Upload folder using huggingface_hub

base/28_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/28_128_e3_3e-5/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "k_proj",
+    "down_proj",
+    "v_proj",
+    "up_proj",
+    "o_proj",
+    "q_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

base/28_128_e3_3e-5/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d612f2e85a95cabb7dd99593dcf894a42834100441275483789423a054a4c978
+size 791751704

base/28_128_e3_3e-5/latest

@@ -0,0 +1 @@
+global_step1128

base/28_128_e3_3e-5/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e57b05c4af2ae71035ea9cca5efff491a7f289c7f4541f11a294455b09466190
+size 16389

base/28_128_e3_3e-5/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db92ba56e65ea750c4a3954ab67c72612447b20dd45c097b1398b719d57e6d8f
+size 16389

base/28_128_e3_3e-5/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff81d0be8664af9fd8e086f7e765ce067c57fc03702b8746df95bf4a1bc5931b
+size 16389

base/28_128_e3_3e-5/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8f4b8d3733241b586f477121980bd1538410112213631a9a686704f8be2cef0
+size 16389

base/28_128_e3_3e-5/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96a4b4b5de4b50e24c72daf26c762e53403b5f6f9c0d1a57cec31a4d5f37ada4
+size 16389

base/28_128_e3_3e-5/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e64206c8b68ad5706e080a379326e88bd43e31ea4d54ca1da86b141fef10362
+size 16389

base/28_128_e3_3e-5/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a005e0e23cd9711ae4ef6ea73c7e1c0d2ac0e7198818651bd23fd64f34b1d2c
+size 16389

base/28_128_e3_3e-5/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeb760c443a4d467025c6e959527b69188bb6061a01f86268b5becb6ec48163f
+size 16389

base/28_128_e3_3e-5/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e81dc1a422171dc9bb907d3080017086e98a3ec3a4de5a809f471bf4754c306
+size 1401

base/28_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/28_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/28_128_e3_3e-5/trainer_state.json

@@ -0,0 +1,1609 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1128,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.013297872340425532,
+      "grad_norm": 1.0066176652908325,
+      "learning_rate": 2.1052631578947366e-06,
+      "loss": 1.2175,
+      "step": 5
+    },
+    {
+      "epoch": 0.026595744680851064,
+      "grad_norm": 0.6157808899879456,
+      "learning_rate": 4.736842105263158e-06,
+      "loss": 1.2256,
+      "step": 10
+    },
+    {
+      "epoch": 0.0398936170212766,
+      "grad_norm": 0.4673263430595398,
+      "learning_rate": 7.3684210526315784e-06,
+      "loss": 1.166,
+      "step": 15
+    },
+    {
+      "epoch": 0.05319148936170213,
+      "grad_norm": 0.4307611584663391,
+      "learning_rate": 9.999999999999999e-06,
+      "loss": 1.203,
+      "step": 20
+    },
+    {
+      "epoch": 0.06648936170212766,
+      "grad_norm": 0.4037454128265381,
+      "learning_rate": 1.263157894736842e-05,
+      "loss": 1.1201,
+      "step": 25
+    },
+    {
+      "epoch": 0.0797872340425532,
+      "grad_norm": 0.4147949516773224,
+      "learning_rate": 1.5263157894736842e-05,
+      "loss": 1.1355,
+      "step": 30
+    },
+    {
+      "epoch": 0.09308510638297872,
+      "grad_norm": 0.3965796232223511,
+      "learning_rate": 1.7894736842105264e-05,
+      "loss": 1.1338,
+      "step": 35
+    },
+    {
+      "epoch": 0.10638297872340426,
+      "grad_norm": 0.3792321979999542,
+      "learning_rate": 2.0526315789473685e-05,
+      "loss": 1.1077,
+      "step": 40
+    },
+    {
+      "epoch": 0.1196808510638298,
+      "grad_norm": 0.39236682653427124,
+      "learning_rate": 2.3157894736842103e-05,
+      "loss": 1.0725,
+      "step": 45
+    },
+    {
+      "epoch": 0.13297872340425532,
+      "grad_norm": 0.4293903112411499,
+      "learning_rate": 2.578947368421053e-05,
+      "loss": 1.0604,
+      "step": 50
+    },
+    {
+      "epoch": 0.14627659574468085,
+      "grad_norm": 0.3851737082004547,
+      "learning_rate": 2.8421052631578946e-05,
+      "loss": 1.0093,
+      "step": 55
+    },
+    {
+      "epoch": 0.1595744680851064,
+      "grad_norm": 0.43541136384010315,
+      "learning_rate": 2.9999741868614275e-05,
+      "loss": 1.0216,
+      "step": 60
+    },
+    {
+      "epoch": 0.17287234042553193,
+      "grad_norm": 0.5105594992637634,
+      "learning_rate": 2.999683799255387e-05,
+      "loss": 0.9983,
+      "step": 65
+    },
+    {
+      "epoch": 0.18617021276595744,
+      "grad_norm": 0.46317487955093384,
+      "learning_rate": 2.9990708202925038e-05,
+      "loss": 1.0195,
+      "step": 70
+    },
+    {
+      "epoch": 0.19946808510638298,
+      "grad_norm": 0.5645899176597595,
+      "learning_rate": 2.9981353818283835e-05,
+      "loss": 0.9699,
+      "step": 75
+    },
+    {
+      "epoch": 0.2127659574468085,
+      "grad_norm": 0.5505724549293518,
+      "learning_rate": 2.996877685081685e-05,
+      "loss": 0.996,
+      "step": 80
+    },
+    {
+      "epoch": 0.22606382978723405,
+      "grad_norm": 0.5268149971961975,
+      "learning_rate": 2.995298000590839e-05,
+      "loss": 0.9291,
+      "step": 85
+    },
+    {
+      "epoch": 0.2393617021276596,
+      "grad_norm": 0.5861604809761047,
+      "learning_rate": 2.99339666815585e-05,
+      "loss": 0.9053,
+      "step": 90
+    },
+    {
+      "epoch": 0.2526595744680851,
+      "grad_norm": 0.5834186673164368,
+      "learning_rate": 2.9911740967652065e-05,
+      "loss": 0.9061,
+      "step": 95
+    },
+    {
+      "epoch": 0.26595744680851063,
+      "grad_norm": 0.6459625363349915,
+      "learning_rate": 2.9886307645079037e-05,
+      "loss": 0.8937,
+      "step": 100
+    },
+    {
+      "epoch": 0.27925531914893614,
+      "grad_norm": 0.6325609683990479,
+      "learning_rate": 2.9857672184706038e-05,
+      "loss": 0.877,
+      "step": 105
+    },
+    {
+      "epoch": 0.2925531914893617,
+      "grad_norm": 0.8154891133308411,
+      "learning_rate": 2.9825840746199534e-05,
+      "loss": 0.8367,
+      "step": 110
+    },
+    {
+      "epoch": 0.3058510638297872,
+      "grad_norm": 0.6585891246795654,
+      "learning_rate": 2.9790820176700872e-05,
+      "loss": 0.8179,
+      "step": 115
+    },
+    {
+      "epoch": 0.3191489361702128,
+      "grad_norm": 0.7468953132629395,
+      "learning_rate": 2.975261800935339e-05,
+      "loss": 0.79,
+      "step": 120
+    },
+    {
+      "epoch": 0.3324468085106383,
+      "grad_norm": 0.7013482451438904,
+      "learning_rate": 2.971124246168202e-05,
+      "loss": 0.8244,
+      "step": 125
+    },
+    {
+      "epoch": 0.34574468085106386,
+      "grad_norm": 0.6997635364532471,
+      "learning_rate": 2.9666702433825614e-05,
+      "loss": 0.7745,
+      "step": 130
+    },
+    {
+      "epoch": 0.35904255319148937,
+      "grad_norm": 0.8057093620300293,
+      "learning_rate": 2.9619007506622506e-05,
+      "loss": 0.699,
+      "step": 135
+    },
+    {
+      "epoch": 0.3723404255319149,
+      "grad_norm": 0.7938418388366699,
+      "learning_rate": 2.956816793954958e-05,
+      "loss": 0.7311,
+      "step": 140
+    },
+    {
+      "epoch": 0.38563829787234044,
+      "grad_norm": 0.9949880242347717,
+      "learning_rate": 2.951419466851542e-05,
+      "loss": 0.7682,
+      "step": 145
+    },
+    {
+      "epoch": 0.39893617021276595,
+      "grad_norm": 0.7443533539772034,
+      "learning_rate": 2.9457099303507904e-05,
+      "loss": 0.7215,
+      "step": 150
+    },
+    {
+      "epoch": 0.4122340425531915,
+      "grad_norm": 0.8264400362968445,
+      "learning_rate": 2.939689412609684e-05,
+      "loss": 0.6583,
+      "step": 155
+    },
+    {
+      "epoch": 0.425531914893617,
+      "grad_norm": 0.9382266402244568,
+      "learning_rate": 2.9333592086792113e-05,
+      "loss": 0.6426,
+      "step": 160
+    },
+    {
+      "epoch": 0.43882978723404253,
+      "grad_norm": 0.9062151908874512,
+      "learning_rate": 2.9267206802257952e-05,
+      "loss": 0.7257,
+      "step": 165
+    },
+    {
+      "epoch": 0.4521276595744681,
+      "grad_norm": 0.8805403113365173,
+      "learning_rate": 2.919775255238392e-05,
+      "loss": 0.6257,
+      "step": 170
+    },
+    {
+      "epoch": 0.4654255319148936,
+      "grad_norm": 0.8294376730918884,
+      "learning_rate": 2.9125244277213176e-05,
+      "loss": 0.692,
+      "step": 175
+    },
+    {
+      "epoch": 0.4787234042553192,
+      "grad_norm": 0.9274476170539856,
+      "learning_rate": 2.9049697573728818e-05,
+      "loss": 0.6255,
+      "step": 180
+    },
+    {
+      "epoch": 0.4920212765957447,
+      "grad_norm": 0.9079066514968872,
+      "learning_rate": 2.8971128692498872e-05,
+      "loss": 0.6185,
+      "step": 185
+    },
+    {
+      "epoch": 0.5053191489361702,
+      "grad_norm": 0.9162901639938354,
+      "learning_rate": 2.8889554534180664e-05,
+      "loss": 0.603,
+      "step": 190
+    },
+    {
+      "epoch": 0.5186170212765957,
+      "grad_norm": 0.8933972120285034,
+      "learning_rate": 2.8804992645885415e-05,
+      "loss": 0.5981,
+      "step": 195
+    },
+    {
+      "epoch": 0.5319148936170213,
+      "grad_norm": 1.043695330619812,
+      "learning_rate": 2.8717461217403726e-05,
+      "loss": 0.5646,
+      "step": 200
+    },
+    {
+      "epoch": 0.5452127659574468,
+      "grad_norm": 2.7653353214263916,
+      "learning_rate": 2.8626979077292856e-05,
+      "loss": 0.5467,
+      "step": 205
+    },
+    {
+      "epoch": 0.5585106382978723,
+      "grad_norm": 0.9881752133369446,
+      "learning_rate": 2.853356568882657e-05,
+      "loss": 0.5299,
+      "step": 210
+    },
+    {
+      "epoch": 0.5718085106382979,
+      "grad_norm": 1.1020253896713257,
+      "learning_rate": 2.843724114580848e-05,
+      "loss": 0.5563,
+      "step": 215
+    },
+    {
+      "epoch": 0.5851063829787234,
+      "grad_norm": 1.0117254257202148,
+      "learning_rate": 2.833802616824972e-05,
+      "loss": 0.6086,
+      "step": 220
+    },
+    {
+      "epoch": 0.598404255319149,
+      "grad_norm": 0.9251948595046997,
+      "learning_rate": 2.8235942097911964e-05,
+      "loss": 0.5559,
+      "step": 225
+    },
+    {
+      "epoch": 0.6117021276595744,
+      "grad_norm": 0.9948434233665466,
+      "learning_rate": 2.8131010893716676e-05,
+      "loss": 0.5554,
+      "step": 230
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.9891300797462463,
+      "learning_rate": 2.8023255127021593e-05,
+      "loss": 0.5282,
+      "step": 235
+    },
+    {
+      "epoch": 0.6382978723404256,
+      "grad_norm": 1.0241698026657104,
+      "learning_rate": 2.7912697976765516e-05,
+      "loss": 0.515,
+      "step": 240
+    },
+    {
+      "epoch": 0.651595744680851,
+      "grad_norm": 1.0021275281906128,
+      "learning_rate": 2.7799363224482334e-05,
+      "loss": 0.5126,
+      "step": 245
+    },
+    {
+      "epoch": 0.6648936170212766,
+      "grad_norm": 1.0727177858352661,
+      "learning_rate": 2.7683275249185507e-05,
+      "loss": 0.4616,
+      "step": 250
+    },
+    {
+      "epoch": 0.6781914893617021,
+      "grad_norm": 1.161234736442566,
+      "learning_rate": 2.7564459022123953e-05,
+      "loss": 0.4936,
+      "step": 255
+    },
+    {
+      "epoch": 0.6914893617021277,
+      "grad_norm": 1.0692106485366821,
+      "learning_rate": 2.744294010141061e-05,
+      "loss": 0.4717,
+      "step": 260
+    },
+    {
+      "epoch": 0.7047872340425532,
+      "grad_norm": 0.9843445420265198,
+      "learning_rate": 2.7318744626524704e-05,
+      "loss": 0.4853,
+      "step": 265
+    },
+    {
+      "epoch": 0.7180851063829787,
+      "grad_norm": 1.016139030456543,
+      "learning_rate": 2.719189931268899e-05,
+      "loss": 0.4736,
+      "step": 270
+    },
+    {
+      "epoch": 0.7313829787234043,
+      "grad_norm": 1.1156508922576904,
+      "learning_rate": 2.7062431445123127e-05,
+      "loss": 0.4802,
+      "step": 275
+    },
+    {
+      "epoch": 0.7446808510638298,
+      "grad_norm": 1.0864167213439941,
+      "learning_rate": 2.6930368873174493e-05,
+      "loss": 0.4611,
+      "step": 280
+    },
+    {
+      "epoch": 0.7579787234042553,
+      "grad_norm": 1.4673291444778442,
+      "learning_rate": 2.6795740004327584e-05,
+      "loss": 0.5026,
+      "step": 285
+    },
+    {
+      "epoch": 0.7712765957446809,
+      "grad_norm": 1.7029342651367188,
+      "learning_rate": 2.665857379809338e-05,
+      "loss": 0.4855,
+      "step": 290
+    },
+    {
+      "epoch": 0.7845744680851063,
+      "grad_norm": 1.1178985834121704,
+      "learning_rate": 2.6518899759780017e-05,
+      "loss": 0.4344,
+      "step": 295
+    },
+    {
+      "epoch": 0.7978723404255319,
+      "grad_norm": 1.3871309757232666,
+      "learning_rate": 2.637674793414596e-05,
+      "loss": 0.4574,
+      "step": 300
+    },
+    {
+      "epoch": 0.8111702127659575,
+      "grad_norm": 1.1455399990081787,
+      "learning_rate": 2.6232148898937223e-05,
+      "loss": 0.4278,
+      "step": 305
+    },
+    {
+      "epoch": 0.824468085106383,
+      "grad_norm": 1.078267216682434,
+      "learning_rate": 2.6085133758309887e-05,
+      "loss": 0.408,
+      "step": 310
+    },
+    {
+      "epoch": 0.8377659574468085,
+      "grad_norm": 1.1813703775405884,
+      "learning_rate": 2.5935734136139407e-05,
+      "loss": 0.3981,
+      "step": 315
+    },
+    {
+      "epoch": 0.851063829787234,
+      "grad_norm": 0.9872785210609436,
+      "learning_rate": 2.5783982169218125e-05,
+      "loss": 0.4317,
+      "step": 320
+    },
+    {
+      "epoch": 0.8643617021276596,
+      "grad_norm": 1.2257829904556274,
+      "learning_rate": 2.5629910500342424e-05,
+      "loss": 0.3995,
+      "step": 325
+    },
+    {
+      "epoch": 0.8776595744680851,
+      "grad_norm": 1.4953336715698242,
+      "learning_rate": 2.5473552271291092e-05,
+      "loss": 0.3682,
+      "step": 330
+    },
+    {
+      "epoch": 0.8909574468085106,
+      "grad_norm": 1.2103192806243896,
+      "learning_rate": 2.531494111569629e-05,
+      "loss": 0.3611,
+      "step": 335
+    },
+    {
+      "epoch": 0.9042553191489362,
+      "grad_norm": 1.1730526685714722,
+      "learning_rate": 2.5154111151808752e-05,
+      "loss": 0.3919,
+      "step": 340
+    },
+    {
+      "epoch": 0.9175531914893617,
+      "grad_norm": 0.9862924814224243,
+      "learning_rate": 2.4991096975158757e-05,
+      "loss": 0.3653,
+      "step": 345
+    },
+    {
+      "epoch": 0.9308510638297872,
+      "grad_norm": 1.0812829732894897,
+      "learning_rate": 2.4825933651114375e-05,
+      "loss": 0.4061,
+      "step": 350
+    },
+    {
+      "epoch": 0.9441489361702128,
+      "grad_norm": 1.1807584762573242,
+      "learning_rate": 2.4658656707338733e-05,
+      "loss": 0.3868,
+      "step": 355
+    },
+    {
+      "epoch": 0.9574468085106383,
+      "grad_norm": 1.1982048749923706,
+      "learning_rate": 2.4489302126147768e-05,
+      "loss": 0.3699,
+      "step": 360
+    },
+    {
+      "epoch": 0.9707446808510638,
+      "grad_norm": 1.0041488409042358,
+      "learning_rate": 2.431790633677019e-05,
+      "loss": 0.3681,
+      "step": 365
+    },
+    {
+      "epoch": 0.9840425531914894,
+      "grad_norm": 1.0690995454788208,
+      "learning_rate": 2.414450620751136e-05,
+      "loss": 0.3565,
+      "step": 370
+    },
+    {
+      "epoch": 0.9973404255319149,
+      "grad_norm": 1.387973666191101,
+      "learning_rate": 2.396913903782268e-05,
+      "loss": 0.3801,
+      "step": 375
+    },
+    {
+      "epoch": 1.0106382978723405,
+      "grad_norm": 1.030849814414978,
+      "learning_rate": 2.379184255027822e-05,
+      "loss": 0.2982,
+      "step": 380
+    },
+    {
+      "epoch": 1.023936170212766,
+      "grad_norm": 1.1483806371688843,
+      "learning_rate": 2.361265488246039e-05,
+      "loss": 0.295,
+      "step": 385
+    },
+    {
+      "epoch": 1.0372340425531914,
+      "grad_norm": 1.1725521087646484,
+      "learning_rate": 2.3431614578756304e-05,
+      "loss": 0.3122,
+      "step": 390
+    },
+    {
+      "epoch": 1.050531914893617,
+      "grad_norm": 1.0999748706817627,
+      "learning_rate": 2.3248760582066605e-05,
+      "loss": 0.2818,
+      "step": 395
+    },
+    {
+      "epoch": 1.0638297872340425,
+      "grad_norm": 1.0982476472854614,
+      "learning_rate": 2.306413222542866e-05,
+      "loss": 0.3156,
+      "step": 400
+    },
+    {
+      "epoch": 1.077127659574468,
+      "grad_norm": 1.0049852132797241,
+      "learning_rate": 2.287776922355573e-05,
+      "loss": 0.2807,
+      "step": 405
+    },
+    {
+      "epoch": 1.0904255319148937,
+      "grad_norm": 1.1379531621932983,
+      "learning_rate": 2.268971166429412e-05,
+      "loss": 0.2751,
+      "step": 410
+    },
+    {
+      "epoch": 1.1037234042553192,
+      "grad_norm": 1.2535802125930786,
+      "learning_rate": 2.25e-05,
+      "loss": 0.2547,
+      "step": 415
+    },
+    {
+      "epoch": 1.1170212765957448,
+      "grad_norm": 1.0478107929229736,
+      "learning_rate": 2.2308675038837887e-05,
+      "loss": 0.2865,
+      "step": 420
+    },
+    {
+      "epoch": 1.1303191489361701,
+      "grad_norm": 1.2182199954986572,
+      "learning_rate": 2.2115777936002533e-05,
+      "loss": 0.2743,
+      "step": 425
+    },
+    {
+      "epoch": 1.1436170212765957,
+      "grad_norm": 1.1040678024291992,
+      "learning_rate": 2.192135018486618e-05,
+      "loss": 0.2575,
+      "step": 430
+    },
+    {
+      "epoch": 1.1569148936170213,
+      "grad_norm": 1.092760682106018,
+      "learning_rate": 2.172543360805308e-05,
+      "loss": 0.2954,
+      "step": 435
+    },
+    {
+      "epoch": 1.1702127659574468,
+      "grad_norm": 1.1354426145553589,
+      "learning_rate": 2.152807034844322e-05,
+      "loss": 0.2607,
+      "step": 440
+    },
+    {
+      "epoch": 1.1835106382978724,
+      "grad_norm": 1.2453972101211548,
+      "learning_rate": 2.1329302860107065e-05,
+      "loss": 0.3025,
+      "step": 445
+    },
+    {
+      "epoch": 1.196808510638298,
+      "grad_norm": 1.1013096570968628,
+      "learning_rate": 2.1129173899173474e-05,
+      "loss": 0.2291,
+      "step": 450
+    },
+    {
+      "epoch": 1.2101063829787235,
+      "grad_norm": 1.1325386762619019,
+      "learning_rate": 2.0927726514632557e-05,
+      "loss": 0.2339,
+      "step": 455
+    },
+    {
+      "epoch": 1.2234042553191489,
+      "grad_norm": 1.1311593055725098,
+      "learning_rate": 2.072500403907559e-05,
+      "loss": 0.2223,
+      "step": 460
+    },
+    {
+      "epoch": 1.2367021276595744,
+      "grad_norm": 1.1409766674041748,
+      "learning_rate": 2.0521050079373895e-05,
+      "loss": 0.1978,
+      "step": 465
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 1.1068118810653687,
+      "learning_rate": 2.0315908507298713e-05,
+      "loss": 0.2327,
+      "step": 470
+    },
+    {
+      "epoch": 1.2632978723404256,
+      "grad_norm": 1.0077366828918457,
+      "learning_rate": 2.0109623450084154e-05,
+      "loss": 0.2354,
+      "step": 475
+    },
+    {
+      "epoch": 1.2765957446808511,
+      "grad_norm": 1.024662971496582,
+      "learning_rate": 1.990223928093511e-05,
+      "loss": 0.2384,
+      "step": 480
+    },
+    {
+      "epoch": 1.2898936170212765,
+      "grad_norm": 1.272299885749817,
+      "learning_rate": 1.9693800609482318e-05,
+      "loss": 0.2209,
+      "step": 485
+    },
+    {
+      "epoch": 1.3031914893617023,
+      "grad_norm": 1.2153682708740234,
+      "learning_rate": 1.9484352272186555e-05,
+      "loss": 0.2509,
+      "step": 490
+    },
+    {
+      "epoch": 1.3164893617021276,
+      "grad_norm": 1.1465338468551636,
+      "learning_rate": 1.9273939322694035e-05,
+      "loss": 0.2604,
+      "step": 495
+    },
+    {
+      "epoch": 1.3297872340425532,
+      "grad_norm": 1.1240887641906738,
+      "learning_rate": 1.906260702214508e-05,
+      "loss": 0.22,
+      "step": 500
+    },
+    {
+      "epoch": 1.3430851063829787,
+      "grad_norm": 1.1756497621536255,
+      "learning_rate": 1.8850400829438157e-05,
+      "loss": 0.2157,
+      "step": 505
+    },
+    {
+      "epoch": 1.3563829787234043,
+      "grad_norm": 1.2772674560546875,
+      "learning_rate": 1.8637366391451414e-05,
+      "loss": 0.2374,
+      "step": 510
+    },
+    {
+      "epoch": 1.3696808510638299,
+      "grad_norm": 1.1081963777542114,
+      "learning_rate": 1.842354953322373e-05,
+      "loss": 0.228,
+      "step": 515
+    },
+    {
+      "epoch": 1.3829787234042552,
+      "grad_norm": 1.0625901222229004,
+      "learning_rate": 1.8208996248097462e-05,
+      "loss": 0.2005,
+      "step": 520
+    },
+    {
+      "epoch": 1.3962765957446808,
+      "grad_norm": 1.2685420513153076,
+      "learning_rate": 1.7993752687825003e-05,
+      "loss": 0.1841,
+      "step": 525
+    },
+    {
+      "epoch": 1.4095744680851063,
+      "grad_norm": 0.9382392764091492,
+      "learning_rate": 1.777786515264123e-05,
+      "loss": 0.2223,
+      "step": 530
+    },
+    {
+      "epoch": 1.422872340425532,
+      "grad_norm": 1.084978461265564,
+      "learning_rate": 1.7561380081304063e-05,
+      "loss": 0.221,
+      "step": 535
+    },
+    {
+      "epoch": 1.4361702127659575,
+      "grad_norm": 1.2145289182662964,
+      "learning_rate": 1.7344344041105177e-05,
+      "loss": 0.2145,
+      "step": 540
+    },
+    {
+      "epoch": 1.449468085106383,
+      "grad_norm": 1.1665339469909668,
+      "learning_rate": 1.7126803717853086e-05,
+      "loss": 0.2208,
+      "step": 545
+    },
+    {
+      "epoch": 1.4627659574468086,
+      "grad_norm": 1.1223244667053223,
+      "learning_rate": 1.6908805905830752e-05,
+      "loss": 0.1803,
+      "step": 550
+    },
+    {
+      "epoch": 1.476063829787234,
+      "grad_norm": 1.1497458219528198,
+      "learning_rate": 1.6690397497729818e-05,
+      "loss": 0.2273,
+      "step": 555
+    },
+    {
+      "epoch": 1.4893617021276595,
+      "grad_norm": 1.3534832000732422,
+      "learning_rate": 1.647162547456372e-05,
+      "loss": 0.1944,
+      "step": 560
+    },
+    {
+      "epoch": 1.502659574468085,
+      "grad_norm": 1.0653460025787354,
+      "learning_rate": 1.6252536895561754e-05,
+      "loss": 0.2122,
+      "step": 565
+    },
+    {
+      "epoch": 1.5159574468085106,
+      "grad_norm": 1.1311181783676147,
+      "learning_rate": 1.6033178888046368e-05,
+      "loss": 0.2128,
+      "step": 570
+    },
+    {
+      "epoch": 1.5292553191489362,
+      "grad_norm": 1.0753321647644043,
+      "learning_rate": 1.5813598637295767e-05,
+      "loss": 0.186,
+      "step": 575
+    },
+    {
+      "epoch": 1.5425531914893615,
+      "grad_norm": 1.0412178039550781,
+      "learning_rate": 1.5593843376394043e-05,
+      "loss": 0.162,
+      "step": 580
+    },
+    {
+      "epoch": 1.5558510638297873,
+      "grad_norm": 1.2652685642242432,
+      "learning_rate": 1.5373960376071095e-05,
+      "loss": 0.1758,
+      "step": 585
+    },
+    {
+      "epoch": 1.5691489361702127,
+      "grad_norm": 1.1265289783477783,
+      "learning_rate": 1.515399693453435e-05,
+      "loss": 0.1493,
+      "step": 590
+    },
+    {
+      "epoch": 1.5824468085106385,
+      "grad_norm": 1.1116056442260742,
+      "learning_rate": 1.493400036729465e-05,
+      "loss": 0.1793,
+      "step": 595
+    },
+    {
+      "epoch": 1.5957446808510638,
+      "grad_norm": 1.1357585191726685,
+      "learning_rate": 1.4714017996988384e-05,
+      "loss": 0.1839,
+      "step": 600
+    },
+    {
+      "epoch": 1.6090425531914894,
+      "grad_norm": 1.2143903970718384,
+      "learning_rate": 1.4494097143198083e-05,
+      "loss": 0.1693,
+      "step": 605
+    },
+    {
+      "epoch": 1.622340425531915,
+      "grad_norm": 0.9241980314254761,
+      "learning_rate": 1.4274285112273701e-05,
+      "loss": 0.1611,
+      "step": 610
+    },
+    {
+      "epoch": 1.6356382978723403,
+      "grad_norm": 1.0376677513122559,
+      "learning_rate": 1.4054629187156702e-05,
+      "loss": 0.1716,
+      "step": 615
+    },
+    {
+      "epoch": 1.648936170212766,
+      "grad_norm": 0.98783278465271,
+      "learning_rate": 1.3835176617209241e-05,
+      "loss": 0.1623,
+      "step": 620
+    },
+    {
+      "epoch": 1.6622340425531914,
+      "grad_norm": 0.9355586171150208,
+      "learning_rate": 1.3615974608050472e-05,
+      "loss": 0.1605,
+      "step": 625
+    },
+    {
+      "epoch": 1.675531914893617,
+      "grad_norm": 1.024682879447937,
+      "learning_rate": 1.3397070311402377e-05,
+      "loss": 0.146,
+      "step": 630
+    },
+    {
+      "epoch": 1.6888297872340425,
+      "grad_norm": 1.1570249795913696,
+      "learning_rate": 1.3178510814947112e-05,
+      "loss": 0.1619,
+      "step": 635
+    },
+    {
+      "epoch": 1.702127659574468,
+      "grad_norm": 1.0723835229873657,
+      "learning_rate": 1.296034313219816e-05,
+      "loss": 0.1438,
+      "step": 640
+    },
+    {
+      "epoch": 1.7154255319148937,
+      "grad_norm": 1.002578616142273,
+      "learning_rate": 1.2742614192387417e-05,
+      "loss": 0.1371,
+      "step": 645
+    },
+    {
+      "epoch": 1.728723404255319,
+      "grad_norm": 0.9965229630470276,
+      "learning_rate": 1.2525370830370447e-05,
+      "loss": 0.1498,
+      "step": 650
+    },
+    {
+      "epoch": 1.7420212765957448,
+      "grad_norm": 1.2053487300872803,
+      "learning_rate": 1.2308659776551985e-05,
+      "loss": 0.1367,
+      "step": 655
+    },
+    {
+      "epoch": 1.7553191489361701,
+      "grad_norm": 1.3584805727005005,
+      "learning_rate": 1.209252764683395e-05,
+      "loss": 0.1494,
+      "step": 660
+    },
+    {
+      "epoch": 1.7686170212765957,
+      "grad_norm": 0.9399833679199219,
+      "learning_rate": 1.1877020932588067e-05,
+      "loss": 0.1318,
+      "step": 665
+    },
+    {
+      "epoch": 1.7819148936170213,
+      "grad_norm": 1.0468209981918335,
+      "learning_rate": 1.1662185990655285e-05,
+      "loss": 0.1135,
+      "step": 670
+    },
+    {
+      "epoch": 1.7952127659574468,
+      "grad_norm": 1.2066107988357544,
+      "learning_rate": 1.1448069033374135e-05,
+      "loss": 0.1455,
+      "step": 675
+    },
+    {
+      "epoch": 1.8085106382978724,
+      "grad_norm": 1.0684256553649902,
+      "learning_rate": 1.1234716118640149e-05,
+      "loss": 0.1394,
+      "step": 680
+    },
+    {
+      "epoch": 1.8218085106382977,
+      "grad_norm": 1.3517318964004517,
+      "learning_rate": 1.1022173139998556e-05,
+      "loss": 0.1386,
+      "step": 685
+    },
+    {
+      "epoch": 1.8351063829787235,
+      "grad_norm": 1.0771692991256714,
+      "learning_rate": 1.0810485816772251e-05,
+      "loss": 0.1367,
+      "step": 690
+    },
+    {
+      "epoch": 1.8484042553191489,
+      "grad_norm": 1.1188035011291504,
+      "learning_rate": 1.0599699684227313e-05,
+      "loss": 0.1415,
+      "step": 695
+    },
+    {
+      "epoch": 1.8617021276595744,
+      "grad_norm": 1.2666881084442139,
+      "learning_rate": 1.0389860083778056e-05,
+      "loss": 0.143,
+      "step": 700
+    },
+    {
+      "epoch": 1.875,
+      "grad_norm": 0.9792928695678711,
+      "learning_rate": 1.0181012153233851e-05,
+      "loss": 0.1164,
+      "step": 705
+    },
+    {
+      "epoch": 1.8882978723404256,
+      "grad_norm": 1.0593355894088745,
+      "learning_rate": 9.973200817089655e-06,
+      "loss": 0.1233,
+      "step": 710
+    },
+    {
+      "epoch": 1.9015957446808511,
+      "grad_norm": 1.142223596572876,
+      "learning_rate": 9.7664707768625e-06,
+      "loss": 0.1272,
+      "step": 715
+    },
+    {
+      "epoch": 1.9148936170212765,
+      "grad_norm": 0.9116256237030029,
+      "learning_rate": 9.560866501475913e-06,
+      "loss": 0.1149,
+      "step": 720
+    },
+    {
+      "epoch": 1.9281914893617023,
+      "grad_norm": 1.1325784921646118,
+      "learning_rate": 9.35643221769436e-06,
+      "loss": 0.1147,
+      "step": 725
+    },
+    {
+      "epoch": 1.9414893617021276,
+      "grad_norm": 0.9534633755683899,
+      "learning_rate": 9.15321190060981e-06,
+      "loss": 0.1087,
+      "step": 730
+    },
+    {
+      "epoch": 1.9547872340425532,
+      "grad_norm": 1.120950698852539,
+      "learning_rate": 8.951249264182403e-06,
+      "loss": 0.1381,
+      "step": 735
+    },
+    {
+      "epoch": 1.9680851063829787,
+      "grad_norm": 0.9316391944885254,
+      "learning_rate": 8.750587751837313e-06,
+      "loss": 0.1332,
+      "step": 740
+    },
+    {
+      "epoch": 1.9813829787234043,
+      "grad_norm": 1.1845289468765259,
+      "learning_rate": 8.551270527119784e-06,
+      "loss": 0.1185,
+      "step": 745
+    },
+    {
+      "epoch": 1.9946808510638299,
+      "grad_norm": 0.8401294946670532,
+      "learning_rate": 8.35334046441041e-06,
+      "loss": 0.122,
+      "step": 750
+    },
+    {
+      "epoch": 2.007978723404255,
+      "grad_norm": 1.1020315885543823,
+      "learning_rate": 8.156840139702554e-06,
+      "loss": 0.0971,
+      "step": 755
+    },
+    {
+      "epoch": 2.021276595744681,
+      "grad_norm": 0.8825281858444214,
+      "learning_rate": 7.961811821444008e-06,
+      "loss": 0.0841,
+      "step": 760
+    },
+    {
+      "epoch": 2.0345744680851063,
+      "grad_norm": 1.0440452098846436,
+      "learning_rate": 7.768297461444766e-06,
+      "loss": 0.0983,
+      "step": 765
+    },
+    {
+      "epoch": 2.047872340425532,
+      "grad_norm": 0.9077974557876587,
+      "learning_rate": 7.576338685852955e-06,
+      "loss": 0.0974,
+      "step": 770
+    },
+    {
+      "epoch": 2.0611702127659575,
+      "grad_norm": 1.0085681676864624,
+      "learning_rate": 7.385976786200765e-06,
+      "loss": 0.0835,
+      "step": 775
+    },
+    {
+      "epoch": 2.074468085106383,
+      "grad_norm": 0.9577857255935669,
+      "learning_rate": 7.197252710522395e-06,
+      "loss": 0.0992,
+      "step": 780
+    },
+    {
+      "epoch": 2.0877659574468086,
+      "grad_norm": 0.8810827136039734,
+      "learning_rate": 7.010207054545873e-06,
+      "loss": 0.0817,
+      "step": 785
+    },
+    {
+      "epoch": 2.101063829787234,
+      "grad_norm": 0.8490446209907532,
+      "learning_rate": 6.8248800529606604e-06,
+      "loss": 0.0935,
+      "step": 790
+    },
+    {
+      "epoch": 2.1143617021276597,
+      "grad_norm": 0.9072813987731934,
+      "learning_rate": 6.641311570762918e-06,
+      "loss": 0.0829,
+      "step": 795
+    },
+    {
+      "epoch": 2.127659574468085,
+      "grad_norm": 1.050277590751648,
+      "learning_rate": 6.4595410946803e-06,
+      "loss": 0.0903,
+      "step": 800
+    },
+    {
+      "epoch": 2.1409574468085104,
+      "grad_norm": 0.9104912877082825,
+      "learning_rate": 6.2796077246781046e-06,
+      "loss": 0.0759,
+      "step": 805
+    },
+    {
+      "epoch": 2.154255319148936,
+      "grad_norm": 1.0978089570999146,
+      "learning_rate": 6.1015501655486365e-06,
+      "loss": 0.0882,
+      "step": 810
+    },
+    {
+      "epoch": 2.1675531914893615,
+      "grad_norm": 0.8886681795120239,
+      "learning_rate": 5.925406718585552e-06,
+      "loss": 0.0873,
+      "step": 815
+    },
+    {
+      "epoch": 2.1808510638297873,
+      "grad_norm": 0.8366052508354187,
+      "learning_rate": 5.751215273345036e-06,
+      "loss": 0.0796,
+      "step": 820
+    },
+    {
+      "epoch": 2.1941489361702127,
+      "grad_norm": 0.9924696087837219,
+      "learning_rate": 5.5790132994954935e-06,
+      "loss": 0.0989,
+      "step": 825
+    },
+    {
+      "epoch": 2.2074468085106385,
+      "grad_norm": 0.7336243987083435,
+      "learning_rate": 5.408837838757588e-06,
+      "loss": 0.0824,
+      "step": 830
+    },
+    {
+      "epoch": 2.220744680851064,
+      "grad_norm": 0.9045805335044861,
+      "learning_rate": 5.240725496936373e-06,
+      "loss": 0.0712,
+      "step": 835
+    },
+    {
+      "epoch": 2.2340425531914896,
+      "grad_norm": 0.8353767395019531,
+      "learning_rate": 5.0747124360471125e-06,
+      "loss": 0.0665,
+      "step": 840
+    },
+    {
+      "epoch": 2.247340425531915,
+      "grad_norm": 1.0857943296432495,
+      "learning_rate": 4.910834366536631e-06,
+      "loss": 0.0745,
+      "step": 845
+    },
+    {
+      "epoch": 2.2606382978723403,
+      "grad_norm": 0.8098640441894531,
+      "learning_rate": 4.74912653960177e-06,
+      "loss": 0.0641,
+      "step": 850
+    },
+    {
+      "epoch": 2.273936170212766,
+      "grad_norm": 0.8824878334999084,
+      "learning_rate": 4.589623739606625e-06,
+      "loss": 0.08,
+      "step": 855
+    },
+    {
+      "epoch": 2.2872340425531914,
+      "grad_norm": 0.8281514644622803,
+      "learning_rate": 4.4323602766002165e-06,
+      "loss": 0.0763,
+      "step": 860
+    },
+    {
+      "epoch": 2.300531914893617,
+      "grad_norm": 0.7379689812660217,
+      "learning_rate": 4.277369978936188e-06,
+      "loss": 0.0727,
+      "step": 865
+    },
+    {
+      "epoch": 2.3138297872340425,
+      "grad_norm": 1.0536364316940308,
+      "learning_rate": 4.1246861859961114e-06,
+      "loss": 0.0802,
+      "step": 870
+    },
+    {
+      "epoch": 2.327127659574468,
+      "grad_norm": 0.7634172439575195,
+      "learning_rate": 3.974341741017978e-06,
+      "loss": 0.0565,
+      "step": 875
+    },
+    {
+      "epoch": 2.3404255319148937,
+      "grad_norm": 0.649806559085846,
+      "learning_rate": 3.826368984031414e-06,
+      "loss": 0.077,
+      "step": 880
+    },
+    {
+      "epoch": 2.353723404255319,
+      "grad_norm": 0.7619276642799377,
+      "learning_rate": 3.6807997449011426e-06,
+      "loss": 0.0853,
+      "step": 885
+    },
+    {
+      "epoch": 2.367021276595745,
+      "grad_norm": 0.8028898239135742,
+      "learning_rate": 3.5376653364801703e-06,
+      "loss": 0.0652,
+      "step": 890
+    },
+    {
+      "epoch": 2.38031914893617,
+      "grad_norm": 0.7436749935150146,
+      "learning_rate": 3.3969965478742038e-06,
+      "loss": 0.0669,
+      "step": 895
+    },
+    {
+      "epoch": 2.393617021276596,
+      "grad_norm": 0.7840447425842285,
+      "learning_rate": 3.258823637818722e-06,
+      "loss": 0.0673,
+      "step": 900
+    },
+    {
+      "epoch": 2.4069148936170213,
+      "grad_norm": 0.8856799602508545,
+      "learning_rate": 3.123176328170131e-06,
+      "loss": 0.0841,
+      "step": 905
+    },
+    {
+      "epoch": 2.420212765957447,
+      "grad_norm": 0.692233681678772,
+      "learning_rate": 2.990083797512401e-06,
+      "loss": 0.0733,
+      "step": 910
+    },
+    {
+      "epoch": 2.4335106382978724,
+      "grad_norm": 0.9426119923591614,
+      "learning_rate": 2.8595746748805805e-06,
+      "loss": 0.0781,
+      "step": 915
+    },
+    {
+      "epoch": 2.4468085106382977,
+      "grad_norm": 0.768660843372345,
+      "learning_rate": 2.7316770336025166e-06,
+      "loss": 0.0664,
+      "step": 920
+    },
+    {
+      "epoch": 2.4601063829787235,
+      "grad_norm": 0.5873117446899414,
+      "learning_rate": 2.60641838526008e-06,
+      "loss": 0.0658,
+      "step": 925
+    },
+    {
+      "epoch": 2.473404255319149,
+      "grad_norm": 1.1409226655960083,
+      "learning_rate": 2.483825673771279e-06,
+      "loss": 0.0734,
+      "step": 930
+    },
+    {
+      "epoch": 2.4867021276595747,
+      "grad_norm": 0.8562684059143066,
+      "learning_rate": 2.363925269594449e-06,
+      "loss": 0.0659,
+      "step": 935
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 0.8828637599945068,
+      "learning_rate": 2.2467429640557903e-06,
+      "loss": 0.0782,
+      "step": 940
+    },
+    {
+      "epoch": 2.5132978723404253,
+      "grad_norm": 0.6943654417991638,
+      "learning_rate": 2.1323039638015024e-06,
+      "loss": 0.0713,
+      "step": 945
+    },
+    {
+      "epoch": 2.526595744680851,
+      "grad_norm": 0.802352786064148,
+      "learning_rate": 2.020632885375684e-06,
+      "loss": 0.0683,
+      "step": 950
+    },
+    {
+      "epoch": 2.5398936170212765,
+      "grad_norm": 0.7754091620445251,
+      "learning_rate": 1.9117537499251416e-06,
+      "loss": 0.0633,
+      "step": 955
+    },
+    {
+      "epoch": 2.5531914893617023,
+      "grad_norm": 0.7612255811691284,
+      "learning_rate": 1.8056899780323016e-06,
+      "loss": 0.0607,
+      "step": 960
+    },
+    {
+      "epoch": 2.5664893617021276,
+      "grad_norm": 0.7471124529838562,
+      "learning_rate": 1.7024643846772981e-06,
+      "loss": 0.0617,
+      "step": 965
+    },
+    {
+      "epoch": 2.579787234042553,
+      "grad_norm": 0.702548623085022,
+      "learning_rate": 1.6020991743303264e-06,
+      "loss": 0.0666,
+      "step": 970
+    },
+    {
+      "epoch": 2.5930851063829787,
+      "grad_norm": 0.5824935436248779,
+      "learning_rate": 1.5046159361753226e-06,
+      "loss": 0.0621,
+      "step": 975
+    },
+    {
+      "epoch": 2.6063829787234045,
+      "grad_norm": 0.569823145866394,
+      "learning_rate": 1.4100356394659863e-06,
+      "loss": 0.0559,
+      "step": 980
+    },
+    {
+      "epoch": 2.61968085106383,
+      "grad_norm": 0.7356989979743958,
+      "learning_rate": 1.318378629015184e-06,
+      "loss": 0.0605,
+      "step": 985
+    },
+    {
+      "epoch": 2.632978723404255,
+      "grad_norm": 0.585648775100708,
+      "learning_rate": 1.229664620818633e-06,
+      "loss": 0.0677,
+      "step": 990
+    },
+    {
+      "epoch": 2.646276595744681,
+      "grad_norm": 0.6193649172782898,
+      "learning_rate": 1.1439126978138769e-06,
+      "loss": 0.0642,
+      "step": 995
+    },
+    {
+      "epoch": 2.6595744680851063,
+      "grad_norm": 0.5839751958847046,
+      "learning_rate": 1.0611413057754221e-06,
+      "loss": 0.0581,
+      "step": 1000
+    },
+    {
+      "epoch": 2.672872340425532,
+      "grad_norm": 0.6062380075454712,
+      "learning_rate": 9.813682493469396e-07,
+      "loss": 0.0638,
+      "step": 1005
+    },
+    {
+      "epoch": 2.6861702127659575,
+      "grad_norm": 0.6278533339500427,
+      "learning_rate": 9.046106882113753e-07,
+      "loss": 0.0554,
+      "step": 1010
+    },
+    {
+      "epoch": 2.699468085106383,
+      "grad_norm": 0.8013328313827515,
+      "learning_rate": 8.308851333997918e-07,
+      "loss": 0.059,
+      "step": 1015
+    },
+    {
+      "epoch": 2.7127659574468086,
+      "grad_norm": 0.6262830495834351,
+      "learning_rate": 7.602074437397455e-07,
+      "loss": 0.0577,
+      "step": 1020
+    },
+    {
+      "epoch": 2.726063829787234,
+      "grad_norm": 0.55390864610672,
+      "learning_rate": 6.925928224439532e-07,
+      "loss": 0.0511,
+      "step": 1025
+    },
+    {
+      "epoch": 2.7393617021276597,
+      "grad_norm": 0.7058159708976746,
+      "learning_rate": 6.280558138399805e-07,
+      "loss": 0.0582,
+      "step": 1030
+    },
+    {
+      "epoch": 2.752659574468085,
+      "grad_norm": 0.659734845161438,
+      "learning_rate": 5.666103002416762e-07,
+      "loss": 0.0572,
+      "step": 1035
+    },
+    {
+      "epoch": 2.7659574468085104,
+      "grad_norm": 0.694625198841095,
+      "learning_rate": 5.082694989629916e-07,
+      "loss": 0.0492,
+      "step": 1040
+    },
+    {
+      "epoch": 2.779255319148936,
+      "grad_norm": 0.7966269254684448,
+      "learning_rate": 4.5304595947485927e-07,
+      "loss": 0.0603,
+      "step": 1045
+    },
+    {
+      "epoch": 2.7925531914893615,
+      "grad_norm": 0.650447428226471,
+      "learning_rate": 4.0095156070571513e-07,
+      "loss": 0.053,
+      "step": 1050
+    },
+    {
+      "epoch": 2.8058510638297873,
+      "grad_norm": 0.7361294031143188,
+      "learning_rate": 3.5199750848627753e-07,
+      "loss": 0.0605,
+      "step": 1055
+    },
+    {
+      "epoch": 2.8191489361702127,
+      "grad_norm": 0.6264920234680176,
+      "learning_rate": 3.0619433313909706e-07,
+      "loss": 0.0743,
+      "step": 1060
+    },
+    {
+      "epoch": 2.8324468085106385,
+      "grad_norm": 0.690265953540802,
+      "learning_rate": 2.635518872134185e-07,
+      "loss": 0.0684,
+      "step": 1065
+    },
+    {
+      "epoch": 2.845744680851064,
+      "grad_norm": 0.6991639137268066,
+      "learning_rate": 2.2407934336583446e-07,
+      "loss": 0.058,
+      "step": 1070
+    },
+    {
+      "epoch": 2.8590425531914896,
+      "grad_norm": 0.7342203259468079,
+      "learning_rate": 1.8778519238719204e-07,
+      "loss": 0.0702,
+      "step": 1075
+    },
+    {
+      "epoch": 2.872340425531915,
+      "grad_norm": 0.5783439874649048,
+      "learning_rate": 1.5467724137617046e-07,
+      "loss": 0.0677,
+      "step": 1080
+    },
+    {
+      "epoch": 2.8856382978723403,
+      "grad_norm": 0.7662835121154785,
+      "learning_rate": 1.2476261205992934e-07,
+      "loss": 0.0546,
+      "step": 1085
+    },
+    {
+      "epoch": 2.898936170212766,
+      "grad_norm": 0.8337989449501038,
+      "learning_rate": 9.804773926217092e-08,
+      "loss": 0.0603,
+      "step": 1090
+    },
+    {
+      "epoch": 2.9122340425531914,
+      "grad_norm": 0.6665154695510864,
+      "learning_rate": 7.453836951897885e-08,
+      "loss": 0.0508,
+      "step": 1095
+    },
+    {
+      "epoch": 2.925531914893617,
+      "grad_norm": 0.5267189145088196,
+      "learning_rate": 5.4239559842695354e-08,
+      "loss": 0.0548,
+      "step": 1100
+    },
+    {
+      "epoch": 2.9388297872340425,
+      "grad_norm": 0.6062968969345093,
+      "learning_rate": 3.715567663412966e-08,
+      "loss": 0.0643,
+      "step": 1105
+    },
+    {
+      "epoch": 2.952127659574468,
+      "grad_norm": 0.8466705679893494,
+      "learning_rate": 2.3290394743317732e-08,
+      "loss": 0.0654,
+      "step": 1110
+    },
+    {
+      "epoch": 2.9654255319148937,
+      "grad_norm": 0.6909672617912292,
+      "learning_rate": 1.2646696679042835e-08,
+      "loss": 0.0503,
+      "step": 1115
+    },
+    {
+      "epoch": 2.978723404255319,
+      "grad_norm": 0.6118969917297363,
+      "learning_rate": 5.2268719672671215e-09,
+      "loss": 0.0578,
+      "step": 1120
+    },
+    {
+      "epoch": 2.992021276595745,
+      "grad_norm": 0.6353937387466431,
+      "learning_rate": 1.0325166586572233e-09,
+      "loss": 0.0658,
+      "step": 1125
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1128,
+  "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": 1.9522232945181983e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

base/28_128_e3_3e-5/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13e873e8bc4d06fd0eb1e39bf49a336e4ba343a6f625d9ba3679c980e6c5dbdd
+size 8273

base/28_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)