Upload folder using huggingface_hub

checkpoint-232/README.md

@@ -0,0 +1,202 @@
+---
+base_model: Qwen/Qwen2.5-7B-Instruct
+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 0.14.0

checkpoint-232/adapter_config.json

@@ -0,0 +1,37 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-7B-Instruct",
+  "bias": "none",
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 512,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 256,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "o_proj",
+    "down_proj",
+    "up_proj",
+    "q_proj",
+    "k_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-232/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c93ab10c4d30fa0ea9cbdb7f9cc28b1cf99e4098fd2f5c2d57515f1ea04c743c
+size 1291899552

checkpoint-232/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoint-232/global_step232/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:934d209e89d542c640c6ab59ee7294a397404432aaa1f1e06b84bfb9c627083b
+size 7751120668

checkpoint-232/global_step232/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ae1dac73f3e554f25cbad86ad35244a95e27edf20a668321cdaeaebbb9ae4d0
+size 1396311192

checkpoint-232/latest

@@ -0,0 +1 @@
+global_step232

checkpoint-232/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5412ace798d3752c044045a9eb737dc2efe514e87260123e64787dde2c78039c
+size 14244

checkpoint-232/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c37220fc18af92831b182917e441e3a0a686ccee61589f8b58ddb65db1457aa
+size 1064

checkpoint-232/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-232/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
+size 11421896

checkpoint-232/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-232/trainer_state.json

@@ -0,0 +1,1657 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0,
+  "eval_steps": 500,
+  "global_step": 232,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.008620689655172414,
+      "grad_norm": 30.156124114990234,
+      "learning_rate": 5.0000000000000004e-08,
+      "loss": 3.946,
+      "step": 1
+    },
+    {
+      "epoch": 0.017241379310344827,
+      "grad_norm": 29.95058250427246,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 4.0398,
+      "step": 2
+    },
+    {
+      "epoch": 0.02586206896551724,
+      "grad_norm": 29.866586685180664,
+      "learning_rate": 1.5000000000000002e-07,
+      "loss": 3.9303,
+      "step": 3
+    },
+    {
+      "epoch": 0.034482758620689655,
+      "grad_norm": 30.012300491333008,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 3.9637,
+      "step": 4
+    },
+    {
+      "epoch": 0.04310344827586207,
+      "grad_norm": 30.934751510620117,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 4.0617,
+      "step": 5
+    },
+    {
+      "epoch": 0.05172413793103448,
+      "grad_norm": 30.031415939331055,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 3.9659,
+      "step": 6
+    },
+    {
+      "epoch": 0.0603448275862069,
+      "grad_norm": 29.62813949584961,
+      "learning_rate": 3.5000000000000004e-07,
+      "loss": 3.9978,
+      "step": 7
+    },
+    {
+      "epoch": 0.06896551724137931,
+      "grad_norm": 31.0523681640625,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 4.0237,
+      "step": 8
+    },
+    {
+      "epoch": 0.07758620689655173,
+      "grad_norm": 28.83756446838379,
+      "learning_rate": 4.5000000000000003e-07,
+      "loss": 3.8214,
+      "step": 9
+    },
+    {
+      "epoch": 0.08620689655172414,
+      "grad_norm": 28.1810359954834,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 3.848,
+      "step": 10
+    },
+    {
+      "epoch": 0.09482758620689655,
+      "grad_norm": 28.367570877075195,
+      "learning_rate": 5.5e-07,
+      "loss": 3.7026,
+      "step": 11
+    },
+    {
+      "epoch": 0.10344827586206896,
+      "grad_norm": 27.66380500793457,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 3.6833,
+      "step": 12
+    },
+    {
+      "epoch": 0.11206896551724138,
+      "grad_norm": 27.496694564819336,
+      "learning_rate": 6.5e-07,
+      "loss": 3.762,
+      "step": 13
+    },
+    {
+      "epoch": 0.1206896551724138,
+      "grad_norm": 28.319055557250977,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 3.8867,
+      "step": 14
+    },
+    {
+      "epoch": 0.12931034482758622,
+      "grad_norm": 26.112581253051758,
+      "learning_rate": 7.5e-07,
+      "loss": 3.6278,
+      "step": 15
+    },
+    {
+      "epoch": 0.13793103448275862,
+      "grad_norm": 25.8245792388916,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 3.7957,
+      "step": 16
+    },
+    {
+      "epoch": 0.14655172413793102,
+      "grad_norm": 23.66245460510254,
+      "learning_rate": 8.500000000000001e-07,
+      "loss": 3.4384,
+      "step": 17
+    },
+    {
+      "epoch": 0.15517241379310345,
+      "grad_norm": 21.69405746459961,
+      "learning_rate": 9.000000000000001e-07,
+      "loss": 3.2602,
+      "step": 18
+    },
+    {
+      "epoch": 0.16379310344827586,
+      "grad_norm": 20.691402435302734,
+      "learning_rate": 9.500000000000001e-07,
+      "loss": 3.2486,
+      "step": 19
+    },
+    {
+      "epoch": 0.1724137931034483,
+      "grad_norm": 20.67167854309082,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 3.2401,
+      "step": 20
+    },
+    {
+      "epoch": 0.1810344827586207,
+      "grad_norm": 19.054428100585938,
+      "learning_rate": 1.0500000000000001e-06,
+      "loss": 3.0375,
+      "step": 21
+    },
+    {
+      "epoch": 0.1896551724137931,
+      "grad_norm": 18.890884399414062,
+      "learning_rate": 1.1e-06,
+      "loss": 2.9677,
+      "step": 22
+    },
+    {
+      "epoch": 0.19827586206896552,
+      "grad_norm": 18.61600112915039,
+      "learning_rate": 1.1500000000000002e-06,
+      "loss": 2.8387,
+      "step": 23
+    },
+    {
+      "epoch": 0.20689655172413793,
+      "grad_norm": 16.910585403442383,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 2.4914,
+      "step": 24
+    },
+    {
+      "epoch": 0.21551724137931033,
+      "grad_norm": 17.708385467529297,
+      "learning_rate": 1.25e-06,
+      "loss": 2.5361,
+      "step": 25
+    },
+    {
+      "epoch": 0.22413793103448276,
+      "grad_norm": 17.07745933532715,
+      "learning_rate": 1.3e-06,
+      "loss": 2.4098,
+      "step": 26
+    },
+    {
+      "epoch": 0.23275862068965517,
+      "grad_norm": 16.244144439697266,
+      "learning_rate": 1.3500000000000002e-06,
+      "loss": 2.2155,
+      "step": 27
+    },
+    {
+      "epoch": 0.2413793103448276,
+      "grad_norm": 16.11887550354004,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 2.0036,
+      "step": 28
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 16.034631729125977,
+      "learning_rate": 1.45e-06,
+      "loss": 1.8617,
+      "step": 29
+    },
+    {
+      "epoch": 0.25862068965517243,
+      "grad_norm": 15.437153816223145,
+      "learning_rate": 1.5e-06,
+      "loss": 1.7413,
+      "step": 30
+    },
+    {
+      "epoch": 0.2672413793103448,
+      "grad_norm": 14.188116073608398,
+      "learning_rate": 1.5500000000000002e-06,
+      "loss": 1.5002,
+      "step": 31
+    },
+    {
+      "epoch": 0.27586206896551724,
+      "grad_norm": 14.200998306274414,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 1.4469,
+      "step": 32
+    },
+    {
+      "epoch": 0.28448275862068967,
+      "grad_norm": 13.209551811218262,
+      "learning_rate": 1.6500000000000003e-06,
+      "loss": 1.3016,
+      "step": 33
+    },
+    {
+      "epoch": 0.29310344827586204,
+      "grad_norm": 12.631085395812988,
+      "learning_rate": 1.7000000000000002e-06,
+      "loss": 1.1522,
+      "step": 34
+    },
+    {
+      "epoch": 0.3017241379310345,
+      "grad_norm": 12.504134178161621,
+      "learning_rate": 1.75e-06,
+      "loss": 1.0058,
+      "step": 35
+    },
+    {
+      "epoch": 0.3103448275862069,
+      "grad_norm": 12.599784851074219,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 0.921,
+      "step": 36
+    },
+    {
+      "epoch": 0.31896551724137934,
+      "grad_norm": 11.577665328979492,
+      "learning_rate": 1.85e-06,
+      "loss": 0.8099,
+      "step": 37
+    },
+    {
+      "epoch": 0.3275862068965517,
+      "grad_norm": 10.465872764587402,
+      "learning_rate": 1.9000000000000002e-06,
+      "loss": 0.6565,
+      "step": 38
+    },
+    {
+      "epoch": 0.33620689655172414,
+      "grad_norm": 10.161813735961914,
+      "learning_rate": 1.9500000000000004e-06,
+      "loss": 0.5712,
+      "step": 39
+    },
+    {
+      "epoch": 0.3448275862068966,
+      "grad_norm": 8.384145736694336,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.4265,
+      "step": 40
+    },
+    {
+      "epoch": 0.35344827586206895,
+      "grad_norm": 7.072062015533447,
+      "learning_rate": 2.05e-06,
+      "loss": 0.409,
+      "step": 41
+    },
+    {
+      "epoch": 0.3620689655172414,
+      "grad_norm": 6.072140693664551,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 0.3263,
+      "step": 42
+    },
+    {
+      "epoch": 0.3706896551724138,
+      "grad_norm": 6.31119441986084,
+      "learning_rate": 2.15e-06,
+      "loss": 0.258,
+      "step": 43
+    },
+    {
+      "epoch": 0.3793103448275862,
+      "grad_norm": 6.619389057159424,
+      "learning_rate": 2.2e-06,
+      "loss": 0.2454,
+      "step": 44
+    },
+    {
+      "epoch": 0.3879310344827586,
+      "grad_norm": 6.546375751495361,
+      "learning_rate": 2.25e-06,
+      "loss": 0.2314,
+      "step": 45
+    },
+    {
+      "epoch": 0.39655172413793105,
+      "grad_norm": 5.219631671905518,
+      "learning_rate": 2.3000000000000004e-06,
+      "loss": 0.2006,
+      "step": 46
+    },
+    {
+      "epoch": 0.4051724137931034,
+      "grad_norm": 3.1164474487304688,
+      "learning_rate": 2.35e-06,
+      "loss": 0.1677,
+      "step": 47
+    },
+    {
+      "epoch": 0.41379310344827586,
+      "grad_norm": 1.872147798538208,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 0.1339,
+      "step": 48
+    },
+    {
+      "epoch": 0.4224137931034483,
+      "grad_norm": 1.4775545597076416,
+      "learning_rate": 2.4500000000000003e-06,
+      "loss": 0.1218,
+      "step": 49
+    },
+    {
+      "epoch": 0.43103448275862066,
+      "grad_norm": 1.2931559085845947,
+      "learning_rate": 2.5e-06,
+      "loss": 0.1151,
+      "step": 50
+    },
+    {
+      "epoch": 0.4396551724137931,
+      "grad_norm": 1.3222297430038452,
+      "learning_rate": 2.55e-06,
+      "loss": 0.1164,
+      "step": 51
+    },
+    {
+      "epoch": 0.4482758620689655,
+      "grad_norm": 1.393062710762024,
+      "learning_rate": 2.6e-06,
+      "loss": 0.1053,
+      "step": 52
+    },
+    {
+      "epoch": 0.45689655172413796,
+      "grad_norm": 1.0838805437088013,
+      "learning_rate": 2.6500000000000005e-06,
+      "loss": 0.105,
+      "step": 53
+    },
+    {
+      "epoch": 0.46551724137931033,
+      "grad_norm": 0.954925537109375,
+      "learning_rate": 2.7000000000000004e-06,
+      "loss": 0.0922,
+      "step": 54
+    },
+    {
+      "epoch": 0.47413793103448276,
+      "grad_norm": 0.7521713972091675,
+      "learning_rate": 2.7500000000000004e-06,
+      "loss": 0.0887,
+      "step": 55
+    },
+    {
+      "epoch": 0.4827586206896552,
+      "grad_norm": 0.7261010408401489,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 0.0997,
+      "step": 56
+    },
+    {
+      "epoch": 0.49137931034482757,
+      "grad_norm": 0.6530802845954895,
+      "learning_rate": 2.85e-06,
+      "loss": 0.0882,
+      "step": 57
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.6899245381355286,
+      "learning_rate": 2.9e-06,
+      "loss": 0.0844,
+      "step": 58
+    },
+    {
+      "epoch": 0.5086206896551724,
+      "grad_norm": 0.6771528124809265,
+      "learning_rate": 2.95e-06,
+      "loss": 0.0884,
+      "step": 59
+    },
+    {
+      "epoch": 0.5172413793103449,
+      "grad_norm": 0.6307985782623291,
+      "learning_rate": 3e-06,
+      "loss": 0.0912,
+      "step": 60
+    },
+    {
+      "epoch": 0.5258620689655172,
+      "grad_norm": 0.5869951844215393,
+      "learning_rate": 3.05e-06,
+      "loss": 0.0875,
+      "step": 61
+    },
+    {
+      "epoch": 0.5344827586206896,
+      "grad_norm": 0.6039404273033142,
+      "learning_rate": 3.1000000000000004e-06,
+      "loss": 0.0768,
+      "step": 62
+    },
+    {
+      "epoch": 0.5431034482758621,
+      "grad_norm": 0.49209344387054443,
+      "learning_rate": 3.1500000000000003e-06,
+      "loss": 0.0815,
+      "step": 63
+    },
+    {
+      "epoch": 0.5517241379310345,
+      "grad_norm": 0.7383344769477844,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 0.0874,
+      "step": 64
+    },
+    {
+      "epoch": 0.5603448275862069,
+      "grad_norm": 0.5552617311477661,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.0762,
+      "step": 65
+    },
+    {
+      "epoch": 0.5689655172413793,
+      "grad_norm": 0.4992441236972809,
+      "learning_rate": 3.3000000000000006e-06,
+      "loss": 0.0793,
+      "step": 66
+    },
+    {
+      "epoch": 0.5775862068965517,
+      "grad_norm": 0.4979636073112488,
+      "learning_rate": 3.3500000000000005e-06,
+      "loss": 0.0792,
+      "step": 67
+    },
+    {
+      "epoch": 0.5862068965517241,
+      "grad_norm": 0.4675934612751007,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 0.0766,
+      "step": 68
+    },
+    {
+      "epoch": 0.5948275862068966,
+      "grad_norm": 0.49495571851730347,
+      "learning_rate": 3.45e-06,
+      "loss": 0.0824,
+      "step": 69
+    },
+    {
+      "epoch": 0.603448275862069,
+      "grad_norm": 0.6044315695762634,
+      "learning_rate": 3.5e-06,
+      "loss": 0.0826,
+      "step": 70
+    },
+    {
+      "epoch": 0.6120689655172413,
+      "grad_norm": 0.4898519217967987,
+      "learning_rate": 3.5500000000000003e-06,
+      "loss": 0.0755,
+      "step": 71
+    },
+    {
+      "epoch": 0.6206896551724138,
+      "grad_norm": 0.4218939244747162,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 0.0762,
+      "step": 72
+    },
+    {
+      "epoch": 0.6293103448275862,
+      "grad_norm": 0.5785802006721497,
+      "learning_rate": 3.65e-06,
+      "loss": 0.0837,
+      "step": 73
+    },
+    {
+      "epoch": 0.6379310344827587,
+      "grad_norm": 0.5505399703979492,
+      "learning_rate": 3.7e-06,
+      "loss": 0.0814,
+      "step": 74
+    },
+    {
+      "epoch": 0.646551724137931,
+      "grad_norm": 0.6062561869621277,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.0725,
+      "step": 75
+    },
+    {
+      "epoch": 0.6551724137931034,
+      "grad_norm": 0.707350492477417,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.0913,
+      "step": 76
+    },
+    {
+      "epoch": 0.6637931034482759,
+      "grad_norm": 0.5091889500617981,
+      "learning_rate": 3.85e-06,
+      "loss": 0.0697,
+      "step": 77
+    },
+    {
+      "epoch": 0.6724137931034483,
+      "grad_norm": 0.4801473319530487,
+      "learning_rate": 3.900000000000001e-06,
+      "loss": 0.073,
+      "step": 78
+    },
+    {
+      "epoch": 0.6810344827586207,
+      "grad_norm": 0.462162047624588,
+      "learning_rate": 3.95e-06,
+      "loss": 0.0805,
+      "step": 79
+    },
+    {
+      "epoch": 0.6896551724137931,
+      "grad_norm": 0.463969349861145,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.0738,
+      "step": 80
+    },
+    {
+      "epoch": 0.6982758620689655,
+      "grad_norm": 0.3980114161968231,
+      "learning_rate": 4.05e-06,
+      "loss": 0.0742,
+      "step": 81
+    },
+    {
+      "epoch": 0.7068965517241379,
+      "grad_norm": 0.3627180755138397,
+      "learning_rate": 4.1e-06,
+      "loss": 0.0683,
+      "step": 82
+    },
+    {
+      "epoch": 0.7155172413793104,
+      "grad_norm": 0.39726322889328003,
+      "learning_rate": 4.15e-06,
+      "loss": 0.0717,
+      "step": 83
+    },
+    {
+      "epoch": 0.7241379310344828,
+      "grad_norm": 0.48898085951805115,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 0.0676,
+      "step": 84
+    },
+    {
+      "epoch": 0.7327586206896551,
+      "grad_norm": 0.6143100261688232,
+      "learning_rate": 4.25e-06,
+      "loss": 0.0708,
+      "step": 85
+    },
+    {
+      "epoch": 0.7413793103448276,
+      "grad_norm": 0.5028995275497437,
+      "learning_rate": 4.3e-06,
+      "loss": 0.0713,
+      "step": 86
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.5576066970825195,
+      "learning_rate": 4.350000000000001e-06,
+      "loss": 0.078,
+      "step": 87
+    },
+    {
+      "epoch": 0.7586206896551724,
+      "grad_norm": 0.37101301550865173,
+      "learning_rate": 4.4e-06,
+      "loss": 0.071,
+      "step": 88
+    },
+    {
+      "epoch": 0.7672413793103449,
+      "grad_norm": 0.442694753408432,
+      "learning_rate": 4.450000000000001e-06,
+      "loss": 0.0791,
+      "step": 89
+    },
+    {
+      "epoch": 0.7758620689655172,
+      "grad_norm": 0.48991039395332336,
+      "learning_rate": 4.5e-06,
+      "loss": 0.0681,
+      "step": 90
+    },
+    {
+      "epoch": 0.7844827586206896,
+      "grad_norm": 0.46367791295051575,
+      "learning_rate": 4.5500000000000005e-06,
+      "loss": 0.0701,
+      "step": 91
+    },
+    {
+      "epoch": 0.7931034482758621,
+      "grad_norm": 0.3454825282096863,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 0.0774,
+      "step": 92
+    },
+    {
+      "epoch": 0.8017241379310345,
+      "grad_norm": 0.35102447867393494,
+      "learning_rate": 4.65e-06,
+      "loss": 0.0659,
+      "step": 93
+    },
+    {
+      "epoch": 0.8103448275862069,
+      "grad_norm": 0.3791246712207794,
+      "learning_rate": 4.7e-06,
+      "loss": 0.0727,
+      "step": 94
+    },
+    {
+      "epoch": 0.8189655172413793,
+      "grad_norm": 0.3911365270614624,
+      "learning_rate": 4.75e-06,
+      "loss": 0.0641,
+      "step": 95
+    },
+    {
+      "epoch": 0.8275862068965517,
+      "grad_norm": 0.35395047068595886,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.0735,
+      "step": 96
+    },
+    {
+      "epoch": 0.8362068965517241,
+      "grad_norm": 0.3499661087989807,
+      "learning_rate": 4.85e-06,
+      "loss": 0.0657,
+      "step": 97
+    },
+    {
+      "epoch": 0.8448275862068966,
+      "grad_norm": 0.34877678751945496,
+      "learning_rate": 4.9000000000000005e-06,
+      "loss": 0.0701,
+      "step": 98
+    },
+    {
+      "epoch": 0.853448275862069,
+      "grad_norm": 0.36436647176742554,
+      "learning_rate": 4.95e-06,
+      "loss": 0.0758,
+      "step": 99
+    },
+    {
+      "epoch": 0.8620689655172413,
+      "grad_norm": 0.4726298749446869,
+      "learning_rate": 5e-06,
+      "loss": 0.0731,
+      "step": 100
+    },
+    {
+      "epoch": 0.8706896551724138,
+      "grad_norm": 0.31892526149749756,
+      "learning_rate": 4.999965269084342e-06,
+      "loss": 0.0635,
+      "step": 101
+    },
+    {
+      "epoch": 0.8793103448275862,
+      "grad_norm": 0.3707881569862366,
+      "learning_rate": 4.999861077302358e-06,
+      "loss": 0.0689,
+      "step": 102
+    },
+    {
+      "epoch": 0.8879310344827587,
+      "grad_norm": 0.3244622051715851,
+      "learning_rate": 4.999687427548989e-06,
+      "loss": 0.0645,
+      "step": 103
+    },
+    {
+      "epoch": 0.896551724137931,
+      "grad_norm": 0.386982262134552,
+      "learning_rate": 4.999444324649045e-06,
+      "loss": 0.0708,
+      "step": 104
+    },
+    {
+      "epoch": 0.9051724137931034,
+      "grad_norm": 0.3552643656730652,
+      "learning_rate": 4.999131775357078e-06,
+      "loss": 0.0766,
+      "step": 105
+    },
+    {
+      "epoch": 0.9137931034482759,
+      "grad_norm": 0.3503671884536743,
+      "learning_rate": 4.998749788357184e-06,
+      "loss": 0.0733,
+      "step": 106
+    },
+    {
+      "epoch": 0.9224137931034483,
+      "grad_norm": 0.4907379746437073,
+      "learning_rate": 4.998298374262771e-06,
+      "loss": 0.0721,
+      "step": 107
+    },
+    {
+      "epoch": 0.9310344827586207,
+      "grad_norm": 0.305571585893631,
+      "learning_rate": 4.997777545616258e-06,
+      "loss": 0.0603,
+      "step": 108
+    },
+    {
+      "epoch": 0.9396551724137931,
+      "grad_norm": 0.3188783824443817,
+      "learning_rate": 4.99718731688873e-06,
+      "loss": 0.0574,
+      "step": 109
+    },
+    {
+      "epoch": 0.9482758620689655,
+      "grad_norm": 0.3277616500854492,
+      "learning_rate": 4.996527704479535e-06,
+      "loss": 0.0671,
+      "step": 110
+    },
+    {
+      "epoch": 0.9568965517241379,
+      "grad_norm": 0.38640740513801575,
+      "learning_rate": 4.995798726715826e-06,
+      "loss": 0.0625,
+      "step": 111
+    },
+    {
+      "epoch": 0.9655172413793104,
+      "grad_norm": 0.5463483929634094,
+      "learning_rate": 4.995000403852057e-06,
+      "loss": 0.0718,
+      "step": 112
+    },
+    {
+      "epoch": 0.9741379310344828,
+      "grad_norm": 0.3974014222621918,
+      "learning_rate": 4.994132758069413e-06,
+      "loss": 0.071,
+      "step": 113
+    },
+    {
+      "epoch": 0.9827586206896551,
+      "grad_norm": 0.3642785847187042,
+      "learning_rate": 4.993195813475202e-06,
+      "loss": 0.0658,
+      "step": 114
+    },
+    {
+      "epoch": 0.9913793103448276,
+      "grad_norm": 0.3118143677711487,
+      "learning_rate": 4.992189596102179e-06,
+      "loss": 0.0633,
+      "step": 115
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.41573596000671387,
+      "learning_rate": 4.991114133907822e-06,
+      "loss": 0.0662,
+      "step": 116
+    },
+    {
+      "epoch": 1.0086206896551724,
+      "grad_norm": 0.29761484265327454,
+      "learning_rate": 4.989969456773562e-06,
+      "loss": 0.0649,
+      "step": 117
+    },
+    {
+      "epoch": 1.0172413793103448,
+      "grad_norm": 0.399275004863739,
+      "learning_rate": 4.988755596503948e-06,
+      "loss": 0.0521,
+      "step": 118
+    },
+    {
+      "epoch": 1.0258620689655173,
+      "grad_norm": 0.37820911407470703,
+      "learning_rate": 4.987472586825762e-06,
+      "loss": 0.0596,
+      "step": 119
+    },
+    {
+      "epoch": 1.0344827586206897,
+      "grad_norm": 0.4371725916862488,
+      "learning_rate": 4.986120463387084e-06,
+      "loss": 0.07,
+      "step": 120
+    },
+    {
+      "epoch": 1.043103448275862,
+      "grad_norm": 0.3165535032749176,
+      "learning_rate": 4.984699263756303e-06,
+      "loss": 0.0613,
+      "step": 121
+    },
+    {
+      "epoch": 1.0517241379310345,
+      "grad_norm": 0.40526920557022095,
+      "learning_rate": 4.983209027421072e-06,
+      "loss": 0.0611,
+      "step": 122
+    },
+    {
+      "epoch": 1.0603448275862069,
+      "grad_norm": 0.39523962140083313,
+      "learning_rate": 4.9816497957872055e-06,
+      "loss": 0.0634,
+      "step": 123
+    },
+    {
+      "epoch": 1.0689655172413792,
+      "grad_norm": 0.33234691619873047,
+      "learning_rate": 4.9800216121775404e-06,
+      "loss": 0.0602,
+      "step": 124
+    },
+    {
+      "epoch": 1.0775862068965518,
+      "grad_norm": 0.3455008268356323,
+      "learning_rate": 4.978324521830721e-06,
+      "loss": 0.0624,
+      "step": 125
+    },
+    {
+      "epoch": 1.0862068965517242,
+      "grad_norm": 0.3968574106693268,
+      "learning_rate": 4.97655857189995e-06,
+      "loss": 0.0525,
+      "step": 126
+    },
+    {
+      "epoch": 1.0948275862068966,
+      "grad_norm": 0.47838935256004333,
+      "learning_rate": 4.974723811451673e-06,
+      "loss": 0.0617,
+      "step": 127
+    },
+    {
+      "epoch": 1.103448275862069,
+      "grad_norm": 0.43055081367492676,
+      "learning_rate": 4.972820291464219e-06,
+      "loss": 0.0484,
+      "step": 128
+    },
+    {
+      "epoch": 1.1120689655172413,
+      "grad_norm": 0.4012243449687958,
+      "learning_rate": 4.97084806482638e-06,
+      "loss": 0.0562,
+      "step": 129
+    },
+    {
+      "epoch": 1.1206896551724137,
+      "grad_norm": 0.586875319480896,
+      "learning_rate": 4.968807186335948e-06,
+      "loss": 0.0633,
+      "step": 130
+    },
+    {
+      "epoch": 1.1293103448275863,
+      "grad_norm": 0.48629340529441833,
+      "learning_rate": 4.966697712698185e-06,
+      "loss": 0.0615,
+      "step": 131
+    },
+    {
+      "epoch": 1.1379310344827587,
+      "grad_norm": 0.41810542345046997,
+      "learning_rate": 4.964519702524251e-06,
+      "loss": 0.0564,
+      "step": 132
+    },
+    {
+      "epoch": 1.146551724137931,
+      "grad_norm": 0.4095732271671295,
+      "learning_rate": 4.962273216329577e-06,
+      "loss": 0.0634,
+      "step": 133
+    },
+    {
+      "epoch": 1.1551724137931034,
+      "grad_norm": 0.3784744441509247,
+      "learning_rate": 4.959958316532181e-06,
+      "loss": 0.0521,
+      "step": 134
+    },
+    {
+      "epoch": 1.1637931034482758,
+      "grad_norm": 0.4087596833705902,
+      "learning_rate": 4.957575067450935e-06,
+      "loss": 0.0569,
+      "step": 135
+    },
+    {
+      "epoch": 1.1724137931034484,
+      "grad_norm": 0.4369741380214691,
+      "learning_rate": 4.955123535303775e-06,
+      "loss": 0.057,
+      "step": 136
+    },
+    {
+      "epoch": 1.1810344827586208,
+      "grad_norm": 0.3851775527000427,
+      "learning_rate": 4.95260378820587e-06,
+      "loss": 0.0521,
+      "step": 137
+    },
+    {
+      "epoch": 1.1896551724137931,
+      "grad_norm": 0.3391599655151367,
+      "learning_rate": 4.950015896167716e-06,
+      "loss": 0.0525,
+      "step": 138
+    },
+    {
+      "epoch": 1.1982758620689655,
+      "grad_norm": 0.4176456332206726,
+      "learning_rate": 4.947359931093202e-06,
+      "loss": 0.0548,
+      "step": 139
+    },
+    {
+      "epoch": 1.206896551724138,
+      "grad_norm": 0.45864036679267883,
+      "learning_rate": 4.944635966777607e-06,
+      "loss": 0.0578,
+      "step": 140
+    },
+    {
+      "epoch": 1.2155172413793103,
+      "grad_norm": 0.4484093487262726,
+      "learning_rate": 4.941844078905551e-06,
+      "loss": 0.0528,
+      "step": 141
+    },
+    {
+      "epoch": 1.2241379310344827,
+      "grad_norm": 0.42641302943229675,
+      "learning_rate": 4.938984345048892e-06,
+      "loss": 0.0519,
+      "step": 142
+    },
+    {
+      "epoch": 1.2327586206896552,
+      "grad_norm": 0.5111292004585266,
+      "learning_rate": 4.936056844664571e-06,
+      "loss": 0.0639,
+      "step": 143
+    },
+    {
+      "epoch": 1.2413793103448276,
+      "grad_norm": 0.3739636540412903,
+      "learning_rate": 4.933061659092401e-06,
+      "loss": 0.055,
+      "step": 144
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 0.3896700143814087,
+      "learning_rate": 4.929998871552814e-06,
+      "loss": 0.0548,
+      "step": 145
+    },
+    {
+      "epoch": 1.2586206896551724,
+      "grad_norm": 0.4158782660961151,
+      "learning_rate": 4.926868567144543e-06,
+      "loss": 0.0572,
+      "step": 146
+    },
+    {
+      "epoch": 1.2672413793103448,
+      "grad_norm": 0.4275132715702057,
+      "learning_rate": 4.923670832842256e-06,
+      "loss": 0.0497,
+      "step": 147
+    },
+    {
+      "epoch": 1.2758620689655173,
+      "grad_norm": 0.3890449106693268,
+      "learning_rate": 4.920405757494147e-06,
+      "loss": 0.0584,
+      "step": 148
+    },
+    {
+      "epoch": 1.2844827586206897,
+      "grad_norm": 0.4569760262966156,
+      "learning_rate": 4.917073431819462e-06,
+      "loss": 0.0618,
+      "step": 149
+    },
+    {
+      "epoch": 1.293103448275862,
+      "grad_norm": 0.4232596457004547,
+      "learning_rate": 4.913673948405977e-06,
+      "loss": 0.0516,
+      "step": 150
+    },
+    {
+      "epoch": 1.3017241379310345,
+      "grad_norm": 0.4578849673271179,
+      "learning_rate": 4.910207401707431e-06,
+      "loss": 0.0538,
+      "step": 151
+    },
+    {
+      "epoch": 1.3103448275862069,
+      "grad_norm": 0.39952200651168823,
+      "learning_rate": 4.906673888040895e-06,
+      "loss": 0.0471,
+      "step": 152
+    },
+    {
+      "epoch": 1.3189655172413794,
+      "grad_norm": 0.38846731185913086,
+      "learning_rate": 4.903073505584102e-06,
+      "loss": 0.0437,
+      "step": 153
+    },
+    {
+      "epoch": 1.3275862068965516,
+      "grad_norm": 0.4084296226501465,
+      "learning_rate": 4.899406354372716e-06,
+      "loss": 0.0532,
+      "step": 154
+    },
+    {
+      "epoch": 1.3362068965517242,
+      "grad_norm": 0.4114966094493866,
+      "learning_rate": 4.895672536297551e-06,
+      "loss": 0.0443,
+      "step": 155
+    },
+    {
+      "epoch": 1.3448275862068966,
+      "grad_norm": 0.6465100049972534,
+      "learning_rate": 4.891872155101746e-06,
+      "loss": 0.0605,
+      "step": 156
+    },
+    {
+      "epoch": 1.353448275862069,
+      "grad_norm": 0.43234798312187195,
+      "learning_rate": 4.888005316377873e-06,
+      "loss": 0.0529,
+      "step": 157
+    },
+    {
+      "epoch": 1.3620689655172413,
+      "grad_norm": 0.408527135848999,
+      "learning_rate": 4.884072127565015e-06,
+      "loss": 0.0485,
+      "step": 158
+    },
+    {
+      "epoch": 1.3706896551724137,
+      "grad_norm": 0.45152318477630615,
+      "learning_rate": 4.880072697945768e-06,
+      "loss": 0.0466,
+      "step": 159
+    },
+    {
+      "epoch": 1.3793103448275863,
+      "grad_norm": 0.4669966995716095,
+      "learning_rate": 4.876007138643216e-06,
+      "loss": 0.0438,
+      "step": 160
+    },
+    {
+      "epoch": 1.3879310344827587,
+      "grad_norm": 0.47117388248443604,
+      "learning_rate": 4.871875562617837e-06,
+      "loss": 0.0464,
+      "step": 161
+    },
+    {
+      "epoch": 1.396551724137931,
+      "grad_norm": 0.6085790991783142,
+      "learning_rate": 4.867678084664365e-06,
+      "loss": 0.0574,
+      "step": 162
+    },
+    {
+      "epoch": 1.4051724137931034,
+      "grad_norm": 0.5736953616142273,
+      "learning_rate": 4.863414821408602e-06,
+      "loss": 0.0507,
+      "step": 163
+    },
+    {
+      "epoch": 1.4137931034482758,
+      "grad_norm": 0.5003281831741333,
+      "learning_rate": 4.8590858913041775e-06,
+      "loss": 0.051,
+      "step": 164
+    },
+    {
+      "epoch": 1.4224137931034484,
+      "grad_norm": 0.4361717104911804,
+      "learning_rate": 4.854691414629258e-06,
+      "loss": 0.0476,
+      "step": 165
+    },
+    {
+      "epoch": 1.4310344827586206,
+      "grad_norm": 0.47907084226608276,
+      "learning_rate": 4.8502315134832e-06,
+      "loss": 0.0495,
+      "step": 166
+    },
+    {
+      "epoch": 1.4396551724137931,
+      "grad_norm": 0.46805888414382935,
+      "learning_rate": 4.8457063117831656e-06,
+      "loss": 0.0474,
+      "step": 167
+    },
+    {
+      "epoch": 1.4482758620689655,
+      "grad_norm": 0.5058819651603699,
+      "learning_rate": 4.8411159352606735e-06,
+      "loss": 0.0541,
+      "step": 168
+    },
+    {
+      "epoch": 1.456896551724138,
+      "grad_norm": 0.4632813334465027,
+      "learning_rate": 4.836460511458107e-06,
+      "loss": 0.0409,
+      "step": 169
+    },
+    {
+      "epoch": 1.4655172413793103,
+      "grad_norm": 0.3654536008834839,
+      "learning_rate": 4.831740169725172e-06,
+      "loss": 0.0466,
+      "step": 170
+    },
+    {
+      "epoch": 1.4741379310344827,
+      "grad_norm": 0.4301295578479767,
+      "learning_rate": 4.8269550412153e-06,
+      "loss": 0.0472,
+      "step": 171
+    },
+    {
+      "epoch": 1.4827586206896552,
+      "grad_norm": 0.37643712759017944,
+      "learning_rate": 4.822105258882007e-06,
+      "loss": 0.0445,
+      "step": 172
+    },
+    {
+      "epoch": 1.4913793103448276,
+      "grad_norm": 0.3928871154785156,
+      "learning_rate": 4.817190957475199e-06,
+      "loss": 0.0427,
+      "step": 173
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 0.39636102318763733,
+      "learning_rate": 4.812212273537426e-06,
+      "loss": 0.0431,
+      "step": 174
+    },
+    {
+      "epoch": 1.5086206896551724,
+      "grad_norm": 0.41640329360961914,
+      "learning_rate": 4.807169345400088e-06,
+      "loss": 0.0442,
+      "step": 175
+    },
+    {
+      "epoch": 1.5172413793103448,
+      "grad_norm": 0.4888840317726135,
+      "learning_rate": 4.802062313179595e-06,
+      "loss": 0.0529,
+      "step": 176
+    },
+    {
+      "epoch": 1.5258620689655173,
+      "grad_norm": 0.4154113829135895,
+      "learning_rate": 4.796891318773472e-06,
+      "loss": 0.0419,
+      "step": 177
+    },
+    {
+      "epoch": 1.5344827586206895,
+      "grad_norm": 0.41872596740722656,
+      "learning_rate": 4.791656505856416e-06,
+      "loss": 0.0411,
+      "step": 178
+    },
+    {
+      "epoch": 1.543103448275862,
+      "grad_norm": 0.4037671983242035,
+      "learning_rate": 4.786358019876301e-06,
+      "loss": 0.0458,
+      "step": 179
+    },
+    {
+      "epoch": 1.5517241379310345,
+      "grad_norm": 0.4356297552585602,
+      "learning_rate": 4.7809960080501464e-06,
+      "loss": 0.0495,
+      "step": 180
+    },
+    {
+      "epoch": 1.5603448275862069,
+      "grad_norm": 0.4112146496772766,
+      "learning_rate": 4.7755706193600135e-06,
+      "loss": 0.0367,
+      "step": 181
+    },
+    {
+      "epoch": 1.5689655172413794,
+      "grad_norm": 0.36699458956718445,
+      "learning_rate": 4.770082004548878e-06,
+      "loss": 0.0427,
+      "step": 182
+    },
+    {
+      "epoch": 1.5775862068965516,
+      "grad_norm": 0.49518612027168274,
+      "learning_rate": 4.764530316116433e-06,
+      "loss": 0.0334,
+      "step": 183
+    },
+    {
+      "epoch": 1.5862068965517242,
+      "grad_norm": 0.3718539774417877,
+      "learning_rate": 4.758915708314858e-06,
+      "loss": 0.0367,
+      "step": 184
+    },
+    {
+      "epoch": 1.5948275862068966,
+      "grad_norm": 0.46727338433265686,
+      "learning_rate": 4.753238337144528e-06,
+      "loss": 0.044,
+      "step": 185
+    },
+    {
+      "epoch": 1.603448275862069,
+      "grad_norm": 0.5345709323883057,
+      "learning_rate": 4.747498360349681e-06,
+      "loss": 0.0401,
+      "step": 186
+    },
+    {
+      "epoch": 1.6120689655172413,
+      "grad_norm": 0.561375617980957,
+      "learning_rate": 4.7416959374140405e-06,
+      "loss": 0.0457,
+      "step": 187
+    },
+    {
+      "epoch": 1.6206896551724137,
+      "grad_norm": 0.4418468773365021,
+      "learning_rate": 4.735831229556374e-06,
+      "loss": 0.0348,
+      "step": 188
+    },
+    {
+      "epoch": 1.6293103448275863,
+      "grad_norm": 0.5142704844474792,
+      "learning_rate": 4.72990439972602e-06,
+      "loss": 0.0391,
+      "step": 189
+    },
+    {
+      "epoch": 1.6379310344827587,
+      "grad_norm": 0.48653194308280945,
+      "learning_rate": 4.72391561259836e-06,
+      "loss": 0.0358,
+      "step": 190
+    },
+    {
+      "epoch": 1.646551724137931,
+      "grad_norm": 0.48498624563217163,
+      "learning_rate": 4.717865034570243e-06,
+      "loss": 0.038,
+      "step": 191
+    },
+    {
+      "epoch": 1.6551724137931034,
+      "grad_norm": 0.4586655795574188,
+      "learning_rate": 4.711752833755362e-06,
+      "loss": 0.0351,
+      "step": 192
+    },
+    {
+      "epoch": 1.6637931034482758,
+      "grad_norm": 0.44023749232292175,
+      "learning_rate": 4.70557917997958e-06,
+      "loss": 0.0277,
+      "step": 193
+    },
+    {
+      "epoch": 1.6724137931034484,
+      "grad_norm": 0.44018658995628357,
+      "learning_rate": 4.6993442447762185e-06,
+      "loss": 0.0382,
+      "step": 194
+    },
+    {
+      "epoch": 1.6810344827586206,
+      "grad_norm": 0.49303099513053894,
+      "learning_rate": 4.693048201381281e-06,
+      "loss": 0.0413,
+      "step": 195
+    },
+    {
+      "epoch": 1.6896551724137931,
+      "grad_norm": 0.572812557220459,
+      "learning_rate": 4.686691224728652e-06,
+      "loss": 0.0389,
+      "step": 196
+    },
+    {
+      "epoch": 1.6982758620689655,
+      "grad_norm": 0.4995497167110443,
+      "learning_rate": 4.680273491445227e-06,
+      "loss": 0.0314,
+      "step": 197
+    },
+    {
+      "epoch": 1.706896551724138,
+      "grad_norm": 0.465404212474823,
+      "learning_rate": 4.673795179846008e-06,
+      "loss": 0.0323,
+      "step": 198
+    },
+    {
+      "epoch": 1.7155172413793105,
+      "grad_norm": 0.45183637738227844,
+      "learning_rate": 4.667256469929149e-06,
+      "loss": 0.0308,
+      "step": 199
+    },
+    {
+      "epoch": 1.7241379310344827,
+      "grad_norm": 0.4891216456890106,
+      "learning_rate": 4.660657543370958e-06,
+      "loss": 0.034,
+      "step": 200
+    },
+    {
+      "epoch": 1.7327586206896552,
+      "grad_norm": 0.5728661417961121,
+      "learning_rate": 4.653998583520844e-06,
+      "loss": 0.0359,
+      "step": 201
+    },
+    {
+      "epoch": 1.7413793103448276,
+      "grad_norm": 0.5465419888496399,
+      "learning_rate": 4.6472797753962255e-06,
+      "loss": 0.0366,
+      "step": 202
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 0.4671058654785156,
+      "learning_rate": 4.640501305677387e-06,
+      "loss": 0.0345,
+      "step": 203
+    },
+    {
+      "epoch": 1.7586206896551724,
+      "grad_norm": 0.4623485803604126,
+      "learning_rate": 4.6336633627023e-06,
+      "loss": 0.0261,
+      "step": 204
+    },
+    {
+      "epoch": 1.7672413793103448,
+      "grad_norm": 0.4471631944179535,
+      "learning_rate": 4.626766136461378e-06,
+      "loss": 0.0244,
+      "step": 205
+    },
+    {
+      "epoch": 1.7758620689655173,
+      "grad_norm": 0.46128782629966736,
+      "learning_rate": 4.61980981859221e-06,
+      "loss": 0.0301,
+      "step": 206
+    },
+    {
+      "epoch": 1.7844827586206895,
+      "grad_norm": 0.45246005058288574,
+      "learning_rate": 4.612794602374226e-06,
+      "loss": 0.0274,
+      "step": 207
+    },
+    {
+      "epoch": 1.793103448275862,
+      "grad_norm": 0.5971981287002563,
+      "learning_rate": 4.605720682723331e-06,
+      "loss": 0.0387,
+      "step": 208
+    },
+    {
+      "epoch": 1.8017241379310345,
+      "grad_norm": 0.4932047128677368,
+      "learning_rate": 4.598588256186491e-06,
+      "loss": 0.0228,
+      "step": 209
+    },
+    {
+      "epoch": 1.8103448275862069,
+      "grad_norm": 0.6208699345588684,
+      "learning_rate": 4.591397520936271e-06,
+      "loss": 0.0307,
+      "step": 210
+    },
+    {
+      "epoch": 1.8189655172413794,
+      "grad_norm": 0.5628708004951477,
+      "learning_rate": 4.584148676765327e-06,
+      "loss": 0.0262,
+      "step": 211
+    },
+    {
+      "epoch": 1.8275862068965516,
+      "grad_norm": 0.6893038749694824,
+      "learning_rate": 4.576841925080853e-06,
+      "loss": 0.0312,
+      "step": 212
+    },
+    {
+      "epoch": 1.8362068965517242,
+      "grad_norm": 0.5523396134376526,
+      "learning_rate": 4.569477468898992e-06,
+      "loss": 0.0218,
+      "step": 213
+    },
+    {
+      "epoch": 1.8448275862068966,
+      "grad_norm": 0.6797971129417419,
+      "learning_rate": 4.562055512839189e-06,
+      "loss": 0.0232,
+      "step": 214
+    },
+    {
+      "epoch": 1.853448275862069,
+      "grad_norm": 0.7214551568031311,
+      "learning_rate": 4.554576263118506e-06,
+      "loss": 0.0326,
+      "step": 215
+    },
+    {
+      "epoch": 1.8620689655172413,
+      "grad_norm": 0.5208979249000549,
+      "learning_rate": 4.547039927545899e-06,
+      "loss": 0.0261,
+      "step": 216
+    },
+    {
+      "epoch": 1.8706896551724137,
+      "grad_norm": 0.5396261215209961,
+      "learning_rate": 4.539446715516434e-06,
+      "loss": 0.0269,
+      "step": 217
+    },
+    {
+      "epoch": 1.8793103448275863,
+      "grad_norm": 0.5789651274681091,
+      "learning_rate": 4.531796838005477e-06,
+      "loss": 0.035,
+      "step": 218
+    },
+    {
+      "epoch": 1.8879310344827587,
+      "grad_norm": 0.5388063192367554,
+      "learning_rate": 4.524090507562828e-06,
+      "loss": 0.0277,
+      "step": 219
+    },
+    {
+      "epoch": 1.896551724137931,
+      "grad_norm": 0.4727953374385834,
+      "learning_rate": 4.516327938306818e-06,
+      "loss": 0.0225,
+      "step": 220
+    },
+    {
+      "epoch": 1.9051724137931034,
+      "grad_norm": 0.4665185809135437,
+      "learning_rate": 4.508509345918357e-06,
+      "loss": 0.0268,
+      "step": 221
+    },
+    {
+      "epoch": 1.9137931034482758,
+      "grad_norm": 0.5534846186637878,
+      "learning_rate": 4.500634947634943e-06,
+      "loss": 0.0277,
+      "step": 222
+    },
+    {
+      "epoch": 1.9224137931034484,
+      "grad_norm": 0.46098095178604126,
+      "learning_rate": 4.492704962244626e-06,
+      "loss": 0.0274,
+      "step": 223
+    },
+    {
+      "epoch": 1.9310344827586206,
+      "grad_norm": 0.49379560351371765,
+      "learning_rate": 4.4847196100799305e-06,
+      "loss": 0.0252,
+      "step": 224
+    },
+    {
+      "epoch": 1.9396551724137931,
+      "grad_norm": 0.4714183807373047,
+      "learning_rate": 4.476679113011729e-06,
+      "loss": 0.0253,
+      "step": 225
+    },
+    {
+      "epoch": 1.9482758620689655,
+      "grad_norm": 0.6230431199073792,
+      "learning_rate": 4.4685836944430815e-06,
+      "loss": 0.0242,
+      "step": 226
+    },
+    {
+      "epoch": 1.956896551724138,
+      "grad_norm": 0.42832595109939575,
+      "learning_rate": 4.46043357930303e-06,
+      "loss": 0.0165,
+      "step": 227
+    },
+    {
+      "epoch": 1.9655172413793105,
+      "grad_norm": 0.38421332836151123,
+      "learning_rate": 4.452228994040341e-06,
+      "loss": 0.0168,
+      "step": 228
+    },
+    {
+      "epoch": 1.9741379310344827,
+      "grad_norm": 0.5493133664131165,
+      "learning_rate": 4.443970166617223e-06,
+      "loss": 0.0193,
+      "step": 229
+    },
+    {
+      "epoch": 1.9827586206896552,
+      "grad_norm": 0.45186540484428406,
+      "learning_rate": 4.435657326502986e-06,
+      "loss": 0.0165,
+      "step": 230
+    },
+    {
+      "epoch": 1.9913793103448276,
+      "grad_norm": 0.8328711986541748,
+      "learning_rate": 4.4272907046676704e-06,
+      "loss": 0.025,
+      "step": 231
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.5378574728965759,
+      "learning_rate": 4.418870533575626e-06,
+      "loss": 0.0166,
+      "step": 232
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 696,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 6,
+  "save_steps": 116,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.1971317692995994e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-232/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b81869dce97233c4dee9b25cc565b7dbf695ff949fafca7ec37b34379228fbff
+size 8056

checkpoint-232/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)

checkpoint-348/README.md

@@ -0,0 +1,202 @@
+---
+base_model: Qwen/Qwen2.5-7B-Instruct
+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 0.14.0

checkpoint-348/adapter_config.json

@@ -0,0 +1,37 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-7B-Instruct",
+  "bias": "none",
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 512,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 256,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "o_proj",
+    "down_proj",
+    "up_proj",
+    "q_proj",
+    "k_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-348/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89aa3baa485b9c720559c74bd65ca4f426c91e86de0be29062005984593a98bf
+size 1291899552

checkpoint-348/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoint-348/global_step348/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bdcf588d61f08912459c3a0a0f7de3f5cd6031ab0ef41df2d7b6c1681a82d740
+size 7751120668

checkpoint-348/global_step348/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eddc93c5d8a4f5abe8744c8e8c3010b480a635e428818fcde418a09692ef2d52
+size 1396311192

checkpoint-348/latest

@@ -0,0 +1 @@
+global_step348

checkpoint-348/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdbb03720420fc8203d8762f7d9124fa376dd664dce57b9a934865404cc4b5e6
+size 14244

checkpoint-348/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ef6bd7c1ff14eb332284d7fea860cf3278bdda6243a0b21d81d031bd4db37c2
+size 1064

checkpoint-348/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-348/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
+size 11421896

checkpoint-348/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-348/trainer_state.json

@@ -0,0 +1,2469 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 348,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.008620689655172414,
+      "grad_norm": 30.156124114990234,
+      "learning_rate": 5.0000000000000004e-08,
+      "loss": 3.946,
+      "step": 1
+    },
+    {
+      "epoch": 0.017241379310344827,
+      "grad_norm": 29.95058250427246,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 4.0398,
+      "step": 2
+    },
+    {
+      "epoch": 0.02586206896551724,
+      "grad_norm": 29.866586685180664,
+      "learning_rate": 1.5000000000000002e-07,
+      "loss": 3.9303,
+      "step": 3
+    },
+    {
+      "epoch": 0.034482758620689655,
+      "grad_norm": 30.012300491333008,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 3.9637,
+      "step": 4
+    },
+    {
+      "epoch": 0.04310344827586207,
+      "grad_norm": 30.934751510620117,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 4.0617,
+      "step": 5
+    },
+    {
+      "epoch": 0.05172413793103448,
+      "grad_norm": 30.031415939331055,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 3.9659,
+      "step": 6
+    },
+    {
+      "epoch": 0.0603448275862069,
+      "grad_norm": 29.62813949584961,
+      "learning_rate": 3.5000000000000004e-07,
+      "loss": 3.9978,
+      "step": 7
+    },
+    {
+      "epoch": 0.06896551724137931,
+      "grad_norm": 31.0523681640625,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 4.0237,
+      "step": 8
+    },
+    {
+      "epoch": 0.07758620689655173,
+      "grad_norm": 28.83756446838379,
+      "learning_rate": 4.5000000000000003e-07,
+      "loss": 3.8214,
+      "step": 9
+    },
+    {
+      "epoch": 0.08620689655172414,
+      "grad_norm": 28.1810359954834,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 3.848,
+      "step": 10
+    },
+    {
+      "epoch": 0.09482758620689655,
+      "grad_norm": 28.367570877075195,
+      "learning_rate": 5.5e-07,
+      "loss": 3.7026,
+      "step": 11
+    },
+    {
+      "epoch": 0.10344827586206896,
+      "grad_norm": 27.66380500793457,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 3.6833,
+      "step": 12
+    },
+    {
+      "epoch": 0.11206896551724138,
+      "grad_norm": 27.496694564819336,
+      "learning_rate": 6.5e-07,
+      "loss": 3.762,
+      "step": 13
+    },
+    {
+      "epoch": 0.1206896551724138,
+      "grad_norm": 28.319055557250977,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 3.8867,
+      "step": 14
+    },
+    {
+      "epoch": 0.12931034482758622,
+      "grad_norm": 26.112581253051758,
+      "learning_rate": 7.5e-07,
+      "loss": 3.6278,
+      "step": 15
+    },
+    {
+      "epoch": 0.13793103448275862,
+      "grad_norm": 25.8245792388916,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 3.7957,
+      "step": 16
+    },
+    {
+      "epoch": 0.14655172413793102,
+      "grad_norm": 23.66245460510254,
+      "learning_rate": 8.500000000000001e-07,
+      "loss": 3.4384,
+      "step": 17
+    },
+    {
+      "epoch": 0.15517241379310345,
+      "grad_norm": 21.69405746459961,
+      "learning_rate": 9.000000000000001e-07,
+      "loss": 3.2602,
+      "step": 18
+    },
+    {
+      "epoch": 0.16379310344827586,
+      "grad_norm": 20.691402435302734,
+      "learning_rate": 9.500000000000001e-07,
+      "loss": 3.2486,
+      "step": 19
+    },
+    {
+      "epoch": 0.1724137931034483,
+      "grad_norm": 20.67167854309082,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 3.2401,
+      "step": 20
+    },
+    {
+      "epoch": 0.1810344827586207,
+      "grad_norm": 19.054428100585938,
+      "learning_rate": 1.0500000000000001e-06,
+      "loss": 3.0375,
+      "step": 21
+    },
+    {
+      "epoch": 0.1896551724137931,
+      "grad_norm": 18.890884399414062,
+      "learning_rate": 1.1e-06,
+      "loss": 2.9677,
+      "step": 22
+    },
+    {
+      "epoch": 0.19827586206896552,
+      "grad_norm": 18.61600112915039,
+      "learning_rate": 1.1500000000000002e-06,
+      "loss": 2.8387,
+      "step": 23
+    },
+    {
+      "epoch": 0.20689655172413793,
+      "grad_norm": 16.910585403442383,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 2.4914,
+      "step": 24
+    },
+    {
+      "epoch": 0.21551724137931033,
+      "grad_norm": 17.708385467529297,
+      "learning_rate": 1.25e-06,
+      "loss": 2.5361,
+      "step": 25
+    },
+    {
+      "epoch": 0.22413793103448276,
+      "grad_norm": 17.07745933532715,
+      "learning_rate": 1.3e-06,
+      "loss": 2.4098,
+      "step": 26
+    },
+    {
+      "epoch": 0.23275862068965517,
+      "grad_norm": 16.244144439697266,
+      "learning_rate": 1.3500000000000002e-06,
+      "loss": 2.2155,
+      "step": 27
+    },
+    {
+      "epoch": 0.2413793103448276,
+      "grad_norm": 16.11887550354004,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 2.0036,
+      "step": 28
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 16.034631729125977,
+      "learning_rate": 1.45e-06,
+      "loss": 1.8617,
+      "step": 29
+    },
+    {
+      "epoch": 0.25862068965517243,
+      "grad_norm": 15.437153816223145,
+      "learning_rate": 1.5e-06,
+      "loss": 1.7413,
+      "step": 30
+    },
+    {
+      "epoch": 0.2672413793103448,
+      "grad_norm": 14.188116073608398,
+      "learning_rate": 1.5500000000000002e-06,
+      "loss": 1.5002,
+      "step": 31
+    },
+    {
+      "epoch": 0.27586206896551724,
+      "grad_norm": 14.200998306274414,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 1.4469,
+      "step": 32
+    },
+    {
+      "epoch": 0.28448275862068967,
+      "grad_norm": 13.209551811218262,
+      "learning_rate": 1.6500000000000003e-06,
+      "loss": 1.3016,
+      "step": 33
+    },
+    {
+      "epoch": 0.29310344827586204,
+      "grad_norm": 12.631085395812988,
+      "learning_rate": 1.7000000000000002e-06,
+      "loss": 1.1522,
+      "step": 34
+    },
+    {
+      "epoch": 0.3017241379310345,
+      "grad_norm": 12.504134178161621,
+      "learning_rate": 1.75e-06,
+      "loss": 1.0058,
+      "step": 35
+    },
+    {
+      "epoch": 0.3103448275862069,
+      "grad_norm": 12.599784851074219,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 0.921,
+      "step": 36
+    },
+    {
+      "epoch": 0.31896551724137934,
+      "grad_norm": 11.577665328979492,
+      "learning_rate": 1.85e-06,
+      "loss": 0.8099,
+      "step": 37
+    },
+    {
+      "epoch": 0.3275862068965517,
+      "grad_norm": 10.465872764587402,
+      "learning_rate": 1.9000000000000002e-06,
+      "loss": 0.6565,
+      "step": 38
+    },
+    {
+      "epoch": 0.33620689655172414,
+      "grad_norm": 10.161813735961914,
+      "learning_rate": 1.9500000000000004e-06,
+      "loss": 0.5712,
+      "step": 39
+    },
+    {
+      "epoch": 0.3448275862068966,
+      "grad_norm": 8.384145736694336,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.4265,
+      "step": 40
+    },
+    {
+      "epoch": 0.35344827586206895,
+      "grad_norm": 7.072062015533447,
+      "learning_rate": 2.05e-06,
+      "loss": 0.409,
+      "step": 41
+    },
+    {
+      "epoch": 0.3620689655172414,
+      "grad_norm": 6.072140693664551,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 0.3263,
+      "step": 42
+    },
+    {
+      "epoch": 0.3706896551724138,
+      "grad_norm": 6.31119441986084,
+      "learning_rate": 2.15e-06,
+      "loss": 0.258,
+      "step": 43
+    },
+    {
+      "epoch": 0.3793103448275862,
+      "grad_norm": 6.619389057159424,
+      "learning_rate": 2.2e-06,
+      "loss": 0.2454,
+      "step": 44
+    },
+    {
+      "epoch": 0.3879310344827586,
+      "grad_norm": 6.546375751495361,
+      "learning_rate": 2.25e-06,
+      "loss": 0.2314,
+      "step": 45
+    },
+    {
+      "epoch": 0.39655172413793105,
+      "grad_norm": 5.219631671905518,
+      "learning_rate": 2.3000000000000004e-06,
+      "loss": 0.2006,
+      "step": 46
+    },
+    {
+      "epoch": 0.4051724137931034,
+      "grad_norm": 3.1164474487304688,
+      "learning_rate": 2.35e-06,
+      "loss": 0.1677,
+      "step": 47
+    },
+    {
+      "epoch": 0.41379310344827586,
+      "grad_norm": 1.872147798538208,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 0.1339,
+      "step": 48
+    },
+    {
+      "epoch": 0.4224137931034483,
+      "grad_norm": 1.4775545597076416,
+      "learning_rate": 2.4500000000000003e-06,
+      "loss": 0.1218,
+      "step": 49
+    },
+    {
+      "epoch": 0.43103448275862066,
+      "grad_norm": 1.2931559085845947,
+      "learning_rate": 2.5e-06,
+      "loss": 0.1151,
+      "step": 50
+    },
+    {
+      "epoch": 0.4396551724137931,
+      "grad_norm": 1.3222297430038452,
+      "learning_rate": 2.55e-06,
+      "loss": 0.1164,
+      "step": 51
+    },
+    {
+      "epoch": 0.4482758620689655,
+      "grad_norm": 1.393062710762024,
+      "learning_rate": 2.6e-06,
+      "loss": 0.1053,
+      "step": 52
+    },
+    {
+      "epoch": 0.45689655172413796,
+      "grad_norm": 1.0838805437088013,
+      "learning_rate": 2.6500000000000005e-06,
+      "loss": 0.105,
+      "step": 53
+    },
+    {
+      "epoch": 0.46551724137931033,
+      "grad_norm": 0.954925537109375,
+      "learning_rate": 2.7000000000000004e-06,
+      "loss": 0.0922,
+      "step": 54
+    },
+    {
+      "epoch": 0.47413793103448276,
+      "grad_norm": 0.7521713972091675,
+      "learning_rate": 2.7500000000000004e-06,
+      "loss": 0.0887,
+      "step": 55
+    },
+    {
+      "epoch": 0.4827586206896552,
+      "grad_norm": 0.7261010408401489,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 0.0997,
+      "step": 56
+    },
+    {
+      "epoch": 0.49137931034482757,
+      "grad_norm": 0.6530802845954895,
+      "learning_rate": 2.85e-06,
+      "loss": 0.0882,
+      "step": 57
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.6899245381355286,
+      "learning_rate": 2.9e-06,
+      "loss": 0.0844,
+      "step": 58
+    },
+    {
+      "epoch": 0.5086206896551724,
+      "grad_norm": 0.6771528124809265,
+      "learning_rate": 2.95e-06,
+      "loss": 0.0884,
+      "step": 59
+    },
+    {
+      "epoch": 0.5172413793103449,
+      "grad_norm": 0.6307985782623291,
+      "learning_rate": 3e-06,
+      "loss": 0.0912,
+      "step": 60
+    },
+    {
+      "epoch": 0.5258620689655172,
+      "grad_norm": 0.5869951844215393,
+      "learning_rate": 3.05e-06,
+      "loss": 0.0875,
+      "step": 61
+    },
+    {
+      "epoch": 0.5344827586206896,
+      "grad_norm": 0.6039404273033142,
+      "learning_rate": 3.1000000000000004e-06,
+      "loss": 0.0768,
+      "step": 62
+    },
+    {
+      "epoch": 0.5431034482758621,
+      "grad_norm": 0.49209344387054443,
+      "learning_rate": 3.1500000000000003e-06,
+      "loss": 0.0815,
+      "step": 63
+    },
+    {
+      "epoch": 0.5517241379310345,
+      "grad_norm": 0.7383344769477844,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 0.0874,
+      "step": 64
+    },
+    {
+      "epoch": 0.5603448275862069,
+      "grad_norm": 0.5552617311477661,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.0762,
+      "step": 65
+    },
+    {
+      "epoch": 0.5689655172413793,
+      "grad_norm": 0.4992441236972809,
+      "learning_rate": 3.3000000000000006e-06,
+      "loss": 0.0793,
+      "step": 66
+    },
+    {
+      "epoch": 0.5775862068965517,
+      "grad_norm": 0.4979636073112488,
+      "learning_rate": 3.3500000000000005e-06,
+      "loss": 0.0792,
+      "step": 67
+    },
+    {
+      "epoch": 0.5862068965517241,
+      "grad_norm": 0.4675934612751007,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 0.0766,
+      "step": 68
+    },
+    {
+      "epoch": 0.5948275862068966,
+      "grad_norm": 0.49495571851730347,
+      "learning_rate": 3.45e-06,
+      "loss": 0.0824,
+      "step": 69
+    },
+    {
+      "epoch": 0.603448275862069,
+      "grad_norm": 0.6044315695762634,
+      "learning_rate": 3.5e-06,
+      "loss": 0.0826,
+      "step": 70
+    },
+    {
+      "epoch": 0.6120689655172413,
+      "grad_norm": 0.4898519217967987,
+      "learning_rate": 3.5500000000000003e-06,
+      "loss": 0.0755,
+      "step": 71
+    },
+    {
+      "epoch": 0.6206896551724138,
+      "grad_norm": 0.4218939244747162,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 0.0762,
+      "step": 72
+    },
+    {
+      "epoch": 0.6293103448275862,
+      "grad_norm": 0.5785802006721497,
+      "learning_rate": 3.65e-06,
+      "loss": 0.0837,
+      "step": 73
+    },
+    {
+      "epoch": 0.6379310344827587,
+      "grad_norm": 0.5505399703979492,
+      "learning_rate": 3.7e-06,
+      "loss": 0.0814,
+      "step": 74
+    },
+    {
+      "epoch": 0.646551724137931,
+      "grad_norm": 0.6062561869621277,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.0725,
+      "step": 75
+    },
+    {
+      "epoch": 0.6551724137931034,
+      "grad_norm": 0.707350492477417,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.0913,
+      "step": 76
+    },
+    {
+      "epoch": 0.6637931034482759,
+      "grad_norm": 0.5091889500617981,
+      "learning_rate": 3.85e-06,
+      "loss": 0.0697,
+      "step": 77
+    },
+    {
+      "epoch": 0.6724137931034483,
+      "grad_norm": 0.4801473319530487,
+      "learning_rate": 3.900000000000001e-06,
+      "loss": 0.073,
+      "step": 78
+    },
+    {
+      "epoch": 0.6810344827586207,
+      "grad_norm": 0.462162047624588,
+      "learning_rate": 3.95e-06,
+      "loss": 0.0805,
+      "step": 79
+    },
+    {
+      "epoch": 0.6896551724137931,
+      "grad_norm": 0.463969349861145,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.0738,
+      "step": 80
+    },
+    {
+      "epoch": 0.6982758620689655,
+      "grad_norm": 0.3980114161968231,
+      "learning_rate": 4.05e-06,
+      "loss": 0.0742,
+      "step": 81
+    },
+    {
+      "epoch": 0.7068965517241379,
+      "grad_norm": 0.3627180755138397,
+      "learning_rate": 4.1e-06,
+      "loss": 0.0683,
+      "step": 82
+    },
+    {
+      "epoch": 0.7155172413793104,
+      "grad_norm": 0.39726322889328003,
+      "learning_rate": 4.15e-06,
+      "loss": 0.0717,
+      "step": 83
+    },
+    {
+      "epoch": 0.7241379310344828,
+      "grad_norm": 0.48898085951805115,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 0.0676,
+      "step": 84
+    },
+    {
+      "epoch": 0.7327586206896551,
+      "grad_norm": 0.6143100261688232,
+      "learning_rate": 4.25e-06,
+      "loss": 0.0708,
+      "step": 85
+    },
+    {
+      "epoch": 0.7413793103448276,
+      "grad_norm": 0.5028995275497437,
+      "learning_rate": 4.3e-06,
+      "loss": 0.0713,
+      "step": 86
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.5576066970825195,
+      "learning_rate": 4.350000000000001e-06,
+      "loss": 0.078,
+      "step": 87
+    },
+    {
+      "epoch": 0.7586206896551724,
+      "grad_norm": 0.37101301550865173,
+      "learning_rate": 4.4e-06,
+      "loss": 0.071,
+      "step": 88
+    },
+    {
+      "epoch": 0.7672413793103449,
+      "grad_norm": 0.442694753408432,
+      "learning_rate": 4.450000000000001e-06,
+      "loss": 0.0791,
+      "step": 89
+    },
+    {
+      "epoch": 0.7758620689655172,
+      "grad_norm": 0.48991039395332336,
+      "learning_rate": 4.5e-06,
+      "loss": 0.0681,
+      "step": 90
+    },
+    {
+      "epoch": 0.7844827586206896,
+      "grad_norm": 0.46367791295051575,
+      "learning_rate": 4.5500000000000005e-06,
+      "loss": 0.0701,
+      "step": 91
+    },
+    {
+      "epoch": 0.7931034482758621,
+      "grad_norm": 0.3454825282096863,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 0.0774,
+      "step": 92
+    },
+    {
+      "epoch": 0.8017241379310345,
+      "grad_norm": 0.35102447867393494,
+      "learning_rate": 4.65e-06,
+      "loss": 0.0659,
+      "step": 93
+    },
+    {
+      "epoch": 0.8103448275862069,
+      "grad_norm": 0.3791246712207794,
+      "learning_rate": 4.7e-06,
+      "loss": 0.0727,
+      "step": 94
+    },
+    {
+      "epoch": 0.8189655172413793,
+      "grad_norm": 0.3911365270614624,
+      "learning_rate": 4.75e-06,
+      "loss": 0.0641,
+      "step": 95
+    },
+    {
+      "epoch": 0.8275862068965517,
+      "grad_norm": 0.35395047068595886,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.0735,
+      "step": 96
+    },
+    {
+      "epoch": 0.8362068965517241,
+      "grad_norm": 0.3499661087989807,
+      "learning_rate": 4.85e-06,
+      "loss": 0.0657,
+      "step": 97
+    },
+    {
+      "epoch": 0.8448275862068966,
+      "grad_norm": 0.34877678751945496,
+      "learning_rate": 4.9000000000000005e-06,
+      "loss": 0.0701,
+      "step": 98
+    },
+    {
+      "epoch": 0.853448275862069,
+      "grad_norm": 0.36436647176742554,
+      "learning_rate": 4.95e-06,
+      "loss": 0.0758,
+      "step": 99
+    },
+    {
+      "epoch": 0.8620689655172413,
+      "grad_norm": 0.4726298749446869,
+      "learning_rate": 5e-06,
+      "loss": 0.0731,
+      "step": 100
+    },
+    {
+      "epoch": 0.8706896551724138,
+      "grad_norm": 0.31892526149749756,
+      "learning_rate": 4.999965269084342e-06,
+      "loss": 0.0635,
+      "step": 101
+    },
+    {
+      "epoch": 0.8793103448275862,
+      "grad_norm": 0.3707881569862366,
+      "learning_rate": 4.999861077302358e-06,
+      "loss": 0.0689,
+      "step": 102
+    },
+    {
+      "epoch": 0.8879310344827587,
+      "grad_norm": 0.3244622051715851,
+      "learning_rate": 4.999687427548989e-06,
+      "loss": 0.0645,
+      "step": 103
+    },
+    {
+      "epoch": 0.896551724137931,
+      "grad_norm": 0.386982262134552,
+      "learning_rate": 4.999444324649045e-06,
+      "loss": 0.0708,
+      "step": 104
+    },
+    {
+      "epoch": 0.9051724137931034,
+      "grad_norm": 0.3552643656730652,
+      "learning_rate": 4.999131775357078e-06,
+      "loss": 0.0766,
+      "step": 105
+    },
+    {
+      "epoch": 0.9137931034482759,
+      "grad_norm": 0.3503671884536743,
+      "learning_rate": 4.998749788357184e-06,
+      "loss": 0.0733,
+      "step": 106
+    },
+    {
+      "epoch": 0.9224137931034483,
+      "grad_norm": 0.4907379746437073,
+      "learning_rate": 4.998298374262771e-06,
+      "loss": 0.0721,
+      "step": 107
+    },
+    {
+      "epoch": 0.9310344827586207,
+      "grad_norm": 0.305571585893631,
+      "learning_rate": 4.997777545616258e-06,
+      "loss": 0.0603,
+      "step": 108
+    },
+    {
+      "epoch": 0.9396551724137931,
+      "grad_norm": 0.3188783824443817,
+      "learning_rate": 4.99718731688873e-06,
+      "loss": 0.0574,
+      "step": 109
+    },
+    {
+      "epoch": 0.9482758620689655,
+      "grad_norm": 0.3277616500854492,
+      "learning_rate": 4.996527704479535e-06,
+      "loss": 0.0671,
+      "step": 110
+    },
+    {
+      "epoch": 0.9568965517241379,
+      "grad_norm": 0.38640740513801575,
+      "learning_rate": 4.995798726715826e-06,
+      "loss": 0.0625,
+      "step": 111
+    },
+    {
+      "epoch": 0.9655172413793104,
+      "grad_norm": 0.5463483929634094,
+      "learning_rate": 4.995000403852057e-06,
+      "loss": 0.0718,
+      "step": 112
+    },
+    {
+      "epoch": 0.9741379310344828,
+      "grad_norm": 0.3974014222621918,
+      "learning_rate": 4.994132758069413e-06,
+      "loss": 0.071,
+      "step": 113
+    },
+    {
+      "epoch": 0.9827586206896551,
+      "grad_norm": 0.3642785847187042,
+      "learning_rate": 4.993195813475202e-06,
+      "loss": 0.0658,
+      "step": 114
+    },
+    {
+      "epoch": 0.9913793103448276,
+      "grad_norm": 0.3118143677711487,
+      "learning_rate": 4.992189596102179e-06,
+      "loss": 0.0633,
+      "step": 115
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.41573596000671387,
+      "learning_rate": 4.991114133907822e-06,
+      "loss": 0.0662,
+      "step": 116
+    },
+    {
+      "epoch": 1.0086206896551724,
+      "grad_norm": 0.29761484265327454,
+      "learning_rate": 4.989969456773562e-06,
+      "loss": 0.0649,
+      "step": 117
+    },
+    {
+      "epoch": 1.0172413793103448,
+      "grad_norm": 0.399275004863739,
+      "learning_rate": 4.988755596503948e-06,
+      "loss": 0.0521,
+      "step": 118
+    },
+    {
+      "epoch": 1.0258620689655173,
+      "grad_norm": 0.37820911407470703,
+      "learning_rate": 4.987472586825762e-06,
+      "loss": 0.0596,
+      "step": 119
+    },
+    {
+      "epoch": 1.0344827586206897,
+      "grad_norm": 0.4371725916862488,
+      "learning_rate": 4.986120463387084e-06,
+      "loss": 0.07,
+      "step": 120
+    },
+    {
+      "epoch": 1.043103448275862,
+      "grad_norm": 0.3165535032749176,
+      "learning_rate": 4.984699263756303e-06,
+      "loss": 0.0613,
+      "step": 121
+    },
+    {
+      "epoch": 1.0517241379310345,
+      "grad_norm": 0.40526920557022095,
+      "learning_rate": 4.983209027421072e-06,
+      "loss": 0.0611,
+      "step": 122
+    },
+    {
+      "epoch": 1.0603448275862069,
+      "grad_norm": 0.39523962140083313,
+      "learning_rate": 4.9816497957872055e-06,
+      "loss": 0.0634,
+      "step": 123
+    },
+    {
+      "epoch": 1.0689655172413792,
+      "grad_norm": 0.33234691619873047,
+      "learning_rate": 4.9800216121775404e-06,
+      "loss": 0.0602,
+      "step": 124
+    },
+    {
+      "epoch": 1.0775862068965518,
+      "grad_norm": 0.3455008268356323,
+      "learning_rate": 4.978324521830721e-06,
+      "loss": 0.0624,
+      "step": 125
+    },
+    {
+      "epoch": 1.0862068965517242,
+      "grad_norm": 0.3968574106693268,
+      "learning_rate": 4.97655857189995e-06,
+      "loss": 0.0525,
+      "step": 126
+    },
+    {
+      "epoch": 1.0948275862068966,
+      "grad_norm": 0.47838935256004333,
+      "learning_rate": 4.974723811451673e-06,
+      "loss": 0.0617,
+      "step": 127
+    },
+    {
+      "epoch": 1.103448275862069,
+      "grad_norm": 0.43055081367492676,
+      "learning_rate": 4.972820291464219e-06,
+      "loss": 0.0484,
+      "step": 128
+    },
+    {
+      "epoch": 1.1120689655172413,
+      "grad_norm": 0.4012243449687958,
+      "learning_rate": 4.97084806482638e-06,
+      "loss": 0.0562,
+      "step": 129
+    },
+    {
+      "epoch": 1.1206896551724137,
+      "grad_norm": 0.586875319480896,
+      "learning_rate": 4.968807186335948e-06,
+      "loss": 0.0633,
+      "step": 130
+    },
+    {
+      "epoch": 1.1293103448275863,
+      "grad_norm": 0.48629340529441833,
+      "learning_rate": 4.966697712698185e-06,
+      "loss": 0.0615,
+      "step": 131
+    },
+    {
+      "epoch": 1.1379310344827587,
+      "grad_norm": 0.41810542345046997,
+      "learning_rate": 4.964519702524251e-06,
+      "loss": 0.0564,
+      "step": 132
+    },
+    {
+      "epoch": 1.146551724137931,
+      "grad_norm": 0.4095732271671295,
+      "learning_rate": 4.962273216329577e-06,
+      "loss": 0.0634,
+      "step": 133
+    },
+    {
+      "epoch": 1.1551724137931034,
+      "grad_norm": 0.3784744441509247,
+      "learning_rate": 4.959958316532181e-06,
+      "loss": 0.0521,
+      "step": 134
+    },
+    {
+      "epoch": 1.1637931034482758,
+      "grad_norm": 0.4087596833705902,
+      "learning_rate": 4.957575067450935e-06,
+      "loss": 0.0569,
+      "step": 135
+    },
+    {
+      "epoch": 1.1724137931034484,
+      "grad_norm": 0.4369741380214691,
+      "learning_rate": 4.955123535303775e-06,
+      "loss": 0.057,
+      "step": 136
+    },
+    {
+      "epoch": 1.1810344827586208,
+      "grad_norm": 0.3851775527000427,
+      "learning_rate": 4.95260378820587e-06,
+      "loss": 0.0521,
+      "step": 137
+    },
+    {
+      "epoch": 1.1896551724137931,
+      "grad_norm": 0.3391599655151367,
+      "learning_rate": 4.950015896167716e-06,
+      "loss": 0.0525,
+      "step": 138
+    },
+    {
+      "epoch": 1.1982758620689655,
+      "grad_norm": 0.4176456332206726,
+      "learning_rate": 4.947359931093202e-06,
+      "loss": 0.0548,
+      "step": 139
+    },
+    {
+      "epoch": 1.206896551724138,
+      "grad_norm": 0.45864036679267883,
+      "learning_rate": 4.944635966777607e-06,
+      "loss": 0.0578,
+      "step": 140
+    },
+    {
+      "epoch": 1.2155172413793103,
+      "grad_norm": 0.4484093487262726,
+      "learning_rate": 4.941844078905551e-06,
+      "loss": 0.0528,
+      "step": 141
+    },
+    {
+      "epoch": 1.2241379310344827,
+      "grad_norm": 0.42641302943229675,
+      "learning_rate": 4.938984345048892e-06,
+      "loss": 0.0519,
+      "step": 142
+    },
+    {
+      "epoch": 1.2327586206896552,
+      "grad_norm": 0.5111292004585266,
+      "learning_rate": 4.936056844664571e-06,
+      "loss": 0.0639,
+      "step": 143
+    },
+    {
+      "epoch": 1.2413793103448276,
+      "grad_norm": 0.3739636540412903,
+      "learning_rate": 4.933061659092401e-06,
+      "loss": 0.055,
+      "step": 144
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 0.3896700143814087,
+      "learning_rate": 4.929998871552814e-06,
+      "loss": 0.0548,
+      "step": 145
+    },
+    {
+      "epoch": 1.2586206896551724,
+      "grad_norm": 0.4158782660961151,
+      "learning_rate": 4.926868567144543e-06,
+      "loss": 0.0572,
+      "step": 146
+    },
+    {
+      "epoch": 1.2672413793103448,
+      "grad_norm": 0.4275132715702057,
+      "learning_rate": 4.923670832842256e-06,
+      "loss": 0.0497,
+      "step": 147
+    },
+    {
+      "epoch": 1.2758620689655173,
+      "grad_norm": 0.3890449106693268,
+      "learning_rate": 4.920405757494147e-06,
+      "loss": 0.0584,
+      "step": 148
+    },
+    {
+      "epoch": 1.2844827586206897,
+      "grad_norm": 0.4569760262966156,
+      "learning_rate": 4.917073431819462e-06,
+      "loss": 0.0618,
+      "step": 149
+    },
+    {
+      "epoch": 1.293103448275862,
+      "grad_norm": 0.4232596457004547,
+      "learning_rate": 4.913673948405977e-06,
+      "loss": 0.0516,
+      "step": 150
+    },
+    {
+      "epoch": 1.3017241379310345,
+      "grad_norm": 0.4578849673271179,
+      "learning_rate": 4.910207401707431e-06,
+      "loss": 0.0538,
+      "step": 151
+    },
+    {
+      "epoch": 1.3103448275862069,
+      "grad_norm": 0.39952200651168823,
+      "learning_rate": 4.906673888040895e-06,
+      "loss": 0.0471,
+      "step": 152
+    },
+    {
+      "epoch": 1.3189655172413794,
+      "grad_norm": 0.38846731185913086,
+      "learning_rate": 4.903073505584102e-06,
+      "loss": 0.0437,
+      "step": 153
+    },
+    {
+      "epoch": 1.3275862068965516,
+      "grad_norm": 0.4084296226501465,
+      "learning_rate": 4.899406354372716e-06,
+      "loss": 0.0532,
+      "step": 154
+    },
+    {
+      "epoch": 1.3362068965517242,
+      "grad_norm": 0.4114966094493866,
+      "learning_rate": 4.895672536297551e-06,
+      "loss": 0.0443,
+      "step": 155
+    },
+    {
+      "epoch": 1.3448275862068966,
+      "grad_norm": 0.6465100049972534,
+      "learning_rate": 4.891872155101746e-06,
+      "loss": 0.0605,
+      "step": 156
+    },
+    {
+      "epoch": 1.353448275862069,
+      "grad_norm": 0.43234798312187195,
+      "learning_rate": 4.888005316377873e-06,
+      "loss": 0.0529,
+      "step": 157
+    },
+    {
+      "epoch": 1.3620689655172413,
+      "grad_norm": 0.408527135848999,
+      "learning_rate": 4.884072127565015e-06,
+      "loss": 0.0485,
+      "step": 158
+    },
+    {
+      "epoch": 1.3706896551724137,
+      "grad_norm": 0.45152318477630615,
+      "learning_rate": 4.880072697945768e-06,
+      "loss": 0.0466,
+      "step": 159
+    },
+    {
+      "epoch": 1.3793103448275863,
+      "grad_norm": 0.4669966995716095,
+      "learning_rate": 4.876007138643216e-06,
+      "loss": 0.0438,
+      "step": 160
+    },
+    {
+      "epoch": 1.3879310344827587,
+      "grad_norm": 0.47117388248443604,
+      "learning_rate": 4.871875562617837e-06,
+      "loss": 0.0464,
+      "step": 161
+    },
+    {
+      "epoch": 1.396551724137931,
+      "grad_norm": 0.6085790991783142,
+      "learning_rate": 4.867678084664365e-06,
+      "loss": 0.0574,
+      "step": 162
+    },
+    {
+      "epoch": 1.4051724137931034,
+      "grad_norm": 0.5736953616142273,
+      "learning_rate": 4.863414821408602e-06,
+      "loss": 0.0507,
+      "step": 163
+    },
+    {
+      "epoch": 1.4137931034482758,
+      "grad_norm": 0.5003281831741333,
+      "learning_rate": 4.8590858913041775e-06,
+      "loss": 0.051,
+      "step": 164
+    },
+    {
+      "epoch": 1.4224137931034484,
+      "grad_norm": 0.4361717104911804,
+      "learning_rate": 4.854691414629258e-06,
+      "loss": 0.0476,
+      "step": 165
+    },
+    {
+      "epoch": 1.4310344827586206,
+      "grad_norm": 0.47907084226608276,
+      "learning_rate": 4.8502315134832e-06,
+      "loss": 0.0495,
+      "step": 166
+    },
+    {
+      "epoch": 1.4396551724137931,
+      "grad_norm": 0.46805888414382935,
+      "learning_rate": 4.8457063117831656e-06,
+      "loss": 0.0474,
+      "step": 167
+    },
+    {
+      "epoch": 1.4482758620689655,
+      "grad_norm": 0.5058819651603699,
+      "learning_rate": 4.8411159352606735e-06,
+      "loss": 0.0541,
+      "step": 168
+    },
+    {
+      "epoch": 1.456896551724138,
+      "grad_norm": 0.4632813334465027,
+      "learning_rate": 4.836460511458107e-06,
+      "loss": 0.0409,
+      "step": 169
+    },
+    {
+      "epoch": 1.4655172413793103,
+      "grad_norm": 0.3654536008834839,
+      "learning_rate": 4.831740169725172e-06,
+      "loss": 0.0466,
+      "step": 170
+    },
+    {
+      "epoch": 1.4741379310344827,
+      "grad_norm": 0.4301295578479767,
+      "learning_rate": 4.8269550412153e-06,
+      "loss": 0.0472,
+      "step": 171
+    },
+    {
+      "epoch": 1.4827586206896552,
+      "grad_norm": 0.37643712759017944,
+      "learning_rate": 4.822105258882007e-06,
+      "loss": 0.0445,
+      "step": 172
+    },
+    {
+      "epoch": 1.4913793103448276,
+      "grad_norm": 0.3928871154785156,
+      "learning_rate": 4.817190957475199e-06,
+      "loss": 0.0427,
+      "step": 173
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 0.39636102318763733,
+      "learning_rate": 4.812212273537426e-06,
+      "loss": 0.0431,
+      "step": 174
+    },
+    {
+      "epoch": 1.5086206896551724,
+      "grad_norm": 0.41640329360961914,
+      "learning_rate": 4.807169345400088e-06,
+      "loss": 0.0442,
+      "step": 175
+    },
+    {
+      "epoch": 1.5172413793103448,
+      "grad_norm": 0.4888840317726135,
+      "learning_rate": 4.802062313179595e-06,
+      "loss": 0.0529,
+      "step": 176
+    },
+    {
+      "epoch": 1.5258620689655173,
+      "grad_norm": 0.4154113829135895,
+      "learning_rate": 4.796891318773472e-06,
+      "loss": 0.0419,
+      "step": 177
+    },
+    {
+      "epoch": 1.5344827586206895,
+      "grad_norm": 0.41872596740722656,
+      "learning_rate": 4.791656505856416e-06,
+      "loss": 0.0411,
+      "step": 178
+    },
+    {
+      "epoch": 1.543103448275862,
+      "grad_norm": 0.4037671983242035,
+      "learning_rate": 4.786358019876301e-06,
+      "loss": 0.0458,
+      "step": 179
+    },
+    {
+      "epoch": 1.5517241379310345,
+      "grad_norm": 0.4356297552585602,
+      "learning_rate": 4.7809960080501464e-06,
+      "loss": 0.0495,
+      "step": 180
+    },
+    {
+      "epoch": 1.5603448275862069,
+      "grad_norm": 0.4112146496772766,
+      "learning_rate": 4.7755706193600135e-06,
+      "loss": 0.0367,
+      "step": 181
+    },
+    {
+      "epoch": 1.5689655172413794,
+      "grad_norm": 0.36699458956718445,
+      "learning_rate": 4.770082004548878e-06,
+      "loss": 0.0427,
+      "step": 182
+    },
+    {
+      "epoch": 1.5775862068965516,
+      "grad_norm": 0.49518612027168274,
+      "learning_rate": 4.764530316116433e-06,
+      "loss": 0.0334,
+      "step": 183
+    },
+    {
+      "epoch": 1.5862068965517242,
+      "grad_norm": 0.3718539774417877,
+      "learning_rate": 4.758915708314858e-06,
+      "loss": 0.0367,
+      "step": 184
+    },
+    {
+      "epoch": 1.5948275862068966,
+      "grad_norm": 0.46727338433265686,
+      "learning_rate": 4.753238337144528e-06,
+      "loss": 0.044,
+      "step": 185
+    },
+    {
+      "epoch": 1.603448275862069,
+      "grad_norm": 0.5345709323883057,
+      "learning_rate": 4.747498360349681e-06,
+      "loss": 0.0401,
+      "step": 186
+    },
+    {
+      "epoch": 1.6120689655172413,
+      "grad_norm": 0.561375617980957,
+      "learning_rate": 4.7416959374140405e-06,
+      "loss": 0.0457,
+      "step": 187
+    },
+    {
+      "epoch": 1.6206896551724137,
+      "grad_norm": 0.4418468773365021,
+      "learning_rate": 4.735831229556374e-06,
+      "loss": 0.0348,
+      "step": 188
+    },
+    {
+      "epoch": 1.6293103448275863,
+      "grad_norm": 0.5142704844474792,
+      "learning_rate": 4.72990439972602e-06,
+      "loss": 0.0391,
+      "step": 189
+    },
+    {
+      "epoch": 1.6379310344827587,
+      "grad_norm": 0.48653194308280945,
+      "learning_rate": 4.72391561259836e-06,
+      "loss": 0.0358,
+      "step": 190
+    },
+    {
+      "epoch": 1.646551724137931,
+      "grad_norm": 0.48498624563217163,
+      "learning_rate": 4.717865034570243e-06,
+      "loss": 0.038,
+      "step": 191
+    },
+    {
+      "epoch": 1.6551724137931034,
+      "grad_norm": 0.4586655795574188,
+      "learning_rate": 4.711752833755362e-06,
+      "loss": 0.0351,
+      "step": 192
+    },
+    {
+      "epoch": 1.6637931034482758,
+      "grad_norm": 0.44023749232292175,
+      "learning_rate": 4.70557917997958e-06,
+      "loss": 0.0277,
+      "step": 193
+    },
+    {
+      "epoch": 1.6724137931034484,
+      "grad_norm": 0.44018658995628357,
+      "learning_rate": 4.6993442447762185e-06,
+      "loss": 0.0382,
+      "step": 194
+    },
+    {
+      "epoch": 1.6810344827586206,
+      "grad_norm": 0.49303099513053894,
+      "learning_rate": 4.693048201381281e-06,
+      "loss": 0.0413,
+      "step": 195
+    },
+    {
+      "epoch": 1.6896551724137931,
+      "grad_norm": 0.572812557220459,
+      "learning_rate": 4.686691224728652e-06,
+      "loss": 0.0389,
+      "step": 196
+    },
+    {
+      "epoch": 1.6982758620689655,
+      "grad_norm": 0.4995497167110443,
+      "learning_rate": 4.680273491445227e-06,
+      "loss": 0.0314,
+      "step": 197
+    },
+    {
+      "epoch": 1.706896551724138,
+      "grad_norm": 0.465404212474823,
+      "learning_rate": 4.673795179846008e-06,
+      "loss": 0.0323,
+      "step": 198
+    },
+    {
+      "epoch": 1.7155172413793105,
+      "grad_norm": 0.45183637738227844,
+      "learning_rate": 4.667256469929149e-06,
+      "loss": 0.0308,
+      "step": 199
+    },
+    {
+      "epoch": 1.7241379310344827,
+      "grad_norm": 0.4891216456890106,
+      "learning_rate": 4.660657543370958e-06,
+      "loss": 0.034,
+      "step": 200
+    },
+    {
+      "epoch": 1.7327586206896552,
+      "grad_norm": 0.5728661417961121,
+      "learning_rate": 4.653998583520844e-06,
+      "loss": 0.0359,
+      "step": 201
+    },
+    {
+      "epoch": 1.7413793103448276,
+      "grad_norm": 0.5465419888496399,
+      "learning_rate": 4.6472797753962255e-06,
+      "loss": 0.0366,
+      "step": 202
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 0.4671058654785156,
+      "learning_rate": 4.640501305677387e-06,
+      "loss": 0.0345,
+      "step": 203
+    },
+    {
+      "epoch": 1.7586206896551724,
+      "grad_norm": 0.4623485803604126,
+      "learning_rate": 4.6336633627023e-06,
+      "loss": 0.0261,
+      "step": 204
+    },
+    {
+      "epoch": 1.7672413793103448,
+      "grad_norm": 0.4471631944179535,
+      "learning_rate": 4.626766136461378e-06,
+      "loss": 0.0244,
+      "step": 205
+    },
+    {
+      "epoch": 1.7758620689655173,
+      "grad_norm": 0.46128782629966736,
+      "learning_rate": 4.61980981859221e-06,
+      "loss": 0.0301,
+      "step": 206
+    },
+    {
+      "epoch": 1.7844827586206895,
+      "grad_norm": 0.45246005058288574,
+      "learning_rate": 4.612794602374226e-06,
+      "loss": 0.0274,
+      "step": 207
+    },
+    {
+      "epoch": 1.793103448275862,
+      "grad_norm": 0.5971981287002563,
+      "learning_rate": 4.605720682723331e-06,
+      "loss": 0.0387,
+      "step": 208
+    },
+    {
+      "epoch": 1.8017241379310345,
+      "grad_norm": 0.4932047128677368,
+      "learning_rate": 4.598588256186491e-06,
+      "loss": 0.0228,
+      "step": 209
+    },
+    {
+      "epoch": 1.8103448275862069,
+      "grad_norm": 0.6208699345588684,
+      "learning_rate": 4.591397520936271e-06,
+      "loss": 0.0307,
+      "step": 210
+    },
+    {
+      "epoch": 1.8189655172413794,
+      "grad_norm": 0.5628708004951477,
+      "learning_rate": 4.584148676765327e-06,
+      "loss": 0.0262,
+      "step": 211
+    },
+    {
+      "epoch": 1.8275862068965516,
+      "grad_norm": 0.6893038749694824,
+      "learning_rate": 4.576841925080853e-06,
+      "loss": 0.0312,
+      "step": 212
+    },
+    {
+      "epoch": 1.8362068965517242,
+      "grad_norm": 0.5523396134376526,
+      "learning_rate": 4.569477468898992e-06,
+      "loss": 0.0218,
+      "step": 213
+    },
+    {
+      "epoch": 1.8448275862068966,
+      "grad_norm": 0.6797971129417419,
+      "learning_rate": 4.562055512839189e-06,
+      "loss": 0.0232,
+      "step": 214
+    },
+    {
+      "epoch": 1.853448275862069,
+      "grad_norm": 0.7214551568031311,
+      "learning_rate": 4.554576263118506e-06,
+      "loss": 0.0326,
+      "step": 215
+    },
+    {
+      "epoch": 1.8620689655172413,
+      "grad_norm": 0.5208979249000549,
+      "learning_rate": 4.547039927545899e-06,
+      "loss": 0.0261,
+      "step": 216
+    },
+    {
+      "epoch": 1.8706896551724137,
+      "grad_norm": 0.5396261215209961,
+      "learning_rate": 4.539446715516434e-06,
+      "loss": 0.0269,
+      "step": 217
+    },
+    {
+      "epoch": 1.8793103448275863,
+      "grad_norm": 0.5789651274681091,
+      "learning_rate": 4.531796838005477e-06,
+      "loss": 0.035,
+      "step": 218
+    },
+    {
+      "epoch": 1.8879310344827587,
+      "grad_norm": 0.5388063192367554,
+      "learning_rate": 4.524090507562828e-06,
+      "loss": 0.0277,
+      "step": 219
+    },
+    {
+      "epoch": 1.896551724137931,
+      "grad_norm": 0.4727953374385834,
+      "learning_rate": 4.516327938306818e-06,
+      "loss": 0.0225,
+      "step": 220
+    },
+    {
+      "epoch": 1.9051724137931034,
+      "grad_norm": 0.4665185809135437,
+      "learning_rate": 4.508509345918357e-06,
+      "loss": 0.0268,
+      "step": 221
+    },
+    {
+      "epoch": 1.9137931034482758,
+      "grad_norm": 0.5534846186637878,
+      "learning_rate": 4.500634947634943e-06,
+      "loss": 0.0277,
+      "step": 222
+    },
+    {
+      "epoch": 1.9224137931034484,
+      "grad_norm": 0.46098095178604126,
+      "learning_rate": 4.492704962244626e-06,
+      "loss": 0.0274,
+      "step": 223
+    },
+    {
+      "epoch": 1.9310344827586206,
+      "grad_norm": 0.49379560351371765,
+      "learning_rate": 4.4847196100799305e-06,
+      "loss": 0.0252,
+      "step": 224
+    },
+    {
+      "epoch": 1.9396551724137931,
+      "grad_norm": 0.4714183807373047,
+      "learning_rate": 4.476679113011729e-06,
+      "loss": 0.0253,
+      "step": 225
+    },
+    {
+      "epoch": 1.9482758620689655,
+      "grad_norm": 0.6230431199073792,
+      "learning_rate": 4.4685836944430815e-06,
+      "loss": 0.0242,
+      "step": 226
+    },
+    {
+      "epoch": 1.956896551724138,
+      "grad_norm": 0.42832595109939575,
+      "learning_rate": 4.46043357930303e-06,
+      "loss": 0.0165,
+      "step": 227
+    },
+    {
+      "epoch": 1.9655172413793105,
+      "grad_norm": 0.38421332836151123,
+      "learning_rate": 4.452228994040341e-06,
+      "loss": 0.0168,
+      "step": 228
+    },
+    {
+      "epoch": 1.9741379310344827,
+      "grad_norm": 0.5493133664131165,
+      "learning_rate": 4.443970166617223e-06,
+      "loss": 0.0193,
+      "step": 229
+    },
+    {
+      "epoch": 1.9827586206896552,
+      "grad_norm": 0.45186540484428406,
+      "learning_rate": 4.435657326502986e-06,
+      "loss": 0.0165,
+      "step": 230
+    },
+    {
+      "epoch": 1.9913793103448276,
+      "grad_norm": 0.8328711986541748,
+      "learning_rate": 4.4272907046676704e-06,
+      "loss": 0.025,
+      "step": 231
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.5378574728965759,
+      "learning_rate": 4.418870533575626e-06,
+      "loss": 0.0166,
+      "step": 232
+    },
+    {
+      "epoch": 2.0086206896551726,
+      "grad_norm": 0.5182176828384399,
+      "learning_rate": 4.410397047179053e-06,
+      "loss": 0.011,
+      "step": 233
+    },
+    {
+      "epoch": 2.0172413793103448,
+      "grad_norm": 0.5283203125,
+      "learning_rate": 4.401870480911505e-06,
+      "loss": 0.0178,
+      "step": 234
+    },
+    {
+      "epoch": 2.0258620689655173,
+      "grad_norm": 0.3926998972892761,
+      "learning_rate": 4.393291071681345e-06,
+      "loss": 0.0121,
+      "step": 235
+    },
+    {
+      "epoch": 2.0344827586206895,
+      "grad_norm": 0.4371504783630371,
+      "learning_rate": 4.384659057865165e-06,
+      "loss": 0.0164,
+      "step": 236
+    },
+    {
+      "epoch": 2.043103448275862,
+      "grad_norm": 0.6982957720756531,
+      "learning_rate": 4.375974679301158e-06,
+      "loss": 0.0192,
+      "step": 237
+    },
+    {
+      "epoch": 2.0517241379310347,
+      "grad_norm": 0.3988925814628601,
+      "learning_rate": 4.367238177282462e-06,
+      "loss": 0.0139,
+      "step": 238
+    },
+    {
+      "epoch": 2.060344827586207,
+      "grad_norm": 0.5101247429847717,
+      "learning_rate": 4.3584497945504465e-06,
+      "loss": 0.0118,
+      "step": 239
+    },
+    {
+      "epoch": 2.0689655172413794,
+      "grad_norm": 0.609987735748291,
+      "learning_rate": 4.349609775287977e-06,
+      "loss": 0.0168,
+      "step": 240
+    },
+    {
+      "epoch": 2.0775862068965516,
+      "grad_norm": 0.37872084975242615,
+      "learning_rate": 4.340718365112623e-06,
+      "loss": 0.009,
+      "step": 241
+    },
+    {
+      "epoch": 2.086206896551724,
+      "grad_norm": 0.5639368891716003,
+      "learning_rate": 4.331775811069837e-06,
+      "loss": 0.0154,
+      "step": 242
+    },
+    {
+      "epoch": 2.0948275862068964,
+      "grad_norm": 0.6034789681434631,
+      "learning_rate": 4.322782361626094e-06,
+      "loss": 0.0174,
+      "step": 243
+    },
+    {
+      "epoch": 2.103448275862069,
+      "grad_norm": 0.46892449259757996,
+      "learning_rate": 4.313738266661979e-06,
+      "loss": 0.0128,
+      "step": 244
+    },
+    {
+      "epoch": 2.1120689655172415,
+      "grad_norm": 0.5814667344093323,
+      "learning_rate": 4.3046437774652525e-06,
+      "loss": 0.0129,
+      "step": 245
+    },
+    {
+      "epoch": 2.1206896551724137,
+      "grad_norm": 0.39165839552879333,
+      "learning_rate": 4.295499146723864e-06,
+      "loss": 0.0133,
+      "step": 246
+    },
+    {
+      "epoch": 2.1293103448275863,
+      "grad_norm": 0.45224517583847046,
+      "learning_rate": 4.286304628518932e-06,
+      "loss": 0.0093,
+      "step": 247
+    },
+    {
+      "epoch": 2.1379310344827585,
+      "grad_norm": 0.7525523900985718,
+      "learning_rate": 4.277060478317687e-06,
+      "loss": 0.0159,
+      "step": 248
+    },
+    {
+      "epoch": 2.146551724137931,
+      "grad_norm": 0.6550068259239197,
+      "learning_rate": 4.267766952966369e-06,
+      "loss": 0.0126,
+      "step": 249
+    },
+    {
+      "epoch": 2.1551724137931036,
+      "grad_norm": 0.5949872136116028,
+      "learning_rate": 4.258424310683094e-06,
+      "loss": 0.01,
+      "step": 250
+    },
+    {
+      "epoch": 2.163793103448276,
+      "grad_norm": 0.44155633449554443,
+      "learning_rate": 4.24903281105068e-06,
+      "loss": 0.0113,
+      "step": 251
+    },
+    {
+      "epoch": 2.1724137931034484,
+      "grad_norm": 0.7045619487762451,
+      "learning_rate": 4.23959271500943e-06,
+      "loss": 0.0108,
+      "step": 252
+    },
+    {
+      "epoch": 2.1810344827586206,
+      "grad_norm": 0.4648723006248474,
+      "learning_rate": 4.230104284849889e-06,
+      "loss": 0.0104,
+      "step": 253
+    },
+    {
+      "epoch": 2.189655172413793,
+      "grad_norm": 0.4972316026687622,
+      "learning_rate": 4.220567784205551e-06,
+      "loss": 0.0082,
+      "step": 254
+    },
+    {
+      "epoch": 2.1982758620689653,
+      "grad_norm": 0.4233075678348541,
+      "learning_rate": 4.210983478045537e-06,
+      "loss": 0.0088,
+      "step": 255
+    },
+    {
+      "epoch": 2.206896551724138,
+      "grad_norm": 0.5844057202339172,
+      "learning_rate": 4.201351632667227e-06,
+      "loss": 0.0101,
+      "step": 256
+    },
+    {
+      "epoch": 2.2155172413793105,
+      "grad_norm": 0.48759040236473083,
+      "learning_rate": 4.191672515688872e-06,
+      "loss": 0.0097,
+      "step": 257
+    },
+    {
+      "epoch": 2.2241379310344827,
+      "grad_norm": 0.48416000604629517,
+      "learning_rate": 4.181946396042146e-06,
+      "loss": 0.0085,
+      "step": 258
+    },
+    {
+      "epoch": 2.2327586206896552,
+      "grad_norm": 0.5503181219100952,
+      "learning_rate": 4.172173543964682e-06,
+      "loss": 0.0122,
+      "step": 259
+    },
+    {
+      "epoch": 2.2413793103448274,
+      "grad_norm": 0.6163179278373718,
+      "learning_rate": 4.162354230992562e-06,
+      "loss": 0.0119,
+      "step": 260
+    },
+    {
+      "epoch": 2.25,
+      "grad_norm": 0.47484147548675537,
+      "learning_rate": 4.1524887299527695e-06,
+      "loss": 0.0142,
+      "step": 261
+    },
+    {
+      "epoch": 2.2586206896551726,
+      "grad_norm": 0.5557327270507812,
+      "learning_rate": 4.142577314955614e-06,
+      "loss": 0.0086,
+      "step": 262
+    },
+    {
+      "epoch": 2.2672413793103448,
+      "grad_norm": 0.3577985465526581,
+      "learning_rate": 4.1326202613871065e-06,
+      "loss": 0.0089,
+      "step": 263
+    },
+    {
+      "epoch": 2.2758620689655173,
+      "grad_norm": 0.33690375089645386,
+      "learning_rate": 4.122617845901322e-06,
+      "loss": 0.0095,
+      "step": 264
+    },
+    {
+      "epoch": 2.2844827586206895,
+      "grad_norm": 0.5704361796379089,
+      "learning_rate": 4.112570346412696e-06,
+      "loss": 0.0078,
+      "step": 265
+    },
+    {
+      "epoch": 2.293103448275862,
+      "grad_norm": 0.4501226544380188,
+      "learning_rate": 4.102478042088315e-06,
+      "loss": 0.0107,
+      "step": 266
+    },
+    {
+      "epoch": 2.3017241379310347,
+      "grad_norm": 0.3546983301639557,
+      "learning_rate": 4.092341213340155e-06,
+      "loss": 0.0087,
+      "step": 267
+    },
+    {
+      "epoch": 2.310344827586207,
+      "grad_norm": 0.6514727473258972,
+      "learning_rate": 4.0821601418172926e-06,
+      "loss": 0.0121,
+      "step": 268
+    },
+    {
+      "epoch": 2.3189655172413794,
+      "grad_norm": 0.5242908596992493,
+      "learning_rate": 4.0719351103980754e-06,
+      "loss": 0.0104,
+      "step": 269
+    },
+    {
+      "epoch": 2.3275862068965516,
+      "grad_norm": 0.40950050950050354,
+      "learning_rate": 4.0616664031822686e-06,
+      "loss": 0.0091,
+      "step": 270
+    },
+    {
+      "epoch": 2.336206896551724,
+      "grad_norm": 0.24205856025218964,
+      "learning_rate": 4.051354305483153e-06,
+      "loss": 0.0043,
+      "step": 271
+    },
+    {
+      "epoch": 2.344827586206897,
+      "grad_norm": 0.35039305686950684,
+      "learning_rate": 4.040999103819606e-06,
+      "loss": 0.0069,
+      "step": 272
+    },
+    {
+      "epoch": 2.353448275862069,
+      "grad_norm": 0.30672788619995117,
+      "learning_rate": 4.030601085908137e-06,
+      "loss": 0.0064,
+      "step": 273
+    },
+    {
+      "epoch": 2.3620689655172415,
+      "grad_norm": 0.26436808705329895,
+      "learning_rate": 4.020160540654892e-06,
+      "loss": 0.0037,
+      "step": 274
+    },
+    {
+      "epoch": 2.3706896551724137,
+      "grad_norm": 0.4178946614265442,
+      "learning_rate": 4.009677758147627e-06,
+      "loss": 0.0041,
+      "step": 275
+    },
+    {
+      "epoch": 2.3793103448275863,
+      "grad_norm": 0.7681214809417725,
+      "learning_rate": 3.999153029647651e-06,
+      "loss": 0.015,
+      "step": 276
+    },
+    {
+      "epoch": 2.3879310344827585,
+      "grad_norm": 0.49876728653907776,
+      "learning_rate": 3.988586647581732e-06,
+      "loss": 0.0082,
+      "step": 277
+    },
+    {
+      "epoch": 2.396551724137931,
+      "grad_norm": 0.5959510207176208,
+      "learning_rate": 3.977978905533966e-06,
+      "loss": 0.0084,
+      "step": 278
+    },
+    {
+      "epoch": 2.405172413793103,
+      "grad_norm": 0.538182258605957,
+      "learning_rate": 3.9673300982376315e-06,
+      "loss": 0.0087,
+      "step": 279
+    },
+    {
+      "epoch": 2.413793103448276,
+      "grad_norm": 0.6760286688804626,
+      "learning_rate": 3.956640521566989e-06,
+      "loss": 0.0123,
+      "step": 280
+    },
+    {
+      "epoch": 2.4224137931034484,
+      "grad_norm": 0.4143922030925751,
+      "learning_rate": 3.945910472529068e-06,
+      "loss": 0.0056,
+      "step": 281
+    },
+    {
+      "epoch": 2.4310344827586206,
+      "grad_norm": 0.4239058792591095,
+      "learning_rate": 3.935140249255412e-06,
+      "loss": 0.009,
+      "step": 282
+    },
+    {
+      "epoch": 2.439655172413793,
+      "grad_norm": 0.495762437582016,
+      "learning_rate": 3.924330150993793e-06,
+      "loss": 0.0091,
+      "step": 283
+    },
+    {
+      "epoch": 2.4482758620689653,
+      "grad_norm": 0.46725037693977356,
+      "learning_rate": 3.913480478099898e-06,
+      "loss": 0.0066,
+      "step": 284
+    },
+    {
+      "epoch": 2.456896551724138,
+      "grad_norm": 0.501814603805542,
+      "learning_rate": 3.902591532028989e-06,
+      "loss": 0.0087,
+      "step": 285
+    },
+    {
+      "epoch": 2.4655172413793105,
+      "grad_norm": 0.38845211267471313,
+      "learning_rate": 3.891663615327518e-06,
+      "loss": 0.0059,
+      "step": 286
+    },
+    {
+      "epoch": 2.4741379310344827,
+      "grad_norm": 0.3252674639225006,
+      "learning_rate": 3.880697031624729e-06,
+      "loss": 0.0049,
+      "step": 287
+    },
+    {
+      "epoch": 2.4827586206896552,
+      "grad_norm": 0.3237666189670563,
+      "learning_rate": 3.869692085624218e-06,
+      "loss": 0.0062,
+      "step": 288
+    },
+    {
+      "epoch": 2.4913793103448274,
+      "grad_norm": 0.4213542342185974,
+      "learning_rate": 3.858649083095463e-06,
+      "loss": 0.0093,
+      "step": 289
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 0.6313378810882568,
+      "learning_rate": 3.8475683308653385e-06,
+      "loss": 0.0082,
+      "step": 290
+    },
+    {
+      "epoch": 2.5086206896551726,
+      "grad_norm": 1.0402452945709229,
+      "learning_rate": 3.836450136809579e-06,
+      "loss": 0.0078,
+      "step": 291
+    },
+    {
+      "epoch": 2.5172413793103448,
+      "grad_norm": 0.6053091287612915,
+      "learning_rate": 3.825294809844234e-06,
+      "loss": 0.0044,
+      "step": 292
+    },
+    {
+      "epoch": 2.5258620689655173,
+      "grad_norm": 0.6727966666221619,
+      "learning_rate": 3.8141026599170792e-06,
+      "loss": 0.0135,
+      "step": 293
+    },
+    {
+      "epoch": 2.5344827586206895,
+      "grad_norm": 0.3446575105190277,
+      "learning_rate": 3.8028739979990072e-06,
+      "loss": 0.0063,
+      "step": 294
+    },
+    {
+      "epoch": 2.543103448275862,
+      "grad_norm": 0.5400390028953552,
+      "learning_rate": 3.791609136075384e-06,
+      "loss": 0.0094,
+      "step": 295
+    },
+    {
+      "epoch": 2.5517241379310347,
+      "grad_norm": 0.2682323157787323,
+      "learning_rate": 3.7803083871373876e-06,
+      "loss": 0.0028,
+      "step": 296
+    },
+    {
+      "epoch": 2.560344827586207,
+      "grad_norm": 0.49711498618125916,
+      "learning_rate": 3.7689720651733034e-06,
+      "loss": 0.0085,
+      "step": 297
+    },
+    {
+      "epoch": 2.5689655172413794,
+      "grad_norm": 0.3905172646045685,
+      "learning_rate": 3.7576004851598052e-06,
+      "loss": 0.0065,
+      "step": 298
+    },
+    {
+      "epoch": 2.5775862068965516,
+      "grad_norm": 0.33470281958580017,
+      "learning_rate": 3.7461939630532018e-06,
+      "loss": 0.0061,
+      "step": 299
+    },
+    {
+      "epoch": 2.586206896551724,
+      "grad_norm": 0.43622592091560364,
+      "learning_rate": 3.734752815780659e-06,
+      "loss": 0.008,
+      "step": 300
+    },
+    {
+      "epoch": 2.594827586206897,
+      "grad_norm": 0.40031933784484863,
+      "learning_rate": 3.7232773612313945e-06,
+      "loss": 0.0107,
+      "step": 301
+    },
+    {
+      "epoch": 2.603448275862069,
+      "grad_norm": 0.30814412236213684,
+      "learning_rate": 3.7117679182478415e-06,
+      "loss": 0.0042,
+      "step": 302
+    },
+    {
+      "epoch": 2.612068965517241,
+      "grad_norm": 0.16828757524490356,
+      "learning_rate": 3.7002248066167977e-06,
+      "loss": 0.0022,
+      "step": 303
+    },
+    {
+      "epoch": 2.6206896551724137,
+      "grad_norm": 0.3726310729980469,
+      "learning_rate": 3.6886483470605293e-06,
+      "loss": 0.0097,
+      "step": 304
+    },
+    {
+      "epoch": 2.6293103448275863,
+      "grad_norm": 0.47232842445373535,
+      "learning_rate": 3.6770388612278713e-06,
+      "loss": 0.0083,
+      "step": 305
+    },
+    {
+      "epoch": 2.637931034482759,
+      "grad_norm": 0.2700427770614624,
+      "learning_rate": 3.66539667168528e-06,
+      "loss": 0.0017,
+      "step": 306
+    },
+    {
+      "epoch": 2.646551724137931,
+      "grad_norm": 0.3362017273902893,
+      "learning_rate": 3.65372210190788e-06,
+      "loss": 0.0054,
+      "step": 307
+    },
+    {
+      "epoch": 2.655172413793103,
+      "grad_norm": 0.3315584063529968,
+      "learning_rate": 3.6420154762704685e-06,
+      "loss": 0.0032,
+      "step": 308
+    },
+    {
+      "epoch": 2.663793103448276,
+      "grad_norm": 0.656471312046051,
+      "learning_rate": 3.630277120038508e-06,
+      "loss": 0.0093,
+      "step": 309
+    },
+    {
+      "epoch": 2.6724137931034484,
+      "grad_norm": 1.8886873722076416,
+      "learning_rate": 3.6185073593590868e-06,
+      "loss": 0.01,
+      "step": 310
+    },
+    {
+      "epoch": 2.6810344827586206,
+      "grad_norm": 0.3265067934989929,
+      "learning_rate": 3.606706521251857e-06,
+      "loss": 0.0037,
+      "step": 311
+    },
+    {
+      "epoch": 2.689655172413793,
+      "grad_norm": 0.39972051978111267,
+      "learning_rate": 3.5948749335999493e-06,
+      "loss": 0.0066,
+      "step": 312
+    },
+    {
+      "epoch": 2.6982758620689653,
+      "grad_norm": 0.48707249760627747,
+      "learning_rate": 3.583012925140862e-06,
+      "loss": 0.0066,
+      "step": 313
+    },
+    {
+      "epoch": 2.706896551724138,
+      "grad_norm": 0.6093474626541138,
+      "learning_rate": 3.571120825457327e-06,
+      "loss": 0.007,
+      "step": 314
+    },
+    {
+      "epoch": 2.7155172413793105,
+      "grad_norm": 0.3256229758262634,
+      "learning_rate": 3.5591989649681534e-06,
+      "loss": 0.0054,
+      "step": 315
+    },
+    {
+      "epoch": 2.7241379310344827,
+      "grad_norm": 0.47791576385498047,
+      "learning_rate": 3.5472476749190465e-06,
+      "loss": 0.0048,
+      "step": 316
+    },
+    {
+      "epoch": 2.7327586206896552,
+      "grad_norm": 0.7635737061500549,
+      "learning_rate": 3.535267287373403e-06,
+      "loss": 0.006,
+      "step": 317
+    },
+    {
+      "epoch": 2.7413793103448274,
+      "grad_norm": 0.6153246164321899,
+      "learning_rate": 3.523258135203087e-06,
+      "loss": 0.0071,
+      "step": 318
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 0.22357748448848724,
+      "learning_rate": 3.5112205520791785e-06,
+      "loss": 0.0036,
+      "step": 319
+    },
+    {
+      "epoch": 2.7586206896551726,
+      "grad_norm": 0.3073960840702057,
+      "learning_rate": 3.4991548724627054e-06,
+      "loss": 0.0028,
+      "step": 320
+    },
+    {
+      "epoch": 2.7672413793103448,
+      "grad_norm": 0.17926068603992462,
+      "learning_rate": 3.48706143159535e-06,
+      "loss": 0.0028,
+      "step": 321
+    },
+    {
+      "epoch": 2.7758620689655173,
+      "grad_norm": 0.43959733843803406,
+      "learning_rate": 3.4749405654901297e-06,
+      "loss": 0.0027,
+      "step": 322
+    },
+    {
+      "epoch": 2.7844827586206895,
+      "grad_norm": 0.396281898021698,
+      "learning_rate": 3.4627926109220684e-06,
+      "loss": 0.0048,
+      "step": 323
+    },
+    {
+      "epoch": 2.793103448275862,
+      "grad_norm": 0.33752575516700745,
+      "learning_rate": 3.450617905418834e-06,
+      "loss": 0.0055,
+      "step": 324
+    },
+    {
+      "epoch": 2.8017241379310347,
+      "grad_norm": 0.3482801914215088,
+      "learning_rate": 3.4384167872513642e-06,
+      "loss": 0.0054,
+      "step": 325
+    },
+    {
+      "epoch": 2.810344827586207,
+      "grad_norm": 0.38212546706199646,
+      "learning_rate": 3.4261895954244613e-06,
+      "loss": 0.0051,
+      "step": 326
+    },
+    {
+      "epoch": 2.8189655172413794,
+      "grad_norm": 0.3641749918460846,
+      "learning_rate": 3.4139366696673815e-06,
+      "loss": 0.0056,
+      "step": 327
+    },
+    {
+      "epoch": 2.8275862068965516,
+      "grad_norm": 0.1609128713607788,
+      "learning_rate": 3.4016583504243892e-06,
+      "loss": 0.0028,
+      "step": 328
+    },
+    {
+      "epoch": 2.836206896551724,
+      "grad_norm": 0.2003130167722702,
+      "learning_rate": 3.3893549788453e-06,
+      "loss": 0.0023,
+      "step": 329
+    },
+    {
+      "epoch": 2.844827586206897,
+      "grad_norm": 0.4341651201248169,
+      "learning_rate": 3.3770268967760026e-06,
+      "loss": 0.006,
+      "step": 330
+    },
+    {
+      "epoch": 2.853448275862069,
+      "grad_norm": 0.28961309790611267,
+      "learning_rate": 3.36467444674896e-06,
+      "loss": 0.0022,
+      "step": 331
+    },
+    {
+      "epoch": 2.862068965517241,
+      "grad_norm": 0.35188788175582886,
+      "learning_rate": 3.3522979719736923e-06,
+      "loss": 0.0079,
+      "step": 332
+    },
+    {
+      "epoch": 2.8706896551724137,
+      "grad_norm": 0.3318072259426117,
+      "learning_rate": 3.339897816327241e-06,
+      "loss": 0.0038,
+      "step": 333
+    },
+    {
+      "epoch": 2.8793103448275863,
+      "grad_norm": 0.6669831871986389,
+      "learning_rate": 3.327474324344614e-06,
+      "loss": 0.0048,
+      "step": 334
+    },
+    {
+      "epoch": 2.887931034482759,
+      "grad_norm": 0.4758647084236145,
+      "learning_rate": 3.3150278412092126e-06,
+      "loss": 0.0031,
+      "step": 335
+    },
+    {
+      "epoch": 2.896551724137931,
+      "grad_norm": 0.45927631855010986,
+      "learning_rate": 3.3025587127432414e-06,
+      "loss": 0.0052,
+      "step": 336
+    },
+    {
+      "epoch": 2.905172413793103,
+      "grad_norm": 0.15748094022274017,
+      "learning_rate": 3.2900672853981e-06,
+      "loss": 0.002,
+      "step": 337
+    },
+    {
+      "epoch": 2.913793103448276,
+      "grad_norm": 0.2490128129720688,
+      "learning_rate": 3.2775539062447566e-06,
+      "loss": 0.0024,
+      "step": 338
+    },
+    {
+      "epoch": 2.9224137931034484,
+      "grad_norm": 0.16914480924606323,
+      "learning_rate": 3.265018922964102e-06,
+      "loss": 0.0035,
+      "step": 339
+    },
+    {
+      "epoch": 2.9310344827586206,
+      "grad_norm": 0.15188859403133392,
+      "learning_rate": 3.2524626838372976e-06,
+      "loss": 0.0013,
+      "step": 340
+    },
+    {
+      "epoch": 2.939655172413793,
+      "grad_norm": 0.8957772254943848,
+      "learning_rate": 3.239885537736085e-06,
+      "loss": 0.0029,
+      "step": 341
+    },
+    {
+      "epoch": 2.9482758620689653,
+      "grad_norm": 0.11430688202381134,
+      "learning_rate": 3.2272878341131074e-06,
+      "loss": 0.0009,
+      "step": 342
+    },
+    {
+      "epoch": 2.956896551724138,
+      "grad_norm": 0.2879055440425873,
+      "learning_rate": 3.2146699229921884e-06,
+      "loss": 0.0034,
+      "step": 343
+    },
+    {
+      "epoch": 2.9655172413793105,
+      "grad_norm": 0.4262220859527588,
+      "learning_rate": 3.202032154958615e-06,
+      "loss": 0.0043,
+      "step": 344
+    },
+    {
+      "epoch": 2.9741379310344827,
+      "grad_norm": 0.26130998134613037,
+      "learning_rate": 3.1893748811493926e-06,
+      "loss": 0.0013,
+      "step": 345
+    },
+    {
+      "epoch": 2.9827586206896552,
+      "grad_norm": 0.29305410385131836,
+      "learning_rate": 3.1766984532434853e-06,
+      "loss": 0.0035,
+      "step": 346
+    },
+    {
+      "epoch": 2.9913793103448274,
+      "grad_norm": 0.3302770256996155,
+      "learning_rate": 3.164003223452055e-06,
+      "loss": 0.0028,
+      "step": 347
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.41331973671913147,
+      "learning_rate": 3.151289544508664e-06,
+      "loss": 0.0042,
+      "step": 348
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 696,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 6,
+  "save_steps": 116,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.795697653949399e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-348/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b81869dce97233c4dee9b25cc565b7dbf695ff949fafca7ec37b34379228fbff
+size 8056

checkpoint-348/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)