Upload folder using huggingface_hub

base/24_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/24_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": [
+    "gate_proj",
+    "o_proj",
+    "down_proj",
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

base/24_128_e3_3e-5/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:530dddc891c7e224c0b12fad2ae67a331c870e6f728d6a6e9af9c598a137d7e2
+size 791751704

base/24_128_e3_3e-5/latest

@@ -0,0 +1 @@
+global_step1062

base/24_128_e3_3e-5/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a933c7e1170ba7c776e9d446905442ad6229b40ed2759e6c98319e94726bfba3
+size 16389

base/24_128_e3_3e-5/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af21ea663cb6f94928e8aebc965bee378011a53be916aa8351adb131a7420ede
+size 16389

base/24_128_e3_3e-5/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6dfd679afb31a230c5d445f1277a9d8fab016979cfb2d8a25bd15ebf7d7d97b2
+size 16389

base/24_128_e3_3e-5/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:326ed974e0392de2e67f7018b26cea76dad706cdcd6c3fad5c3861a92dc5e71e
+size 16389

base/24_128_e3_3e-5/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c3410f844d16ad6981281d5a799be775b07210d917fbd807d10a5d403858cbe
+size 16389

base/24_128_e3_3e-5/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c47e316b5ced04e0ea8f57440d63d0efd405da08e2355f3b0e3e3cb6838a08a6
+size 16389

base/24_128_e3_3e-5/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:763bc1980cd8a72938476db51164b550246087c189610dfb5189c60c85de2544
+size 16389

base/24_128_e3_3e-5/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:913a7e1327ee834e6f2255046ee7f1a83512cd9a52d22041ddc54f05c90e63d6
+size 16389

base/24_128_e3_3e-5/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f856041c237130bf67247a1537bd13fd856019bb7f2cb164c8a02664132d693c
+size 1401

base/24_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/24_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/24_128_e3_3e-5/trainer_state.json

@@ -0,0 +1,1518 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1062,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.014144271570014143,
+      "grad_norm": 0.9338141083717346,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 1.2596,
+      "step": 5
+    },
+    {
+      "epoch": 0.028288543140028287,
+      "grad_norm": 0.5998201966285706,
+      "learning_rate": 4.9999999999999996e-06,
+      "loss": 1.3113,
+      "step": 10
+    },
+    {
+      "epoch": 0.042432814710042434,
+      "grad_norm": 0.48028072714805603,
+      "learning_rate": 7.777777777777777e-06,
+      "loss": 1.3119,
+      "step": 15
+    },
+    {
+      "epoch": 0.056577086280056574,
+      "grad_norm": 0.5012581944465637,
+      "learning_rate": 1.0555555555555555e-05,
+      "loss": 1.311,
+      "step": 20
+    },
+    {
+      "epoch": 0.07072135785007072,
+      "grad_norm": 0.4757620692253113,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.283,
+      "step": 25
+    },
+    {
+      "epoch": 0.08486562942008487,
+      "grad_norm": 0.4793965518474579,
+      "learning_rate": 1.6111111111111115e-05,
+      "loss": 1.2888,
+      "step": 30
+    },
+    {
+      "epoch": 0.09900990099009901,
+      "grad_norm": 0.43736374378204346,
+      "learning_rate": 1.888888888888889e-05,
+      "loss": 1.255,
+      "step": 35
+    },
+    {
+      "epoch": 0.11315417256011315,
+      "grad_norm": 0.40398138761520386,
+      "learning_rate": 2.1666666666666667e-05,
+      "loss": 1.2099,
+      "step": 40
+    },
+    {
+      "epoch": 0.1272984441301273,
+      "grad_norm": 0.4287651479244232,
+      "learning_rate": 2.4444444444444445e-05,
+      "loss": 1.208,
+      "step": 45
+    },
+    {
+      "epoch": 0.14144271570014144,
+      "grad_norm": 0.4603016972541809,
+      "learning_rate": 2.7222222222222223e-05,
+      "loss": 1.1863,
+      "step": 50
+    },
+    {
+      "epoch": 0.15558698727015557,
+      "grad_norm": 0.4060690402984619,
+      "learning_rate": 3e-05,
+      "loss": 1.0807,
+      "step": 55
+    },
+    {
+      "epoch": 0.16973125884016974,
+      "grad_norm": 0.4641093611717224,
+      "learning_rate": 2.9998178743291113e-05,
+      "loss": 1.1251,
+      "step": 60
+    },
+    {
+      "epoch": 0.18387553041018387,
+      "grad_norm": 0.5613163113594055,
+      "learning_rate": 2.9992715415427915e-05,
+      "loss": 1.1053,
+      "step": 65
+    },
+    {
+      "epoch": 0.19801980198019803,
+      "grad_norm": 0.5369486808776855,
+      "learning_rate": 2.998361134309341e-05,
+      "loss": 1.0841,
+      "step": 70
+    },
+    {
+      "epoch": 0.21216407355021216,
+      "grad_norm": 0.5142983794212341,
+      "learning_rate": 2.997086873706798e-05,
+      "loss": 1.0497,
+      "step": 75
+    },
+    {
+      "epoch": 0.2263083451202263,
+      "grad_norm": 0.5890116691589355,
+      "learning_rate": 2.995449069169251e-05,
+      "loss": 1.0288,
+      "step": 80
+    },
+    {
+      "epoch": 0.24045261669024046,
+      "grad_norm": 0.6341537237167358,
+      "learning_rate": 2.9934481184117005e-05,
+      "loss": 0.9607,
+      "step": 85
+    },
+    {
+      "epoch": 0.2545968882602546,
+      "grad_norm": 0.6208008527755737,
+      "learning_rate": 2.9910845073334793e-05,
+      "loss": 0.9995,
+      "step": 90
+    },
+    {
+      "epoch": 0.26874115983026875,
+      "grad_norm": 0.5995469689369202,
+      "learning_rate": 2.9883588099002583e-05,
+      "loss": 0.9933,
+      "step": 95
+    },
+    {
+      "epoch": 0.2828854314002829,
+      "grad_norm": 0.8123688101768494,
+      "learning_rate": 2.9852716880046687e-05,
+      "loss": 1.0237,
+      "step": 100
+    },
+    {
+      "epoch": 0.297029702970297,
+      "grad_norm": 0.6606298089027405,
+      "learning_rate": 2.981823891305572e-05,
+      "loss": 0.9322,
+      "step": 105
+    },
+    {
+      "epoch": 0.31117397454031115,
+      "grad_norm": 0.678730845451355,
+      "learning_rate": 2.9780162570460182e-05,
+      "loss": 0.9408,
+      "step": 110
+    },
+    {
+      "epoch": 0.32531824611032534,
+      "grad_norm": 0.7530872225761414,
+      "learning_rate": 2.9738497098499325e-05,
+      "loss": 0.8753,
+      "step": 115
+    },
+    {
+      "epoch": 0.33946251768033947,
+      "grad_norm": 0.8183106184005737,
+      "learning_rate": 2.969325261497586e-05,
+      "loss": 0.8502,
+      "step": 120
+    },
+    {
+      "epoch": 0.3536067892503536,
+      "grad_norm": 0.7925098538398743,
+      "learning_rate": 2.9644440106799003e-05,
+      "loss": 0.8898,
+      "step": 125
+    },
+    {
+      "epoch": 0.36775106082036774,
+      "grad_norm": 0.9719401597976685,
+      "learning_rate": 2.9592071427316492e-05,
+      "loss": 0.8522,
+      "step": 130
+    },
+    {
+      "epoch": 0.38189533239038187,
+      "grad_norm": 1.0798102617263794,
+      "learning_rate": 2.9536159293436172e-05,
+      "loss": 0.8656,
+      "step": 135
+    },
+    {
+      "epoch": 0.39603960396039606,
+      "grad_norm": 0.9090424180030823,
+      "learning_rate": 2.9476717282537897e-05,
+      "loss": 0.8121,
+      "step": 140
+    },
+    {
+      "epoch": 0.4101838755304102,
+      "grad_norm": 0.9594690203666687,
+      "learning_rate": 2.9413759829176497e-05,
+      "loss": 0.8172,
+      "step": 145
+    },
+    {
+      "epoch": 0.4243281471004243,
+      "grad_norm": 0.8140248656272888,
+      "learning_rate": 2.934730222157653e-05,
+      "loss": 0.7469,
+      "step": 150
+    },
+    {
+      "epoch": 0.43847241867043846,
+      "grad_norm": 1.3013801574707031,
+      "learning_rate": 2.927736059791984e-05,
+      "loss": 0.7695,
+      "step": 155
+    },
+    {
+      "epoch": 0.4526166902404526,
+      "grad_norm": 1.080292820930481,
+      "learning_rate": 2.9203951942426587e-05,
+      "loss": 0.7836,
+      "step": 160
+    },
+    {
+      "epoch": 0.4667609618104668,
+      "grad_norm": 0.8857566714286804,
+      "learning_rate": 2.9127094081230952e-05,
+      "loss": 0.7534,
+      "step": 165
+    },
+    {
+      "epoch": 0.4809052333804809,
+      "grad_norm": 0.975486695766449,
+      "learning_rate": 2.9046805678052312e-05,
+      "loss": 0.7059,
+      "step": 170
+    },
+    {
+      "epoch": 0.49504950495049505,
+      "grad_norm": 0.8915780782699585,
+      "learning_rate": 2.8963106229663064e-05,
+      "loss": 0.7638,
+      "step": 175
+    },
+    {
+      "epoch": 0.5091937765205092,
+      "grad_norm": 1.0145010948181152,
+      "learning_rate": 2.8876016061154125e-05,
+      "loss": 0.698,
+      "step": 180
+    },
+    {
+      "epoch": 0.5233380480905233,
+      "grad_norm": 0.9601012468338013,
+      "learning_rate": 2.8785556320999308e-05,
+      "loss": 0.6783,
+      "step": 185
+    },
+    {
+      "epoch": 0.5374823196605375,
+      "grad_norm": 1.096042275428772,
+      "learning_rate": 2.8691748975919784e-05,
+      "loss": 0.6536,
+      "step": 190
+    },
+    {
+      "epoch": 0.5516265912305516,
+      "grad_norm": 1.0635658502578735,
+      "learning_rate": 2.8594616805549752e-05,
+      "loss": 0.6427,
+      "step": 195
+    },
+    {
+      "epoch": 0.5657708628005658,
+      "grad_norm": 1.121099591255188,
+      "learning_rate": 2.84941833969048e-05,
+      "loss": 0.6255,
+      "step": 200
+    },
+    {
+      "epoch": 0.57991513437058,
+      "grad_norm": 1.0187793970108032,
+      "learning_rate": 2.8390473138654175e-05,
+      "loss": 0.6564,
+      "step": 205
+    },
+    {
+      "epoch": 0.594059405940594,
+      "grad_norm": 1.0827370882034302,
+      "learning_rate": 2.828351121519835e-05,
+      "loss": 0.611,
+      "step": 210
+    },
+    {
+      "epoch": 0.6082036775106082,
+      "grad_norm": 1.049991250038147,
+      "learning_rate": 2.817332360055343e-05,
+      "loss": 0.5932,
+      "step": 215
+    },
+    {
+      "epoch": 0.6223479490806223,
+      "grad_norm": 1.0529643297195435,
+      "learning_rate": 2.8059937052043724e-05,
+      "loss": 0.6175,
+      "step": 220
+    },
+    {
+      "epoch": 0.6364922206506365,
+      "grad_norm": 1.067714810371399,
+      "learning_rate": 2.7943379103804196e-05,
+      "loss": 0.5887,
+      "step": 225
+    },
+    {
+      "epoch": 0.6506364922206507,
+      "grad_norm": 1.1245330572128296,
+      "learning_rate": 2.7823678060094198e-05,
+      "loss": 0.5658,
+      "step": 230
+    },
+    {
+      "epoch": 0.6647807637906648,
+      "grad_norm": 1.0193228721618652,
+      "learning_rate": 2.770086298842426e-05,
+      "loss": 0.5381,
+      "step": 235
+    },
+    {
+      "epoch": 0.6789250353606789,
+      "grad_norm": 1.012141227722168,
+      "learning_rate": 2.7574963712497486e-05,
+      "loss": 0.6174,
+      "step": 240
+    },
+    {
+      "epoch": 0.693069306930693,
+      "grad_norm": 1.1761394739151,
+      "learning_rate": 2.7446010804967314e-05,
+      "loss": 0.577,
+      "step": 245
+    },
+    {
+      "epoch": 0.7072135785007072,
+      "grad_norm": 1.1795742511749268,
+      "learning_rate": 2.7314035580013487e-05,
+      "loss": 0.563,
+      "step": 250
+    },
+    {
+      "epoch": 0.7213578500707214,
+      "grad_norm": 1.019864797592163,
+      "learning_rate": 2.717907008573785e-05,
+      "loss": 0.5822,
+      "step": 255
+    },
+    {
+      "epoch": 0.7355021216407355,
+      "grad_norm": 1.274475336074829,
+      "learning_rate": 2.7041147096381988e-05,
+      "loss": 0.5201,
+      "step": 260
+    },
+    {
+      "epoch": 0.7496463932107497,
+      "grad_norm": 1.3784422874450684,
+      "learning_rate": 2.6900300104368527e-05,
+      "loss": 0.5397,
+      "step": 265
+    },
+    {
+      "epoch": 0.7637906647807637,
+      "grad_norm": 1.1281862258911133,
+      "learning_rate": 2.6756563312168022e-05,
+      "loss": 0.5323,
+      "step": 270
+    },
+    {
+      "epoch": 0.7779349363507779,
+      "grad_norm": 1.314391851425171,
+      "learning_rate": 2.6609971623993412e-05,
+      "loss": 0.4568,
+      "step": 275
+    },
+    {
+      "epoch": 0.7920792079207921,
+      "grad_norm": 1.223934531211853,
+      "learning_rate": 2.6460560637324113e-05,
+      "loss": 0.4579,
+      "step": 280
+    },
+    {
+      "epoch": 0.8062234794908062,
+      "grad_norm": 1.1018980741500854,
+      "learning_rate": 2.6308366634261697e-05,
+      "loss": 0.4582,
+      "step": 285
+    },
+    {
+      "epoch": 0.8203677510608204,
+      "grad_norm": 1.1452151536941528,
+      "learning_rate": 2.6153426572719393e-05,
+      "loss": 0.5036,
+      "step": 290
+    },
+    {
+      "epoch": 0.8345120226308345,
+      "grad_norm": 1.1809228658676147,
+      "learning_rate": 2.5995778077447393e-05,
+      "loss": 0.4506,
+      "step": 295
+    },
+    {
+      "epoch": 0.8486562942008486,
+      "grad_norm": 1.306628942489624,
+      "learning_rate": 2.5835459430896334e-05,
+      "loss": 0.4625,
+      "step": 300
+    },
+    {
+      "epoch": 0.8628005657708628,
+      "grad_norm": 1.1720468997955322,
+      "learning_rate": 2.5672509563920953e-05,
+      "loss": 0.4858,
+      "step": 305
+    },
+    {
+      "epoch": 0.8769448373408769,
+      "grad_norm": 1.2971402406692505,
+      "learning_rate": 2.5506968046326374e-05,
+      "loss": 0.4482,
+      "step": 310
+    },
+    {
+      "epoch": 0.8910891089108911,
+      "grad_norm": 0.9836746454238892,
+      "learning_rate": 2.5338875077259207e-05,
+      "loss": 0.4743,
+      "step": 315
+    },
+    {
+      "epoch": 0.9052333804809052,
+      "grad_norm": 1.1179490089416504,
+      "learning_rate": 2.5168271475445795e-05,
+      "loss": 0.4469,
+      "step": 320
+    },
+    {
+      "epoch": 0.9193776520509194,
+      "grad_norm": 1.301911473274231,
+      "learning_rate": 2.499519866928006e-05,
+      "loss": 0.4434,
+      "step": 325
+    },
+    {
+      "epoch": 0.9335219236209336,
+      "grad_norm": 1.2816864252090454,
+      "learning_rate": 2.4819698686763245e-05,
+      "loss": 0.4484,
+      "step": 330
+    },
+    {
+      "epoch": 0.9476661951909476,
+      "grad_norm": 1.0469902753829956,
+      "learning_rate": 2.464181414529809e-05,
+      "loss": 0.4616,
+      "step": 335
+    },
+    {
+      "epoch": 0.9618104667609618,
+      "grad_norm": 1.2096281051635742,
+      "learning_rate": 2.4461588241339877e-05,
+      "loss": 0.4239,
+      "step": 340
+    },
+    {
+      "epoch": 0.9759547383309759,
+      "grad_norm": 1.1392852067947388,
+      "learning_rate": 2.4279064739906824e-05,
+      "loss": 0.4087,
+      "step": 345
+    },
+    {
+      "epoch": 0.9900990099009901,
+      "grad_norm": 1.468208909034729,
+      "learning_rate": 2.4094287963952468e-05,
+      "loss": 0.4462,
+      "step": 350
+    },
+    {
+      "epoch": 1.002828854314003,
+      "grad_norm": 1.2137415409088135,
+      "learning_rate": 2.3907302783602522e-05,
+      "loss": 0.3975,
+      "step": 355
+    },
+    {
+      "epoch": 1.016973125884017,
+      "grad_norm": 1.1279393434524536,
+      "learning_rate": 2.3718154605258885e-05,
+      "loss": 0.3556,
+      "step": 360
+    },
+    {
+      "epoch": 1.031117397454031,
+      "grad_norm": 1.1613800525665283,
+      "learning_rate": 2.3526889360573387e-05,
+      "loss": 0.3459,
+      "step": 365
+    },
+    {
+      "epoch": 1.0452616690240453,
+      "grad_norm": 1.5729336738586426,
+      "learning_rate": 2.3333553495294033e-05,
+      "loss": 0.3471,
+      "step": 370
+    },
+    {
+      "epoch": 1.0594059405940595,
+      "grad_norm": 1.4145255088806152,
+      "learning_rate": 2.3138193957986392e-05,
+      "loss": 0.3502,
+      "step": 375
+    },
+    {
+      "epoch": 1.0735502121640736,
+      "grad_norm": 1.2861406803131104,
+      "learning_rate": 2.2940858188632853e-05,
+      "loss": 0.3574,
+      "step": 380
+    },
+    {
+      "epoch": 1.0876944837340876,
+      "grad_norm": 1.2673900127410889,
+      "learning_rate": 2.27415941071126e-05,
+      "loss": 0.3136,
+      "step": 385
+    },
+    {
+      "epoch": 1.1018387553041018,
+      "grad_norm": 1.2568155527114868,
+      "learning_rate": 2.2540450101565002e-05,
+      "loss": 0.3468,
+      "step": 390
+    },
+    {
+      "epoch": 1.115983026874116,
+      "grad_norm": 1.3373382091522217,
+      "learning_rate": 2.233747501663934e-05,
+      "loss": 0.3319,
+      "step": 395
+    },
+    {
+      "epoch": 1.1301272984441302,
+      "grad_norm": 1.3243392705917358,
+      "learning_rate": 2.213271814163363e-05,
+      "loss": 0.354,
+      "step": 400
+    },
+    {
+      "epoch": 1.1442715700141444,
+      "grad_norm": 1.1834981441497803,
+      "learning_rate": 2.192622919852551e-05,
+      "loss": 0.3007,
+      "step": 405
+    },
+    {
+      "epoch": 1.1584158415841583,
+      "grad_norm": 1.3145798444747925,
+      "learning_rate": 2.171805832989804e-05,
+      "loss": 0.2841,
+      "step": 410
+    },
+    {
+      "epoch": 1.1725601131541725,
+      "grad_norm": 1.2424591779708862,
+      "learning_rate": 2.1508256086763372e-05,
+      "loss": 0.3398,
+      "step": 415
+    },
+    {
+      "epoch": 1.1867043847241867,
+      "grad_norm": 1.3383370637893677,
+      "learning_rate": 2.1296873416287216e-05,
+      "loss": 0.3213,
+      "step": 420
+    },
+    {
+      "epoch": 1.200848656294201,
+      "grad_norm": 1.335883378982544,
+      "learning_rate": 2.108396164941713e-05,
+      "loss": 0.3185,
+      "step": 425
+    },
+    {
+      "epoch": 1.214992927864215,
+      "grad_norm": 1.2965710163116455,
+      "learning_rate": 2.0869572488417626e-05,
+      "loss": 0.2877,
+      "step": 430
+    },
+    {
+      "epoch": 1.229137199434229,
+      "grad_norm": 1.2683219909667969,
+      "learning_rate": 2.0653757994315078e-05,
+      "loss": 0.2864,
+      "step": 435
+    },
+    {
+      "epoch": 1.2432814710042432,
+      "grad_norm": 1.278365135192871,
+      "learning_rate": 2.0436570574255526e-05,
+      "loss": 0.2609,
+      "step": 440
+    },
+    {
+      "epoch": 1.2574257425742574,
+      "grad_norm": 1.2500219345092773,
+      "learning_rate": 2.0218062968778407e-05,
+      "loss": 0.2867,
+      "step": 445
+    },
+    {
+      "epoch": 1.2715700141442716,
+      "grad_norm": 1.2469351291656494,
+      "learning_rate": 1.999828823900939e-05,
+      "loss": 0.2962,
+      "step": 450
+    },
+    {
+      "epoch": 1.2857142857142856,
+      "grad_norm": 1.203505039215088,
+      "learning_rate": 1.977729975377527e-05,
+      "loss": 0.28,
+      "step": 455
+    },
+    {
+      "epoch": 1.2998585572842998,
+      "grad_norm": 1.1468207836151123,
+      "learning_rate": 1.9555151176644223e-05,
+      "loss": 0.2663,
+      "step": 460
+    },
+    {
+      "epoch": 1.314002828854314,
+      "grad_norm": 1.4789674282073975,
+      "learning_rate": 1.933189645289445e-05,
+      "loss": 0.2709,
+      "step": 465
+    },
+    {
+      "epoch": 1.3281471004243282,
+      "grad_norm": 1.365321397781372,
+      "learning_rate": 1.91075897964144e-05,
+      "loss": 0.2823,
+      "step": 470
+    },
+    {
+      "epoch": 1.3422913719943423,
+      "grad_norm": 1.3452250957489014,
+      "learning_rate": 1.888228567653781e-05,
+      "loss": 0.2599,
+      "step": 475
+    },
+    {
+      "epoch": 1.3564356435643563,
+      "grad_norm": 1.245861291885376,
+      "learning_rate": 1.8656038804816662e-05,
+      "loss": 0.2645,
+      "step": 480
+    },
+    {
+      "epoch": 1.3705799151343705,
+      "grad_norm": 1.340573787689209,
+      "learning_rate": 1.8428904121735345e-05,
+      "loss": 0.2729,
+      "step": 485
+    },
+    {
+      "epoch": 1.3847241867043847,
+      "grad_norm": 1.5999422073364258,
+      "learning_rate": 1.8200936783369252e-05,
+      "loss": 0.2205,
+      "step": 490
+    },
+    {
+      "epoch": 1.3988684582743989,
+      "grad_norm": 1.0805375576019287,
+      "learning_rate": 1.7972192147990966e-05,
+      "loss": 0.2342,
+      "step": 495
+    },
+    {
+      "epoch": 1.413012729844413,
+      "grad_norm": 1.2994537353515625,
+      "learning_rate": 1.7742725762627395e-05,
+      "loss": 0.2354,
+      "step": 500
+    },
+    {
+      "epoch": 1.427157001414427,
+      "grad_norm": 1.3901698589324951,
+      "learning_rate": 1.7512593349571046e-05,
+      "loss": 0.2331,
+      "step": 505
+    },
+    {
+      "epoch": 1.4413012729844414,
+      "grad_norm": 1.3277639150619507,
+      "learning_rate": 1.7281850792848752e-05,
+      "loss": 0.2386,
+      "step": 510
+    },
+    {
+      "epoch": 1.4554455445544554,
+      "grad_norm": 1.1629115343093872,
+      "learning_rate": 1.7050554124651103e-05,
+      "loss": 0.2115,
+      "step": 515
+    },
+    {
+      "epoch": 1.4695898161244696,
+      "grad_norm": 1.1724551916122437,
+      "learning_rate": 1.6818759511725922e-05,
+      "loss": 0.23,
+      "step": 520
+    },
+    {
+      "epoch": 1.4837340876944838,
+      "grad_norm": 1.0543526411056519,
+      "learning_rate": 1.658652324173907e-05,
+      "loss": 0.2076,
+      "step": 525
+    },
+    {
+      "epoch": 1.4978783592644977,
+      "grad_norm": 1.2712016105651855,
+      "learning_rate": 1.6353901709605838e-05,
+      "loss": 0.1885,
+      "step": 530
+    },
+    {
+      "epoch": 1.5120226308345122,
+      "grad_norm": 1.18239426612854,
+      "learning_rate": 1.6120951403796367e-05,
+      "loss": 0.2065,
+      "step": 535
+    },
+    {
+      "epoch": 1.5261669024045261,
+      "grad_norm": 1.333791732788086,
+      "learning_rate": 1.5887728892618292e-05,
+      "loss": 0.1853,
+      "step": 540
+    },
+    {
+      "epoch": 1.5403111739745403,
+      "grad_norm": 1.2601056098937988,
+      "learning_rate": 1.5654290810480043e-05,
+      "loss": 0.2001,
+      "step": 545
+    },
+    {
+      "epoch": 1.5544554455445545,
+      "grad_norm": 1.3857390880584717,
+      "learning_rate": 1.5420693844138036e-05,
+      "loss": 0.2135,
+      "step": 550
+    },
+    {
+      "epoch": 1.5685997171145685,
+      "grad_norm": 1.197896122932434,
+      "learning_rate": 1.5186994718931227e-05,
+      "loss": 0.206,
+      "step": 555
+    },
+    {
+      "epoch": 1.5827439886845829,
+      "grad_norm": 1.4663914442062378,
+      "learning_rate": 1.4953250185006236e-05,
+      "loss": 0.1854,
+      "step": 560
+    },
+    {
+      "epoch": 1.5968882602545968,
+      "grad_norm": 1.130748987197876,
+      "learning_rate": 1.471951700353647e-05,
+      "loss": 0.1999,
+      "step": 565
+    },
+    {
+      "epoch": 1.611032531824611,
+      "grad_norm": 1.272182583808899,
+      "learning_rate": 1.4485851932938575e-05,
+      "loss": 0.2086,
+      "step": 570
+    },
+    {
+      "epoch": 1.6251768033946252,
+      "grad_norm": 1.2622343301773071,
+      "learning_rate": 1.4252311715089542e-05,
+      "loss": 0.2056,
+      "step": 575
+    },
+    {
+      "epoch": 1.6393210749646392,
+      "grad_norm": 1.158294677734375,
+      "learning_rate": 1.4018953061547854e-05,
+      "loss": 0.1947,
+      "step": 580
+    },
+    {
+      "epoch": 1.6534653465346536,
+      "grad_norm": 1.2073551416397095,
+      "learning_rate": 1.3785832639781951e-05,
+      "loss": 0.1833,
+      "step": 585
+    },
+    {
+      "epoch": 1.6676096181046676,
+      "grad_norm": 1.1043719053268433,
+      "learning_rate": 1.355300705940945e-05,
+      "loss": 0.1889,
+      "step": 590
+    },
+    {
+      "epoch": 1.6817538896746818,
+      "grad_norm": 1.0225766897201538,
+      "learning_rate": 1.3320532858450382e-05,
+      "loss": 0.1775,
+      "step": 595
+    },
+    {
+      "epoch": 1.695898161244696,
+      "grad_norm": 1.2476178407669067,
+      "learning_rate": 1.3088466489597838e-05,
+      "loss": 0.1851,
+      "step": 600
+    },
+    {
+      "epoch": 1.71004243281471,
+      "grad_norm": 1.2418464422225952,
+      "learning_rate": 1.2856864306509303e-05,
+      "loss": 0.1868,
+      "step": 605
+    },
+    {
+      "epoch": 1.7241867043847243,
+      "grad_norm": 1.302118182182312,
+      "learning_rate": 1.2625782550122076e-05,
+      "loss": 0.1735,
+      "step": 610
+    },
+    {
+      "epoch": 1.7383309759547383,
+      "grad_norm": 1.1017781496047974,
+      "learning_rate": 1.2395277334996045e-05,
+      "loss": 0.1704,
+      "step": 615
+    },
+    {
+      "epoch": 1.7524752475247525,
+      "grad_norm": 1.327065348625183,
+      "learning_rate": 1.2165404635687139e-05,
+      "loss": 0.1688,
+      "step": 620
+    },
+    {
+      "epoch": 1.7666195190947667,
+      "grad_norm": 1.2876105308532715,
+      "learning_rate": 1.1936220273154797e-05,
+      "loss": 0.1861,
+      "step": 625
+    },
+    {
+      "epoch": 1.7807637906647806,
+      "grad_norm": 1.127238392829895,
+      "learning_rate": 1.1707779901206733e-05,
+      "loss": 0.1633,
+      "step": 630
+    },
+    {
+      "epoch": 1.794908062234795,
+      "grad_norm": 1.2322150468826294,
+      "learning_rate": 1.1480138992984276e-05,
+      "loss": 0.1598,
+      "step": 635
+    },
+    {
+      "epoch": 1.809052333804809,
+      "grad_norm": 1.243147373199463,
+      "learning_rate": 1.1253352827491594e-05,
+      "loss": 0.1373,
+      "step": 640
+    },
+    {
+      "epoch": 1.8231966053748232,
+      "grad_norm": 0.9856458902359009,
+      "learning_rate": 1.102747647617209e-05,
+      "loss": 0.1462,
+      "step": 645
+    },
+    {
+      "epoch": 1.8373408769448374,
+      "grad_norm": 1.614333152770996,
+      "learning_rate": 1.080256478953512e-05,
+      "loss": 0.1847,
+      "step": 650
+    },
+    {
+      "epoch": 1.8514851485148514,
+      "grad_norm": 1.194786548614502,
+      "learning_rate": 1.0578672383836437e-05,
+      "loss": 0.1216,
+      "step": 655
+    },
+    {
+      "epoch": 1.8656294200848658,
+      "grad_norm": 1.1689525842666626,
+      "learning_rate": 1.0355853627815506e-05,
+      "loss": 0.1593,
+      "step": 660
+    },
+    {
+      "epoch": 1.8797736916548797,
+      "grad_norm": 1.1279332637786865,
+      "learning_rate": 1.0134162629492895e-05,
+      "loss": 0.1506,
+      "step": 665
+    },
+    {
+      "epoch": 1.893917963224894,
+      "grad_norm": 1.4175636768341064,
+      "learning_rate": 9.913653223031e-06,
+      "loss": 0.1796,
+      "step": 670
+    },
+    {
+      "epoch": 1.908062234794908,
+      "grad_norm": 1.0475925207138062,
+      "learning_rate": 9.69437895566128e-06,
+      "loss": 0.1445,
+      "step": 675
+    },
+    {
+      "epoch": 1.922206506364922,
+      "grad_norm": 1.3626298904418945,
+      "learning_rate": 9.476393074681134e-06,
+      "loss": 0.1608,
+      "step": 680
+    },
+    {
+      "epoch": 1.9363507779349365,
+      "grad_norm": 1.0629104375839233,
+      "learning_rate": 9.259748514523654e-06,
+      "loss": 0.109,
+      "step": 685
+    },
+    {
+      "epoch": 1.9504950495049505,
+      "grad_norm": 1.1413698196411133,
+      "learning_rate": 9.044497883903326e-06,
+      "loss": 0.1339,
+      "step": 690
+    },
+    {
+      "epoch": 1.9646393210749646,
+      "grad_norm": 1.0946385860443115,
+      "learning_rate": 8.83069345304083e-06,
+      "loss": 0.1405,
+      "step": 695
+    },
+    {
+      "epoch": 1.9787835926449788,
+      "grad_norm": 1.299659013748169,
+      "learning_rate": 8.618387140970047e-06,
+      "loss": 0.1719,
+      "step": 700
+    },
+    {
+      "epoch": 1.9929278642149928,
+      "grad_norm": 1.2637230157852173,
+      "learning_rate": 8.407630502930325e-06,
+      "loss": 0.1275,
+      "step": 705
+    },
+    {
+      "epoch": 2.005657708628006,
+      "grad_norm": 1.1021050214767456,
+      "learning_rate": 8.198474717847146e-06,
+      "loss": 0.1197,
+      "step": 710
+    },
+    {
+      "epoch": 2.01980198019802,
+      "grad_norm": 1.1758499145507812,
+      "learning_rate": 7.99097057590407e-06,
+      "loss": 0.1179,
+      "step": 715
+    },
+    {
+      "epoch": 2.033946251768034,
+      "grad_norm": 1.1342909336090088,
+      "learning_rate": 7.785168466209188e-06,
+      "loss": 0.1073,
+      "step": 720
+    },
+    {
+      "epoch": 2.048090523338048,
+      "grad_norm": 1.1724259853363037,
+      "learning_rate": 7.581118364558889e-06,
+      "loss": 0.1067,
+      "step": 725
+    },
+    {
+      "epoch": 2.062234794908062,
+      "grad_norm": 1.0283716917037964,
+      "learning_rate": 7.378869821302062e-06,
+      "loss": 0.0921,
+      "step": 730
+    },
+    {
+      "epoch": 2.0763790664780766,
+      "grad_norm": 1.1380531787872314,
+      "learning_rate": 7.17847194930753e-06,
+      "loss": 0.1023,
+      "step": 735
+    },
+    {
+      "epoch": 2.0905233380480905,
+      "grad_norm": 1.0300986766815186,
+      "learning_rate": 6.9799734120378105e-06,
+      "loss": 0.0927,
+      "step": 740
+    },
+    {
+      "epoch": 2.1046676096181045,
+      "grad_norm": 1.0661970376968384,
+      "learning_rate": 6.783422411731932e-06,
+      "loss": 0.1056,
+      "step": 745
+    },
+    {
+      "epoch": 2.118811881188119,
+      "grad_norm": 1.1393872499465942,
+      "learning_rate": 6.58886667770028e-06,
+      "loss": 0.1022,
+      "step": 750
+    },
+    {
+      "epoch": 2.132956152758133,
+      "grad_norm": 0.9001080989837646,
+      "learning_rate": 6.3963534547343126e-06,
+      "loss": 0.0889,
+      "step": 755
+    },
+    {
+      "epoch": 2.1471004243281473,
+      "grad_norm": 0.9404515027999878,
+      "learning_rate": 6.205929491633869e-06,
+      "loss": 0.0985,
+      "step": 760
+    },
+    {
+      "epoch": 2.1612446958981613,
+      "grad_norm": 1.0070072412490845,
+      "learning_rate": 6.017641029854996e-06,
+      "loss": 0.0879,
+      "step": 765
+    },
+    {
+      "epoch": 2.1753889674681752,
+      "grad_norm": 0.9131155014038086,
+      "learning_rate": 5.831533792280926e-06,
+      "loss": 0.0867,
+      "step": 770
+    },
+    {
+      "epoch": 2.1895332390381896,
+      "grad_norm": 1.0125209093093872,
+      "learning_rate": 5.647652972118998e-06,
+      "loss": 0.0939,
+      "step": 775
+    },
+    {
+      "epoch": 2.2036775106082036,
+      "grad_norm": 0.9427091479301453,
+      "learning_rate": 5.46604322192618e-06,
+      "loss": 0.0974,
+      "step": 780
+    },
+    {
+      "epoch": 2.217821782178218,
+      "grad_norm": 1.0026077032089233,
+      "learning_rate": 5.286748642765945e-06,
+      "loss": 0.0968,
+      "step": 785
+    },
+    {
+      "epoch": 2.231966053748232,
+      "grad_norm": 0.8422291278839111,
+      "learning_rate": 5.109812773498968e-06,
+      "loss": 0.0931,
+      "step": 790
+    },
+    {
+      "epoch": 2.246110325318246,
+      "grad_norm": 0.9946248531341553,
+      "learning_rate": 4.935278580210451e-06,
+      "loss": 0.0882,
+      "step": 795
+    },
+    {
+      "epoch": 2.2602545968882604,
+      "grad_norm": 0.808124840259552,
+      "learning_rate": 4.763188445776447e-06,
+      "loss": 0.0896,
+      "step": 800
+    },
+    {
+      "epoch": 2.2743988684582743,
+      "grad_norm": 0.6636917591094971,
+      "learning_rate": 4.593584159571875e-06,
+      "loss": 0.0872,
+      "step": 805
+    },
+    {
+      "epoch": 2.2885431400282887,
+      "grad_norm": 0.8471669554710388,
+      "learning_rate": 4.426506907322624e-06,
+      "loss": 0.0772,
+      "step": 810
+    },
+    {
+      "epoch": 2.3026874115983027,
+      "grad_norm": 0.7714807987213135,
+      "learning_rate": 4.261997261104223e-06,
+      "loss": 0.0829,
+      "step": 815
+    },
+    {
+      "epoch": 2.3168316831683167,
+      "grad_norm": 0.9270514249801636,
+      "learning_rate": 4.100095169489597e-06,
+      "loss": 0.0959,
+      "step": 820
+    },
+    {
+      "epoch": 2.330975954738331,
+      "grad_norm": 0.920443058013916,
+      "learning_rate": 3.940839947848141e-06,
+      "loss": 0.0818,
+      "step": 825
+    },
+    {
+      "epoch": 2.345120226308345,
+      "grad_norm": 0.8559970855712891,
+      "learning_rate": 3.7842702687986374e-06,
+      "loss": 0.0964,
+      "step": 830
+    },
+    {
+      "epoch": 2.3592644978783595,
+      "grad_norm": 1.0089553594589233,
+      "learning_rate": 3.630424152818203e-06,
+      "loss": 0.0843,
+      "step": 835
+    },
+    {
+      "epoch": 2.3734087694483734,
+      "grad_norm": 0.9114897847175598,
+      "learning_rate": 3.479338959009625e-06,
+      "loss": 0.0778,
+      "step": 840
+    },
+    {
+      "epoch": 2.3875530410183874,
+      "grad_norm": 0.8136684894561768,
+      "learning_rate": 3.331051376029279e-06,
+      "loss": 0.0911,
+      "step": 845
+    },
+    {
+      "epoch": 2.401697312588402,
+      "grad_norm": 0.9038828015327454,
+      "learning_rate": 3.1855974131778696e-06,
+      "loss": 0.0807,
+      "step": 850
+    },
+    {
+      "epoch": 2.4158415841584158,
+      "grad_norm": 0.9237598776817322,
+      "learning_rate": 3.0430123916561677e-06,
+      "loss": 0.0726,
+      "step": 855
+    },
+    {
+      "epoch": 2.42998585572843,
+      "grad_norm": 1.0363318920135498,
+      "learning_rate": 2.90333093598776e-06,
+      "loss": 0.0894,
+      "step": 860
+    },
+    {
+      "epoch": 2.444130127298444,
+      "grad_norm": 0.9020030498504639,
+      "learning_rate": 2.7665869656110974e-06,
+      "loss": 0.088,
+      "step": 865
+    },
+    {
+      "epoch": 2.458274398868458,
+      "grad_norm": 0.8941312432289124,
+      "learning_rate": 2.6328136866426294e-06,
+      "loss": 0.0834,
+      "step": 870
+    },
+    {
+      "epoch": 2.4724186704384725,
+      "grad_norm": 0.8324545621871948,
+      "learning_rate": 2.502043583813268e-06,
+      "loss": 0.0793,
+      "step": 875
+    },
+    {
+      "epoch": 2.4865629420084865,
+      "grad_norm": 0.8406239748001099,
+      "learning_rate": 2.3743084125799515e-06,
+      "loss": 0.0861,
+      "step": 880
+    },
+    {
+      "epoch": 2.500707213578501,
+      "grad_norm": 1.0227402448654175,
+      "learning_rate": 2.2496391914143634e-06,
+      "loss": 0.0872,
+      "step": 885
+    },
+    {
+      "epoch": 2.514851485148515,
+      "grad_norm": 0.6563328504562378,
+      "learning_rate": 2.1280661942705554e-06,
+      "loss": 0.0807,
+      "step": 890
+    },
+    {
+      "epoch": 2.528995756718529,
+      "grad_norm": 0.7643815875053406,
+      "learning_rate": 2.0096189432334194e-06,
+      "loss": 0.0911,
+      "step": 895
+    },
+    {
+      "epoch": 2.5431400282885432,
+      "grad_norm": 0.6065130829811096,
+      "learning_rate": 1.8943262013497082e-06,
+      "loss": 0.0773,
+      "step": 900
+    },
+    {
+      "epoch": 2.557284299858557,
+      "grad_norm": 0.8953626751899719,
+      "learning_rate": 1.782215965643364e-06,
+      "loss": 0.0747,
+      "step": 905
+    },
+    {
+      "epoch": 2.571428571428571,
+      "grad_norm": 0.7425594329833984,
+      "learning_rate": 1.6733154603169177e-06,
+      "loss": 0.0729,
+      "step": 910
+    },
+    {
+      "epoch": 2.5855728429985856,
+      "grad_norm": 0.8329071998596191,
+      "learning_rate": 1.5676511301404861e-06,
+      "loss": 0.0783,
+      "step": 915
+    },
+    {
+      "epoch": 2.5997171145685996,
+      "grad_norm": 0.8714070916175842,
+      "learning_rate": 1.465248634030103e-06,
+      "loss": 0.0729,
+      "step": 920
+    },
+    {
+      "epoch": 2.613861386138614,
+      "grad_norm": 0.7767753005027771,
+      "learning_rate": 1.3661328388168358e-06,
+      "loss": 0.0716,
+      "step": 925
+    },
+    {
+      "epoch": 2.628005657708628,
+      "grad_norm": 0.9768275022506714,
+      "learning_rate": 1.2703278132082936e-06,
+      "loss": 0.0761,
+      "step": 930
+    },
+    {
+      "epoch": 2.6421499292786423,
+      "grad_norm": 0.804628312587738,
+      "learning_rate": 1.177856821943884e-06,
+      "loss": 0.0829,
+      "step": 935
+    },
+    {
+      "epoch": 2.6562942008486563,
+      "grad_norm": 0.6891278624534607,
+      "learning_rate": 1.0887423201453778e-06,
+      "loss": 0.0781,
+      "step": 940
+    },
+    {
+      "epoch": 2.6704384724186703,
+      "grad_norm": 0.9018495082855225,
+      "learning_rate": 1.0030059478640025e-06,
+      "loss": 0.0649,
+      "step": 945
+    },
+    {
+      "epoch": 2.6845827439886847,
+      "grad_norm": 0.8080662488937378,
+      "learning_rate": 9.206685248255248e-07,
+      "loss": 0.076,
+      "step": 950
+    },
+    {
+      "epoch": 2.6987270155586986,
+      "grad_norm": 0.8211700916290283,
+      "learning_rate": 8.417500453744864e-07,
+      "loss": 0.079,
+      "step": 955
+    },
+    {
+      "epoch": 2.7128712871287126,
+      "grad_norm": 0.9703523516654968,
+      "learning_rate": 7.662696736189129e-07,
+      "loss": 0.0827,
+      "step": 960
+    },
+    {
+      "epoch": 2.727015558698727,
+      "grad_norm": 0.6461146473884583,
+      "learning_rate": 6.942457387765977e-07,
+      "loss": 0.0708,
+      "step": 965
+    },
+    {
+      "epoch": 2.741159830268741,
+      "grad_norm": 0.6331745386123657,
+      "learning_rate": 6.256957307241396e-07,
+      "loss": 0.0734,
+      "step": 970
+    },
+    {
+      "epoch": 2.7553041018387554,
+      "grad_norm": 0.8518006801605225,
+      "learning_rate": 5.606362957498195e-07,
+      "loss": 0.0711,
+      "step": 975
+    },
+    {
+      "epoch": 2.7694483734087694,
+      "grad_norm": 0.6589148640632629,
+      "learning_rate": 4.990832325112898e-07,
+      "loss": 0.0785,
+      "step": 980
+    },
+    {
+      "epoch": 2.783592644978784,
+      "grad_norm": 0.8413832783699036,
+      "learning_rate": 4.4105148819913564e-07,
+      "loss": 0.0689,
+      "step": 985
+    },
+    {
+      "epoch": 2.7977369165487977,
+      "grad_norm": 0.7082400918006897,
+      "learning_rate": 3.865551549071772e-07,
+      "loss": 0.0772,
+      "step": 990
+    },
+    {
+      "epoch": 2.8118811881188117,
+      "grad_norm": 0.7935705780982971,
+      "learning_rate": 3.3560746621043193e-07,
+      "loss": 0.0671,
+      "step": 995
+    },
+    {
+      "epoch": 2.826025459688826,
+      "grad_norm": 0.6121994853019714,
+      "learning_rate": 2.8822079395154357e-07,
+      "loss": 0.069,
+      "step": 1000
+    },
+    {
+      "epoch": 2.84016973125884,
+      "grad_norm": 0.7138044238090515,
+      "learning_rate": 2.4440664523648017e-07,
+      "loss": 0.0494,
+      "step": 1005
+    },
+    {
+      "epoch": 2.854314002828854,
+      "grad_norm": 0.8106822967529297,
+      "learning_rate": 2.041756596402161e-07,
+      "loss": 0.0717,
+      "step": 1010
+    },
+    {
+      "epoch": 2.8684582743988685,
+      "grad_norm": 0.6301187872886658,
+      "learning_rate": 1.6753760662307217e-07,
+      "loss": 0.0687,
+      "step": 1015
+    },
+    {
+      "epoch": 2.882602545968883,
+      "grad_norm": 0.7379102110862732,
+      "learning_rate": 1.3450138315836303e-07,
+      "loss": 0.0702,
+      "step": 1020
+    },
+    {
+      "epoch": 2.896746817538897,
+      "grad_norm": 0.7840865850448608,
+      "learning_rate": 1.050750115719057e-07,
+      "loss": 0.0652,
+      "step": 1025
+    },
+    {
+      "epoch": 2.910891089108911,
+      "grad_norm": 0.6759986877441406,
+      "learning_rate": 7.926563759392436e-08,
+      "loss": 0.0576,
+      "step": 1030
+    },
+    {
+      "epoch": 2.9250353606789252,
+      "grad_norm": 0.7382733821868896,
+      "learning_rate": 5.7079528623816824e-08,
+      "loss": 0.0698,
+      "step": 1035
+    },
+    {
+      "epoch": 2.939179632248939,
+      "grad_norm": 0.7347772717475891,
+      "learning_rate": 3.85220722082269e-08,
+      "loss": 0.079,
+      "step": 1040
+    },
+    {
+      "epoch": 2.953323903818953,
+      "grad_norm": 0.7667756080627441,
+      "learning_rate": 2.3597774732750932e-08,
+      "loss": 0.0833,
+      "step": 1045
+    },
+    {
+      "epoch": 2.9674681753889676,
+      "grad_norm": 0.7999646663665771,
+      "learning_rate": 1.2310260327646483e-08,
+      "loss": 0.059,
+      "step": 1050
+    },
+    {
+      "epoch": 2.9816124469589815,
+      "grad_norm": 0.6131082773208618,
+      "learning_rate": 4.662269987756318e-09,
+      "loss": 0.0685,
+      "step": 1055
+    },
+    {
+      "epoch": 2.9957567185289955,
+      "grad_norm": 0.694364070892334,
+      "learning_rate": 6.556609069091613e-10,
+      "loss": 0.0738,
+      "step": 1060
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1062,
+  "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": 1.790533321055273e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

base/24_128_e3_3e-5/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fffb8a6b4a03684b60972d97fcf9e1c8e1c2ef2cc2a362fccbf0fd0a311458e
+size 8273

base/24_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)