step": 1020
+    },
+    {
+      "epoch": 2.6214833759590794,
+      "grad_norm": 0.6328378915786743,
+      "learning_rate": 1.3048592691153137e-06,
+      "loss": 0.0755,
+      "step": 1025
+    },
+    {
+      "epoch": 2.634271099744246,
+      "grad_norm": 0.6597533822059631,
+      "learning_rate": 1.2199414695067001e-06,
+      "loss": 0.0567,
+      "step": 1030
+    },
+    {
+      "epoch": 2.6470588235294117,
+      "grad_norm": 1.0632027387619019,
+      "learning_rate": 1.1377634344285826e-06,
+      "loss": 0.0702,
+      "step": 1035
+    },
+    {
+      "epoch": 2.659846547314578,
+      "grad_norm": 0.7278709411621094,
+      "learning_rate": 1.0583415025971693e-06,
+      "loss": 0.0689,
+      "step": 1040
+    },
+    {
+      "epoch": 2.6726342710997444,
+      "grad_norm": 0.5549436807632446,
+      "learning_rate": 9.816914647575653e-07,
+      "loss": 0.0735,
+      "step": 1045
+    },
+    {
+      "epoch": 2.6854219948849103,
+      "grad_norm": 0.7885385155677795,
+      "learning_rate": 9.078285605442365e-07,
+      "loss": 0.0793,
+      "step": 1050
+    },
+    {
+      "epoch": 2.6982097186700766,
+      "grad_norm": 0.7415198087692261,
+      "learning_rate": 8.36767475451054e-07,
+      "loss": 0.0719,
+      "step": 1055
+    },
+    {
+      "epoch": 2.710997442455243,
+      "grad_norm": 0.8270477652549744,
+      "learning_rate": 7.685223379115075e-07,
+      "loss": 0.0828,
+      "step": 1060
+    },
+    {
+      "epoch": 2.7237851662404093,
+      "grad_norm": 0.791292130947113,
+      "learning_rate": 7.031067164896776e-07,
+      "loss": 0.0697,
+      "step": 1065
+    },
+    {
+      "epoch": 2.7365728900255757,
+      "grad_norm": 0.6786651611328125,
+      "learning_rate": 6.405336171825222e-07,
+      "loss": 0.0713,
+      "step": 1070
+    },
+    {
+      "epoch": 2.7493606138107416,
+      "grad_norm": 0.7264712452888489,
+      "learning_rate": 5.808154808340077e-07,
+      "loss": 0.0593,
+      "step": 1075
+    },
+    {
+      "epoch": 2.762148337595908,
+      "grad_norm": 0.6654142737388611,
+      "learning_rate": 5.239641806616119e-07,
+      "loss": 0.0565,
+      "step": 1080
+    },
+    {
+      "epoch": 2.7749360613810743,
+      "grad_norm": 0.7833060622215271,
+      "learning_rate": 4.6999101989568136e-07,
+      "loss": 0.064,
+      "step": 1085
+    },
+    {
+      "epoch": 2.78772378516624,
+      "grad_norm": 0.7088398337364197,
+      "learning_rate": 4.1890672953210475e-07,
+      "loss": 0.0592,
+      "step": 1090
+    },
+    {
+      "epoch": 2.8005115089514065,
+      "grad_norm": 0.6554706692695618,
+      "learning_rate": 3.70721466198774e-07,
+      "loss": 0.0701,
+      "step": 1095
+    },
+    {
+      "epoch": 2.813299232736573,
+      "grad_norm": 0.8148102164268494,
+      "learning_rate": 3.2544481013622673e-07,
+      "loss": 0.0709,
+      "step": 1100
+    },
+    {
+      "epoch": 2.8260869565217392,
+      "grad_norm": 0.7193168997764587,
+      "learning_rate": 2.8308576329290125e-07,
+      "loss": 0.0585,
+      "step": 1105
+    },
+    {
+      "epoch": 2.8388746803069056,
+      "grad_norm": 0.7357467412948608,
+      "learning_rate": 2.436527475353517e-07,
+      "loss": 0.0698,
+      "step": 1110
+    },
+    {
+      "epoch": 2.8516624040920715,
+      "grad_norm": 0.7839689254760742,
+      "learning_rate": 2.0715360297381746e-07,
+      "loss": 0.0711,
+      "step": 1115
+    },
+    {
+      "epoch": 2.864450127877238,
+      "grad_norm": 0.8969751000404358,
+      "learning_rate": 1.735955864034233e-07,
+      "loss": 0.0668,
+      "step": 1120
+    },
+    {
+      "epoch": 2.877237851662404,
+      "grad_norm": 0.6244706511497498,
+      "learning_rate": 1.4298536986139865e-07,
+      "loss": 0.0594,
+      "step": 1125
+    },
+    {
+      "epoch": 2.89002557544757,
+      "grad_norm": 0.6994004845619202,
+      "learning_rate": 1.1532903930053018e-07,
+      "loss": 0.076,
+      "step": 1130
+    },
+    {
+      "epoch": 2.9028132992327365,
+      "grad_norm": 0.5827573537826538,
+      "learning_rate": 9.063209337913492e-08,
+      "loss": 0.0723,
+      "step": 1135
+    },
+    {
+      "epoch": 2.915601023017903,
+      "grad_norm": 0.6875705718994141,
+      "learning_rate": 6.889944236782631e-08,
+      "loss": 0.0635,
+      "step": 1140
+    },
+    {
+      "epoch": 2.928388746803069,
+      "grad_norm": 0.7700561881065369,
+      "learning_rate": 5.0135407173245163e-08,
+      "loss": 0.0644,
+      "step": 1145
+    },
+    {
+      "epoch": 2.9411764705882355,
+      "grad_norm": 0.7268701791763306,
+      "learning_rate": 3.434371847897022e-08,
+      "loss": 0.0764,
+      "step": 1150
+    },
+    {
+      "epoch": 2.9539641943734014,
+      "grad_norm": 0.7423122525215149,
+      "learning_rate": 2.1527516003781443e-08,
+      "loss": 0.0698,
+      "step": 1155
+    },
+    {
+      "epoch": 2.9667519181585678,
+      "grad_norm": 0.6876771450042725,
+      "learning_rate": 1.1689347877419377e-08,
+      "loss": 0.0812,
+      "step": 1160
+    },
+    {
+      "epoch": 2.979539641943734,
+      "grad_norm": 0.7034068703651428,
+      "learning_rate": 4.831170133960394e-09,
+      "loss": 0.0619,
+      "step": 1165
+    },
+    {
+      "epoch": 2.9923273657289,
+      "grad_norm": 0.6960203051567078,
+      "learning_rate": 9.543463229177984e-10,
+      "loss": 0.0648,
+      "step": 1170
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1173,
+  "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": 2.0001498052585062e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

base/25_128_e3_3e-5/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ace55c5c2bc06dc4a92923577e6e10ea1e278ac9ec9c69581ecf8b2be72d6fb
+size 8273

base/25_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)