Upload folder using huggingface_hub

checkpoint-308/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-308/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": [
+    "gate_proj",
+    "o_proj",
+    "v_proj",
+    "up_proj",
+    "k_proj",
+    "down_proj",
+    "q_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-308/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1975d8cc1a4882345e0c72fe3104006a9e9f52cc88e20dbbd493cc7ec9db922
+size 1291899552

checkpoint-308/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-308/global_step308/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:c719ce3634ab3d27a78d2ca6f4fc0bdf204c309641b910f84773a2fab65a240b
+size 7751120668

checkpoint-308/global_step308/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b653e2df0d5657902c9e5a1f800ce843b4ad12a0c7b5354cc6b00583aa131381
+size 1396311192

checkpoint-308/latest

@@ -0,0 +1 @@
+global_step308

checkpoint-308/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60c6c6e18dd3e4bd258d8301c2686e8fb775f618bbb39f866b2f8d952a585d9d
+size 14244

checkpoint-308/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81a97ae2a5653bfcff109db5411f527d6e70a93b850d6d0ffe5b654ee8648f75
+size 1064

checkpoint-308/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-308/tokenizer.json

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

checkpoint-308/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-308/trainer_state.json

@@ -0,0 +1,2189 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 308,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.012987012987012988,
+      "grad_norm": 30.513944625854492,
+      "learning_rate": 5.0000000000000004e-08,
+      "loss": 4.1013,
+      "step": 1
+    },
+    {
+      "epoch": 0.025974025974025976,
+      "grad_norm": 30.965682983398438,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 4.1757,
+      "step": 2
+    },
+    {
+      "epoch": 0.03896103896103896,
+      "grad_norm": 30.834516525268555,
+      "learning_rate": 1.5000000000000002e-07,
+      "loss": 4.1371,
+      "step": 3
+    },
+    {
+      "epoch": 0.05194805194805195,
+      "grad_norm": 30.98870849609375,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 4.0543,
+      "step": 4
+    },
+    {
+      "epoch": 0.06493506493506493,
+      "grad_norm": 30.404874801635742,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 4.09,
+      "step": 5
+    },
+    {
+      "epoch": 0.07792207792207792,
+      "grad_norm": 31.433774948120117,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 4.1777,
+      "step": 6
+    },
+    {
+      "epoch": 0.09090909090909091,
+      "grad_norm": 30.850692749023438,
+      "learning_rate": 3.5000000000000004e-07,
+      "loss": 4.142,
+      "step": 7
+    },
+    {
+      "epoch": 0.1038961038961039,
+      "grad_norm": 31.00360870361328,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 4.1115,
+      "step": 8
+    },
+    {
+      "epoch": 0.11688311688311688,
+      "grad_norm": 31.221616744995117,
+      "learning_rate": 4.5000000000000003e-07,
+      "loss": 4.1585,
+      "step": 9
+    },
+    {
+      "epoch": 0.12987012987012986,
+      "grad_norm": 30.560623168945312,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 4.1155,
+      "step": 10
+    },
+    {
+      "epoch": 0.14285714285714285,
+      "grad_norm": 30.089527130126953,
+      "learning_rate": 5.5e-07,
+      "loss": 4.0638,
+      "step": 11
+    },
+    {
+      "epoch": 0.15584415584415584,
+      "grad_norm": 30.662569046020508,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 4.0152,
+      "step": 12
+    },
+    {
+      "epoch": 0.16883116883116883,
+      "grad_norm": 29.455503463745117,
+      "learning_rate": 6.5e-07,
+      "loss": 3.9501,
+      "step": 13
+    },
+    {
+      "epoch": 0.18181818181818182,
+      "grad_norm": 28.851863861083984,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 3.9066,
+      "step": 14
+    },
+    {
+      "epoch": 0.19480519480519481,
+      "grad_norm": 27.41941261291504,
+      "learning_rate": 7.5e-07,
+      "loss": 3.816,
+      "step": 15
+    },
+    {
+      "epoch": 0.2077922077922078,
+      "grad_norm": 25.754905700683594,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 3.6715,
+      "step": 16
+    },
+    {
+      "epoch": 0.22077922077922077,
+      "grad_norm": 24.467802047729492,
+      "learning_rate": 8.500000000000001e-07,
+      "loss": 3.6226,
+      "step": 17
+    },
+    {
+      "epoch": 0.23376623376623376,
+      "grad_norm": 23.671884536743164,
+      "learning_rate": 9.000000000000001e-07,
+      "loss": 3.5756,
+      "step": 18
+    },
+    {
+      "epoch": 0.24675324675324675,
+      "grad_norm": 21.28356170654297,
+      "learning_rate": 9.500000000000001e-07,
+      "loss": 3.3822,
+      "step": 19
+    },
+    {
+      "epoch": 0.2597402597402597,
+      "grad_norm": 20.54037857055664,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 3.3799,
+      "step": 20
+    },
+    {
+      "epoch": 0.2727272727272727,
+      "grad_norm": 18.18222427368164,
+      "learning_rate": 1.0500000000000001e-06,
+      "loss": 3.1964,
+      "step": 21
+    },
+    {
+      "epoch": 0.2857142857142857,
+      "grad_norm": 17.782703399658203,
+      "learning_rate": 1.1e-06,
+      "loss": 3.1745,
+      "step": 22
+    },
+    {
+      "epoch": 0.2987012987012987,
+      "grad_norm": 15.965900421142578,
+      "learning_rate": 1.1500000000000002e-06,
+      "loss": 2.9002,
+      "step": 23
+    },
+    {
+      "epoch": 0.3116883116883117,
+      "grad_norm": 16.00725746154785,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 2.9123,
+      "step": 24
+    },
+    {
+      "epoch": 0.3246753246753247,
+      "grad_norm": 15.301984786987305,
+      "learning_rate": 1.25e-06,
+      "loss": 2.6531,
+      "step": 25
+    },
+    {
+      "epoch": 0.33766233766233766,
+      "grad_norm": 16.359207153320312,
+      "learning_rate": 1.3e-06,
+      "loss": 2.6089,
+      "step": 26
+    },
+    {
+      "epoch": 0.35064935064935066,
+      "grad_norm": 17.74176025390625,
+      "learning_rate": 1.3500000000000002e-06,
+      "loss": 2.4805,
+      "step": 27
+    },
+    {
+      "epoch": 0.36363636363636365,
+      "grad_norm": 18.702547073364258,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 2.3312,
+      "step": 28
+    },
+    {
+      "epoch": 0.37662337662337664,
+      "grad_norm": 18.44041633605957,
+      "learning_rate": 1.45e-06,
+      "loss": 2.2273,
+      "step": 29
+    },
+    {
+      "epoch": 0.38961038961038963,
+      "grad_norm": 15.867247581481934,
+      "learning_rate": 1.5e-06,
+      "loss": 2.0514,
+      "step": 30
+    },
+    {
+      "epoch": 0.4025974025974026,
+      "grad_norm": 13.97642707824707,
+      "learning_rate": 1.5500000000000002e-06,
+      "loss": 1.8689,
+      "step": 31
+    },
+    {
+      "epoch": 0.4155844155844156,
+      "grad_norm": 13.599323272705078,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 1.7907,
+      "step": 32
+    },
+    {
+      "epoch": 0.42857142857142855,
+      "grad_norm": 13.204086303710938,
+      "learning_rate": 1.6500000000000003e-06,
+      "loss": 1.6537,
+      "step": 33
+    },
+    {
+      "epoch": 0.44155844155844154,
+      "grad_norm": 13.255966186523438,
+      "learning_rate": 1.7000000000000002e-06,
+      "loss": 1.5693,
+      "step": 34
+    },
+    {
+      "epoch": 0.45454545454545453,
+      "grad_norm": 13.00086784362793,
+      "learning_rate": 1.75e-06,
+      "loss": 1.3916,
+      "step": 35
+    },
+    {
+      "epoch": 0.4675324675324675,
+      "grad_norm": 12.996549606323242,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 1.2734,
+      "step": 36
+    },
+    {
+      "epoch": 0.4805194805194805,
+      "grad_norm": 12.506110191345215,
+      "learning_rate": 1.85e-06,
+      "loss": 1.1444,
+      "step": 37
+    },
+    {
+      "epoch": 0.4935064935064935,
+      "grad_norm": 12.379636764526367,
+      "learning_rate": 1.9000000000000002e-06,
+      "loss": 1.033,
+      "step": 38
+    },
+    {
+      "epoch": 0.5064935064935064,
+      "grad_norm": 12.38772201538086,
+      "learning_rate": 1.9500000000000004e-06,
+      "loss": 0.8932,
+      "step": 39
+    },
+    {
+      "epoch": 0.5194805194805194,
+      "grad_norm": 12.370598793029785,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.7914,
+      "step": 40
+    },
+    {
+      "epoch": 0.5324675324675324,
+      "grad_norm": 11.510255813598633,
+      "learning_rate": 2.05e-06,
+      "loss": 0.6336,
+      "step": 41
+    },
+    {
+      "epoch": 0.5454545454545454,
+      "grad_norm": 10.088979721069336,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 0.5115,
+      "step": 42
+    },
+    {
+      "epoch": 0.5584415584415584,
+      "grad_norm": 9.241965293884277,
+      "learning_rate": 2.15e-06,
+      "loss": 0.416,
+      "step": 43
+    },
+    {
+      "epoch": 0.5714285714285714,
+      "grad_norm": 7.457204341888428,
+      "learning_rate": 2.2e-06,
+      "loss": 0.2947,
+      "step": 44
+    },
+    {
+      "epoch": 0.5844155844155844,
+      "grad_norm": 6.164954662322998,
+      "learning_rate": 2.25e-06,
+      "loss": 0.2202,
+      "step": 45
+    },
+    {
+      "epoch": 0.5974025974025974,
+      "grad_norm": 4.587556838989258,
+      "learning_rate": 2.3000000000000004e-06,
+      "loss": 0.1721,
+      "step": 46
+    },
+    {
+      "epoch": 0.6103896103896104,
+      "grad_norm": 3.2366952896118164,
+      "learning_rate": 2.35e-06,
+      "loss": 0.1451,
+      "step": 47
+    },
+    {
+      "epoch": 0.6233766233766234,
+      "grad_norm": 2.397469997406006,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 0.1116,
+      "step": 48
+    },
+    {
+      "epoch": 0.6363636363636364,
+      "grad_norm": 1.872839093208313,
+      "learning_rate": 2.4500000000000003e-06,
+      "loss": 0.1,
+      "step": 49
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 1.2464176416397095,
+      "learning_rate": 2.5e-06,
+      "loss": 0.1024,
+      "step": 50
+    },
+    {
+      "epoch": 0.6623376623376623,
+      "grad_norm": 1.0786665678024292,
+      "learning_rate": 2.55e-06,
+      "loss": 0.087,
+      "step": 51
+    },
+    {
+      "epoch": 0.6753246753246753,
+      "grad_norm": 1.10042142868042,
+      "learning_rate": 2.6e-06,
+      "loss": 0.0838,
+      "step": 52
+    },
+    {
+      "epoch": 0.6883116883116883,
+      "grad_norm": 0.9031299948692322,
+      "learning_rate": 2.6500000000000005e-06,
+      "loss": 0.0804,
+      "step": 53
+    },
+    {
+      "epoch": 0.7012987012987013,
+      "grad_norm": 0.8072167634963989,
+      "learning_rate": 2.7000000000000004e-06,
+      "loss": 0.0786,
+      "step": 54
+    },
+    {
+      "epoch": 0.7142857142857143,
+      "grad_norm": 0.6948329210281372,
+      "learning_rate": 2.7500000000000004e-06,
+      "loss": 0.0773,
+      "step": 55
+    },
+    {
+      "epoch": 0.7272727272727273,
+      "grad_norm": 0.6062957644462585,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 0.0756,
+      "step": 56
+    },
+    {
+      "epoch": 0.7402597402597403,
+      "grad_norm": 0.4804394841194153,
+      "learning_rate": 2.85e-06,
+      "loss": 0.0697,
+      "step": 57
+    },
+    {
+      "epoch": 0.7532467532467533,
+      "grad_norm": 0.5353919863700867,
+      "learning_rate": 2.9e-06,
+      "loss": 0.0697,
+      "step": 58
+    },
+    {
+      "epoch": 0.7662337662337663,
+      "grad_norm": 0.4344547390937805,
+      "learning_rate": 2.95e-06,
+      "loss": 0.0704,
+      "step": 59
+    },
+    {
+      "epoch": 0.7792207792207793,
+      "grad_norm": 0.574044942855835,
+      "learning_rate": 3e-06,
+      "loss": 0.0778,
+      "step": 60
+    },
+    {
+      "epoch": 0.7922077922077922,
+      "grad_norm": 0.5392284393310547,
+      "learning_rate": 3.05e-06,
+      "loss": 0.0658,
+      "step": 61
+    },
+    {
+      "epoch": 0.8051948051948052,
+      "grad_norm": 0.5098194479942322,
+      "learning_rate": 3.1000000000000004e-06,
+      "loss": 0.0661,
+      "step": 62
+    },
+    {
+      "epoch": 0.8181818181818182,
+      "grad_norm": 0.3733481168746948,
+      "learning_rate": 3.1500000000000003e-06,
+      "loss": 0.0667,
+      "step": 63
+    },
+    {
+      "epoch": 0.8311688311688312,
+      "grad_norm": 0.3714415431022644,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 0.0635,
+      "step": 64
+    },
+    {
+      "epoch": 0.8441558441558441,
+      "grad_norm": 0.37473437190055847,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.0661,
+      "step": 65
+    },
+    {
+      "epoch": 0.8571428571428571,
+      "grad_norm": 0.35993579030036926,
+      "learning_rate": 3.3000000000000006e-06,
+      "loss": 0.0637,
+      "step": 66
+    },
+    {
+      "epoch": 0.8701298701298701,
+      "grad_norm": 0.41175705194473267,
+      "learning_rate": 3.3500000000000005e-06,
+      "loss": 0.0646,
+      "step": 67
+    },
+    {
+      "epoch": 0.8831168831168831,
+      "grad_norm": 0.29564622044563293,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 0.0627,
+      "step": 68
+    },
+    {
+      "epoch": 0.8961038961038961,
+      "grad_norm": 0.3018590211868286,
+      "learning_rate": 3.45e-06,
+      "loss": 0.0613,
+      "step": 69
+    },
+    {
+      "epoch": 0.9090909090909091,
+      "grad_norm": 0.499183714389801,
+      "learning_rate": 3.5e-06,
+      "loss": 0.0664,
+      "step": 70
+    },
+    {
+      "epoch": 0.922077922077922,
+      "grad_norm": 0.3450847864151001,
+      "learning_rate": 3.5500000000000003e-06,
+      "loss": 0.0589,
+      "step": 71
+    },
+    {
+      "epoch": 0.935064935064935,
+      "grad_norm": 0.26693543791770935,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 0.0633,
+      "step": 72
+    },
+    {
+      "epoch": 0.948051948051948,
+      "grad_norm": 0.2867635190486908,
+      "learning_rate": 3.65e-06,
+      "loss": 0.062,
+      "step": 73
+    },
+    {
+      "epoch": 0.961038961038961,
+      "grad_norm": 0.3328873813152313,
+      "learning_rate": 3.7e-06,
+      "loss": 0.0587,
+      "step": 74
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 0.33611243963241577,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.058,
+      "step": 75
+    },
+    {
+      "epoch": 0.987012987012987,
+      "grad_norm": 0.2983446419239044,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.0632,
+      "step": 76
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.3246895372867584,
+      "learning_rate": 3.85e-06,
+      "loss": 0.0583,
+      "step": 77
+    },
+    {
+      "epoch": 1.0129870129870129,
+      "grad_norm": 0.2912401556968689,
+      "learning_rate": 3.900000000000001e-06,
+      "loss": 0.0609,
+      "step": 78
+    },
+    {
+      "epoch": 1.025974025974026,
+      "grad_norm": 0.37205109000205994,
+      "learning_rate": 3.95e-06,
+      "loss": 0.0646,
+      "step": 79
+    },
+    {
+      "epoch": 1.0389610389610389,
+      "grad_norm": 0.2430678755044937,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.0567,
+      "step": 80
+    },
+    {
+      "epoch": 1.051948051948052,
+      "grad_norm": 0.32996001839637756,
+      "learning_rate": 4.05e-06,
+      "loss": 0.0517,
+      "step": 81
+    },
+    {
+      "epoch": 1.0649350649350648,
+      "grad_norm": 0.274720698595047,
+      "learning_rate": 4.1e-06,
+      "loss": 0.0579,
+      "step": 82
+    },
+    {
+      "epoch": 1.077922077922078,
+      "grad_norm": 0.3047677278518677,
+      "learning_rate": 4.15e-06,
+      "loss": 0.0575,
+      "step": 83
+    },
+    {
+      "epoch": 1.0909090909090908,
+      "grad_norm": 0.2987304925918579,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 0.0576,
+      "step": 84
+    },
+    {
+      "epoch": 1.103896103896104,
+      "grad_norm": 0.32784077525138855,
+      "learning_rate": 4.25e-06,
+      "loss": 0.0607,
+      "step": 85
+    },
+    {
+      "epoch": 1.1168831168831168,
+      "grad_norm": 0.254140704870224,
+      "learning_rate": 4.3e-06,
+      "loss": 0.0582,
+      "step": 86
+    },
+    {
+      "epoch": 1.12987012987013,
+      "grad_norm": 0.2915174663066864,
+      "learning_rate": 4.350000000000001e-06,
+      "loss": 0.0614,
+      "step": 87
+    },
+    {
+      "epoch": 1.1428571428571428,
+      "grad_norm": 0.2874818444252014,
+      "learning_rate": 4.4e-06,
+      "loss": 0.0541,
+      "step": 88
+    },
+    {
+      "epoch": 1.155844155844156,
+      "grad_norm": 0.27614182233810425,
+      "learning_rate": 4.450000000000001e-06,
+      "loss": 0.0556,
+      "step": 89
+    },
+    {
+      "epoch": 1.1688311688311688,
+      "grad_norm": 0.3335096836090088,
+      "learning_rate": 4.5e-06,
+      "loss": 0.0601,
+      "step": 90
+    },
+    {
+      "epoch": 1.1818181818181819,
+      "grad_norm": 0.2302844524383545,
+      "learning_rate": 4.5500000000000005e-06,
+      "loss": 0.056,
+      "step": 91
+    },
+    {
+      "epoch": 1.1948051948051948,
+      "grad_norm": 0.22703984379768372,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 0.0577,
+      "step": 92
+    },
+    {
+      "epoch": 1.2077922077922079,
+      "grad_norm": 0.2960239350795746,
+      "learning_rate": 4.65e-06,
+      "loss": 0.0597,
+      "step": 93
+    },
+    {
+      "epoch": 1.2207792207792207,
+      "grad_norm": 0.20548498630523682,
+      "learning_rate": 4.7e-06,
+      "loss": 0.0531,
+      "step": 94
+    },
+    {
+      "epoch": 1.2337662337662338,
+      "grad_norm": 0.2792559564113617,
+      "learning_rate": 4.75e-06,
+      "loss": 0.0552,
+      "step": 95
+    },
+    {
+      "epoch": 1.2467532467532467,
+      "grad_norm": 0.25889474153518677,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.053,
+      "step": 96
+    },
+    {
+      "epoch": 1.2597402597402598,
+      "grad_norm": 0.2224734127521515,
+      "learning_rate": 4.85e-06,
+      "loss": 0.0528,
+      "step": 97
+    },
+    {
+      "epoch": 1.2727272727272727,
+      "grad_norm": 0.2264651507139206,
+      "learning_rate": 4.9000000000000005e-06,
+      "loss": 0.0545,
+      "step": 98
+    },
+    {
+      "epoch": 1.2857142857142856,
+      "grad_norm": 0.29424387216567993,
+      "learning_rate": 4.95e-06,
+      "loss": 0.0571,
+      "step": 99
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 0.3200651705265045,
+      "learning_rate": 5e-06,
+      "loss": 0.0605,
+      "step": 100
+    },
+    {
+      "epoch": 1.3116883116883118,
+      "grad_norm": 0.284705251455307,
+      "learning_rate": 4.999905856597241e-06,
+      "loss": 0.0512,
+      "step": 101
+    },
+    {
+      "epoch": 1.3246753246753247,
+      "grad_norm": 0.25781720876693726,
+      "learning_rate": 4.999623433479346e-06,
+      "loss": 0.053,
+      "step": 102
+    },
+    {
+      "epoch": 1.3376623376623376,
+      "grad_norm": 0.2585068643093109,
+      "learning_rate": 4.999152751916936e-06,
+      "loss": 0.0557,
+      "step": 103
+    },
+    {
+      "epoch": 1.3506493506493507,
+      "grad_norm": 0.24417804181575775,
+      "learning_rate": 4.99849384735926e-06,
+      "loss": 0.0599,
+      "step": 104
+    },
+    {
+      "epoch": 1.3636363636363638,
+      "grad_norm": 0.2860206067562103,
+      "learning_rate": 4.997646769431532e-06,
+      "loss": 0.053,
+      "step": 105
+    },
+    {
+      "epoch": 1.3766233766233766,
+      "grad_norm": 0.22996841371059418,
+      "learning_rate": 4.9966115819311926e-06,
+      "loss": 0.0544,
+      "step": 106
+    },
+    {
+      "epoch": 1.3896103896103895,
+      "grad_norm": 0.2630089819431305,
+      "learning_rate": 4.9953883628231e-06,
+      "loss": 0.0509,
+      "step": 107
+    },
+    {
+      "epoch": 1.4025974025974026,
+      "grad_norm": 0.2588723599910736,
+      "learning_rate": 4.99397720423366e-06,
+      "loss": 0.0521,
+      "step": 108
+    },
+    {
+      "epoch": 1.4155844155844157,
+      "grad_norm": 0.3795599937438965,
+      "learning_rate": 4.992378212443891e-06,
+      "loss": 0.0527,
+      "step": 109
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 0.3013223111629486,
+      "learning_rate": 4.990591507881416e-06,
+      "loss": 0.0523,
+      "step": 110
+    },
+    {
+      "epoch": 1.4415584415584415,
+      "grad_norm": 0.24418188631534576,
+      "learning_rate": 4.988617225111392e-06,
+      "loss": 0.0494,
+      "step": 111
+    },
+    {
+      "epoch": 1.4545454545454546,
+      "grad_norm": 0.3064674735069275,
+      "learning_rate": 4.986455512826377e-06,
+      "loss": 0.0438,
+      "step": 112
+    },
+    {
+      "epoch": 1.4675324675324675,
+      "grad_norm": 0.2957422137260437,
+      "learning_rate": 4.984106533835132e-06,
+      "loss": 0.0496,
+      "step": 113
+    },
+    {
+      "epoch": 1.4805194805194806,
+      "grad_norm": 0.21359635889530182,
+      "learning_rate": 4.981570465050357e-06,
+      "loss": 0.0508,
+      "step": 114
+    },
+    {
+      "epoch": 1.4935064935064934,
+      "grad_norm": 0.20495343208312988,
+      "learning_rate": 4.978847497475369e-06,
+      "loss": 0.0535,
+      "step": 115
+    },
+    {
+      "epoch": 1.5064935064935066,
+      "grad_norm": 0.26068198680877686,
+      "learning_rate": 4.975937836189712e-06,
+      "loss": 0.0485,
+      "step": 116
+    },
+    {
+      "epoch": 1.5194805194805194,
+      "grad_norm": 0.26806536316871643,
+      "learning_rate": 4.97284170033372e-06,
+      "loss": 0.0505,
+      "step": 117
+    },
+    {
+      "epoch": 1.5324675324675323,
+      "grad_norm": 0.24916547536849976,
+      "learning_rate": 4.969559323092004e-06,
+      "loss": 0.0507,
+      "step": 118
+    },
+    {
+      "epoch": 1.5454545454545454,
+      "grad_norm": 0.26345163583755493,
+      "learning_rate": 4.966090951675893e-06,
+      "loss": 0.049,
+      "step": 119
+    },
+    {
+      "epoch": 1.5584415584415585,
+      "grad_norm": 0.2529750168323517,
+      "learning_rate": 4.962436847304818e-06,
+      "loss": 0.0471,
+      "step": 120
+    },
+    {
+      "epoch": 1.5714285714285714,
+      "grad_norm": 0.2879778742790222,
+      "learning_rate": 4.958597285186635e-06,
+      "loss": 0.047,
+      "step": 121
+    },
+    {
+      "epoch": 1.5844155844155843,
+      "grad_norm": 0.24268512427806854,
+      "learning_rate": 4.954572554496897e-06,
+      "loss": 0.0448,
+      "step": 122
+    },
+    {
+      "epoch": 1.5974025974025974,
+      "grad_norm": 0.3030378520488739,
+      "learning_rate": 4.950362958357078e-06,
+      "loss": 0.0483,
+      "step": 123
+    },
+    {
+      "epoch": 1.6103896103896105,
+      "grad_norm": 0.3552945852279663,
+      "learning_rate": 4.945968813811743e-06,
+      "loss": 0.0518,
+      "step": 124
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 0.32233837246894836,
+      "learning_rate": 4.941390451804668e-06,
+      "loss": 0.0488,
+      "step": 125
+    },
+    {
+      "epoch": 1.6363636363636362,
+      "grad_norm": 0.24810399115085602,
+      "learning_rate": 4.936628217153914e-06,
+      "loss": 0.0487,
+      "step": 126
+    },
+    {
+      "epoch": 1.6493506493506493,
+      "grad_norm": 0.2843495011329651,
+      "learning_rate": 4.931682468525863e-06,
+      "loss": 0.0485,
+      "step": 127
+    },
+    {
+      "epoch": 1.6623376623376624,
+      "grad_norm": 0.4453549087047577,
+      "learning_rate": 4.9265535784081965e-06,
+      "loss": 0.0441,
+      "step": 128
+    },
+    {
+      "epoch": 1.6753246753246753,
+      "grad_norm": 0.2711978852748871,
+      "learning_rate": 4.92124193308185e-06,
+      "loss": 0.0456,
+      "step": 129
+    },
+    {
+      "epoch": 1.6883116883116882,
+      "grad_norm": 0.2913718521595001,
+      "learning_rate": 4.915747932591916e-06,
+      "loss": 0.0478,
+      "step": 130
+    },
+    {
+      "epoch": 1.7012987012987013,
+      "grad_norm": 0.26739779114723206,
+      "learning_rate": 4.910071990717516e-06,
+      "loss": 0.0418,
+      "step": 131
+    },
+    {
+      "epoch": 1.7142857142857144,
+      "grad_norm": 0.33316293358802795,
+      "learning_rate": 4.9042145349406335e-06,
+      "loss": 0.0466,
+      "step": 132
+    },
+    {
+      "epoch": 1.7272727272727273,
+      "grad_norm": 0.3216155469417572,
+      "learning_rate": 4.898176006413925e-06,
+      "loss": 0.043,
+      "step": 133
+    },
+    {
+      "epoch": 1.7402597402597402,
+      "grad_norm": 0.3745043873786926,
+      "learning_rate": 4.891956859927489e-06,
+      "loss": 0.0471,
+      "step": 134
+    },
+    {
+      "epoch": 1.7532467532467533,
+      "grad_norm": 0.2518353760242462,
+      "learning_rate": 4.885557563874614e-06,
+      "loss": 0.045,
+      "step": 135
+    },
+    {
+      "epoch": 1.7662337662337664,
+      "grad_norm": 0.29494860768318176,
+      "learning_rate": 4.8789786002165055e-06,
+      "loss": 0.0437,
+      "step": 136
+    },
+    {
+      "epoch": 1.7792207792207793,
+      "grad_norm": 0.2658509612083435,
+      "learning_rate": 4.872220464445983e-06,
+      "loss": 0.0523,
+      "step": 137
+    },
+    {
+      "epoch": 1.7922077922077921,
+      "grad_norm": 0.3264024257659912,
+      "learning_rate": 4.865283665550167e-06,
+      "loss": 0.047,
+      "step": 138
+    },
+    {
+      "epoch": 1.8051948051948052,
+      "grad_norm": 0.29048386216163635,
+      "learning_rate": 4.8581687259721375e-06,
+      "loss": 0.0513,
+      "step": 139
+    },
+    {
+      "epoch": 1.8181818181818183,
+      "grad_norm": 0.21431702375411987,
+      "learning_rate": 4.850876181571592e-06,
+      "loss": 0.0426,
+      "step": 140
+    },
+    {
+      "epoch": 1.8311688311688312,
+      "grad_norm": 0.29871106147766113,
+      "learning_rate": 4.843406581584487e-06,
+      "loss": 0.0498,
+      "step": 141
+    },
+    {
+      "epoch": 1.844155844155844,
+      "grad_norm": 0.2374885231256485,
+      "learning_rate": 4.83576048858167e-06,
+      "loss": 0.0423,
+      "step": 142
+    },
+    {
+      "epoch": 1.8571428571428572,
+      "grad_norm": 0.32642969489097595,
+      "learning_rate": 4.8279384784265124e-06,
+      "loss": 0.043,
+      "step": 143
+    },
+    {
+      "epoch": 1.87012987012987,
+      "grad_norm": 0.3235689103603363,
+      "learning_rate": 4.8199411402315356e-06,
+      "loss": 0.0431,
+      "step": 144
+    },
+    {
+      "epoch": 1.883116883116883,
+      "grad_norm": 0.25325465202331543,
+      "learning_rate": 4.811769076314044e-06,
+      "loss": 0.0405,
+      "step": 145
+    },
+    {
+      "epoch": 1.896103896103896,
+      "grad_norm": 0.25112220644950867,
+      "learning_rate": 4.803422902150762e-06,
+      "loss": 0.0399,
+      "step": 146
+    },
+    {
+      "epoch": 1.9090909090909092,
+      "grad_norm": 0.2481043040752411,
+      "learning_rate": 4.794903246331477e-06,
+      "loss": 0.0405,
+      "step": 147
+    },
+    {
+      "epoch": 1.922077922077922,
+      "grad_norm": 0.28779706358909607,
+      "learning_rate": 4.786210750511701e-06,
+      "loss": 0.0407,
+      "step": 148
+    },
+    {
+      "epoch": 1.935064935064935,
+      "grad_norm": 0.2663511335849762,
+      "learning_rate": 4.777346069364343e-06,
+      "loss": 0.0404,
+      "step": 149
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 0.27789220213890076,
+      "learning_rate": 4.7683098705303995e-06,
+      "loss": 0.0384,
+      "step": 150
+    },
+    {
+      "epoch": 1.9610389610389611,
+      "grad_norm": 0.34537413716316223,
+      "learning_rate": 4.7591028345686765e-06,
+      "loss": 0.0466,
+      "step": 151
+    },
+    {
+      "epoch": 1.974025974025974,
+      "grad_norm": 0.3259763717651367,
+      "learning_rate": 4.749725654904529e-06,
+      "loss": 0.0398,
+      "step": 152
+    },
+    {
+      "epoch": 1.987012987012987,
+      "grad_norm": 0.3018707036972046,
+      "learning_rate": 4.740179037777639e-06,
+      "loss": 0.0415,
+      "step": 153
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.27019912004470825,
+      "learning_rate": 4.730463702188824e-06,
+      "loss": 0.0428,
+      "step": 154
+    },
+    {
+      "epoch": 2.012987012987013,
+      "grad_norm": 0.34202930331230164,
+      "learning_rate": 4.720580379845884e-06,
+      "loss": 0.04,
+      "step": 155
+    },
+    {
+      "epoch": 2.0259740259740258,
+      "grad_norm": 0.23893073201179504,
+      "learning_rate": 4.710529815108496e-06,
+      "loss": 0.0352,
+      "step": 156
+    },
+    {
+      "epoch": 2.038961038961039,
+      "grad_norm": 0.3122200071811676,
+      "learning_rate": 4.700312764932151e-06,
+      "loss": 0.0369,
+      "step": 157
+    },
+    {
+      "epoch": 2.051948051948052,
+      "grad_norm": 0.37174174189567566,
+      "learning_rate": 4.689929998811145e-06,
+      "loss": 0.035,
+      "step": 158
+    },
+    {
+      "epoch": 2.064935064935065,
+      "grad_norm": 0.30702632665634155,
+      "learning_rate": 4.679382298720625e-06,
+      "loss": 0.0335,
+      "step": 159
+    },
+    {
+      "epoch": 2.0779220779220777,
+      "grad_norm": 0.2560974955558777,
+      "learning_rate": 4.668670459057693e-06,
+      "loss": 0.0307,
+      "step": 160
+    },
+    {
+      "epoch": 2.090909090909091,
+      "grad_norm": 0.25084158778190613,
+      "learning_rate": 4.657795286581576e-06,
+      "loss": 0.033,
+      "step": 161
+    },
+    {
+      "epoch": 2.103896103896104,
+      "grad_norm": 0.25736576318740845,
+      "learning_rate": 4.64675760035287e-06,
+      "loss": 0.0336,
+      "step": 162
+    },
+    {
+      "epoch": 2.116883116883117,
+      "grad_norm": 0.29564720392227173,
+      "learning_rate": 4.635558231671846e-06,
+      "loss": 0.0325,
+      "step": 163
+    },
+    {
+      "epoch": 2.1298701298701297,
+      "grad_norm": 0.32640376687049866,
+      "learning_rate": 4.624198024015845e-06,
+      "loss": 0.0336,
+      "step": 164
+    },
+    {
+      "epoch": 2.142857142857143,
+      "grad_norm": 0.27138862013816833,
+      "learning_rate": 4.612677832975751e-06,
+      "loss": 0.0324,
+      "step": 165
+    },
+    {
+      "epoch": 2.155844155844156,
+      "grad_norm": 0.28909239172935486,
+      "learning_rate": 4.6009985261915536e-06,
+      "loss": 0.0351,
+      "step": 166
+    },
+    {
+      "epoch": 2.168831168831169,
+      "grad_norm": 0.26905709505081177,
+      "learning_rate": 4.5891609832869964e-06,
+      "loss": 0.0348,
+      "step": 167
+    },
+    {
+      "epoch": 2.1818181818181817,
+      "grad_norm": 0.3421840965747833,
+      "learning_rate": 4.577166095803336e-06,
+      "loss": 0.0362,
+      "step": 168
+    },
+    {
+      "epoch": 2.1948051948051948,
+      "grad_norm": 0.37504997849464417,
+      "learning_rate": 4.565014767132191e-06,
+      "loss": 0.038,
+      "step": 169
+    },
+    {
+      "epoch": 2.207792207792208,
+      "grad_norm": 0.3792576789855957,
+      "learning_rate": 4.552707912447504e-06,
+      "loss": 0.0345,
+      "step": 170
+    },
+    {
+      "epoch": 2.220779220779221,
+      "grad_norm": 0.38099661469459534,
+      "learning_rate": 4.540246458636619e-06,
+      "loss": 0.0265,
+      "step": 171
+    },
+    {
+      "epoch": 2.2337662337662336,
+      "grad_norm": 0.40823256969451904,
+      "learning_rate": 4.527631344230466e-06,
+      "loss": 0.0368,
+      "step": 172
+    },
+    {
+      "epoch": 2.2467532467532467,
+      "grad_norm": 0.28929030895233154,
+      "learning_rate": 4.514863519332882e-06,
+      "loss": 0.0282,
+      "step": 173
+    },
+    {
+      "epoch": 2.25974025974026,
+      "grad_norm": 0.32313477993011475,
+      "learning_rate": 4.501943945549054e-06,
+      "loss": 0.0292,
+      "step": 174
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 0.4433124363422394,
+      "learning_rate": 4.488873595913092e-06,
+      "loss": 0.0298,
+      "step": 175
+    },
+    {
+      "epoch": 2.2857142857142856,
+      "grad_norm": 0.32401469349861145,
+      "learning_rate": 4.475653454814746e-06,
+      "loss": 0.0296,
+      "step": 176
+    },
+    {
+      "epoch": 2.2987012987012987,
+      "grad_norm": 0.35543468594551086,
+      "learning_rate": 4.4622845179252735e-06,
+      "loss": 0.0313,
+      "step": 177
+    },
+    {
+      "epoch": 2.311688311688312,
+      "grad_norm": 0.37388646602630615,
+      "learning_rate": 4.44876779212244e-06,
+      "loss": 0.0305,
+      "step": 178
+    },
+    {
+      "epoch": 2.324675324675325,
+      "grad_norm": 0.33474496006965637,
+      "learning_rate": 4.435104295414697e-06,
+      "loss": 0.0289,
+      "step": 179
+    },
+    {
+      "epoch": 2.3376623376623376,
+      "grad_norm": 0.33111104369163513,
+      "learning_rate": 4.421295056864501e-06,
+      "loss": 0.0312,
+      "step": 180
+    },
+    {
+      "epoch": 2.3506493506493507,
+      "grad_norm": 0.3842385709285736,
+      "learning_rate": 4.407341116510818e-06,
+      "loss": 0.0278,
+      "step": 181
+    },
+    {
+      "epoch": 2.3636363636363638,
+      "grad_norm": 0.4244345724582672,
+      "learning_rate": 4.3932435252907914e-06,
+      "loss": 0.0275,
+      "step": 182
+    },
+    {
+      "epoch": 2.3766233766233764,
+      "grad_norm": 0.5842180252075195,
+      "learning_rate": 4.379003344960585e-06,
+      "loss": 0.035,
+      "step": 183
+    },
+    {
+      "epoch": 2.3896103896103895,
+      "grad_norm": 0.35604268312454224,
+      "learning_rate": 4.364621648015426e-06,
+      "loss": 0.0275,
+      "step": 184
+    },
+    {
+      "epoch": 2.4025974025974026,
+      "grad_norm": 0.41529580950737,
+      "learning_rate": 4.3500995176088235e-06,
+      "loss": 0.0315,
+      "step": 185
+    },
+    {
+      "epoch": 2.4155844155844157,
+      "grad_norm": 0.4051007926464081,
+      "learning_rate": 4.335438047470996e-06,
+      "loss": 0.0271,
+      "step": 186
+    },
+    {
+      "epoch": 2.4285714285714284,
+      "grad_norm": 0.33859074115753174,
+      "learning_rate": 4.320638341826494e-06,
+      "loss": 0.0271,
+      "step": 187
+    },
+    {
+      "epoch": 2.4415584415584415,
+      "grad_norm": 0.33747461438179016,
+      "learning_rate": 4.305701515311037e-06,
+      "loss": 0.0235,
+      "step": 188
+    },
+    {
+      "epoch": 2.4545454545454546,
+      "grad_norm": 0.3473111093044281,
+      "learning_rate": 4.290628692887564e-06,
+      "loss": 0.0284,
+      "step": 189
+    },
+    {
+      "epoch": 2.4675324675324677,
+      "grad_norm": 0.3321944773197174,
+      "learning_rate": 4.27542100976151e-06,
+      "loss": 0.024,
+      "step": 190
+    },
+    {
+      "epoch": 2.4805194805194803,
+      "grad_norm": 0.3281101584434509,
+      "learning_rate": 4.260079611295303e-06,
+      "loss": 0.0262,
+      "step": 191
+    },
+    {
+      "epoch": 2.4935064935064934,
+      "grad_norm": 0.4142085611820221,
+      "learning_rate": 4.244605652922108e-06,
+      "loss": 0.0218,
+      "step": 192
+    },
+    {
+      "epoch": 2.5064935064935066,
+      "grad_norm": 0.3764965832233429,
+      "learning_rate": 4.229000300058802e-06,
+      "loss": 0.0238,
+      "step": 193
+    },
+    {
+      "epoch": 2.5194805194805197,
+      "grad_norm": 0.4414701461791992,
+      "learning_rate": 4.2132647280182e-06,
+      "loss": 0.0267,
+      "step": 194
+    },
+    {
+      "epoch": 2.5324675324675323,
+      "grad_norm": 0.4074808657169342,
+      "learning_rate": 4.197400121920539e-06,
+      "loss": 0.022,
+      "step": 195
+    },
+    {
+      "epoch": 2.5454545454545454,
+      "grad_norm": 0.5768939852714539,
+      "learning_rate": 4.1814076766042206e-06,
+      "loss": 0.0311,
+      "step": 196
+    },
+    {
+      "epoch": 2.5584415584415585,
+      "grad_norm": 0.35587024688720703,
+      "learning_rate": 4.165288596535821e-06,
+      "loss": 0.0246,
+      "step": 197
+    },
+    {
+      "epoch": 2.571428571428571,
+      "grad_norm": 0.36500850319862366,
+      "learning_rate": 4.149044095719377e-06,
+      "loss": 0.0195,
+      "step": 198
+    },
+    {
+      "epoch": 2.5844155844155843,
+      "grad_norm": 0.3946477472782135,
+      "learning_rate": 4.132675397604956e-06,
+      "loss": 0.0208,
+      "step": 199
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 0.3371339440345764,
+      "learning_rate": 4.11618373499651e-06,
+      "loss": 0.0204,
+      "step": 200
+    },
+    {
+      "epoch": 2.6103896103896105,
+      "grad_norm": 0.41154706478118896,
+      "learning_rate": 4.099570349959025e-06,
+      "loss": 0.0223,
+      "step": 201
+    },
+    {
+      "epoch": 2.6233766233766236,
+      "grad_norm": 0.39301300048828125,
+      "learning_rate": 4.082836493724981e-06,
+      "loss": 0.0204,
+      "step": 202
+    },
+    {
+      "epoch": 2.6363636363636362,
+      "grad_norm": 0.3674089014530182,
+      "learning_rate": 4.065983426600113e-06,
+      "loss": 0.0195,
+      "step": 203
+    },
+    {
+      "epoch": 2.6493506493506493,
+      "grad_norm": 0.40296265482902527,
+      "learning_rate": 4.0490124178684884e-06,
+      "loss": 0.021,
+      "step": 204
+    },
+    {
+      "epoch": 2.6623376623376624,
+      "grad_norm": 0.4691706895828247,
+      "learning_rate": 4.031924745696916e-06,
+      "loss": 0.0179,
+      "step": 205
+    },
+    {
+      "epoch": 2.675324675324675,
+      "grad_norm": 0.44443100690841675,
+      "learning_rate": 4.014721697038678e-06,
+      "loss": 0.0242,
+      "step": 206
+    },
+    {
+      "epoch": 2.688311688311688,
+      "grad_norm": 0.31682541966438293,
+      "learning_rate": 3.997404567536606e-06,
+      "loss": 0.0193,
+      "step": 207
+    },
+    {
+      "epoch": 2.7012987012987013,
+      "grad_norm": 0.33720874786376953,
+      "learning_rate": 3.979974661425497e-06,
+      "loss": 0.0212,
+      "step": 208
+    },
+    {
+      "epoch": 2.7142857142857144,
+      "grad_norm": 0.41582977771759033,
+      "learning_rate": 3.962433291433889e-06,
+      "loss": 0.0242,
+      "step": 209
+    },
+    {
+      "epoch": 2.7272727272727275,
+      "grad_norm": 0.4688783884048462,
+      "learning_rate": 3.944781778685189e-06,
+      "loss": 0.0218,
+      "step": 210
+    },
+    {
+      "epoch": 2.74025974025974,
+      "grad_norm": 0.3696669638156891,
+      "learning_rate": 3.927021452598177e-06,
+      "loss": 0.0209,
+      "step": 211
+    },
+    {
+      "epoch": 2.7532467532467533,
+      "grad_norm": 0.36734238266944885,
+      "learning_rate": 3.909153650786878e-06,
+      "loss": 0.0165,
+      "step": 212
+    },
+    {
+      "epoch": 2.7662337662337664,
+      "grad_norm": 0.33928048610687256,
+      "learning_rate": 3.891179718959822e-06,
+      "loss": 0.0137,
+      "step": 213
+    },
+    {
+      "epoch": 2.779220779220779,
+      "grad_norm": 0.656510055065155,
+      "learning_rate": 3.873101010818692e-06,
+      "loss": 0.0258,
+      "step": 214
+    },
+    {
+      "epoch": 2.792207792207792,
+      "grad_norm": 0.5539777278900146,
+      "learning_rate": 3.8549188879563685e-06,
+      "loss": 0.0208,
+      "step": 215
+    },
+    {
+      "epoch": 2.8051948051948052,
+      "grad_norm": 0.4914684593677521,
+      "learning_rate": 3.836634719754385e-06,
+      "loss": 0.0192,
+      "step": 216
+    },
+    {
+      "epoch": 2.8181818181818183,
+      "grad_norm": 0.38955259323120117,
+      "learning_rate": 3.818249883279791e-06,
+      "loss": 0.0185,
+      "step": 217
+    },
+    {
+      "epoch": 2.8311688311688314,
+      "grad_norm": 0.45634275674819946,
+      "learning_rate": 3.7997657631814366e-06,
+      "loss": 0.0202,
+      "step": 218
+    },
+    {
+      "epoch": 2.844155844155844,
+      "grad_norm": 0.41473978757858276,
+      "learning_rate": 3.781183751585693e-06,
+      "loss": 0.0133,
+      "step": 219
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 0.6281463503837585,
+      "learning_rate": 3.762505247991601e-06,
+      "loss": 0.0188,
+      "step": 220
+    },
+    {
+      "epoch": 2.87012987012987,
+      "grad_norm": 0.34762486815452576,
+      "learning_rate": 3.7437316591654726e-06,
+      "loss": 0.0199,
+      "step": 221
+    },
+    {
+      "epoch": 2.883116883116883,
+      "grad_norm": 0.43128135800361633,
+      "learning_rate": 3.724864399034932e-06,
+      "loss": 0.0175,
+      "step": 222
+    },
+    {
+      "epoch": 2.896103896103896,
+      "grad_norm": 0.4040665328502655,
+      "learning_rate": 3.7059048885824367e-06,
+      "loss": 0.0119,
+      "step": 223
+    },
+    {
+      "epoch": 2.909090909090909,
+      "grad_norm": 0.4182071089744568,
+      "learning_rate": 3.686854555738249e-06,
+      "loss": 0.0171,
+      "step": 224
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 0.38667839765548706,
+      "learning_rate": 3.6677148352728947e-06,
+      "loss": 0.0155,
+      "step": 225
+    },
+    {
+      "epoch": 2.935064935064935,
+      "grad_norm": 0.3983423709869385,
+      "learning_rate": 3.6484871686891044e-06,
+      "loss": 0.013,
+      "step": 226
+    },
+    {
+      "epoch": 2.948051948051948,
+      "grad_norm": 0.48213276267051697,
+      "learning_rate": 3.629173004113245e-06,
+      "loss": 0.0131,
+      "step": 227
+    },
+    {
+      "epoch": 2.961038961038961,
+      "grad_norm": 0.4538249969482422,
+      "learning_rate": 3.609773796186256e-06,
+      "loss": 0.0137,
+      "step": 228
+    },
+    {
+      "epoch": 2.974025974025974,
+      "grad_norm": 0.4226680397987366,
+      "learning_rate": 3.590291005954094e-06,
+      "loss": 0.0144,
+      "step": 229
+    },
+    {
+      "epoch": 2.987012987012987,
+      "grad_norm": 0.603485107421875,
+      "learning_rate": 3.570726100757693e-06,
+      "loss": 0.0179,
+      "step": 230
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.5114914774894714,
+      "learning_rate": 3.5510805541224536e-06,
+      "loss": 0.0134,
+      "step": 231
+    },
+    {
+      "epoch": 3.012987012987013,
+      "grad_norm": 0.5039042830467224,
+      "learning_rate": 3.5313558456472623e-06,
+      "loss": 0.0113,
+      "step": 232
+    },
+    {
+      "epoch": 3.0259740259740258,
+      "grad_norm": 0.37786880135536194,
+      "learning_rate": 3.511553460893059e-06,
+      "loss": 0.0089,
+      "step": 233
+    },
+    {
+      "epoch": 3.038961038961039,
+      "grad_norm": 0.31840819120407104,
+      "learning_rate": 3.4916748912709506e-06,
+      "loss": 0.0078,
+      "step": 234
+    },
+    {
+      "epoch": 3.051948051948052,
+      "grad_norm": 0.3462812006473541,
+      "learning_rate": 3.4717216339298852e-06,
+      "loss": 0.0095,
+      "step": 235
+    },
+    {
+      "epoch": 3.064935064935065,
+      "grad_norm": 0.28717392683029175,
+      "learning_rate": 3.4516951916438974e-06,
+      "loss": 0.0043,
+      "step": 236
+    },
+    {
+      "epoch": 3.0779220779220777,
+      "grad_norm": 0.4881342649459839,
+      "learning_rate": 3.4315970726989244e-06,
+      "loss": 0.0109,
+      "step": 237
+    },
+    {
+      "epoch": 3.090909090909091,
+      "grad_norm": 0.34244024753570557,
+      "learning_rate": 3.4114287907792115e-06,
+      "loss": 0.0047,
+      "step": 238
+    },
+    {
+      "epoch": 3.103896103896104,
+      "grad_norm": 0.3604426085948944,
+      "learning_rate": 3.3911918648533094e-06,
+      "loss": 0.0059,
+      "step": 239
+    },
+    {
+      "epoch": 3.116883116883117,
+      "grad_norm": 0.34059545397758484,
+      "learning_rate": 3.3708878190596724e-06,
+      "loss": 0.0037,
+      "step": 240
+    },
+    {
+      "epoch": 3.1298701298701297,
+      "grad_norm": 0.350000262260437,
+      "learning_rate": 3.3505181825918685e-06,
+      "loss": 0.0059,
+      "step": 241
+    },
+    {
+      "epoch": 3.142857142857143,
+      "grad_norm": 0.4721887409687042,
+      "learning_rate": 3.33008448958341e-06,
+      "loss": 0.0082,
+      "step": 242
+    },
+    {
+      "epoch": 3.155844155844156,
+      "grad_norm": 0.3560850918292999,
+      "learning_rate": 3.3095882789922102e-06,
+      "loss": 0.007,
+      "step": 243
+    },
+    {
+      "epoch": 3.168831168831169,
+      "grad_norm": 0.385452002286911,
+      "learning_rate": 3.289031094484675e-06,
+      "loss": 0.0047,
+      "step": 244
+    },
+    {
+      "epoch": 3.1818181818181817,
+      "grad_norm": 0.3492503762245178,
+      "learning_rate": 3.2684144843194453e-06,
+      "loss": 0.0053,
+      "step": 245
+    },
+    {
+      "epoch": 3.1948051948051948,
+      "grad_norm": 0.3584858179092407,
+      "learning_rate": 3.2477400012307885e-06,
+      "loss": 0.0077,
+      "step": 246
+    },
+    {
+      "epoch": 3.207792207792208,
+      "grad_norm": 0.2692736089229584,
+      "learning_rate": 3.2270092023116564e-06,
+      "loss": 0.0031,
+      "step": 247
+    },
+    {
+      "epoch": 3.220779220779221,
+      "grad_norm": 0.33909231424331665,
+      "learning_rate": 3.206223648896409e-06,
+      "loss": 0.0051,
+      "step": 248
+    },
+    {
+      "epoch": 3.2337662337662336,
+      "grad_norm": 0.47461092472076416,
+      "learning_rate": 3.1853849064432296e-06,
+      "loss": 0.0053,
+      "step": 249
+    },
+    {
+      "epoch": 3.2467532467532467,
+      "grad_norm": 0.46319177746772766,
+      "learning_rate": 3.1644945444162155e-06,
+      "loss": 0.0076,
+      "step": 250
+    },
+    {
+      "epoch": 3.25974025974026,
+      "grad_norm": 0.3297309875488281,
+      "learning_rate": 3.14355413616718e-06,
+      "loss": 0.0035,
+      "step": 251
+    },
+    {
+      "epoch": 3.2727272727272725,
+      "grad_norm": 0.3955172598361969,
+      "learning_rate": 3.1225652588171534e-06,
+      "loss": 0.0047,
+      "step": 252
+    },
+    {
+      "epoch": 3.2857142857142856,
+      "grad_norm": 0.2394171506166458,
+      "learning_rate": 3.1015294931376035e-06,
+      "loss": 0.0025,
+      "step": 253
+    },
+    {
+      "epoch": 3.2987012987012987,
+      "grad_norm": 0.24900038540363312,
+      "learning_rate": 3.080448423431377e-06,
+      "loss": 0.0028,
+      "step": 254
+    },
+    {
+      "epoch": 3.311688311688312,
+      "grad_norm": 0.3579304814338684,
+      "learning_rate": 3.059323637413385e-06,
+      "loss": 0.0031,
+      "step": 255
+    },
+    {
+      "epoch": 3.324675324675325,
+      "grad_norm": 0.2939773499965668,
+      "learning_rate": 3.0381567260910166e-06,
+      "loss": 0.0037,
+      "step": 256
+    },
+    {
+      "epoch": 3.3376623376623376,
+      "grad_norm": 0.39939820766448975,
+      "learning_rate": 3.016949283644319e-06,
+      "loss": 0.0055,
+      "step": 257
+    },
+    {
+      "epoch": 3.3506493506493507,
+      "grad_norm": 0.4479775130748749,
+      "learning_rate": 2.9957029073059276e-06,
+      "loss": 0.0038,
+      "step": 258
+    },
+    {
+      "epoch": 3.3636363636363638,
+      "grad_norm": 0.3758586049079895,
+      "learning_rate": 2.9744191972407754e-06,
+      "loss": 0.0034,
+      "step": 259
+    },
+    {
+      "epoch": 3.3766233766233764,
+      "grad_norm": 0.31011316180229187,
+      "learning_rate": 2.9530997564255728e-06,
+      "loss": 0.0022,
+      "step": 260
+    },
+    {
+      "epoch": 3.3896103896103895,
+      "grad_norm": 0.5502382516860962,
+      "learning_rate": 2.931746190528082e-06,
+      "loss": 0.0021,
+      "step": 261
+    },
+    {
+      "epoch": 3.4025974025974026,
+      "grad_norm": 0.285674512386322,
+      "learning_rate": 2.9103601077861875e-06,
+      "loss": 0.0027,
+      "step": 262
+    },
+    {
+      "epoch": 3.4155844155844157,
+      "grad_norm": 0.3391030430793762,
+      "learning_rate": 2.888943118886771e-06,
+      "loss": 0.0037,
+      "step": 263
+    },
+    {
+      "epoch": 3.4285714285714284,
+      "grad_norm": 0.313784122467041,
+      "learning_rate": 2.8674968368444004e-06,
+      "loss": 0.002,
+      "step": 264
+    },
+    {
+      "epoch": 3.4415584415584415,
+      "grad_norm": 0.2796953618526459,
+      "learning_rate": 2.8460228768798507e-06,
+      "loss": 0.0026,
+      "step": 265
+    },
+    {
+      "epoch": 3.4545454545454546,
+      "grad_norm": 0.3058723509311676,
+      "learning_rate": 2.8245228562984518e-06,
+      "loss": 0.0025,
+      "step": 266
+    },
+    {
+      "epoch": 3.4675324675324677,
+      "grad_norm": 0.54490727186203,
+      "learning_rate": 2.80299839436828e-06,
+      "loss": 0.0043,
+      "step": 267
+    },
+    {
+      "epoch": 3.4805194805194803,
+      "grad_norm": 0.41025906801223755,
+      "learning_rate": 2.781451112198208e-06,
+      "loss": 0.0027,
+      "step": 268
+    },
+    {
+      "epoch": 3.4935064935064934,
+      "grad_norm": 0.35861679911613464,
+      "learning_rate": 2.7598826326158045e-06,
+      "loss": 0.004,
+      "step": 269
+    },
+    {
+      "epoch": 3.5064935064935066,
+      "grad_norm": 0.40969038009643555,
+      "learning_rate": 2.738294580045119e-06,
+      "loss": 0.0031,
+      "step": 270
+    },
+    {
+      "epoch": 3.5194805194805197,
+      "grad_norm": 0.4820692241191864,
+      "learning_rate": 2.7166885803843347e-06,
+      "loss": 0.0027,
+      "step": 271
+    },
+    {
+      "epoch": 3.5324675324675323,
+      "grad_norm": 0.3031913936138153,
+      "learning_rate": 2.695066260883313e-06,
+      "loss": 0.0013,
+      "step": 272
+    },
+    {
+      "epoch": 3.5454545454545454,
+      "grad_norm": 0.3935272693634033,
+      "learning_rate": 2.67342925002104e-06,
+      "loss": 0.0027,
+      "step": 273
+    },
+    {
+      "epoch": 3.5584415584415585,
+      "grad_norm": 0.2262854278087616,
+      "learning_rate": 2.65177917738298e-06,
+      "loss": 0.0019,
+      "step": 274
+    },
+    {
+      "epoch": 3.571428571428571,
+      "grad_norm": 0.2690446972846985,
+      "learning_rate": 2.6301176735383382e-06,
+      "loss": 0.0014,
+      "step": 275
+    },
+    {
+      "epoch": 3.5844155844155843,
+      "grad_norm": 0.33330366015434265,
+      "learning_rate": 2.6084463699172594e-06,
+      "loss": 0.0025,
+      "step": 276
+    },
+    {
+      "epoch": 3.5974025974025974,
+      "grad_norm": 0.3452131748199463,
+      "learning_rate": 2.586766898687955e-06,
+      "loss": 0.0018,
+      "step": 277
+    },
+    {
+      "epoch": 3.6103896103896105,
+      "grad_norm": 0.3789026737213135,
+      "learning_rate": 2.56508089263378e-06,
+      "loss": 0.001,
+      "step": 278
+    },
+    {
+      "epoch": 3.6233766233766236,
+      "grad_norm": 0.3677704334259033,
+      "learning_rate": 2.5433899850302552e-06,
+      "loss": 0.0013,
+      "step": 279
+    },
+    {
+      "epoch": 3.6363636363636362,
+      "grad_norm": 0.4359056055545807,
+      "learning_rate": 2.521695809522061e-06,
+      "loss": 0.0032,
+      "step": 280
+    },
+    {
+      "epoch": 3.6493506493506493,
+      "grad_norm": 0.2561199367046356,
+      "learning_rate": 2.5e-06,
+      "loss": 0.0018,
+      "step": 281
+    },
+    {
+      "epoch": 3.6623376623376624,
+      "grad_norm": 0.39096924662590027,
+      "learning_rate": 2.4783041904779386e-06,
+      "loss": 0.0031,
+      "step": 282
+    },
+    {
+      "epoch": 3.675324675324675,
+      "grad_norm": 0.3999404013156891,
+      "learning_rate": 2.4566100149697456e-06,
+      "loss": 0.0019,
+      "step": 283
+    },
+    {
+      "epoch": 3.688311688311688,
+      "grad_norm": 0.8379652500152588,
+      "learning_rate": 2.4349191073662203e-06,
+      "loss": 0.004,
+      "step": 284
+    },
+    {
+      "epoch": 3.7012987012987013,
+      "grad_norm": 0.3195594847202301,
+      "learning_rate": 2.4132331013120454e-06,
+      "loss": 0.0011,
+      "step": 285
+    },
+    {
+      "epoch": 3.7142857142857144,
+      "grad_norm": 0.2558881342411041,
+      "learning_rate": 2.3915536300827414e-06,
+      "loss": 0.001,
+      "step": 286
+    },
+    {
+      "epoch": 3.7272727272727275,
+      "grad_norm": 0.284496933221817,
+      "learning_rate": 2.3698823264616635e-06,
+      "loss": 0.0018,
+      "step": 287
+    },
+    {
+      "epoch": 3.74025974025974,
+      "grad_norm": 0.20427702367305756,
+      "learning_rate": 2.348220822617021e-06,
+      "loss": 0.0014,
+      "step": 288
+    },
+    {
+      "epoch": 3.7532467532467533,
+      "grad_norm": 0.32599779963493347,
+      "learning_rate": 2.32657074997896e-06,
+      "loss": 0.002,
+      "step": 289
+    },
+    {
+      "epoch": 3.7662337662337664,
+      "grad_norm": 0.3040351867675781,
+      "learning_rate": 2.3049337391166884e-06,
+      "loss": 0.0011,
+      "step": 290
+    },
+    {
+      "epoch": 3.779220779220779,
+      "grad_norm": 0.2709053158760071,
+      "learning_rate": 2.2833114196156657e-06,
+      "loss": 0.0008,
+      "step": 291
+    },
+    {
+      "epoch": 3.792207792207792,
+      "grad_norm": 0.2777629792690277,
+      "learning_rate": 2.261705419954882e-06,
+      "loss": 0.0014,
+      "step": 292
+    },
+    {
+      "epoch": 3.8051948051948052,
+      "grad_norm": 0.3083540201187134,
+      "learning_rate": 2.2401173673841963e-06,
+      "loss": 0.0016,
+      "step": 293
+    },
+    {
+      "epoch": 3.8181818181818183,
+      "grad_norm": 0.1168127954006195,
+      "learning_rate": 2.2185488878017934e-06,
+      "loss": 0.0005,
+      "step": 294
+    },
+    {
+      "epoch": 3.8311688311688314,
+      "grad_norm": 0.25492948293685913,
+      "learning_rate": 2.1970016056317202e-06,
+      "loss": 0.001,
+      "step": 295
+    },
+    {
+      "epoch": 3.844155844155844,
+      "grad_norm": 0.3559608459472656,
+      "learning_rate": 2.1754771437015495e-06,
+      "loss": 0.0009,
+      "step": 296
+    },
+    {
+      "epoch": 3.857142857142857,
+      "grad_norm": 0.43275198340415955,
+      "learning_rate": 2.1539771231201497e-06,
+      "loss": 0.002,
+      "step": 297
+    },
+    {
+      "epoch": 3.87012987012987,
+      "grad_norm": 0.35539472103118896,
+      "learning_rate": 2.1325031631555996e-06,
+      "loss": 0.0012,
+      "step": 298
+    },
+    {
+      "epoch": 3.883116883116883,
+      "grad_norm": 0.2857511341571808,
+      "learning_rate": 2.11105688111323e-06,
+      "loss": 0.0017,
+      "step": 299
+    },
+    {
+      "epoch": 3.896103896103896,
+      "grad_norm": 0.22510217130184174,
+      "learning_rate": 2.0896398922138124e-06,
+      "loss": 0.0007,
+      "step": 300
+    },
+    {
+      "epoch": 3.909090909090909,
+      "grad_norm": 0.15990178287029266,
+      "learning_rate": 2.0682538094719183e-06,
+      "loss": 0.0011,
+      "step": 301
+    },
+    {
+      "epoch": 3.9220779220779223,
+      "grad_norm": 0.11588839441537857,
+      "learning_rate": 2.0469002435744285e-06,
+      "loss": 0.0007,
+      "step": 302
+    },
+    {
+      "epoch": 3.935064935064935,
+      "grad_norm": 0.36557283997535706,
+      "learning_rate": 2.0255808027592263e-06,
+      "loss": 0.0008,
+      "step": 303
+    },
+    {
+      "epoch": 3.948051948051948,
+      "grad_norm": 0.2867473363876343,
+      "learning_rate": 2.004297092694073e-06,
+      "loss": 0.001,
+      "step": 304
+    },
+    {
+      "epoch": 3.961038961038961,
+      "grad_norm": 0.270842969417572,
+      "learning_rate": 1.9830507163556815e-06,
+      "loss": 0.0008,
+      "step": 305
+    },
+    {
+      "epoch": 3.974025974025974,
+      "grad_norm": 0.24662533402442932,
+      "learning_rate": 1.9618432739089843e-06,
+      "loss": 0.0006,
+      "step": 306
+    },
+    {
+      "epoch": 3.987012987012987,
+      "grad_norm": 0.12252429127693176,
+      "learning_rate": 1.9406763625866155e-06,
+      "loss": 0.0004,
+      "step": 307
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 0.24726521968841553,
+      "learning_rate": 1.9195515765686237e-06,
+      "loss": 0.0004,
+      "step": 308
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 462,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 6,
+  "save_steps": 77,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.88470616386306e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-308/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92007cb7b6ef33ad8c015d3ab5f86bb507fe352d75f5b7187cbf5d8f57b8447a
+size 7992

checkpoint-308/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-385/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-385/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": [
+    "gate_proj",
+    "o_proj",
+    "v_proj",
+    "up_proj",
+    "k_proj",
+    "down_proj",
+    "q_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-385/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db023f51489cc9d0bb40a257beb0c6bca9a83871789641a30c134855defa3567
+size 1291899552

checkpoint-385/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-385/global_step385/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19877bf1b7f70871798b21bc99374bd1dfa16da4f0bdef71d1aabadd43e1f6ed
+size 7751120668

checkpoint-385/global_step385/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac45db1bdae68226d19ce5279d2360eab77286b2391fe41bb84d892b61b08ba6
+size 1396311192

checkpoint-385/latest

@@ -0,0 +1 @@
+global_step385

checkpoint-385/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55676dea98ea045a11ea22e347f99175a9be466c320c600671bd838743d0be8c
+size 14244

checkpoint-385/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ad2c42e997ca01bc69d654c08c3bceef9bc110a0b3a6ea9f3ae2a076d3d6c734
+size 1064

checkpoint-385/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-385/tokenizer.json

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

checkpoint-385/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-385/trainer_state.json

@@ -0,0 +1,2728 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.0,
+  "eval_steps": 500,
+  "global_step": 385,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.012987012987012988,
+      "grad_norm": 30.513944625854492,
+      "learning_rate": 5.0000000000000004e-08,
+      "loss": 4.1013,
+      "step": 1
+    },
+    {
+      "epoch": 0.025974025974025976,
+      "grad_norm": 30.965682983398438,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 4.1757,
+      "step": 2
+    },
+    {
+      "epoch": 0.03896103896103896,
+      "grad_norm": 30.834516525268555,
+      "learning_rate": 1.5000000000000002e-07,
+      "loss": 4.1371,
+      "step": 3
+    },
+    {
+      "epoch": 0.05194805194805195,
+      "grad_norm": 30.98870849609375,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 4.0543,
+      "step": 4
+    },
+    {
+      "epoch": 0.06493506493506493,
+      "grad_norm": 30.404874801635742,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 4.09,
+      "step": 5
+    },
+    {
+      "epoch": 0.07792207792207792,
+      "grad_norm": 31.433774948120117,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 4.1777,
+      "step": 6
+    },
+    {
+      "epoch": 0.09090909090909091,
+      "grad_norm": 30.850692749023438,
+      "learning_rate": 3.5000000000000004e-07,
+      "loss": 4.142,
+      "step": 7
+    },
+    {
+      "epoch": 0.1038961038961039,
+      "grad_norm": 31.00360870361328,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 4.1115,
+      "step": 8
+    },
+    {
+      "epoch": 0.11688311688311688,
+      "grad_norm": 31.221616744995117,
+      "learning_rate": 4.5000000000000003e-07,
+      "loss": 4.1585,
+      "step": 9
+    },
+    {
+      "epoch": 0.12987012987012986,
+      "grad_norm": 30.560623168945312,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 4.1155,
+      "step": 10
+    },
+    {
+      "epoch": 0.14285714285714285,
+      "grad_norm": 30.089527130126953,
+      "learning_rate": 5.5e-07,
+      "loss": 4.0638,
+      "step": 11
+    },
+    {
+      "epoch": 0.15584415584415584,
+      "grad_norm": 30.662569046020508,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 4.0152,
+      "step": 12
+    },
+    {
+      "epoch": 0.16883116883116883,
+      "grad_norm": 29.455503463745117,
+      "learning_rate": 6.5e-07,
+      "loss": 3.9501,
+      "step": 13
+    },
+    {
+      "epoch": 0.18181818181818182,
+      "grad_norm": 28.851863861083984,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 3.9066,
+      "step": 14
+    },
+    {
+      "epoch": 0.19480519480519481,
+      "grad_norm": 27.41941261291504,
+      "learning_rate": 7.5e-07,
+      "loss": 3.816,
+      "step": 15
+    },
+    {
+      "epoch": 0.2077922077922078,
+      "grad_norm": 25.754905700683594,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 3.6715,
+      "step": 16
+    },
+    {
+      "epoch": 0.22077922077922077,
+      "grad_norm": 24.467802047729492,
+      "learning_rate": 8.500000000000001e-07,
+      "loss": 3.6226,
+      "step": 17
+    },
+    {
+      "epoch": 0.23376623376623376,
+      "grad_norm": 23.671884536743164,
+      "learning_rate": 9.000000000000001e-07,
+      "loss": 3.5756,
+      "step": 18
+    },
+    {
+      "epoch": 0.24675324675324675,
+      "grad_norm": 21.28356170654297,
+      "learning_rate": 9.500000000000001e-07,
+      "loss": 3.3822,
+      "step": 19
+    },
+    {
+      "epoch": 0.2597402597402597,
+      "grad_norm": 20.54037857055664,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 3.3799,
+      "step": 20
+    },
+    {
+      "epoch": 0.2727272727272727,
+      "grad_norm": 18.18222427368164,
+      "learning_rate": 1.0500000000000001e-06,
+      "loss": 3.1964,
+      "step": 21
+    },
+    {
+      "epoch": 0.2857142857142857,
+      "grad_norm": 17.782703399658203,
+      "learning_rate": 1.1e-06,
+      "loss": 3.1745,
+      "step": 22
+    },
+    {
+      "epoch": 0.2987012987012987,
+      "grad_norm": 15.965900421142578,
+      "learning_rate": 1.1500000000000002e-06,
+      "loss": 2.9002,
+      "step": 23
+    },
+    {
+      "epoch": 0.3116883116883117,
+      "grad_norm": 16.00725746154785,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 2.9123,
+      "step": 24
+    },
+    {
+      "epoch": 0.3246753246753247,
+      "grad_norm": 15.301984786987305,
+      "learning_rate": 1.25e-06,
+      "loss": 2.6531,
+      "step": 25
+    },
+    {
+      "epoch": 0.33766233766233766,
+      "grad_norm": 16.359207153320312,
+      "learning_rate": 1.3e-06,
+      "loss": 2.6089,
+      "step": 26
+    },
+    {
+      "epoch": 0.35064935064935066,
+      "grad_norm": 17.74176025390625,
+      "learning_rate": 1.3500000000000002e-06,
+      "loss": 2.4805,
+      "step": 27
+    },
+    {
+      "epoch": 0.36363636363636365,
+      "grad_norm": 18.702547073364258,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 2.3312,
+      "step": 28
+    },
+    {
+      "epoch": 0.37662337662337664,
+      "grad_norm": 18.44041633605957,
+      "learning_rate": 1.45e-06,
+      "loss": 2.2273,
+      "step": 29
+    },
+    {
+      "epoch": 0.38961038961038963,
+      "grad_norm": 15.867247581481934,
+      "learning_rate": 1.5e-06,
+      "loss": 2.0514,
+      "step": 30
+    },
+    {
+      "epoch": 0.4025974025974026,
+      "grad_norm": 13.97642707824707,
+      "learning_rate": 1.5500000000000002e-06,
+      "loss": 1.8689,
+      "step": 31
+    },
+    {
+      "epoch": 0.4155844155844156,
+      "grad_norm": 13.599323272705078,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 1.7907,
+      "step": 32
+    },
+    {
+      "epoch": 0.42857142857142855,
+      "grad_norm": 13.204086303710938,
+      "learning_rate": 1.6500000000000003e-06,
+      "loss": 1.6537,
+      "step": 33
+    },
+    {
+      "epoch": 0.44155844155844154,
+      "grad_norm": 13.255966186523438,
+      "learning_rate": 1.7000000000000002e-06,
+      "loss": 1.5693,
+      "step": 34
+    },
+    {
+      "epoch": 0.45454545454545453,
+      "grad_norm": 13.00086784362793,
+      "learning_rate": 1.75e-06,
+      "loss": 1.3916,
+      "step": 35
+    },
+    {
+      "epoch": 0.4675324675324675,
+      "grad_norm": 12.996549606323242,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 1.2734,
+      "step": 36
+    },
+    {
+      "epoch": 0.4805194805194805,
+      "grad_norm": 12.506110191345215,
+      "learning_rate": 1.85e-06,
+      "loss": 1.1444,
+      "step": 37
+    },
+    {
+      "epoch": 0.4935064935064935,
+      "grad_norm": 12.379636764526367,
+      "learning_rate": 1.9000000000000002e-06,
+      "loss": 1.033,
+      "step": 38
+    },
+    {
+      "epoch": 0.5064935064935064,
+      "grad_norm": 12.38772201538086,
+      "learning_rate": 1.9500000000000004e-06,
+      "loss": 0.8932,
+      "step": 39
+    },
+    {
+      "epoch": 0.5194805194805194,
+      "grad_norm": 12.370598793029785,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.7914,
+      "step": 40
+    },
+    {
+      "epoch": 0.5324675324675324,
+      "grad_norm": 11.510255813598633,
+      "learning_rate": 2.05e-06,
+      "loss": 0.6336,
+      "step": 41
+    },
+    {
+      "epoch": 0.5454545454545454,
+      "grad_norm": 10.088979721069336,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 0.5115,
+      "step": 42
+    },
+    {
+      "epoch": 0.5584415584415584,
+      "grad_norm": 9.241965293884277,
+      "learning_rate": 2.15e-06,
+      "loss": 0.416,
+      "step": 43
+    },
+    {
+      "epoch": 0.5714285714285714,
+      "grad_norm": 7.457204341888428,
+      "learning_rate": 2.2e-06,
+      "loss": 0.2947,
+      "step": 44
+    },
+    {
+      "epoch": 0.5844155844155844,
+      "grad_norm": 6.164954662322998,
+      "learning_rate": 2.25e-06,
+      "loss": 0.2202,
+      "step": 45
+    },
+    {
+      "epoch": 0.5974025974025974,
+      "grad_norm": 4.587556838989258,
+      "learning_rate": 2.3000000000000004e-06,
+      "loss": 0.1721,
+      "step": 46
+    },
+    {
+      "epoch": 0.6103896103896104,
+      "grad_norm": 3.2366952896118164,
+      "learning_rate": 2.35e-06,
+      "loss": 0.1451,
+      "step": 47
+    },
+    {
+      "epoch": 0.6233766233766234,
+      "grad_norm": 2.397469997406006,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 0.1116,
+      "step": 48
+    },
+    {
+      "epoch": 0.6363636363636364,
+      "grad_norm": 1.872839093208313,
+      "learning_rate": 2.4500000000000003e-06,
+      "loss": 0.1,
+      "step": 49
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 1.2464176416397095,
+      "learning_rate": 2.5e-06,
+      "loss": 0.1024,
+      "step": 50
+    },
+    {
+      "epoch": 0.6623376623376623,
+      "grad_norm": 1.0786665678024292,
+      "learning_rate": 2.55e-06,
+      "loss": 0.087,
+      "step": 51
+    },
+    {
+      "epoch": 0.6753246753246753,
+      "grad_norm": 1.10042142868042,
+      "learning_rate": 2.6e-06,
+      "loss": 0.0838,
+      "step": 52
+    },
+    {
+      "epoch": 0.6883116883116883,
+      "grad_norm": 0.9031299948692322,
+      "learning_rate": 2.6500000000000005e-06,
+      "loss": 0.0804,
+      "step": 53
+    },
+    {
+      "epoch": 0.7012987012987013,
+      "grad_norm": 0.8072167634963989,
+      "learning_rate": 2.7000000000000004e-06,
+      "loss": 0.0786,
+      "step": 54
+    },
+    {
+      "epoch": 0.7142857142857143,
+      "grad_norm": 0.6948329210281372,
+      "learning_rate": 2.7500000000000004e-06,
+      "loss": 0.0773,
+      "step": 55
+    },
+    {
+      "epoch": 0.7272727272727273,
+      "grad_norm": 0.6062957644462585,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 0.0756,
+      "step": 56
+    },
+    {
+      "epoch": 0.7402597402597403,
+      "grad_norm": 0.4804394841194153,
+      "learning_rate": 2.85e-06,
+      "loss": 0.0697,
+      "step": 57
+    },
+    {
+      "epoch": 0.7532467532467533,
+      "grad_norm": 0.5353919863700867,
+      "learning_rate": 2.9e-06,
+      "loss": 0.0697,
+      "step": 58
+    },
+    {
+      "epoch": 0.7662337662337663,
+      "grad_norm": 0.4344547390937805,
+      "learning_rate": 2.95e-06,
+      "loss": 0.0704,
+      "step": 59
+    },
+    {
+      "epoch": 0.7792207792207793,
+      "grad_norm": 0.574044942855835,
+      "learning_rate": 3e-06,
+      "loss": 0.0778,
+      "step": 60
+    },
+    {
+      "epoch": 0.7922077922077922,
+      "grad_norm": 0.5392284393310547,
+      "learning_rate": 3.05e-06,
+      "loss": 0.0658,
+      "step": 61
+    },
+    {
+      "epoch": 0.8051948051948052,
+      "grad_norm": 0.5098194479942322,
+      "learning_rate": 3.1000000000000004e-06,
+      "loss": 0.0661,
+      "step": 62
+    },
+    {
+      "epoch": 0.8181818181818182,
+      "grad_norm": 0.3733481168746948,
+      "learning_rate": 3.1500000000000003e-06,
+      "loss": 0.0667,
+      "step": 63
+    },
+    {
+      "epoch": 0.8311688311688312,
+      "grad_norm": 0.3714415431022644,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 0.0635,
+      "step": 64
+    },
+    {
+      "epoch": 0.8441558441558441,
+      "grad_norm": 0.37473437190055847,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.0661,
+      "step": 65
+    },
+    {
+      "epoch": 0.8571428571428571,
+      "grad_norm": 0.35993579030036926,
+      "learning_rate": 3.3000000000000006e-06,
+      "loss": 0.0637,
+      "step": 66
+    },
+    {
+      "epoch": 0.8701298701298701,
+      "grad_norm": 0.41175705194473267,
+      "learning_rate": 3.3500000000000005e-06,
+      "loss": 0.0646,
+      "step": 67
+    },
+    {
+      "epoch": 0.8831168831168831,
+      "grad_norm": 0.29564622044563293,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 0.0627,
+      "step": 68
+    },
+    {
+      "epoch": 0.8961038961038961,
+      "grad_norm": 0.3018590211868286,
+      "learning_rate": 3.45e-06,
+      "loss": 0.0613,
+      "step": 69
+    },
+    {
+      "epoch": 0.9090909090909091,
+      "grad_norm": 0.499183714389801,
+      "learning_rate": 3.5e-06,
+      "loss": 0.0664,
+      "step": 70
+    },
+    {
+      "epoch": 0.922077922077922,
+      "grad_norm": 0.3450847864151001,
+      "learning_rate": 3.5500000000000003e-06,
+      "loss": 0.0589,
+      "step": 71
+    },
+    {
+      "epoch": 0.935064935064935,
+      "grad_norm": 0.26693543791770935,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 0.0633,
+      "step": 72
+    },
+    {
+      "epoch": 0.948051948051948,
+      "grad_norm": 0.2867635190486908,
+      "learning_rate": 3.65e-06,
+      "loss": 0.062,
+      "step": 73
+    },
+    {
+      "epoch": 0.961038961038961,
+      "grad_norm": 0.3328873813152313,
+      "learning_rate": 3.7e-06,
+      "loss": 0.0587,
+      "step": 74
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 0.33611243963241577,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.058,
+      "step": 75
+    },
+    {
+      "epoch": 0.987012987012987,
+      "grad_norm": 0.2983446419239044,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.0632,
+      "step": 76
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.3246895372867584,
+      "learning_rate": 3.85e-06,
+      "loss": 0.0583,
+      "step": 77
+    },
+    {
+      "epoch": 1.0129870129870129,
+      "grad_norm": 0.2912401556968689,
+      "learning_rate": 3.900000000000001e-06,
+      "loss": 0.0609,
+      "step": 78
+    },
+    {
+      "epoch": 1.025974025974026,
+      "grad_norm": 0.37205109000205994,
+      "learning_rate": 3.95e-06,
+      "loss": 0.0646,
+      "step": 79
+    },
+    {
+      "epoch": 1.0389610389610389,
+      "grad_norm": 0.2430678755044937,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.0567,
+      "step": 80
+    },
+    {
+      "epoch": 1.051948051948052,
+      "grad_norm": 0.32996001839637756,
+      "learning_rate": 4.05e-06,
+      "loss": 0.0517,
+      "step": 81
+    },
+    {
+      "epoch": 1.0649350649350648,
+      "grad_norm": 0.274720698595047,
+      "learning_rate": 4.1e-06,
+      "loss": 0.0579,
+      "step": 82
+    },
+    {
+      "epoch": 1.077922077922078,
+      "grad_norm": 0.3047677278518677,
+      "learning_rate": 4.15e-06,
+      "loss": 0.0575,
+      "step": 83
+    },
+    {
+      "epoch": 1.0909090909090908,
+      "grad_norm": 0.2987304925918579,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 0.0576,
+      "step": 84
+    },
+    {
+      "epoch": 1.103896103896104,
+      "grad_norm": 0.32784077525138855,
+      "learning_rate": 4.25e-06,
+      "loss": 0.0607,
+      "step": 85
+    },
+    {
+      "epoch": 1.1168831168831168,
+      "grad_norm": 0.254140704870224,
+      "learning_rate": 4.3e-06,
+      "loss": 0.0582,
+      "step": 86
+    },
+    {
+      "epoch": 1.12987012987013,
+      "grad_norm": 0.2915174663066864,
+      "learning_rate": 4.350000000000001e-06,
+      "loss": 0.0614,
+      "step": 87
+    },
+    {
+      "epoch": 1.1428571428571428,
+      "grad_norm": 0.2874818444252014,
+      "learning_rate": 4.4e-06,
+      "loss": 0.0541,
+      "step": 88
+    },
+    {
+      "epoch": 1.155844155844156,
+      "grad_norm": 0.27614182233810425,
+      "learning_rate": 4.450000000000001e-06,
+      "loss": 0.0556,
+      "step": 89
+    },
+    {
+      "epoch": 1.1688311688311688,
+      "grad_norm": 0.3335096836090088,
+      "learning_rate": 4.5e-06,
+      "loss": 0.0601,
+      "step": 90
+    },
+    {
+      "epoch": 1.1818181818181819,
+      "grad_norm": 0.2302844524383545,
+      "learning_rate": 4.5500000000000005e-06,
+      "loss": 0.056,
+      "step": 91
+    },
+    {
+      "epoch": 1.1948051948051948,
+      "grad_norm": 0.22703984379768372,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 0.0577,
+      "step": 92
+    },
+    {
+      "epoch": 1.2077922077922079,
+      "grad_norm": 0.2960239350795746,
+      "learning_rate": 4.65e-06,
+      "loss": 0.0597,
+      "step": 93
+    },
+    {
+      "epoch": 1.2207792207792207,
+      "grad_norm": 0.20548498630523682,
+      "learning_rate": 4.7e-06,
+      "loss": 0.0531,
+      "step": 94
+    },
+    {
+      "epoch": 1.2337662337662338,
+      "grad_norm": 0.2792559564113617,
+      "learning_rate": 4.75e-06,
+      "loss": 0.0552,
+      "step": 95
+    },
+    {
+      "epoch": 1.2467532467532467,
+      "grad_norm": 0.25889474153518677,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.053,
+      "step": 96
+    },
+    {
+      "epoch": 1.2597402597402598,
+      "grad_norm": 0.2224734127521515,
+      "learning_rate": 4.85e-06,
+      "loss": 0.0528,
+      "step": 97
+    },
+    {
+      "epoch": 1.2727272727272727,
+      "grad_norm": 0.2264651507139206,
+      "learning_rate": 4.9000000000000005e-06,
+      "loss": 0.0545,
+      "step": 98
+    },
+    {
+      "epoch": 1.2857142857142856,
+      "grad_norm": 0.29424387216567993,
+      "learning_rate": 4.95e-06,
+      "loss": 0.0571,
+      "step": 99
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 0.3200651705265045,
+      "learning_rate": 5e-06,
+      "loss": 0.0605,
+      "step": 100
+    },
+    {
+      "epoch": 1.3116883116883118,
+      "grad_norm": 0.284705251455307,
+      "learning_rate": 4.999905856597241e-06,
+      "loss": 0.0512,
+      "step": 101
+    },
+    {
+      "epoch": 1.3246753246753247,
+      "grad_norm": 0.25781720876693726,
+      "learning_rate": 4.999623433479346e-06,
+      "loss": 0.053,
+      "step": 102
+    },
+    {
+      "epoch": 1.3376623376623376,
+      "grad_norm": 0.2585068643093109,
+      "learning_rate": 4.999152751916936e-06,
+      "loss": 0.0557,
+      "step": 103
+    },
+    {
+      "epoch": 1.3506493506493507,
+      "grad_norm": 0.24417804181575775,
+      "learning_rate": 4.99849384735926e-06,
+      "loss": 0.0599,
+      "step": 104
+    },
+    {
+      "epoch": 1.3636363636363638,
+      "grad_norm": 0.2860206067562103,
+      "learning_rate": 4.997646769431532e-06,
+      "loss": 0.053,
+      "step": 105
+    },
+    {
+      "epoch": 1.3766233766233766,
+      "grad_norm": 0.22996841371059418,
+      "learning_rate": 4.9966115819311926e-06,
+      "loss": 0.0544,
+      "step": 106
+    },
+    {
+      "epoch": 1.3896103896103895,
+      "grad_norm": 0.2630089819431305,
+      "learning_rate": 4.9953883628231e-06,
+      "loss": 0.0509,
+      "step": 107
+    },
+    {
+      "epoch": 1.4025974025974026,
+      "grad_norm": 0.2588723599910736,
+      "learning_rate": 4.99397720423366e-06,
+      "loss": 0.0521,
+      "step": 108
+    },
+    {
+      "epoch": 1.4155844155844157,
+      "grad_norm": 0.3795599937438965,
+      "learning_rate": 4.992378212443891e-06,
+      "loss": 0.0527,
+      "step": 109
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 0.3013223111629486,
+      "learning_rate": 4.990591507881416e-06,
+      "loss": 0.0523,
+      "step": 110
+    },
+    {
+      "epoch": 1.4415584415584415,
+      "grad_norm": 0.24418188631534576,
+      "learning_rate": 4.988617225111392e-06,
+      "loss": 0.0494,
+      "step": 111
+    },
+    {
+      "epoch": 1.4545454545454546,
+      "grad_norm": 0.3064674735069275,
+      "learning_rate": 4.986455512826377e-06,
+      "loss": 0.0438,
+      "step": 112
+    },
+    {
+      "epoch": 1.4675324675324675,
+      "grad_norm": 0.2957422137260437,
+      "learning_rate": 4.984106533835132e-06,
+      "loss": 0.0496,
+      "step": 113
+    },
+    {
+      "epoch": 1.4805194805194806,
+      "grad_norm": 0.21359635889530182,
+      "learning_rate": 4.981570465050357e-06,
+      "loss": 0.0508,
+      "step": 114
+    },
+    {
+      "epoch": 1.4935064935064934,
+      "grad_norm": 0.20495343208312988,
+      "learning_rate": 4.978847497475369e-06,
+      "loss": 0.0535,
+      "step": 115
+    },
+    {
+      "epoch": 1.5064935064935066,
+      "grad_norm": 0.26068198680877686,
+      "learning_rate": 4.975937836189712e-06,
+      "loss": 0.0485,
+      "step": 116
+    },
+    {
+      "epoch": 1.5194805194805194,
+      "grad_norm": 0.26806536316871643,
+      "learning_rate": 4.97284170033372e-06,
+      "loss": 0.0505,
+      "step": 117
+    },
+    {
+      "epoch": 1.5324675324675323,
+      "grad_norm": 0.24916547536849976,
+      "learning_rate": 4.969559323092004e-06,
+      "loss": 0.0507,
+      "step": 118
+    },
+    {
+      "epoch": 1.5454545454545454,
+      "grad_norm": 0.26345163583755493,
+      "learning_rate": 4.966090951675893e-06,
+      "loss": 0.049,
+      "step": 119
+    },
+    {
+      "epoch": 1.5584415584415585,
+      "grad_norm": 0.2529750168323517,
+      "learning_rate": 4.962436847304818e-06,
+      "loss": 0.0471,
+      "step": 120
+    },
+    {
+      "epoch": 1.5714285714285714,
+      "grad_norm": 0.2879778742790222,
+      "learning_rate": 4.958597285186635e-06,
+      "loss": 0.047,
+      "step": 121
+    },
+    {
+      "epoch": 1.5844155844155843,
+      "grad_norm": 0.24268512427806854,
+      "learning_rate": 4.954572554496897e-06,
+      "loss": 0.0448,
+      "step": 122
+    },
+    {
+      "epoch": 1.5974025974025974,
+      "grad_norm": 0.3030378520488739,
+      "learning_rate": 4.950362958357078e-06,
+      "loss": 0.0483,
+      "step": 123
+    },
+    {
+      "epoch": 1.6103896103896105,
+      "grad_norm": 0.3552945852279663,
+      "learning_rate": 4.945968813811743e-06,
+      "loss": 0.0518,
+      "step": 124
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 0.32233837246894836,
+      "learning_rate": 4.941390451804668e-06,
+      "loss": 0.0488,
+      "step": 125
+    },
+    {
+      "epoch": 1.6363636363636362,
+      "grad_norm": 0.24810399115085602,
+      "learning_rate": 4.936628217153914e-06,
+      "loss": 0.0487,
+      "step": 126
+    },
+    {
+      "epoch": 1.6493506493506493,
+      "grad_norm": 0.2843495011329651,
+      "learning_rate": 4.931682468525863e-06,
+      "loss": 0.0485,
+      "step": 127
+    },
+    {
+      "epoch": 1.6623376623376624,
+      "grad_norm": 0.4453549087047577,
+      "learning_rate": 4.9265535784081965e-06,
+      "loss": 0.0441,
+      "step": 128
+    },
+    {
+      "epoch": 1.6753246753246753,
+      "grad_norm": 0.2711978852748871,
+      "learning_rate": 4.92124193308185e-06,
+      "loss": 0.0456,
+      "step": 129
+    },
+    {
+      "epoch": 1.6883116883116882,
+      "grad_norm": 0.2913718521595001,
+      "learning_rate": 4.915747932591916e-06,
+      "loss": 0.0478,
+      "step": 130
+    },
+    {
+      "epoch": 1.7012987012987013,
+      "grad_norm": 0.26739779114723206,
+      "learning_rate": 4.910071990717516e-06,
+      "loss": 0.0418,
+      "step": 131
+    },
+    {
+      "epoch": 1.7142857142857144,
+      "grad_norm": 0.33316293358802795,
+      "learning_rate": 4.9042145349406335e-06,
+      "loss": 0.0466,
+      "step": 132
+    },
+    {
+      "epoch": 1.7272727272727273,
+      "grad_norm": 0.3216155469417572,
+      "learning_rate": 4.898176006413925e-06,
+      "loss": 0.043,
+      "step": 133
+    },
+    {
+      "epoch": 1.7402597402597402,
+      "grad_norm": 0.3745043873786926,
+      "learning_rate": 4.891956859927489e-06,
+      "loss": 0.0471,
+      "step": 134
+    },
+    {
+      "epoch": 1.7532467532467533,
+      "grad_norm": 0.2518353760242462,
+      "learning_rate": 4.885557563874614e-06,
+      "loss": 0.045,
+      "step": 135
+    },
+    {
+      "epoch": 1.7662337662337664,
+      "grad_norm": 0.29494860768318176,
+      "learning_rate": 4.8789786002165055e-06,
+      "loss": 0.0437,
+      "step": 136
+    },
+    {
+      "epoch": 1.7792207792207793,
+      "grad_norm": 0.2658509612083435,
+      "learning_rate": 4.872220464445983e-06,
+      "loss": 0.0523,
+      "step": 137
+    },
+    {
+      "epoch": 1.7922077922077921,
+      "grad_norm": 0.3264024257659912,
+      "learning_rate": 4.865283665550167e-06,
+      "loss": 0.047,
+      "step": 138
+    },
+    {
+      "epoch": 1.8051948051948052,
+      "grad_norm": 0.29048386216163635,
+      "learning_rate": 4.8581687259721375e-06,
+      "loss": 0.0513,
+      "step": 139
+    },
+    {
+      "epoch": 1.8181818181818183,
+      "grad_norm": 0.21431702375411987,
+      "learning_rate": 4.850876181571592e-06,
+      "loss": 0.0426,
+      "step": 140
+    },
+    {
+      "epoch": 1.8311688311688312,
+      "grad_norm": 0.29871106147766113,
+      "learning_rate": 4.843406581584487e-06,
+      "loss": 0.0498,
+      "step": 141
+    },
+    {
+      "epoch": 1.844155844155844,
+      "grad_norm": 0.2374885231256485,
+      "learning_rate": 4.83576048858167e-06,
+      "loss": 0.0423,
+      "step": 142
+    },
+    {
+      "epoch": 1.8571428571428572,
+      "grad_norm": 0.32642969489097595,
+      "learning_rate": 4.8279384784265124e-06,
+      "loss": 0.043,
+      "step": 143
+    },
+    {
+      "epoch": 1.87012987012987,
+      "grad_norm": 0.3235689103603363,
+      "learning_rate": 4.8199411402315356e-06,
+      "loss": 0.0431,
+      "step": 144
+    },
+    {
+      "epoch": 1.883116883116883,
+      "grad_norm": 0.25325465202331543,
+      "learning_rate": 4.811769076314044e-06,
+      "loss": 0.0405,
+      "step": 145
+    },
+    {
+      "epoch": 1.896103896103896,
+      "grad_norm": 0.25112220644950867,
+      "learning_rate": 4.803422902150762e-06,
+      "loss": 0.0399,
+      "step": 146
+    },
+    {
+      "epoch": 1.9090909090909092,
+      "grad_norm": 0.2481043040752411,
+      "learning_rate": 4.794903246331477e-06,
+      "loss": 0.0405,
+      "step": 147
+    },
+    {
+      "epoch": 1.922077922077922,
+      "grad_norm": 0.28779706358909607,
+      "learning_rate": 4.786210750511701e-06,
+      "loss": 0.0407,
+      "step": 148
+    },
+    {
+      "epoch": 1.935064935064935,
+      "grad_norm": 0.2663511335849762,
+      "learning_rate": 4.777346069364343e-06,
+      "loss": 0.0404,
+      "step": 149
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 0.27789220213890076,
+      "learning_rate": 4.7683098705303995e-06,
+      "loss": 0.0384,
+      "step": 150
+    },
+    {
+      "epoch": 1.9610389610389611,
+      "grad_norm": 0.34537413716316223,
+      "learning_rate": 4.7591028345686765e-06,
+      "loss": 0.0466,
+      "step": 151
+    },
+    {
+      "epoch": 1.974025974025974,
+      "grad_norm": 0.3259763717651367,
+      "learning_rate": 4.749725654904529e-06,
+      "loss": 0.0398,
+      "step": 152
+    },
+    {
+      "epoch": 1.987012987012987,
+      "grad_norm": 0.3018707036972046,
+      "learning_rate": 4.740179037777639e-06,
+      "loss": 0.0415,
+      "step": 153
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.27019912004470825,
+      "learning_rate": 4.730463702188824e-06,
+      "loss": 0.0428,
+      "step": 154
+    },
+    {
+      "epoch": 2.012987012987013,
+      "grad_norm": 0.34202930331230164,
+      "learning_rate": 4.720580379845884e-06,
+      "loss": 0.04,
+      "step": 155
+    },
+    {
+      "epoch": 2.0259740259740258,
+      "grad_norm": 0.23893073201179504,
+      "learning_rate": 4.710529815108496e-06,
+      "loss": 0.0352,
+      "step": 156
+    },
+    {
+      "epoch": 2.038961038961039,
+      "grad_norm": 0.3122200071811676,
+      "learning_rate": 4.700312764932151e-06,
+      "loss": 0.0369,
+      "step": 157
+    },
+    {
+      "epoch": 2.051948051948052,
+      "grad_norm": 0.37174174189567566,
+      "learning_rate": 4.689929998811145e-06,
+      "loss": 0.035,
+      "step": 158
+    },
+    {
+      "epoch": 2.064935064935065,
+      "grad_norm": 0.30702632665634155,
+      "learning_rate": 4.679382298720625e-06,
+      "loss": 0.0335,
+      "step": 159
+    },
+    {
+      "epoch": 2.0779220779220777,
+      "grad_norm": 0.2560974955558777,
+      "learning_rate": 4.668670459057693e-06,
+      "loss": 0.0307,
+      "step": 160
+    },
+    {
+      "epoch": 2.090909090909091,
+      "grad_norm": 0.25084158778190613,
+      "learning_rate": 4.657795286581576e-06,
+      "loss": 0.033,
+      "step": 161
+    },
+    {
+      "epoch": 2.103896103896104,
+      "grad_norm": 0.25736576318740845,
+      "learning_rate": 4.64675760035287e-06,
+      "loss": 0.0336,
+      "step": 162
+    },
+    {
+      "epoch": 2.116883116883117,
+      "grad_norm": 0.29564720392227173,
+      "learning_rate": 4.635558231671846e-06,
+      "loss": 0.0325,
+      "step": 163
+    },
+    {
+      "epoch": 2.1298701298701297,
+      "grad_norm": 0.32640376687049866,
+      "learning_rate": 4.624198024015845e-06,
+      "loss": 0.0336,
+      "step": 164
+    },
+    {
+      "epoch": 2.142857142857143,
+      "grad_norm": 0.27138862013816833,
+      "learning_rate": 4.612677832975751e-06,
+      "loss": 0.0324,
+      "step": 165
+    },
+    {
+      "epoch": 2.155844155844156,
+      "grad_norm": 0.28909239172935486,
+      "learning_rate": 4.6009985261915536e-06,
+      "loss": 0.0351,
+      "step": 166
+    },
+    {
+      "epoch": 2.168831168831169,
+      "grad_norm": 0.26905709505081177,
+      "learning_rate": 4.5891609832869964e-06,
+      "loss": 0.0348,
+      "step": 167
+    },
+    {
+      "epoch": 2.1818181818181817,
+      "grad_norm": 0.3421840965747833,
+      "learning_rate": 4.577166095803336e-06,
+      "loss": 0.0362,
+      "step": 168
+    },
+    {
+      "epoch": 2.1948051948051948,
+      "grad_norm": 0.37504997849464417,
+      "learning_rate": 4.565014767132191e-06,
+      "loss": 0.038,
+      "step": 169
+    },
+    {
+      "epoch": 2.207792207792208,
+      "grad_norm": 0.3792576789855957,
+      "learning_rate": 4.552707912447504e-06,
+      "loss": 0.0345,
+      "step": 170
+    },
+    {
+      "epoch": 2.220779220779221,
+      "grad_norm": 0.38099661469459534,
+      "learning_rate": 4.540246458636619e-06,
+      "loss": 0.0265,
+      "step": 171
+    },
+    {
+      "epoch": 2.2337662337662336,
+      "grad_norm": 0.40823256969451904,
+      "learning_rate": 4.527631344230466e-06,
+      "loss": 0.0368,
+      "step": 172
+    },
+    {
+      "epoch": 2.2467532467532467,
+      "grad_norm": 0.28929030895233154,
+      "learning_rate": 4.514863519332882e-06,
+      "loss": 0.0282,
+      "step": 173
+    },
+    {
+      "epoch": 2.25974025974026,
+      "grad_norm": 0.32313477993011475,
+      "learning_rate": 4.501943945549054e-06,
+      "loss": 0.0292,
+      "step": 174
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 0.4433124363422394,
+      "learning_rate": 4.488873595913092e-06,
+      "loss": 0.0298,
+      "step": 175
+    },
+    {
+      "epoch": 2.2857142857142856,
+      "grad_norm": 0.32401469349861145,
+      "learning_rate": 4.475653454814746e-06,
+      "loss": 0.0296,
+      "step": 176
+    },
+    {
+      "epoch": 2.2987012987012987,
+      "grad_norm": 0.35543468594551086,
+      "learning_rate": 4.4622845179252735e-06,
+      "loss": 0.0313,
+      "step": 177
+    },
+    {
+      "epoch": 2.311688311688312,
+      "grad_norm": 0.37388646602630615,
+      "learning_rate": 4.44876779212244e-06,
+      "loss": 0.0305,
+      "step": 178
+    },
+    {
+      "epoch": 2.324675324675325,
+      "grad_norm": 0.33474496006965637,
+      "learning_rate": 4.435104295414697e-06,
+      "loss": 0.0289,
+      "step": 179
+    },
+    {
+      "epoch": 2.3376623376623376,
+      "grad_norm": 0.33111104369163513,
+      "learning_rate": 4.421295056864501e-06,
+      "loss": 0.0312,
+      "step": 180
+    },
+    {
+      "epoch": 2.3506493506493507,
+      "grad_norm": 0.3842385709285736,
+      "learning_rate": 4.407341116510818e-06,
+      "loss": 0.0278,
+      "step": 181
+    },
+    {
+      "epoch": 2.3636363636363638,
+      "grad_norm": 0.4244345724582672,
+      "learning_rate": 4.3932435252907914e-06,
+      "loss": 0.0275,
+      "step": 182
+    },
+    {
+      "epoch": 2.3766233766233764,
+      "grad_norm": 0.5842180252075195,
+      "learning_rate": 4.379003344960585e-06,
+      "loss": 0.035,
+      "step": 183
+    },
+    {
+      "epoch": 2.3896103896103895,
+      "grad_norm": 0.35604268312454224,
+      "learning_rate": 4.364621648015426e-06,
+      "loss": 0.0275,
+      "step": 184
+    },
+    {
+      "epoch": 2.4025974025974026,
+      "grad_norm": 0.41529580950737,
+      "learning_rate": 4.3500995176088235e-06,
+      "loss": 0.0315,
+      "step": 185
+    },
+    {
+      "epoch": 2.4155844155844157,
+      "grad_norm": 0.4051007926464081,
+      "learning_rate": 4.335438047470996e-06,
+      "loss": 0.0271,
+      "step": 186
+    },
+    {
+      "epoch": 2.4285714285714284,
+      "grad_norm": 0.33859074115753174,
+      "learning_rate": 4.320638341826494e-06,
+      "loss": 0.0271,
+      "step": 187
+    },
+    {
+      "epoch": 2.4415584415584415,
+      "grad_norm": 0.33747461438179016,
+      "learning_rate": 4.305701515311037e-06,
+      "loss": 0.0235,
+      "step": 188
+    },
+    {
+      "epoch": 2.4545454545454546,
+      "grad_norm": 0.3473111093044281,
+      "learning_rate": 4.290628692887564e-06,
+      "loss": 0.0284,
+      "step": 189
+    },
+    {
+      "epoch": 2.4675324675324677,
+      "grad_norm": 0.3321944773197174,
+      "learning_rate": 4.27542100976151e-06,
+      "loss": 0.024,
+      "step": 190
+    },
+    {
+      "epoch": 2.4805194805194803,
+      "grad_norm": 0.3281101584434509,
+      "learning_rate": 4.260079611295303e-06,
+      "loss": 0.0262,
+      "step": 191
+    },
+    {
+      "epoch": 2.4935064935064934,
+      "grad_norm": 0.4142085611820221,
+      "learning_rate": 4.244605652922108e-06,
+      "loss": 0.0218,
+      "step": 192
+    },
+    {
+      "epoch": 2.5064935064935066,
+      "grad_norm": 0.3764965832233429,
+      "learning_rate": 4.229000300058802e-06,
+      "loss": 0.0238,
+      "step": 193
+    },
+    {
+      "epoch": 2.5194805194805197,
+      "grad_norm": 0.4414701461791992,
+      "learning_rate": 4.2132647280182e-06,
+      "loss": 0.0267,
+      "step": 194
+    },
+    {
+      "epoch": 2.5324675324675323,
+      "grad_norm": 0.4074808657169342,
+      "learning_rate": 4.197400121920539e-06,
+      "loss": 0.022,
+      "step": 195
+    },
+    {
+      "epoch": 2.5454545454545454,
+      "grad_norm": 0.5768939852714539,
+      "learning_rate": 4.1814076766042206e-06,
+      "loss": 0.0311,
+      "step": 196
+    },
+    {
+      "epoch": 2.5584415584415585,
+      "grad_norm": 0.35587024688720703,
+      "learning_rate": 4.165288596535821e-06,
+      "loss": 0.0246,
+      "step": 197
+    },
+    {
+      "epoch": 2.571428571428571,
+      "grad_norm": 0.36500850319862366,
+      "learning_rate": 4.149044095719377e-06,
+      "loss": 0.0195,
+      "step": 198
+    },
+    {
+      "epoch": 2.5844155844155843,
+      "grad_norm": 0.3946477472782135,
+      "learning_rate": 4.132675397604956e-06,
+      "loss": 0.0208,
+      "step": 199
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 0.3371339440345764,
+      "learning_rate": 4.11618373499651e-06,
+      "loss": 0.0204,
+      "step": 200
+    },
+    {
+      "epoch": 2.6103896103896105,
+      "grad_norm": 0.41154706478118896,
+      "learning_rate": 4.099570349959025e-06,
+      "loss": 0.0223,
+      "step": 201
+    },
+    {
+      "epoch": 2.6233766233766236,
+      "grad_norm": 0.39301300048828125,
+      "learning_rate": 4.082836493724981e-06,
+      "loss": 0.0204,
+      "step": 202
+    },
+    {
+      "epoch": 2.6363636363636362,
+      "grad_norm": 0.3674089014530182,
+      "learning_rate": 4.065983426600113e-06,
+      "loss": 0.0195,
+      "step": 203
+    },
+    {
+      "epoch": 2.6493506493506493,
+      "grad_norm": 0.40296265482902527,
+      "learning_rate": 4.0490124178684884e-06,
+      "loss": 0.021,
+      "step": 204
+    },
+    {
+      "epoch": 2.6623376623376624,
+      "grad_norm": 0.4691706895828247,
+      "learning_rate": 4.031924745696916e-06,
+      "loss": 0.0179,
+      "step": 205
+    },
+    {
+      "epoch": 2.675324675324675,
+      "grad_norm": 0.44443100690841675,
+      "learning_rate": 4.014721697038678e-06,
+      "loss": 0.0242,
+      "step": 206
+    },
+    {
+      "epoch": 2.688311688311688,
+      "grad_norm": 0.31682541966438293,
+      "learning_rate": 3.997404567536606e-06,
+      "loss": 0.0193,
+      "step": 207
+    },
+    {
+      "epoch": 2.7012987012987013,
+      "grad_norm": 0.33720874786376953,
+      "learning_rate": 3.979974661425497e-06,
+      "loss": 0.0212,
+      "step": 208
+    },
+    {
+      "epoch": 2.7142857142857144,
+      "grad_norm": 0.41582977771759033,
+      "learning_rate": 3.962433291433889e-06,
+      "loss": 0.0242,
+      "step": 209
+    },
+    {
+      "epoch": 2.7272727272727275,
+      "grad_norm": 0.4688783884048462,
+      "learning_rate": 3.944781778685189e-06,
+      "loss": 0.0218,
+      "step": 210
+    },
+    {
+      "epoch": 2.74025974025974,
+      "grad_norm": 0.3696669638156891,
+      "learning_rate": 3.927021452598177e-06,
+      "loss": 0.0209,
+      "step": 211
+    },
+    {
+      "epoch": 2.7532467532467533,
+      "grad_norm": 0.36734238266944885,
+      "learning_rate": 3.909153650786878e-06,
+      "loss": 0.0165,
+      "step": 212
+    },
+    {
+      "epoch": 2.7662337662337664,
+      "grad_norm": 0.33928048610687256,
+      "learning_rate": 3.891179718959822e-06,
+      "loss": 0.0137,
+      "step": 213
+    },
+    {
+      "epoch": 2.779220779220779,
+      "grad_norm": 0.656510055065155,
+      "learning_rate": 3.873101010818692e-06,
+      "loss": 0.0258,
+      "step": 214
+    },
+    {
+      "epoch": 2.792207792207792,
+      "grad_norm": 0.5539777278900146,
+      "learning_rate": 3.8549188879563685e-06,
+      "loss": 0.0208,
+      "step": 215
+    },
+    {
+      "epoch": 2.8051948051948052,
+      "grad_norm": 0.4914684593677521,
+      "learning_rate": 3.836634719754385e-06,
+      "loss": 0.0192,
+      "step": 216
+    },
+    {
+      "epoch": 2.8181818181818183,
+      "grad_norm": 0.38955259323120117,
+      "learning_rate": 3.818249883279791e-06,
+      "loss": 0.0185,
+      "step": 217
+    },
+    {
+      "epoch": 2.8311688311688314,
+      "grad_norm": 0.45634275674819946,
+      "learning_rate": 3.7997657631814366e-06,
+      "loss": 0.0202,
+      "step": 218
+    },
+    {
+      "epoch": 2.844155844155844,
+      "grad_norm": 0.41473978757858276,
+      "learning_rate": 3.781183751585693e-06,
+      "loss": 0.0133,
+      "step": 219
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 0.6281463503837585,
+      "learning_rate": 3.762505247991601e-06,
+      "loss": 0.0188,
+      "step": 220
+    },
+    {
+      "epoch": 2.87012987012987,
+      "grad_norm": 0.34762486815452576,
+      "learning_rate": 3.7437316591654726e-06,
+      "loss": 0.0199,
+      "step": 221
+    },
+    {
+      "epoch": 2.883116883116883,
+      "grad_norm": 0.43128135800361633,
+      "learning_rate": 3.724864399034932e-06,
+      "loss": 0.0175,
+      "step": 222
+    },
+    {
+      "epoch": 2.896103896103896,
+      "grad_norm": 0.4040665328502655,
+      "learning_rate": 3.7059048885824367e-06,
+      "loss": 0.0119,
+      "step": 223
+    },
+    {
+      "epoch": 2.909090909090909,
+      "grad_norm": 0.4182071089744568,
+      "learning_rate": 3.686854555738249e-06,
+      "loss": 0.0171,
+      "step": 224
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 0.38667839765548706,
+      "learning_rate": 3.6677148352728947e-06,
+      "loss": 0.0155,
+      "step": 225
+    },
+    {
+      "epoch": 2.935064935064935,
+      "grad_norm": 0.3983423709869385,
+      "learning_rate": 3.6484871686891044e-06,
+      "loss": 0.013,
+      "step": 226
+    },
+    {
+      "epoch": 2.948051948051948,
+      "grad_norm": 0.48213276267051697,
+      "learning_rate": 3.629173004113245e-06,
+      "loss": 0.0131,
+      "step": 227
+    },
+    {
+      "epoch": 2.961038961038961,
+      "grad_norm": 0.4538249969482422,
+      "learning_rate": 3.609773796186256e-06,
+      "loss": 0.0137,
+      "step": 228
+    },
+    {
+      "epoch": 2.974025974025974,
+      "grad_norm": 0.4226680397987366,
+      "learning_rate": 3.590291005954094e-06,
+      "loss": 0.0144,
+      "step": 229
+    },
+    {
+      "epoch": 2.987012987012987,
+      "grad_norm": 0.603485107421875,
+      "learning_rate": 3.570726100757693e-06,
+      "loss": 0.0179,
+      "step": 230
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.5114914774894714,
+      "learning_rate": 3.5510805541224536e-06,
+      "loss": 0.0134,
+      "step": 231
+    },
+    {
+      "epoch": 3.012987012987013,
+      "grad_norm": 0.5039042830467224,
+      "learning_rate": 3.5313558456472623e-06,
+      "loss": 0.0113,
+      "step": 232
+    },
+    {
+      "epoch": 3.0259740259740258,
+      "grad_norm": 0.37786880135536194,
+      "learning_rate": 3.511553460893059e-06,
+      "loss": 0.0089,
+      "step": 233
+    },
+    {
+      "epoch": 3.038961038961039,
+      "grad_norm": 0.31840819120407104,
+      "learning_rate": 3.4916748912709506e-06,
+      "loss": 0.0078,
+      "step": 234
+    },
+    {
+      "epoch": 3.051948051948052,
+      "grad_norm": 0.3462812006473541,
+      "learning_rate": 3.4717216339298852e-06,
+      "loss": 0.0095,
+      "step": 235
+    },
+    {
+      "epoch": 3.064935064935065,
+      "grad_norm": 0.28717392683029175,
+      "learning_rate": 3.4516951916438974e-06,
+      "loss": 0.0043,
+      "step": 236
+    },
+    {
+      "epoch": 3.0779220779220777,
+      "grad_norm": 0.4881342649459839,
+      "learning_rate": 3.4315970726989244e-06,
+      "loss": 0.0109,
+      "step": 237
+    },
+    {
+      "epoch": 3.090909090909091,
+      "grad_norm": 0.34244024753570557,
+      "learning_rate": 3.4114287907792115e-06,
+      "loss": 0.0047,
+      "step": 238
+    },
+    {
+      "epoch": 3.103896103896104,
+      "grad_norm": 0.3604426085948944,
+      "learning_rate": 3.3911918648533094e-06,
+      "loss": 0.0059,
+      "step": 239
+    },
+    {
+      "epoch": 3.116883116883117,
+      "grad_norm": 0.34059545397758484,
+      "learning_rate": 3.3708878190596724e-06,
+      "loss": 0.0037,
+      "step": 240
+    },
+    {
+      "epoch": 3.1298701298701297,
+      "grad_norm": 0.350000262260437,
+      "learning_rate": 3.3505181825918685e-06,
+      "loss": 0.0059,
+      "step": 241
+    },
+    {
+      "epoch": 3.142857142857143,
+      "grad_norm": 0.4721887409687042,
+      "learning_rate": 3.33008448958341e-06,
+      "loss": 0.0082,
+      "step": 242
+    },
+    {
+      "epoch": 3.155844155844156,
+      "grad_norm": 0.3560850918292999,
+      "learning_rate": 3.3095882789922102e-06,
+      "loss": 0.007,
+      "step": 243
+    },
+    {
+      "epoch": 3.168831168831169,
+      "grad_norm": 0.385452002286911,
+      "learning_rate": 3.289031094484675e-06,
+      "loss": 0.0047,
+      "step": 244
+    },
+    {
+      "epoch": 3.1818181818181817,
+      "grad_norm": 0.3492503762245178,
+      "learning_rate": 3.2684144843194453e-06,
+      "loss": 0.0053,
+      "step": 245
+    },
+    {
+      "epoch": 3.1948051948051948,
+      "grad_norm": 0.3584858179092407,
+      "learning_rate": 3.2477400012307885e-06,
+      "loss": 0.0077,
+      "step": 246
+    },
+    {
+      "epoch": 3.207792207792208,
+      "grad_norm": 0.2692736089229584,
+      "learning_rate": 3.2270092023116564e-06,
+      "loss": 0.0031,
+      "step": 247
+    },
+    {
+      "epoch": 3.220779220779221,
+      "grad_norm": 0.33909231424331665,
+      "learning_rate": 3.206223648896409e-06,
+      "loss": 0.0051,
+      "step": 248
+    },
+    {
+      "epoch": 3.2337662337662336,
+      "grad_norm": 0.47461092472076416,
+      "learning_rate": 3.1853849064432296e-06,
+      "loss": 0.0053,
+      "step": 249
+    },
+    {
+      "epoch": 3.2467532467532467,
+      "grad_norm": 0.46319177746772766,
+      "learning_rate": 3.1644945444162155e-06,
+      "loss": 0.0076,
+      "step": 250
+    },
+    {
+      "epoch": 3.25974025974026,
+      "grad_norm": 0.3297309875488281,
+      "learning_rate": 3.14355413616718e-06,
+      "loss": 0.0035,
+      "step": 251
+    },
+    {
+      "epoch": 3.2727272727272725,
+      "grad_norm": 0.3955172598361969,
+      "learning_rate": 3.1225652588171534e-06,
+      "loss": 0.0047,
+      "step": 252
+    },
+    {
+      "epoch": 3.2857142857142856,
+      "grad_norm": 0.2394171506166458,
+      "learning_rate": 3.1015294931376035e-06,
+      "loss": 0.0025,
+      "step": 253
+    },
+    {
+      "epoch": 3.2987012987012987,
+      "grad_norm": 0.24900038540363312,
+      "learning_rate": 3.080448423431377e-06,
+      "loss": 0.0028,
+      "step": 254
+    },
+    {
+      "epoch": 3.311688311688312,
+      "grad_norm": 0.3579304814338684,
+      "learning_rate": 3.059323637413385e-06,
+      "loss": 0.0031,
+      "step": 255
+    },
+    {
+      "epoch": 3.324675324675325,
+      "grad_norm": 0.2939773499965668,
+      "learning_rate": 3.0381567260910166e-06,
+      "loss": 0.0037,
+      "step": 256
+    },
+    {
+      "epoch": 3.3376623376623376,
+      "grad_norm": 0.39939820766448975,
+      "learning_rate": 3.016949283644319e-06,
+      "loss": 0.0055,
+      "step": 257
+    },
+    {
+      "epoch": 3.3506493506493507,
+      "grad_norm": 0.4479775130748749,
+      "learning_rate": 2.9957029073059276e-06,
+      "loss": 0.0038,
+      "step": 258
+    },
+    {
+      "epoch": 3.3636363636363638,
+      "grad_norm": 0.3758586049079895,
+      "learning_rate": 2.9744191972407754e-06,
+      "loss": 0.0034,
+      "step": 259
+    },
+    {
+      "epoch": 3.3766233766233764,
+      "grad_norm": 0.31011316180229187,
+      "learning_rate": 2.9530997564255728e-06,
+      "loss": 0.0022,
+      "step": 260
+    },
+    {
+      "epoch": 3.3896103896103895,
+      "grad_norm": 0.5502382516860962,
+      "learning_rate": 2.931746190528082e-06,
+      "loss": 0.0021,
+      "step": 261
+    },
+    {
+      "epoch": 3.4025974025974026,
+      "grad_norm": 0.285674512386322,
+      "learning_rate": 2.9103601077861875e-06,
+      "loss": 0.0027,
+      "step": 262
+    },
+    {
+      "epoch": 3.4155844155844157,
+      "grad_norm": 0.3391030430793762,
+      "learning_rate": 2.888943118886771e-06,
+      "loss": 0.0037,
+      "step": 263
+    },
+    {
+      "epoch": 3.4285714285714284,
+      "grad_norm": 0.313784122467041,
+      "learning_rate": 2.8674968368444004e-06,
+      "loss": 0.002,
+      "step": 264
+    },
+    {
+      "epoch": 3.4415584415584415,
+      "grad_norm": 0.2796953618526459,
+      "learning_rate": 2.8460228768798507e-06,
+      "loss": 0.0026,
+      "step": 265
+    },
+    {
+      "epoch": 3.4545454545454546,
+      "grad_norm": 0.3058723509311676,
+      "learning_rate": 2.8245228562984518e-06,
+      "loss": 0.0025,
+      "step": 266
+    },
+    {
+      "epoch": 3.4675324675324677,
+      "grad_norm": 0.54490727186203,
+      "learning_rate": 2.80299839436828e-06,
+      "loss": 0.0043,
+      "step": 267
+    },
+    {
+      "epoch": 3.4805194805194803,
+      "grad_norm": 0.41025906801223755,
+      "learning_rate": 2.781451112198208e-06,
+      "loss": 0.0027,
+      "step": 268
+    },
+    {
+      "epoch": 3.4935064935064934,
+      "grad_norm": 0.35861679911613464,
+      "learning_rate": 2.7598826326158045e-06,
+      "loss": 0.004,
+      "step": 269
+    },
+    {
+      "epoch": 3.5064935064935066,
+      "grad_norm": 0.40969038009643555,
+      "learning_rate": 2.738294580045119e-06,
+      "loss": 0.0031,
+      "step": 270
+    },
+    {
+      "epoch": 3.5194805194805197,
+      "grad_norm": 0.4820692241191864,
+      "learning_rate": 2.7166885803843347e-06,
+      "loss": 0.0027,
+      "step": 271
+    },
+    {
+      "epoch": 3.5324675324675323,
+      "grad_norm": 0.3031913936138153,
+      "learning_rate": 2.695066260883313e-06,
+      "loss": 0.0013,
+      "step": 272
+    },
+    {
+      "epoch": 3.5454545454545454,
+      "grad_norm": 0.3935272693634033,
+      "learning_rate": 2.67342925002104e-06,
+      "loss": 0.0027,
+      "step": 273
+    },
+    {
+      "epoch": 3.5584415584415585,
+      "grad_norm": 0.2262854278087616,
+      "learning_rate": 2.65177917738298e-06,
+      "loss": 0.0019,
+      "step": 274
+    },
+    {
+      "epoch": 3.571428571428571,
+      "grad_norm": 0.2690446972846985,
+      "learning_rate": 2.6301176735383382e-06,
+      "loss": 0.0014,
+      "step": 275
+    },
+    {
+      "epoch": 3.5844155844155843,
+      "grad_norm": 0.33330366015434265,
+      "learning_rate": 2.6084463699172594e-06,
+      "loss": 0.0025,
+      "step": 276
+    },
+    {
+      "epoch": 3.5974025974025974,
+      "grad_norm": 0.3452131748199463,
+      "learning_rate": 2.586766898687955e-06,
+      "loss": 0.0018,
+      "step": 277
+    },
+    {
+      "epoch": 3.6103896103896105,
+      "grad_norm": 0.3789026737213135,
+      "learning_rate": 2.56508089263378e-06,
+      "loss": 0.001,
+      "step": 278
+    },
+    {
+      "epoch": 3.6233766233766236,
+      "grad_norm": 0.3677704334259033,
+      "learning_rate": 2.5433899850302552e-06,
+      "loss": 0.0013,
+      "step": 279
+    },
+    {
+      "epoch": 3.6363636363636362,
+      "grad_norm": 0.4359056055545807,
+      "learning_rate": 2.521695809522061e-06,
+      "loss": 0.0032,
+      "step": 280
+    },
+    {
+      "epoch": 3.6493506493506493,
+      "grad_norm": 0.2561199367046356,
+      "learning_rate": 2.5e-06,
+      "loss": 0.0018,
+      "step": 281
+    },
+    {
+      "epoch": 3.6623376623376624,
+      "grad_norm": 0.39096924662590027,
+      "learning_rate": 2.4783041904779386e-06,
+      "loss": 0.0031,
+      "step": 282
+    },
+    {
+      "epoch": 3.675324675324675,
+      "grad_norm": 0.3999404013156891,
+      "learning_rate": 2.4566100149697456e-06,
+      "loss": 0.0019,
+      "step": 283
+    },
+    {
+      "epoch": 3.688311688311688,
+      "grad_norm": 0.8379652500152588,
+      "learning_rate": 2.4349191073662203e-06,
+      "loss": 0.004,
+      "step": 284
+    },
+    {
+      "epoch": 3.7012987012987013,
+      "grad_norm": 0.3195594847202301,
+      "learning_rate": 2.4132331013120454e-06,
+      "loss": 0.0011,
+      "step": 285
+    },
+    {
+      "epoch": 3.7142857142857144,
+      "grad_norm": 0.2558881342411041,
+      "learning_rate": 2.3915536300827414e-06,
+      "loss": 0.001,
+      "step": 286
+    },
+    {
+      "epoch": 3.7272727272727275,
+      "grad_norm": 0.284496933221817,
+      "learning_rate": 2.3698823264616635e-06,
+      "loss": 0.0018,
+      "step": 287
+    },
+    {
+      "epoch": 3.74025974025974,
+      "grad_norm": 0.20427702367305756,
+      "learning_rate": 2.348220822617021e-06,
+      "loss": 0.0014,
+      "step": 288
+    },
+    {
+      "epoch": 3.7532467532467533,
+      "grad_norm": 0.32599779963493347,
+      "learning_rate": 2.32657074997896e-06,
+      "loss": 0.002,
+      "step": 289
+    },
+    {
+      "epoch": 3.7662337662337664,
+      "grad_norm": 0.3040351867675781,
+      "learning_rate": 2.3049337391166884e-06,
+      "loss": 0.0011,
+      "step": 290
+    },
+    {
+      "epoch": 3.779220779220779,
+      "grad_norm": 0.2709053158760071,
+      "learning_rate": 2.2833114196156657e-06,
+      "loss": 0.0008,
+      "step": 291
+    },
+    {
+      "epoch": 3.792207792207792,
+      "grad_norm": 0.2777629792690277,
+      "learning_rate": 2.261705419954882e-06,
+      "loss": 0.0014,
+      "step": 292
+    },
+    {
+      "epoch": 3.8051948051948052,
+      "grad_norm": 0.3083540201187134,
+      "learning_rate": 2.2401173673841963e-06,
+      "loss": 0.0016,
+      "step": 293
+    },
+    {
+      "epoch": 3.8181818181818183,
+      "grad_norm": 0.1168127954006195,
+      "learning_rate": 2.2185488878017934e-06,
+      "loss": 0.0005,
+      "step": 294
+    },
+    {
+      "epoch": 3.8311688311688314,
+      "grad_norm": 0.25492948293685913,
+      "learning_rate": 2.1970016056317202e-06,
+      "loss": 0.001,
+      "step": 295
+    },
+    {
+      "epoch": 3.844155844155844,
+      "grad_norm": 0.3559608459472656,
+      "learning_rate": 2.1754771437015495e-06,
+      "loss": 0.0009,
+      "step": 296
+    },
+    {
+      "epoch": 3.857142857142857,
+      "grad_norm": 0.43275198340415955,
+      "learning_rate": 2.1539771231201497e-06,
+      "loss": 0.002,
+      "step": 297
+    },
+    {
+      "epoch": 3.87012987012987,
+      "grad_norm": 0.35539472103118896,
+      "learning_rate": 2.1325031631555996e-06,
+      "loss": 0.0012,
+      "step": 298
+    },
+    {
+      "epoch": 3.883116883116883,
+      "grad_norm": 0.2857511341571808,
+      "learning_rate": 2.11105688111323e-06,
+      "loss": 0.0017,
+      "step": 299
+    },
+    {
+      "epoch": 3.896103896103896,
+      "grad_norm": 0.22510217130184174,
+      "learning_rate": 2.0896398922138124e-06,
+      "loss": 0.0007,
+      "step": 300
+    },
+    {
+      "epoch": 3.909090909090909,
+      "grad_norm": 0.15990178287029266,
+      "learning_rate": 2.0682538094719183e-06,
+      "loss": 0.0011,
+      "step": 301
+    },
+    {
+      "epoch": 3.9220779220779223,
+      "grad_norm": 0.11588839441537857,
+      "learning_rate": 2.0469002435744285e-06,
+      "loss": 0.0007,
+      "step": 302
+    },
+    {
+      "epoch": 3.935064935064935,
+      "grad_norm": 0.36557283997535706,
+      "learning_rate": 2.0255808027592263e-06,
+      "loss": 0.0008,
+      "step": 303
+    },
+    {
+      "epoch": 3.948051948051948,
+      "grad_norm": 0.2867473363876343,
+      "learning_rate": 2.004297092694073e-06,
+      "loss": 0.001,
+      "step": 304
+    },
+    {
+      "epoch": 3.961038961038961,
+      "grad_norm": 0.270842969417572,
+      "learning_rate": 1.9830507163556815e-06,
+      "loss": 0.0008,
+      "step": 305
+    },
+    {
+      "epoch": 3.974025974025974,
+      "grad_norm": 0.24662533402442932,
+      "learning_rate": 1.9618432739089843e-06,
+      "loss": 0.0006,
+      "step": 306
+    },
+    {
+      "epoch": 3.987012987012987,
+      "grad_norm": 0.12252429127693176,
+      "learning_rate": 1.9406763625866155e-06,
+      "loss": 0.0004,
+      "step": 307
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 0.24726521968841553,
+      "learning_rate": 1.9195515765686237e-06,
+      "loss": 0.0004,
+      "step": 308
+    },
+    {
+      "epoch": 4.012987012987013,
+      "grad_norm": 0.04255533963441849,
+      "learning_rate": 1.8984705068623976e-06,
+      "loss": 0.0002,
+      "step": 309
+    },
+    {
+      "epoch": 4.025974025974026,
+      "grad_norm": 0.0485837422311306,
+      "learning_rate": 1.8774347411828472e-06,
+      "loss": 0.0002,
+      "step": 310
+    },
+    {
+      "epoch": 4.038961038961039,
+      "grad_norm": 0.06497793644666672,
+      "learning_rate": 1.8564458638328203e-06,
+      "loss": 0.0005,
+      "step": 311
+    },
+    {
+      "epoch": 4.0519480519480515,
+      "grad_norm": 0.06974362581968307,
+      "learning_rate": 1.835505455583786e-06,
+      "loss": 0.0002,
+      "step": 312
+    },
+    {
+      "epoch": 4.064935064935065,
+      "grad_norm": 0.10802063345909119,
+      "learning_rate": 1.8146150935567712e-06,
+      "loss": 0.0004,
+      "step": 313
+    },
+    {
+      "epoch": 4.077922077922078,
+      "grad_norm": 0.04839176684617996,
+      "learning_rate": 1.7937763511035904e-06,
+      "loss": 0.0002,
+      "step": 314
+    },
+    {
+      "epoch": 4.090909090909091,
+      "grad_norm": 0.16950847208499908,
+      "learning_rate": 1.7729907976883443e-06,
+      "loss": 0.0006,
+      "step": 315
+    },
+    {
+      "epoch": 4.103896103896104,
+      "grad_norm": 0.060468755662441254,
+      "learning_rate": 1.7522599987692113e-06,
+      "loss": 0.0001,
+      "step": 316
+    },
+    {
+      "epoch": 4.116883116883117,
+      "grad_norm": 0.054468460381031036,
+      "learning_rate": 1.7315855156805558e-06,
+      "loss": 0.0002,
+      "step": 317
+    },
+    {
+      "epoch": 4.12987012987013,
+      "grad_norm": 0.05310998484492302,
+      "learning_rate": 1.7109689055153261e-06,
+      "loss": 0.0002,
+      "step": 318
+    },
+    {
+      "epoch": 4.142857142857143,
+      "grad_norm": 0.1378607451915741,
+      "learning_rate": 1.690411721007791e-06,
+      "loss": 0.0004,
+      "step": 319
+    },
+    {
+      "epoch": 4.1558441558441555,
+      "grad_norm": 0.01266263797879219,
+      "learning_rate": 1.6699155104165903e-06,
+      "loss": 0.0001,
+      "step": 320
+    },
+    {
+      "epoch": 4.1688311688311686,
+      "grad_norm": 0.17714223265647888,
+      "learning_rate": 1.6494818174081317e-06,
+      "loss": 0.001,
+      "step": 321
+    },
+    {
+      "epoch": 4.181818181818182,
+      "grad_norm": 0.03800169378519058,
+      "learning_rate": 1.6291121809403287e-06,
+      "loss": 0.0002,
+      "step": 322
+    },
+    {
+      "epoch": 4.194805194805195,
+      "grad_norm": 0.17321327328681946,
+      "learning_rate": 1.6088081351466908e-06,
+      "loss": 0.0008,
+      "step": 323
+    },
+    {
+      "epoch": 4.207792207792208,
+      "grad_norm": 0.019961990416049957,
+      "learning_rate": 1.588571209220789e-06,
+      "loss": 0.0001,
+      "step": 324
+    },
+    {
+      "epoch": 4.220779220779221,
+      "grad_norm": 0.040765196084976196,
+      "learning_rate": 1.5684029273010762e-06,
+      "loss": 0.0002,
+      "step": 325
+    },
+    {
+      "epoch": 4.233766233766234,
+      "grad_norm": 0.06156046316027641,
+      "learning_rate": 1.5483048083561036e-06,
+      "loss": 0.0003,
+      "step": 326
+    },
+    {
+      "epoch": 4.246753246753247,
+      "grad_norm": 0.06152096390724182,
+      "learning_rate": 1.5282783660701154e-06,
+      "loss": 0.0003,
+      "step": 327
+    },
+    {
+      "epoch": 4.259740259740259,
+      "grad_norm": 0.010597504675388336,
+      "learning_rate": 1.5083251087290506e-06,
+      "loss": 0.0001,
+      "step": 328
+    },
+    {
+      "epoch": 4.2727272727272725,
+      "grad_norm": 0.0316089391708374,
+      "learning_rate": 1.4884465391069415e-06,
+      "loss": 0.0001,
+      "step": 329
+    },
+    {
+      "epoch": 4.285714285714286,
+      "grad_norm": 0.02659175731241703,
+      "learning_rate": 1.4686441543527374e-06,
+      "loss": 0.0002,
+      "step": 330
+    },
+    {
+      "epoch": 4.298701298701299,
+      "grad_norm": 0.02682899683713913,
+      "learning_rate": 1.4489194458775468e-06,
+      "loss": 0.0002,
+      "step": 331
+    },
+    {
+      "epoch": 4.311688311688312,
+      "grad_norm": 0.01274723932147026,
+      "learning_rate": 1.4292738992423066e-06,
+      "loss": 0.0001,
+      "step": 332
+    },
+    {
+      "epoch": 4.324675324675325,
+      "grad_norm": 0.10610726475715637,
+      "learning_rate": 1.409708994045907e-06,
+      "loss": 0.0005,
+      "step": 333
+    },
+    {
+      "epoch": 4.337662337662338,
+      "grad_norm": 0.018574735149741173,
+      "learning_rate": 1.3902262038137449e-06,
+      "loss": 0.0001,
+      "step": 334
+    },
+    {
+      "epoch": 4.35064935064935,
+      "grad_norm": 0.03301926702260971,
+      "learning_rate": 1.3708269958867565e-06,
+      "loss": 0.0002,
+      "step": 335
+    },
+    {
+      "epoch": 4.363636363636363,
+      "grad_norm": 0.05397050827741623,
+      "learning_rate": 1.3515128313108966e-06,
+      "loss": 0.0001,
+      "step": 336
+    },
+    {
+      "epoch": 4.376623376623376,
+      "grad_norm": 0.5025430917739868,
+      "learning_rate": 1.3322851647271057e-06,
+      "loss": 0.0008,
+      "step": 337
+    },
+    {
+      "epoch": 4.3896103896103895,
+      "grad_norm": 0.06098180264234543,
+      "learning_rate": 1.3131454442617523e-06,
+      "loss": 0.0002,
+      "step": 338
+    },
+    {
+      "epoch": 4.402597402597403,
+      "grad_norm": 0.02098267152905464,
+      "learning_rate": 1.2940951114175637e-06,
+      "loss": 0.0001,
+      "step": 339
+    },
+    {
+      "epoch": 4.415584415584416,
+      "grad_norm": 0.044198423624038696,
+      "learning_rate": 1.275135600965068e-06,
+      "loss": 0.0002,
+      "step": 340
+    },
+    {
+      "epoch": 4.428571428571429,
+      "grad_norm": 0.014930199831724167,
+      "learning_rate": 1.2562683408345279e-06,
+      "loss": 0.0002,
+      "step": 341
+    },
+    {
+      "epoch": 4.441558441558442,
+      "grad_norm": 0.33325693011283875,
+      "learning_rate": 1.237494752008399e-06,
+      "loss": 0.0006,
+      "step": 342
+    },
+    {
+      "epoch": 4.454545454545454,
+      "grad_norm": 0.031100880354642868,
+      "learning_rate": 1.2188162484143077e-06,
+      "loss": 0.0002,
+      "step": 343
+    },
+    {
+      "epoch": 4.467532467532467,
+      "grad_norm": 0.03873937577009201,
+      "learning_rate": 1.2002342368185638e-06,
+      "loss": 0.0001,
+      "step": 344
+    },
+    {
+      "epoch": 4.48051948051948,
+      "grad_norm": 0.22393281757831573,
+      "learning_rate": 1.18175011672021e-06,
+      "loss": 0.0005,
+      "step": 345
+    },
+    {
+      "epoch": 4.4935064935064934,
+      "grad_norm": 0.06489652395248413,
+      "learning_rate": 1.163365280245615e-06,
+      "loss": 0.0002,
+      "step": 346
+    },
+    {
+      "epoch": 4.5064935064935066,
+      "grad_norm": 0.04202171042561531,
+      "learning_rate": 1.1450811120436319e-06,
+      "loss": 0.0002,
+      "step": 347
+    },
+    {
+      "epoch": 4.51948051948052,
+      "grad_norm": 0.03509971871972084,
+      "learning_rate": 1.1268989891813085e-06,
+      "loss": 0.0002,
+      "step": 348
+    },
+    {
+      "epoch": 4.532467532467533,
+      "grad_norm": 0.013964281417429447,
+      "learning_rate": 1.1088202810401789e-06,
+      "loss": 0.0001,
+      "step": 349
+    },
+    {
+      "epoch": 4.545454545454545,
+      "grad_norm": 0.025696441531181335,
+      "learning_rate": 1.0908463492131227e-06,
+      "loss": 0.0001,
+      "step": 350
+    },
+    {
+      "epoch": 4.558441558441558,
+      "grad_norm": 0.03067847713828087,
+      "learning_rate": 1.0729785474018243e-06,
+      "loss": 0.0002,
+      "step": 351
+    },
+    {
+      "epoch": 4.571428571428571,
+      "grad_norm": 0.06406523287296295,
+      "learning_rate": 1.0552182213148119e-06,
+      "loss": 0.0002,
+      "step": 352
+    },
+    {
+      "epoch": 4.584415584415584,
+      "grad_norm": 0.02622838132083416,
+      "learning_rate": 1.0375667085661115e-06,
+      "loss": 0.0001,
+      "step": 353
+    },
+    {
+      "epoch": 4.597402597402597,
+      "grad_norm": 0.026939135044813156,
+      "learning_rate": 1.020025338574504e-06,
+      "loss": 0.0001,
+      "step": 354
+    },
+    {
+      "epoch": 4.6103896103896105,
+      "grad_norm": 0.009263522922992706,
+      "learning_rate": 1.0025954324633949e-06,
+      "loss": 0.0001,
+      "step": 355
+    },
+    {
+      "epoch": 4.623376623376624,
+      "grad_norm": 0.1577068418264389,
+      "learning_rate": 9.852783029613224e-07,
+      "loss": 0.0002,
+      "step": 356
+    },
+    {
+      "epoch": 4.636363636363637,
+      "grad_norm": 0.026724133640527725,
+      "learning_rate": 9.680752543030844e-07,
+      "loss": 0.0001,
+      "step": 357
+    },
+    {
+      "epoch": 4.64935064935065,
+      "grad_norm": 0.017666451632976532,
+      "learning_rate": 9.509875821315126e-07,
+      "loss": 0.0001,
+      "step": 358
+    },
+    {
+      "epoch": 4.662337662337662,
+      "grad_norm": 0.035038791596889496,
+      "learning_rate": 9.340165733998877e-07,
+      "loss": 0.0001,
+      "step": 359
+    },
+    {
+      "epoch": 4.675324675324675,
+      "grad_norm": 0.05781812220811844,
+      "learning_rate": 9.171635062750189e-07,
+      "loss": 0.0002,
+      "step": 360
+    },
+    {
+      "epoch": 4.688311688311688,
+      "grad_norm": 0.021358780562877655,
+      "learning_rate": 9.004296500409759e-07,
+      "loss": 0.0001,
+      "step": 361
+    },
+    {
+      "epoch": 4.701298701298701,
+      "grad_norm": 0.014494797214865685,
+      "learning_rate": 8.838162650034912e-07,
+      "loss": 0.0001,
+      "step": 362
+    },
+    {
+      "epoch": 4.714285714285714,
+      "grad_norm": 0.01157679595053196,
+      "learning_rate": 8.673246023950449e-07,
+      "loss": 0.0001,
+      "step": 363
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.015167619101703167,
+      "learning_rate": 8.509559042806237e-07,
+      "loss": 0.0001,
+      "step": 364
+    },
+    {
+      "epoch": 4.740259740259741,
+      "grad_norm": 0.01940612867474556,
+      "learning_rate": 8.347114034641807e-07,
+      "loss": 0.0001,
+      "step": 365
+    },
+    {
+      "epoch": 4.753246753246753,
+      "grad_norm": 0.07376569509506226,
+      "learning_rate": 8.185923233957802e-07,
+      "loss": 0.0001,
+      "step": 366
+    },
+    {
+      "epoch": 4.766233766233766,
+      "grad_norm": 0.015078631229698658,
+      "learning_rate": 8.025998780794622e-07,
+      "loss": 0.0001,
+      "step": 367
+    },
+    {
+      "epoch": 4.779220779220779,
+      "grad_norm": 0.03011564165353775,
+      "learning_rate": 7.867352719818008e-07,
+      "loss": 0.0002,
+      "step": 368
+    },
+    {
+      "epoch": 4.792207792207792,
+      "grad_norm": 0.017493847757577896,
+      "learning_rate": 7.709996999411984e-07,
+      "loss": 0.0001,
+      "step": 369
+    },
+    {
+      "epoch": 4.805194805194805,
+      "grad_norm": 0.02849338948726654,
+      "learning_rate": 7.553943470778927e-07,
+      "loss": 0.0001,
+      "step": 370
+    },
+    {
+      "epoch": 4.818181818181818,
+      "grad_norm": 0.038560837507247925,
+      "learning_rate": 7.399203887046977e-07,
+      "loss": 0.0001,
+      "step": 371
+    },
+    {
+      "epoch": 4.8311688311688314,
+      "grad_norm": 0.020657677203416824,
+      "learning_rate": 7.245789902384908e-07,
+      "loss": 0.0001,
+      "step": 372
+    },
+    {
+      "epoch": 4.8441558441558445,
+      "grad_norm": 0.030862174928188324,
+      "learning_rate": 7.093713071124361e-07,
+      "loss": 0.0001,
+      "step": 373
+    },
+    {
+      "epoch": 4.857142857142857,
+      "grad_norm": 0.02041393704712391,
+      "learning_rate": 6.942984846889639e-07,
+      "loss": 0.0002,
+      "step": 374
+    },
+    {
+      "epoch": 4.87012987012987,
+      "grad_norm": 0.04283998906612396,
+      "learning_rate": 6.793616581735063e-07,
+      "loss": 0.0002,
+      "step": 375
+    },
+    {
+      "epoch": 4.883116883116883,
+      "grad_norm": 0.010590342804789543,
+      "learning_rate": 6.645619525290043e-07,
+      "loss": 0.0001,
+      "step": 376
+    },
+    {
+      "epoch": 4.896103896103896,
+      "grad_norm": 0.012111359275877476,
+      "learning_rate": 6.499004823911772e-07,
+      "loss": 0.0001,
+      "step": 377
+    },
+    {
+      "epoch": 4.909090909090909,
+      "grad_norm": 0.05435957387089729,
+      "learning_rate": 6.353783519845752e-07,
+      "loss": 0.0002,
+      "step": 378
+    },
+    {
+      "epoch": 4.922077922077922,
+      "grad_norm": 0.030580980703234673,
+      "learning_rate": 6.209966550394162e-07,
+      "loss": 0.0001,
+      "step": 379
+    },
+    {
+      "epoch": 4.935064935064935,
+      "grad_norm": 0.026306739076972008,
+      "learning_rate": 6.067564747092095e-07,
+      "loss": 0.0001,
+      "step": 380
+    },
+    {
+      "epoch": 4.948051948051948,
+      "grad_norm": 0.01105985976755619,
+      "learning_rate": 5.926588834891823e-07,
+      "loss": 0.0001,
+      "step": 381
+    },
+    {
+      "epoch": 4.961038961038961,
+      "grad_norm": 0.011411640793085098,
+      "learning_rate": 5.787049431354996e-07,
+      "loss": 0.0001,
+      "step": 382
+    },
+    {
+      "epoch": 4.974025974025974,
+      "grad_norm": 0.013728387653827667,
+      "learning_rate": 5.648957045853043e-07,
+      "loss": 0.0001,
+      "step": 383
+    },
+    {
+      "epoch": 4.987012987012987,
+      "grad_norm": 0.016256002709269524,
+      "learning_rate": 5.512322078775603e-07,
+      "loss": 0.0001,
+      "step": 384
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 0.01369263231754303,
+      "learning_rate": 5.377154820747271e-07,
+      "loss": 0.0001,
+      "step": 385
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 462,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 6,
+  "save_steps": 77,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.604015918768456e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-385/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92007cb7b6ef33ad8c015d3ab5f86bb507fe352d75f5b7187cbf5d8f57b8447a
+size 7992

checkpoint-385/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)