Upload folder using huggingface_hub

base/15_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/15_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",
+    "down_proj",
+    "q_proj",
+    "o_proj",
+    "up_proj",
+    "k_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

base/15_128_e3_3e-5/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5b26f2205e02a739e134c982485f36bb86a7b780c585ea574bc9a257bac125e
+size 791751704

base/15_128_e3_3e-5/latest

@@ -0,0 +1 @@
+global_step1206

base/15_128_e3_3e-5/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd52e3a3ac4a576cd66a559ae5a4a6843a74fb71f23326bb91f6abe6ae4d11b2
+size 16389

base/15_128_e3_3e-5/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96323865b5ec1949c259088006adfa2ff6a602d8a139b3190ba4968a3cdd5bff
+size 16389

base/15_128_e3_3e-5/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fae56551bb09e83cbb9548e3b92b05006cac494a25b918fd2b90b084614a68d
+size 16389

base/15_128_e3_3e-5/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4dd8b0de2c6f20c23b9846de13eaabcfb40d759f9a63c5230915cee00fd3203a
+size 16389

base/15_128_e3_3e-5/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:139df4232ea55ce295713aed916dac6d82dc4085ca7b9afa6deed208f881956b
+size 16389

base/15_128_e3_3e-5/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55bfa5062d42288a17132d21a7b7b52e7901b8dd0cd29053b5ae88cb5b3d45cc
+size 16389

base/15_128_e3_3e-5/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec841e3969abafc61ae47bd033680b77581c9acf8c084c3a8d5c1e889120e86a
+size 16389

base/15_128_e3_3e-5/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97b20088adcc791aa9057b4fe6cb752b83a1d6b49aff7a1fb00c55278258105e
+size 16389

base/15_128_e3_3e-5/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c15cbd142e69709606504fa4915f2c35e10aac34ede6b3eea570a384d000e192
+size 1401

base/15_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/15_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/15_128_e3_3e-5/trainer_state.json

@@ -0,0 +1,1721 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1206,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.012453300124533,
+      "grad_norm": 0.8944751620292664,
+      "learning_rate": 1.9672131147540985e-06,
+      "loss": 1.274,
+      "step": 5
+    },
+    {
+      "epoch": 0.024906600249066,
+      "grad_norm": 0.6717709302902222,
+      "learning_rate": 4.426229508196722e-06,
+      "loss": 1.2455,
+      "step": 10
+    },
+    {
+      "epoch": 0.037359900373599,
+      "grad_norm": 0.47737956047058105,
+      "learning_rate": 6.885245901639345e-06,
+      "loss": 1.277,
+      "step": 15
+    },
+    {
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.5123236775398254,
+      "learning_rate": 9.344262295081968e-06,
+      "loss": 1.287,
+      "step": 20
+    },
+    {
+      "epoch": 0.062266500622665005,
+      "grad_norm": 0.6041452288627625,
+      "learning_rate": 1.180327868852459e-05,
+      "loss": 1.2063,
+      "step": 25
+    },
+    {
+      "epoch": 0.074719800747198,
+      "grad_norm": 0.41610515117645264,
+      "learning_rate": 1.4262295081967213e-05,
+      "loss": 1.2335,
+      "step": 30
+    },
+    {
+      "epoch": 0.08717310087173101,
+      "grad_norm": 0.41696736216545105,
+      "learning_rate": 1.6721311475409834e-05,
+      "loss": 1.2678,
+      "step": 35
+    },
+    {
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.5723266005516052,
+      "learning_rate": 1.9180327868852462e-05,
+      "loss": 1.2012,
+      "step": 40
+    },
+    {
+      "epoch": 0.11207970112079702,
+      "grad_norm": 0.5067686438560486,
+      "learning_rate": 2.1639344262295084e-05,
+      "loss": 1.1862,
+      "step": 45
+    },
+    {
+      "epoch": 0.12453300124533001,
+      "grad_norm": 0.4886264503002167,
+      "learning_rate": 2.4098360655737705e-05,
+      "loss": 1.1781,
+      "step": 50
+    },
+    {
+      "epoch": 0.136986301369863,
+      "grad_norm": 0.43660783767700195,
+      "learning_rate": 2.6557377049180327e-05,
+      "loss": 1.2137,
+      "step": 55
+    },
+    {
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.442975252866745,
+      "learning_rate": 2.901639344262295e-05,
+      "loss": 1.1715,
+      "step": 60
+    },
+    {
+      "epoch": 0.16189290161892902,
+      "grad_norm": 0.5017277598381042,
+      "learning_rate": 2.9999491852149543e-05,
+      "loss": 1.1159,
+      "step": 65
+    },
+    {
+      "epoch": 0.17434620174346202,
+      "grad_norm": 0.4889864921569824,
+      "learning_rate": 2.999638662885322e-05,
+      "loss": 1.0831,
+      "step": 70
+    },
+    {
+      "epoch": 0.18679950186799502,
+      "grad_norm": 0.5795808434486389,
+      "learning_rate": 2.9990459070319718e-05,
+      "loss": 1.0808,
+      "step": 75
+    },
+    {
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.48043161630630493,
+      "learning_rate": 2.9981710292121587e-05,
+      "loss": 1.0435,
+      "step": 80
+    },
+    {
+      "epoch": 0.21170610211706103,
+      "grad_norm": 0.5148184299468994,
+      "learning_rate": 2.9970141940787794e-05,
+      "loss": 1.0211,
+      "step": 85
+    },
+    {
+      "epoch": 0.22415940224159403,
+      "grad_norm": 0.6004700660705566,
+      "learning_rate": 2.9955756193493843e-05,
+      "loss": 1.0386,
+      "step": 90
+    },
+    {
+      "epoch": 0.23661270236612703,
+      "grad_norm": 0.610205352306366,
+      "learning_rate": 2.9938555757652027e-05,
+      "loss": 0.9959,
+      "step": 95
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.6105800271034241,
+      "learning_rate": 2.991854387040189e-05,
+      "loss": 1.0207,
+      "step": 100
+    },
+    {
+      "epoch": 0.261519302615193,
+      "grad_norm": 0.6438499689102173,
+      "learning_rate": 2.9895724298000995e-05,
+      "loss": 0.9658,
+      "step": 105
+    },
+    {
+      "epoch": 0.273972602739726,
+      "grad_norm": 0.664660632610321,
+      "learning_rate": 2.9870101335116107e-05,
+      "loss": 0.9458,
+      "step": 110
+    },
+    {
+      "epoch": 0.286425902864259,
+      "grad_norm": 0.5920573472976685,
+      "learning_rate": 2.9841679804014938e-05,
+      "loss": 0.9278,
+      "step": 115
+    },
+    {
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.7403149604797363,
+      "learning_rate": 2.981046505365859e-05,
+      "loss": 0.9712,
+      "step": 120
+    },
+    {
+      "epoch": 0.31133250311332505,
+      "grad_norm": 0.6950231790542603,
+      "learning_rate": 2.9776462958694873e-05,
+      "loss": 0.9423,
+      "step": 125
+    },
+    {
+      "epoch": 0.32378580323785805,
+      "grad_norm": 0.867012619972229,
+      "learning_rate": 2.9739679918352686e-05,
+      "loss": 0.9127,
+      "step": 130
+    },
+    {
+      "epoch": 0.33623910336239105,
+      "grad_norm": 0.715686559677124,
+      "learning_rate": 2.9700122855237685e-05,
+      "loss": 0.8783,
+      "step": 135
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9141703844070435,
+      "learning_rate": 2.965779921402944e-05,
+      "loss": 0.8842,
+      "step": 140
+    },
+    {
+      "epoch": 0.36114570361145704,
+      "grad_norm": 0.6886006593704224,
+      "learning_rate": 2.961271696008033e-05,
+      "loss": 0.9081,
+      "step": 145
+    },
+    {
+      "epoch": 0.37359900373599003,
+      "grad_norm": 0.7516241073608398,
+      "learning_rate": 2.9564884577916463e-05,
+      "loss": 0.8338,
+      "step": 150
+    },
+    {
+      "epoch": 0.386052303860523,
+      "grad_norm": 0.8223601579666138,
+      "learning_rate": 2.951431106964088e-05,
+      "loss": 0.7842,
+      "step": 155
+    },
+    {
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8259574770927429,
+      "learning_rate": 2.9461005953239347e-05,
+      "loss": 0.8422,
+      "step": 160
+    },
+    {
+      "epoch": 0.410958904109589,
+      "grad_norm": 1.0828617811203003,
+      "learning_rate": 2.9404979260789064e-05,
+      "loss": 0.7767,
+      "step": 165
+    },
+    {
+      "epoch": 0.42341220423412207,
+      "grad_norm": 1.814573049545288,
+      "learning_rate": 2.934624153657061e-05,
+      "loss": 0.8412,
+      "step": 170
+    },
+    {
+      "epoch": 0.43586550435865506,
+      "grad_norm": 1.0583600997924805,
+      "learning_rate": 2.9284803835083507e-05,
+      "loss": 0.7083,
+      "step": 175
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "grad_norm": 1.0985431671142578,
+      "learning_rate": 2.9220677718965747e-05,
+      "loss": 0.7787,
+      "step": 180
+    },
+    {
+      "epoch": 0.46077210460772106,
+      "grad_norm": 0.9476515054702759,
+      "learning_rate": 2.9153875256817696e-05,
+      "loss": 0.7584,
+      "step": 185
+    },
+    {
+      "epoch": 0.47322540473225405,
+      "grad_norm": 0.8766788840293884,
+      "learning_rate": 2.9084409020930767e-05,
+      "loss": 0.7568,
+      "step": 190
+    },
+    {
+      "epoch": 0.48567870485678705,
+      "grad_norm": 1.102367639541626,
+      "learning_rate": 2.9012292084921306e-05,
+      "loss": 0.73,
+      "step": 195
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9269284605979919,
+      "learning_rate": 2.893753802127012e-05,
+      "loss": 0.713,
+      "step": 200
+    },
+    {
+      "epoch": 0.5105853051058531,
+      "grad_norm": 1.029188871383667,
+      "learning_rate": 2.8860160898768123e-05,
+      "loss": 0.6807,
+      "step": 205
+    },
+    {
+      "epoch": 0.523038605230386,
+      "grad_norm": 1.0745360851287842,
+      "learning_rate": 2.8780175279868577e-05,
+      "loss": 0.7191,
+      "step": 210
+    },
+    {
+      "epoch": 0.5354919053549191,
+      "grad_norm": 0.9979711771011353,
+      "learning_rate": 2.8697596217946426e-05,
+      "loss": 0.6793,
+      "step": 215
+    },
+    {
+      "epoch": 0.547945205479452,
+      "grad_norm": 1.4455933570861816,
+      "learning_rate": 2.861243925446523e-05,
+      "loss": 0.7008,
+      "step": 220
+    },
+    {
+      "epoch": 0.5603985056039851,
+      "grad_norm": 0.9922125339508057,
+      "learning_rate": 2.8524720416052243e-05,
+      "loss": 0.6996,
+      "step": 225
+    },
+    {
+      "epoch": 0.572851805728518,
+      "grad_norm": 1.0485483407974243,
+      "learning_rate": 2.84344562114822e-05,
+      "loss": 0.6297,
+      "step": 230
+    },
+    {
+      "epoch": 0.5853051058530511,
+      "grad_norm": 1.1109188795089722,
+      "learning_rate": 2.8341663628570328e-05,
+      "loss": 0.6691,
+      "step": 235
+    },
+    {
+      "epoch": 0.597758405977584,
+      "grad_norm": 1.066172480583191,
+      "learning_rate": 2.824636013097524e-05,
+      "loss": 0.5782,
+      "step": 240
+    },
+    {
+      "epoch": 0.6102117061021171,
+      "grad_norm": 1.0541414022445679,
+      "learning_rate": 2.8148563654912257e-05,
+      "loss": 0.5966,
+      "step": 245
+    },
+    {
+      "epoch": 0.6226650062266501,
+      "grad_norm": 1.110661268234253,
+      "learning_rate": 2.8048292605777766e-05,
+      "loss": 0.644,
+      "step": 250
+    },
+    {
+      "epoch": 0.635118306351183,
+      "grad_norm": 0.9957571029663086,
+      "learning_rate": 2.7945565854685348e-05,
+      "loss": 0.5983,
+      "step": 255
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "grad_norm": 1.0196036100387573,
+      "learning_rate": 2.7840402734914182e-05,
+      "loss": 0.5671,
+      "step": 260
+    },
+    {
+      "epoch": 0.660024906600249,
+      "grad_norm": 0.9648929238319397,
+      "learning_rate": 2.773282303827052e-05,
+      "loss": 0.6237,
+      "step": 265
+    },
+    {
+      "epoch": 0.6724782067247821,
+      "grad_norm": 0.9666902422904968,
+      "learning_rate": 2.762284701136283e-05,
+      "loss": 0.5698,
+      "step": 270
+    },
+    {
+      "epoch": 0.684931506849315,
+      "grad_norm": 1.0487078428268433,
+      "learning_rate": 2.7510495351791397e-05,
+      "loss": 0.5657,
+      "step": 275
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "grad_norm": 1.111111044883728,
+      "learning_rate": 2.739578920425297e-05,
+      "loss": 0.5536,
+      "step": 280
+    },
+    {
+      "epoch": 0.709838107098381,
+      "grad_norm": 1.348073124885559,
+      "learning_rate": 2.727875015656135e-05,
+      "loss": 0.5315,
+      "step": 285
+    },
+    {
+      "epoch": 0.7222914072229141,
+      "grad_norm": 1.1467576026916504,
+      "learning_rate": 2.7159400235584507e-05,
+      "loss": 0.5204,
+      "step": 290
+    },
+    {
+      "epoch": 0.7347447073474471,
+      "grad_norm": 1.1384685039520264,
+      "learning_rate": 2.703776190309914e-05,
+      "loss": 0.5306,
+      "step": 295
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "grad_norm": 1.138477087020874,
+      "learning_rate": 2.691385805156329e-05,
+      "loss": 0.5101,
+      "step": 300
+    },
+    {
+      "epoch": 0.7596513075965131,
+      "grad_norm": 1.0870301723480225,
+      "learning_rate": 2.6787711999808026e-05,
+      "loss": 0.5384,
+      "step": 305
+    },
+    {
+      "epoch": 0.772104607721046,
+      "grad_norm": 1.0281298160552979,
+      "learning_rate": 2.6659347488648763e-05,
+      "loss": 0.4771,
+      "step": 310
+    },
+    {
+      "epoch": 0.7845579078455791,
+      "grad_norm": 1.0905897617340088,
+      "learning_rate": 2.6528788676417238e-05,
+      "loss": 0.5443,
+      "step": 315
+    },
+    {
+      "epoch": 0.797011207970112,
+      "grad_norm": 1.1924077272415161,
+      "learning_rate": 2.6396060134414883e-05,
+      "loss": 0.5149,
+      "step": 320
+    },
+    {
+      "epoch": 0.8094645080946451,
+      "grad_norm": 9.401925086975098,
+      "learning_rate": 2.6261186842288482e-05,
+      "loss": 0.4952,
+      "step": 325
+    },
+    {
+      "epoch": 0.821917808219178,
+      "grad_norm": 1.137606143951416,
+      "learning_rate": 2.6124194183328992e-05,
+      "loss": 0.4787,
+      "step": 330
+    },
+    {
+      "epoch": 0.8343711083437111,
+      "grad_norm": 1.1087591648101807,
+      "learning_rate": 2.5985107939694346e-05,
+      "loss": 0.4712,
+      "step": 335
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "grad_norm": 1.3555240631103516,
+      "learning_rate": 2.5843954287557253e-05,
+      "loss": 0.4705,
+      "step": 340
+    },
+    {
+      "epoch": 0.8592777085927771,
+      "grad_norm": 1.2785403728485107,
+      "learning_rate": 2.5700759792178813e-05,
+      "loss": 0.4348,
+      "step": 345
+    },
+    {
+      "epoch": 0.8717310087173101,
+      "grad_norm": 1.3248988389968872,
+      "learning_rate": 2.5555551402908896e-05,
+      "loss": 0.4803,
+      "step": 350
+    },
+    {
+      "epoch": 0.8841843088418431,
+      "grad_norm": 1.4196927547454834,
+      "learning_rate": 2.5408356448114255e-05,
+      "loss": 0.4269,
+      "step": 355
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "grad_norm": 1.309814691543579,
+      "learning_rate": 2.5259202630035296e-05,
+      "loss": 0.5176,
+      "step": 360
+    },
+    {
+      "epoch": 0.9090909090909091,
+      "grad_norm": 1.0402923822402954,
+      "learning_rate": 2.51081180195725e-05,
+      "loss": 0.4238,
+      "step": 365
+    },
+    {
+      "epoch": 0.9215442092154421,
+      "grad_norm": 1.1925145387649536,
+      "learning_rate": 2.4955131051003427e-05,
+      "loss": 0.4255,
+      "step": 370
+    },
+    {
+      "epoch": 0.933997509339975,
+      "grad_norm": 1.1271957159042358,
+      "learning_rate": 2.4800270516631376e-05,
+      "loss": 0.4591,
+      "step": 375
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "grad_norm": 1.1804146766662598,
+      "learning_rate": 2.4643565561366644e-05,
+      "loss": 0.3939,
+      "step": 380
+    },
+    {
+      "epoch": 0.958904109589041,
+      "grad_norm": 1.1274124383926392,
+      "learning_rate": 2.4485045677241415e-05,
+      "loss": 0.4647,
+      "step": 385
+    },
+    {
+      "epoch": 0.9713574097135741,
+      "grad_norm": 1.6224528551101685,
+      "learning_rate": 2.4324740697859326e-05,
+      "loss": 0.4043,
+      "step": 390
+    },
+    {
+      "epoch": 0.9838107098381071,
+      "grad_norm": 1.3642631769180298,
+      "learning_rate": 2.4162680792780775e-05,
+      "loss": 0.4184,
+      "step": 395
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.0766072273254395,
+      "learning_rate": 2.399889646184494e-05,
+      "loss": 0.4339,
+      "step": 400
+    },
+    {
+      "epoch": 1.0074719800747198,
+      "grad_norm": 1.2490921020507812,
+      "learning_rate": 2.3833418529429728e-05,
+      "loss": 0.3731,
+      "step": 405
+    },
+    {
+      "epoch": 1.0199252801992529,
+      "grad_norm": 1.2552846670150757,
+      "learning_rate": 2.366627813865055e-05,
+      "loss": 0.3498,
+      "step": 410
+    },
+    {
+      "epoch": 1.0323785803237857,
+      "grad_norm": 1.2313730716705322,
+      "learning_rate": 2.349750674549918e-05,
+      "loss": 0.3278,
+      "step": 415
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1826484203338623,
+      "learning_rate": 2.332713611292371e-05,
+      "loss": 0.3841,
+      "step": 420
+    },
+    {
+      "epoch": 1.0572851805728518,
+      "grad_norm": 1.4367692470550537,
+      "learning_rate": 2.3155198304850694e-05,
+      "loss": 0.3324,
+      "step": 425
+    },
+    {
+      "epoch": 1.0697384806973849,
+      "grad_norm": 1.1179848909378052,
+      "learning_rate": 2.2981725680150745e-05,
+      "loss": 0.3253,
+      "step": 430
+    },
+    {
+      "epoch": 1.0821917808219177,
+      "grad_norm": 1.0725845098495483,
+      "learning_rate": 2.2806750886548508e-05,
+      "loss": 0.3536,
+      "step": 435
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "grad_norm": 1.3300111293792725,
+      "learning_rate": 2.2630306854478335e-05,
+      "loss": 0.3581,
+      "step": 440
+    },
+    {
+      "epoch": 1.1070983810709838,
+      "grad_norm": 1.1115226745605469,
+      "learning_rate": 2.245242679088679e-05,
+      "loss": 0.2967,
+      "step": 445
+    },
+    {
+      "epoch": 1.1195516811955168,
+      "grad_norm": 1.1758500337600708,
+      "learning_rate": 2.2273144172982985e-05,
+      "loss": 0.3087,
+      "step": 450
+    },
+    {
+      "epoch": 1.13200498132005,
+      "grad_norm": 1.0533242225646973,
+      "learning_rate": 2.2092492741938222e-05,
+      "loss": 0.3551,
+      "step": 455
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.530187726020813,
+      "learning_rate": 2.1910506496535816e-05,
+      "loss": 0.3268,
+      "step": 460
+    },
+    {
+      "epoch": 1.1569115815691158,
+      "grad_norm": 1.400524616241455,
+      "learning_rate": 2.1727219686772494e-05,
+      "loss": 0.35,
+      "step": 465
+    },
+    {
+      "epoch": 1.1693648816936488,
+      "grad_norm": 1.238639235496521,
+      "learning_rate": 2.154266680741253e-05,
+      "loss": 0.3046,
+      "step": 470
+    },
+    {
+      "epoch": 1.1818181818181819,
+      "grad_norm": 1.1537262201309204,
+      "learning_rate": 2.1356882591495795e-05,
+      "loss": 0.2677,
+      "step": 475
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.2166029214859009,
+      "learning_rate": 2.116990200380093e-05,
+      "loss": 0.2958,
+      "step": 480
+    },
+    {
+      "epoch": 1.2067247820672478,
+      "grad_norm": 1.3575838804244995,
+      "learning_rate": 2.0981760234264983e-05,
+      "loss": 0.2913,
+      "step": 485
+    },
+    {
+      "epoch": 1.2191780821917808,
+      "grad_norm": 1.247124195098877,
+      "learning_rate": 2.07924926913606e-05,
+      "loss": 0.2742,
+      "step": 490
+    },
+    {
+      "epoch": 1.2316313823163139,
+      "grad_norm": 1.1005693674087524,
+      "learning_rate": 2.0602134995432124e-05,
+      "loss": 0.2629,
+      "step": 495
+    },
+    {
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.2406929731369019,
+      "learning_rate": 2.0410722971991802e-05,
+      "loss": 0.2576,
+      "step": 500
+    },
+    {
+      "epoch": 1.25653798256538,
+      "grad_norm": 1.3006343841552734,
+      "learning_rate": 2.0218292644977396e-05,
+      "loss": 0.2729,
+      "step": 505
+    },
+    {
+      "epoch": 1.2689912826899128,
+      "grad_norm": 1.0782970190048218,
+      "learning_rate": 2.002488022997244e-05,
+      "loss": 0.3292,
+      "step": 510
+    },
+    {
+      "epoch": 1.2814445828144458,
+      "grad_norm": 1.2004520893096924,
+      "learning_rate": 1.9830522127390428e-05,
+      "loss": 0.2666,
+      "step": 515
+    },
+    {
+      "epoch": 1.293897882938979,
+      "grad_norm": 1.1653896570205688,
+      "learning_rate": 1.963525491562421e-05,
+      "loss": 0.2538,
+      "step": 520
+    },
+    {
+      "epoch": 1.3063511830635117,
+      "grad_norm": 1.273256540298462,
+      "learning_rate": 1.943911534416193e-05,
+      "loss": 0.2774,
+      "step": 525
+    },
+    {
+      "epoch": 1.3188044831880448,
+      "grad_norm": 1.3534176349639893,
+      "learning_rate": 1.924214032667069e-05,
+      "loss": 0.2627,
+      "step": 530
+    },
+    {
+      "epoch": 1.3312577833125778,
+      "grad_norm": 1.168590784072876,
+      "learning_rate": 1.9044366934049408e-05,
+      "loss": 0.2835,
+      "step": 535
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "grad_norm": 1.3265107870101929,
+      "learning_rate": 1.8845832387451995e-05,
+      "loss": 0.2532,
+      "step": 540
+    },
+    {
+      "epoch": 1.356164383561644,
+      "grad_norm": 1.1510742902755737,
+      "learning_rate": 1.8646574051282337e-05,
+      "loss": 0.272,
+      "step": 545
+    },
+    {
+      "epoch": 1.3686176836861768,
+      "grad_norm": 1.464789867401123,
+      "learning_rate": 1.844662942616224e-05,
+      "loss": 0.2493,
+      "step": 550
+    },
+    {
+      "epoch": 1.3810709838107098,
+      "grad_norm": 1.117134928703308,
+      "learning_rate": 1.8246036141873786e-05,
+      "loss": 0.2529,
+      "step": 555
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "grad_norm": 1.3581222295761108,
+      "learning_rate": 1.804483195027739e-05,
+      "loss": 0.2313,
+      "step": 560
+    },
+    {
+      "epoch": 1.405977584059776,
+      "grad_norm": 1.217405080795288,
+      "learning_rate": 1.7843054718206818e-05,
+      "loss": 0.2226,
+      "step": 565
+    },
+    {
+      "epoch": 1.4184308841843087,
+      "grad_norm": 1.166867971420288,
+      "learning_rate": 1.7640742420342672e-05,
+      "loss": 0.2628,
+      "step": 570
+    },
+    {
+      "epoch": 1.4308841843088418,
+      "grad_norm": 1.0479912757873535,
+      "learning_rate": 1.7437933132065452e-05,
+      "loss": 0.2081,
+      "step": 575
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "grad_norm": 1.218416690826416,
+      "learning_rate": 1.7234665022289777e-05,
+      "loss": 0.2243,
+      "step": 580
+    },
+    {
+      "epoch": 1.455790784557908,
+      "grad_norm": 1.2621955871582031,
+      "learning_rate": 1.7030976346280924e-05,
+      "loss": 0.221,
+      "step": 585
+    },
+    {
+      "epoch": 1.468244084682441,
+      "grad_norm": 1.1466797590255737,
+      "learning_rate": 1.6826905438455174e-05,
+      "loss": 0.2478,
+      "step": 590
+    },
+    {
+      "epoch": 1.4806973848069738,
+      "grad_norm": 1.202735185623169,
+      "learning_rate": 1.662249070516523e-05,
+      "loss": 0.2087,
+      "step": 595
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "grad_norm": 1.2896475791931152,
+      "learning_rate": 1.641777061747209e-05,
+      "loss": 0.2189,
+      "step": 600
+    },
+    {
+      "epoch": 1.5056039850560399,
+      "grad_norm": 1.2872331142425537,
+      "learning_rate": 1.621278370390476e-05,
+      "loss": 0.2063,
+      "step": 605
+    },
+    {
+      "epoch": 1.5180572851805727,
+      "grad_norm": 1.446150779724121,
+      "learning_rate": 1.6007568543209153e-05,
+      "loss": 0.2256,
+      "step": 610
+    },
+    {
+      "epoch": 1.5305105853051058,
+      "grad_norm": 1.5150728225708008,
+      "learning_rate": 1.5802163757087513e-05,
+      "loss": 0.2091,
+      "step": 615
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "grad_norm": 1.2465964555740356,
+      "learning_rate": 1.5596608002929793e-05,
+      "loss": 0.216,
+      "step": 620
+    },
+    {
+      "epoch": 1.5554171855541719,
+      "grad_norm": 1.2056424617767334,
+      "learning_rate": 1.539093996653829e-05,
+      "loss": 0.2046,
+      "step": 625
+    },
+    {
+      "epoch": 1.567870485678705,
+      "grad_norm": 1.0530568361282349,
+      "learning_rate": 1.518519835484691e-05,
+      "loss": 0.1954,
+      "step": 630
+    },
+    {
+      "epoch": 1.580323785803238,
+      "grad_norm": 1.110557198524475,
+      "learning_rate": 1.4979421888636532e-05,
+      "loss": 0.1845,
+      "step": 635
+    },
+    {
+      "epoch": 1.592777085927771,
+      "grad_norm": 1.1743559837341309,
+      "learning_rate": 1.4773649295247668e-05,
+      "loss": 0.2158,
+      "step": 640
+    },
+    {
+      "epoch": 1.6052303860523038,
+      "grad_norm": 1.06714928150177,
+      "learning_rate": 1.4567919301291976e-05,
+      "loss": 0.1906,
+      "step": 645
+    },
+    {
+      "epoch": 1.6176836861768369,
+      "grad_norm": 1.2848938703536987,
+      "learning_rate": 1.4362270625363852e-05,
+      "loss": 0.2313,
+      "step": 650
+    },
+    {
+      "epoch": 1.6301369863013697,
+      "grad_norm": 1.4657275676727295,
+      "learning_rate": 1.415674197075355e-05,
+      "loss": 0.2226,
+      "step": 655
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "grad_norm": 1.2392781972885132,
+      "learning_rate": 1.3951372018163197e-05,
+      "loss": 0.2002,
+      "step": 660
+    },
+    {
+      "epoch": 1.6550435865504358,
+      "grad_norm": 1.1880520582199097,
+      "learning_rate": 1.3746199418427044e-05,
+      "loss": 0.185,
+      "step": 665
+    },
+    {
+      "epoch": 1.6674968866749689,
+      "grad_norm": 1.2727371454238892,
+      "learning_rate": 1.3541262785237321e-05,
+      "loss": 0.1928,
+      "step": 670
+    },
+    {
+      "epoch": 1.679950186799502,
+      "grad_norm": 1.1105854511260986,
+      "learning_rate": 1.3336600687877124e-05,
+      "loss": 0.1741,
+      "step": 675
+    },
+    {
+      "epoch": 1.692403486924035,
+      "grad_norm": 1.1599812507629395,
+      "learning_rate": 1.313225164396162e-05,
+      "loss": 0.1808,
+      "step": 680
+    },
+    {
+      "epoch": 1.704856787048568,
+      "grad_norm": 1.2341241836547852,
+      "learning_rate": 1.2928254112189e-05,
+      "loss": 0.199,
+      "step": 685
+    },
+    {
+      "epoch": 1.7173100871731009,
+      "grad_norm": 1.2572351694107056,
+      "learning_rate": 1.272464648510251e-05,
+      "loss": 0.1445,
+      "step": 690
+    },
+    {
+      "epoch": 1.729763387297634,
+      "grad_norm": 1.2169501781463623,
+      "learning_rate": 1.2521467081864945e-05,
+      "loss": 0.1776,
+      "step": 695
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "grad_norm": 1.2215276956558228,
+      "learning_rate": 1.2318754141046936e-05,
+      "loss": 0.1894,
+      "step": 700
+    },
+    {
+      "epoch": 1.7546699875466998,
+      "grad_norm": 1.1602585315704346,
+      "learning_rate": 1.211654581343039e-05,
+      "loss": 0.1796,
+      "step": 705
+    },
+    {
+      "epoch": 1.7671232876712328,
+      "grad_norm": 1.2874171733856201,
+      "learning_rate": 1.1914880154828514e-05,
+      "loss": 0.1825,
+      "step": 710
+    },
+    {
+      "epoch": 1.7795765877957659,
+      "grad_norm": 1.1928547620773315,
+      "learning_rate": 1.1713795118923659e-05,
+      "loss": 0.1824,
+      "step": 715
+    },
+    {
+      "epoch": 1.792029887920299,
+      "grad_norm": 1.1241790056228638,
+      "learning_rate": 1.1513328550124379e-05,
+      "loss": 0.1731,
+      "step": 720
+    },
+    {
+      "epoch": 1.804483188044832,
+      "grad_norm": 1.611409068107605,
+      "learning_rate": 1.1313518176443099e-05,
+      "loss": 0.164,
+      "step": 725
+    },
+    {
+      "epoch": 1.816936488169365,
+      "grad_norm": 1.0602563619613647,
+      "learning_rate": 1.1114401602395647e-05,
+      "loss": 0.1291,
+      "step": 730
+    },
+    {
+      "epoch": 1.8293897882938979,
+      "grad_norm": 1.3375521898269653,
+      "learning_rate": 1.0916016301924056e-05,
+      "loss": 0.1754,
+      "step": 735
+    },
+    {
+      "epoch": 1.841843088418431,
+      "grad_norm": 1.0828105211257935,
+      "learning_rate": 1.071839961134393e-05,
+      "loss": 0.1542,
+      "step": 740
+    },
+    {
+      "epoch": 1.8542963885429637,
+      "grad_norm": 1.1736550331115723,
+      "learning_rate": 1.0521588722317707e-05,
+      "loss": 0.1401,
+      "step": 745
+    },
+    {
+      "epoch": 1.8667496886674968,
+      "grad_norm": 1.207003116607666,
+      "learning_rate": 1.0325620674855147e-05,
+      "loss": 0.1417,
+      "step": 750
+    },
+    {
+      "epoch": 1.8792029887920298,
+      "grad_norm": 1.0486218929290771,
+      "learning_rate": 1.0130532350342381e-05,
+      "loss": 0.1476,
+      "step": 755
+    },
+    {
+      "epoch": 1.891656288916563,
+      "grad_norm": 1.122843623161316,
+      "learning_rate": 9.936360464600769e-06,
+      "loss": 0.1388,
+      "step": 760
+    },
+    {
+      "epoch": 1.904109589041096,
+      "grad_norm": 1.038020133972168,
+      "learning_rate": 9.74314156097697e-06,
+      "loss": 0.1517,
+      "step": 765
+    },
+    {
+      "epoch": 1.916562889165629,
+      "grad_norm": 1.0586464405059814,
+      "learning_rate": 9.550912003465442e-06,
+      "loss": 0.1473,
+      "step": 770
+    },
+    {
+      "epoch": 1.929016189290162,
+      "grad_norm": 1.0590243339538574,
+      "learning_rate": 9.359707969864688e-06,
+      "loss": 0.1245,
+      "step": 775
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "grad_norm": 1.3519401550292969,
+      "learning_rate": 9.16956544496857e-06,
+      "loss": 0.1321,
+      "step": 780
+    },
+    {
+      "epoch": 1.953922789539228,
+      "grad_norm": 1.2438642978668213,
+      "learning_rate": 8.980520213793934e-06,
+      "loss": 0.1307,
+      "step": 785
+    },
+    {
+      "epoch": 1.9663760896637608,
+      "grad_norm": 1.066774606704712,
+      "learning_rate": 8.792607854845829e-06,
+      "loss": 0.1322,
+      "step": 790
+    },
+    {
+      "epoch": 1.9788293897882938,
+      "grad_norm": 1.2269333600997925,
+      "learning_rate": 8.605863733421594e-06,
+      "loss": 0.1367,
+      "step": 795
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "grad_norm": 1.1303420066833496,
+      "learning_rate": 8.420322994955074e-06,
+      "loss": 0.1433,
+      "step": 800
+    },
+    {
+      "epoch": 2.0024906600249066,
+      "grad_norm": 0.9883002042770386,
+      "learning_rate": 8.236020558402222e-06,
+      "loss": 0.1283,
+      "step": 805
+    },
+    {
+      "epoch": 2.0149439601494397,
+      "grad_norm": 1.2297390699386597,
+      "learning_rate": 8.052991109669306e-06,
+      "loss": 0.1104,
+      "step": 810
+    },
+    {
+      "epoch": 2.0273972602739727,
+      "grad_norm": 1.0766546726226807,
+      "learning_rate": 7.87126909508499e-06,
+      "loss": 0.103,
+      "step": 815
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.9960305094718933,
+      "learning_rate": 7.690888714917507e-06,
+      "loss": 0.1164,
+      "step": 820
+    },
+    {
+      "epoch": 2.052303860523039,
+      "grad_norm": 1.1033507585525513,
+      "learning_rate": 7.511883916938109e-06,
+      "loss": 0.1049,
+      "step": 825
+    },
+    {
+      "epoch": 2.0647571606475714,
+      "grad_norm": 1.1227325201034546,
+      "learning_rate": 7.334288390032098e-06,
+      "loss": 0.1026,
+      "step": 830
+    },
+    {
+      "epoch": 2.0772104607721045,
+      "grad_norm": 1.0751335620880127,
+      "learning_rate": 7.158135557858515e-06,
+      "loss": 0.0996,
+      "step": 835
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.909429132938385,
+      "learning_rate": 6.983458572559782e-06,
+      "loss": 0.0982,
+      "step": 840
+    },
+    {
+      "epoch": 2.1021170610211706,
+      "grad_norm": 0.8205214738845825,
+      "learning_rate": 6.81029030852244e-06,
+      "loss": 0.1015,
+      "step": 845
+    },
+    {
+      "epoch": 2.1145703611457036,
+      "grad_norm": 0.9863373041152954,
+      "learning_rate": 6.63866335619015e-06,
+      "loss": 0.1038,
+      "step": 850
+    },
+    {
+      "epoch": 2.1270236612702367,
+      "grad_norm": 1.0564978122711182,
+      "learning_rate": 6.468610015930143e-06,
+      "loss": 0.0926,
+      "step": 855
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.8808130621910095,
+      "learning_rate": 6.3001622919542495e-06,
+      "loss": 0.0897,
+      "step": 860
+    },
+    {
+      "epoch": 2.151930261519303,
+      "grad_norm": 1.1858986616134644,
+      "learning_rate": 6.133351886295691e-06,
+      "loss": 0.0819,
+      "step": 865
+    },
+    {
+      "epoch": 2.1643835616438354,
+      "grad_norm": 1.0272358655929565,
+      "learning_rate": 5.9682101928426966e-06,
+      "loss": 0.109,
+      "step": 870
+    },
+    {
+      "epoch": 2.1768368617683684,
+      "grad_norm": 1.053252100944519,
+      "learning_rate": 5.804768291430174e-06,
+      "loss": 0.1022,
+      "step": 875
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8846246600151062,
+      "learning_rate": 5.643056941990433e-06,
+      "loss": 0.0954,
+      "step": 880
+    },
+    {
+      "epoch": 2.2017434620174345,
+      "grad_norm": 1.0628403425216675,
+      "learning_rate": 5.483106578764136e-06,
+      "loss": 0.0836,
+      "step": 885
+    },
+    {
+      "epoch": 2.2141967621419676,
+      "grad_norm": 1.0350326299667358,
+      "learning_rate": 5.324947304572553e-06,
+      "loss": 0.0873,
+      "step": 890
+    },
+    {
+      "epoch": 2.2266500622665006,
+      "grad_norm": 1.087530255317688,
+      "learning_rate": 5.1686088851521685e-06,
+      "loss": 0.0914,
+      "step": 895
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "grad_norm": 1.190813660621643,
+      "learning_rate": 5.014120743552749e-06,
+      "loss": 0.098,
+      "step": 900
+    },
+    {
+      "epoch": 2.2515566625155667,
+      "grad_norm": 0.953442394733429,
+      "learning_rate": 4.861511954599883e-06,
+      "loss": 0.0861,
+      "step": 905
+    },
+    {
+      "epoch": 2.2640099626401,
+      "grad_norm": 1.108999490737915,
+      "learning_rate": 4.710811239423083e-06,
+      "loss": 0.0852,
+      "step": 910
+    },
+    {
+      "epoch": 2.276463262764633,
+      "grad_norm": 0.9217025637626648,
+      "learning_rate": 4.5620469600504355e-06,
+      "loss": 0.0823,
+      "step": 915
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.9060125946998596,
+      "learning_rate": 4.415247114070834e-06,
+      "loss": 0.0901,
+      "step": 920
+    },
+    {
+      "epoch": 2.3013698630136985,
+      "grad_norm": 0.8266595602035522,
+      "learning_rate": 4.270439329364799e-06,
+      "loss": 0.083,
+      "step": 925
+    },
+    {
+      "epoch": 2.3138231631382316,
+      "grad_norm": 0.975999116897583,
+      "learning_rate": 4.1276508589048986e-06,
+      "loss": 0.0988,
+      "step": 930
+    },
+    {
+      "epoch": 2.3262764632627646,
+      "grad_norm": 0.9437015652656555,
+      "learning_rate": 3.986908575626699e-06,
+      "loss": 0.088,
+      "step": 935
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.8634318709373474,
+      "learning_rate": 3.848238967371265e-06,
+      "loss": 0.0819,
+      "step": 940
+    },
+    {
+      "epoch": 2.3511830635118307,
+      "grad_norm": 1.3089238405227661,
+      "learning_rate": 3.7116681319001018e-06,
+      "loss": 0.0926,
+      "step": 945
+    },
+    {
+      "epoch": 2.3636363636363638,
+      "grad_norm": 0.9045596718788147,
+      "learning_rate": 3.5772217719835384e-06,
+      "loss": 0.0799,
+      "step": 950
+    },
+    {
+      "epoch": 2.376089663760897,
+      "grad_norm": 0.8352115154266357,
+      "learning_rate": 3.444925190563445e-06,
+      "loss": 0.0822,
+      "step": 955
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7647663950920105,
+      "learning_rate": 3.3148032859911844e-06,
+      "loss": 0.0726,
+      "step": 960
+    },
+    {
+      "epoch": 2.4009962640099625,
+      "grad_norm": 0.7704339623451233,
+      "learning_rate": 3.186880547341727e-06,
+      "loss": 0.0773,
+      "step": 965
+    },
+    {
+      "epoch": 2.4134495641344955,
+      "grad_norm": 0.7999119758605957,
+      "learning_rate": 3.0611810498047742e-06,
+      "loss": 0.0742,
+      "step": 970
+    },
+    {
+      "epoch": 2.4259028642590286,
+      "grad_norm": 0.9019289016723633,
+      "learning_rate": 2.937728450153789e-06,
+      "loss": 0.0875,
+      "step": 975
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8594040870666504,
+      "learning_rate": 2.816545982293752e-06,
+      "loss": 0.0789,
+      "step": 980
+    },
+    {
+      "epoch": 2.4508094645080947,
+      "grad_norm": 0.8093627095222473,
+      "learning_rate": 2.6976564528885422e-06,
+      "loss": 0.0795,
+      "step": 985
+    },
+    {
+      "epoch": 2.4632627646326277,
+      "grad_norm": 0.7957439422607422,
+      "learning_rate": 2.5810822370686804e-06,
+      "loss": 0.0655,
+      "step": 990
+    },
+    {
+      "epoch": 2.4757160647571608,
+      "grad_norm": 0.8996023535728455,
+      "learning_rate": 2.466845274220316e-06,
+      "loss": 0.0698,
+      "step": 995
+    },
+    {
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.8167709112167358,
+      "learning_rate": 2.3549670638562016e-06,
+      "loss": 0.0746,
+      "step": 1000
+    },
+    {
+      "epoch": 2.500622665006227,
+      "grad_norm": 0.7774357795715332,
+      "learning_rate": 2.2454686615694785e-06,
+      "loss": 0.0865,
+      "step": 1005
+    },
+    {
+      "epoch": 2.51307596513076,
+      "grad_norm": 0.8335087299346924,
+      "learning_rate": 2.138370675070977e-06,
+      "loss": 0.087,
+      "step": 1010
+    },
+    {
+      "epoch": 2.5255292652552925,
+      "grad_norm": 0.9378474354743958,
+      "learning_rate": 2.0336932603108355e-06,
+      "loss": 0.0748,
+      "step": 1015
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7766979932785034,
+      "learning_rate": 1.9314561176851235e-06,
+      "loss": 0.0859,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5504358655043586,
+      "grad_norm": 0.773000180721283,
+      "learning_rate": 1.8316784883282105e-06,
+      "loss": 0.0625,
+      "step": 1025
+    },
+    {
+      "epoch": 2.5628891656288917,
+      "grad_norm": 0.7652465105056763,
+      "learning_rate": 1.7343791504915684e-06,
+      "loss": 0.0795,
+      "step": 1030
+    },
+    {
+      "epoch": 2.5753424657534247,
+      "grad_norm": 0.717429518699646,
+      "learning_rate": 1.6395764160096678e-06,
+      "loss": 0.08,
+      "step": 1035
+    },
+    {
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.9643293619155884,
+      "learning_rate": 1.547288126853697e-06,
+      "loss": 0.0796,
+      "step": 1040
+    },
+    {
+      "epoch": 2.6002490660024904,
+      "grad_norm": 0.8420007824897766,
+      "learning_rate": 1.4575316517736714e-06,
+      "loss": 0.0971,
+      "step": 1045
+    },
+    {
+      "epoch": 2.6127023661270234,
+      "grad_norm": 0.9652625918388367,
+      "learning_rate": 1.370323883029615e-06,
+      "loss": 0.0961,
+      "step": 1050
+    },
+    {
+      "epoch": 2.6251556662515565,
+      "grad_norm": 0.7456346750259399,
+      "learning_rate": 1.2856812332124274e-06,
+      "loss": 0.0697,
+      "step": 1055
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.8239288926124573,
+      "learning_rate": 1.2036196321550096e-06,
+      "loss": 0.0782,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6500622665006226,
+      "grad_norm": 0.8267461657524109,
+      "learning_rate": 1.1241545239342609e-06,
+      "loss": 0.0755,
+      "step": 1065
+    },
+    {
+      "epoch": 2.6625155666251556,
+      "grad_norm": 0.8509014844894409,
+      "learning_rate": 1.0473008639644814e-06,
+      "loss": 0.079,
+      "step": 1070
+    },
+    {
+      "epoch": 2.6749688667496887,
+      "grad_norm": 0.8588144779205322,
+      "learning_rate": 9.730731161827528e-07,
+      "loss": 0.0808,
+      "step": 1075
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6590099930763245,
+      "learning_rate": 9.014852503268045e-07,
+      "loss": 0.0631,
+      "step": 1080
+    },
+    {
+      "epoch": 2.699875466998755,
+      "grad_norm": 0.607670247554779,
+      "learning_rate": 8.325507393059101e-07,
+      "loss": 0.0741,
+      "step": 1085
+    },
+    {
+      "epoch": 2.712328767123288,
+      "grad_norm": 0.6613425612449646,
+      "learning_rate": 7.662825566652442e-07,
+      "loss": 0.0702,
+      "step": 1090
+    },
+    {
+      "epoch": 2.724782067247821,
+      "grad_norm": 0.7123258113861084,
+      "learning_rate": 7.026931741442783e-07,
+      "loss": 0.079,
+      "step": 1095
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6819207668304443,
+      "learning_rate": 6.417945593295638e-07,
+      "loss": 0.059,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7496886674968866,
+      "grad_norm": 0.7601874470710754,
+      "learning_rate": 5.835981734024348e-07,
+      "loss": 0.0694,
+      "step": 1105
+    },
+    {
+      "epoch": 2.7621419676214196,
+      "grad_norm": 0.8470236659049988,
+      "learning_rate": 5.281149689819981e-07,
+      "loss": 0.079,
+      "step": 1110
+    },
+    {
+      "epoch": 2.7745952677459527,
+      "grad_norm": 0.6963379979133606,
+      "learning_rate": 4.7535538806383006e-07,
+      "loss": 0.0689,
+      "step": 1115
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9559626579284668,
+      "learning_rate": 4.2532936005479585e-07,
+      "loss": 0.0711,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7995018679950188,
+      "grad_norm": 0.5686436295509338,
+      "learning_rate": 3.7804629990431884e-07,
+      "loss": 0.0622,
+      "step": 1125
+    },
+    {
+      "epoch": 2.811955168119552,
+      "grad_norm": 0.6661945581436157,
+      "learning_rate": 3.335151063324765e-07,
+      "loss": 0.0756,
+      "step": 1130
+    },
+    {
+      "epoch": 2.8244084682440844,
+      "grad_norm": 0.7337989807128906,
+      "learning_rate": 2.917441601552534e-07,
+      "loss": 0.0618,
+      "step": 1135
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.5880770087242126,
+      "learning_rate": 2.527413227072628e-07,
+      "loss": 0.0626,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8493150684931505,
+      "grad_norm": 0.6966707706451416,
+      "learning_rate": 2.165139343622352e-07,
+      "loss": 0.0706,
+      "step": 1145
+    },
+    {
+      "epoch": 2.8617683686176836,
+      "grad_norm": 0.6662957072257996,
+      "learning_rate": 1.830688131515551e-07,
+      "loss": 0.0689,
+      "step": 1150
+    },
+    {
+      "epoch": 2.8742216687422166,
+      "grad_norm": 0.7926638126373291,
+      "learning_rate": 1.5241225348109898e-07,
+      "loss": 0.0782,
+      "step": 1155
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.5801357626914978,
+      "learning_rate": 1.2455002494661972e-07,
+      "loss": 0.0761,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8991282689912827,
+      "grad_norm": 0.681266188621521,
+      "learning_rate": 9.948737124790331e-08,
+      "loss": 0.0608,
+      "step": 1165
+    },
+    {
+      "epoch": 2.911581569115816,
+      "grad_norm": 0.7630389928817749,
+      "learning_rate": 7.722900920190179e-08,
+      "loss": 0.07,
+      "step": 1170
+    },
+    {
+      "epoch": 2.924034869240349,
+      "grad_norm": 0.7673495411872864,
+      "learning_rate": 5.777912785502493e-08,
+      "loss": 0.0754,
+      "step": 1175
+    },
+    {
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7313214540481567,
+      "learning_rate": 4.114138769474918e-08,
+      "loss": 0.079,
+      "step": 1180
+    },
+    {
+      "epoch": 2.948941469489415,
+      "grad_norm": 0.7274189591407776,
+      "learning_rate": 2.731891996071878e-08,
+      "loss": 0.0717,
+      "step": 1185
+    },
+    {
+      "epoch": 2.9613947696139475,
+      "grad_norm": 0.8052839636802673,
+      "learning_rate": 1.6314326055440475e-08,
+      "loss": 0.0775,
+      "step": 1190
+    },
+    {
+      "epoch": 2.9738480697384806,
+      "grad_norm": 0.6404252648353577,
+      "learning_rate": 8.129677054693474e-09,
+      "loss": 0.0585,
+      "step": 1195
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6112473607063293,
+      "learning_rate": 2.7665133177545708e-09,
+      "loss": 0.0585,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9987546699875467,
+      "grad_norm": 0.6919467449188232,
+      "learning_rate": 2.2584419750504293e-10,
+      "loss": 0.0758,
+      "step": 1205
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1206,
+  "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": 2.118099997403644e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

base/15_128_e3_3e-5/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a620bf404acd91a84b0a51b85c980db32da4383e8ff43ce781767ff017227b9f
+size 8273

base/15_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)