Upload folder using huggingface_hub

checkpoint-296/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-296/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": [
+    "up_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj",
+    "k_proj",
+    "v_proj",
+    "q_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-296/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d38b7b7167f0eed66be4f587d58729a59d05155ef177bf6a47440903e5257360
+size 1291899552

checkpoint-296/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-296/global_step294/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:f6dbb482014eaf6dee81b23dc4420ba1d1dd93e4dba61357c8ccd9790cd3144d
+size 7751120668

checkpoint-296/global_step294/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d4315f64576183cb26b98625c6845aa0b2a60c413b5aecd9a52d33e400cc3e0
+size 1396311192

checkpoint-296/latest

@@ -0,0 +1 @@
+global_step294

checkpoint-296/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3a7d01b5f8accea40aee757137e35d2dc168ef6677dd6269159c11688eeb927
+size 14244

checkpoint-296/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4f1193a6c77d2212f2618afc120fddb487bece245042bef7ebff23653813529
+size 1064

checkpoint-296/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-296/tokenizer.json

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

checkpoint-296/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-296/trainer_state.json

@@ -0,0 +1,2105 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.9551569506726456,
+  "eval_steps": 500,
+  "global_step": 296,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.013452914798206279,
+      "grad_norm": 95.4926986694336,
+      "learning_rate": 5.0000000000000004e-08,
+      "loss": 12.2856,
+      "step": 1
+    },
+    {
+      "epoch": 0.026905829596412557,
+      "grad_norm": 93.69285583496094,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 12.2383,
+      "step": 2
+    },
+    {
+      "epoch": 0.04035874439461883,
+      "grad_norm": 95.09840393066406,
+      "learning_rate": 1.5000000000000002e-07,
+      "loss": 12.1293,
+      "step": 3
+    },
+    {
+      "epoch": 0.053811659192825115,
+      "grad_norm": 95.04216766357422,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 12.1453,
+      "step": 4
+    },
+    {
+      "epoch": 0.06726457399103139,
+      "grad_norm": 93.44210052490234,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 12.165,
+      "step": 5
+    },
+    {
+      "epoch": 0.08071748878923767,
+      "grad_norm": 93.28514862060547,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 12.063,
+      "step": 6
+    },
+    {
+      "epoch": 0.09417040358744394,
+      "grad_norm": 95.3654556274414,
+      "learning_rate": 3.5000000000000004e-07,
+      "loss": 11.9676,
+      "step": 7
+    },
+    {
+      "epoch": 0.10762331838565023,
+      "grad_norm": 96.05154418945312,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 12.0911,
+      "step": 8
+    },
+    {
+      "epoch": 0.1210762331838565,
+      "grad_norm": 97.69881439208984,
+      "learning_rate": 4.5000000000000003e-07,
+      "loss": 12.0717,
+      "step": 9
+    },
+    {
+      "epoch": 0.13452914798206278,
+      "grad_norm": 95.55254364013672,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 11.9752,
+      "step": 10
+    },
+    {
+      "epoch": 0.14798206278026907,
+      "grad_norm": 95.9182357788086,
+      "learning_rate": 5.5e-07,
+      "loss": 11.9413,
+      "step": 11
+    },
+    {
+      "epoch": 0.16143497757847533,
+      "grad_norm": 95.40771484375,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 11.7523,
+      "step": 12
+    },
+    {
+      "epoch": 0.17488789237668162,
+      "grad_norm": 94.40055847167969,
+      "learning_rate": 6.5e-07,
+      "loss": 11.6384,
+      "step": 13
+    },
+    {
+      "epoch": 0.18834080717488788,
+      "grad_norm": 93.58352661132812,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 11.4293,
+      "step": 14
+    },
+    {
+      "epoch": 0.20179372197309417,
+      "grad_norm": 94.48737335205078,
+      "learning_rate": 7.5e-07,
+      "loss": 11.1445,
+      "step": 15
+    },
+    {
+      "epoch": 0.21524663677130046,
+      "grad_norm": 92.44265747070312,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 10.7705,
+      "step": 16
+    },
+    {
+      "epoch": 0.22869955156950672,
+      "grad_norm": 90.97422790527344,
+      "learning_rate": 8.500000000000001e-07,
+      "loss": 10.3754,
+      "step": 17
+    },
+    {
+      "epoch": 0.242152466367713,
+      "grad_norm": 88.54856872558594,
+      "learning_rate": 9.000000000000001e-07,
+      "loss": 10.0019,
+      "step": 18
+    },
+    {
+      "epoch": 0.2556053811659193,
+      "grad_norm": 88.39138793945312,
+      "learning_rate": 9.500000000000001e-07,
+      "loss": 9.5093,
+      "step": 19
+    },
+    {
+      "epoch": 0.26905829596412556,
+      "grad_norm": 86.55109405517578,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 9.2342,
+      "step": 20
+    },
+    {
+      "epoch": 0.2825112107623318,
+      "grad_norm": 80.62335205078125,
+      "learning_rate": 1.0500000000000001e-06,
+      "loss": 8.597,
+      "step": 21
+    },
+    {
+      "epoch": 0.29596412556053814,
+      "grad_norm": 73.67768859863281,
+      "learning_rate": 1.1e-06,
+      "loss": 8.1122,
+      "step": 22
+    },
+    {
+      "epoch": 0.3094170403587444,
+      "grad_norm": 64.57353210449219,
+      "learning_rate": 1.1500000000000002e-06,
+      "loss": 7.6455,
+      "step": 23
+    },
+    {
+      "epoch": 0.32286995515695066,
+      "grad_norm": 55.2818603515625,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 7.2493,
+      "step": 24
+    },
+    {
+      "epoch": 0.336322869955157,
+      "grad_norm": 48.274452209472656,
+      "learning_rate": 1.25e-06,
+      "loss": 7.0377,
+      "step": 25
+    },
+    {
+      "epoch": 0.34977578475336324,
+      "grad_norm": 42.7370491027832,
+      "learning_rate": 1.3e-06,
+      "loss": 6.5782,
+      "step": 26
+    },
+    {
+      "epoch": 0.3632286995515695,
+      "grad_norm": 39.297462463378906,
+      "learning_rate": 1.3500000000000002e-06,
+      "loss": 6.2558,
+      "step": 27
+    },
+    {
+      "epoch": 0.37668161434977576,
+      "grad_norm": 37.91667938232422,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 5.9809,
+      "step": 28
+    },
+    {
+      "epoch": 0.3901345291479821,
+      "grad_norm": 37.87322998046875,
+      "learning_rate": 1.45e-06,
+      "loss": 5.7268,
+      "step": 29
+    },
+    {
+      "epoch": 0.40358744394618834,
+      "grad_norm": 36.48906707763672,
+      "learning_rate": 1.5e-06,
+      "loss": 5.449,
+      "step": 30
+    },
+    {
+      "epoch": 0.4170403587443946,
+      "grad_norm": 36.38510513305664,
+      "learning_rate": 1.5500000000000002e-06,
+      "loss": 5.1884,
+      "step": 31
+    },
+    {
+      "epoch": 0.4304932735426009,
+      "grad_norm": 35.656829833984375,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 4.899,
+      "step": 32
+    },
+    {
+      "epoch": 0.4439461883408072,
+      "grad_norm": 34.09960174560547,
+      "learning_rate": 1.6500000000000003e-06,
+      "loss": 4.5842,
+      "step": 33
+    },
+    {
+      "epoch": 0.45739910313901344,
+      "grad_norm": 32.74240493774414,
+      "learning_rate": 1.7000000000000002e-06,
+      "loss": 4.3009,
+      "step": 34
+    },
+    {
+      "epoch": 0.47085201793721976,
+      "grad_norm": 31.867507934570312,
+      "learning_rate": 1.75e-06,
+      "loss": 3.9865,
+      "step": 35
+    },
+    {
+      "epoch": 0.484304932735426,
+      "grad_norm": 30.741374969482422,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 3.6916,
+      "step": 36
+    },
+    {
+      "epoch": 0.4977578475336323,
+      "grad_norm": 27.8775577545166,
+      "learning_rate": 1.85e-06,
+      "loss": 3.3719,
+      "step": 37
+    },
+    {
+      "epoch": 0.5112107623318386,
+      "grad_norm": 25.97083282470703,
+      "learning_rate": 1.9000000000000002e-06,
+      "loss": 3.0907,
+      "step": 38
+    },
+    {
+      "epoch": 0.5246636771300448,
+      "grad_norm": 23.62006950378418,
+      "learning_rate": 1.9500000000000004e-06,
+      "loss": 2.8336,
+      "step": 39
+    },
+    {
+      "epoch": 0.5381165919282511,
+      "grad_norm": 23.80520248413086,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 2.5717,
+      "step": 40
+    },
+    {
+      "epoch": 0.5515695067264574,
+      "grad_norm": 25.32924461364746,
+      "learning_rate": 2.05e-06,
+      "loss": 2.3658,
+      "step": 41
+    },
+    {
+      "epoch": 0.5650224215246636,
+      "grad_norm": 26.20570182800293,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 2.2443,
+      "step": 42
+    },
+    {
+      "epoch": 0.57847533632287,
+      "grad_norm": 24.581693649291992,
+      "learning_rate": 2.15e-06,
+      "loss": 1.926,
+      "step": 43
+    },
+    {
+      "epoch": 0.5919282511210763,
+      "grad_norm": 24.414310455322266,
+      "learning_rate": 2.2e-06,
+      "loss": 1.7034,
+      "step": 44
+    },
+    {
+      "epoch": 0.6053811659192825,
+      "grad_norm": 22.691083908081055,
+      "learning_rate": 2.25e-06,
+      "loss": 1.4857,
+      "step": 45
+    },
+    {
+      "epoch": 0.6188340807174888,
+      "grad_norm": 20.669803619384766,
+      "learning_rate": 2.3000000000000004e-06,
+      "loss": 1.2415,
+      "step": 46
+    },
+    {
+      "epoch": 0.6322869955156951,
+      "grad_norm": 20.149641036987305,
+      "learning_rate": 2.35e-06,
+      "loss": 0.997,
+      "step": 47
+    },
+    {
+      "epoch": 0.6457399103139013,
+      "grad_norm": 18.632596969604492,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 0.7552,
+      "step": 48
+    },
+    {
+      "epoch": 0.6591928251121076,
+      "grad_norm": 16.93793296813965,
+      "learning_rate": 2.4500000000000003e-06,
+      "loss": 0.5883,
+      "step": 49
+    },
+    {
+      "epoch": 0.672645739910314,
+      "grad_norm": 14.432519912719727,
+      "learning_rate": 2.5e-06,
+      "loss": 0.4382,
+      "step": 50
+    },
+    {
+      "epoch": 0.6860986547085202,
+      "grad_norm": 11.829660415649414,
+      "learning_rate": 2.55e-06,
+      "loss": 0.2983,
+      "step": 51
+    },
+    {
+      "epoch": 0.6995515695067265,
+      "grad_norm": 8.680500030517578,
+      "learning_rate": 2.6e-06,
+      "loss": 0.1988,
+      "step": 52
+    },
+    {
+      "epoch": 0.7130044843049327,
+      "grad_norm": 6.53156852722168,
+      "learning_rate": 2.6500000000000005e-06,
+      "loss": 0.1589,
+      "step": 53
+    },
+    {
+      "epoch": 0.726457399103139,
+      "grad_norm": 2.9756624698638916,
+      "learning_rate": 2.7000000000000004e-06,
+      "loss": 0.0686,
+      "step": 54
+    },
+    {
+      "epoch": 0.7399103139013453,
+      "grad_norm": 5.545580863952637,
+      "learning_rate": 2.7500000000000004e-06,
+      "loss": 0.0865,
+      "step": 55
+    },
+    {
+      "epoch": 0.7533632286995515,
+      "grad_norm": 4.045405387878418,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 0.0949,
+      "step": 56
+    },
+    {
+      "epoch": 0.7668161434977578,
+      "grad_norm": 1.6688120365142822,
+      "learning_rate": 2.85e-06,
+      "loss": 0.0396,
+      "step": 57
+    },
+    {
+      "epoch": 0.7802690582959642,
+      "grad_norm": 2.4520657062530518,
+      "learning_rate": 2.9e-06,
+      "loss": 0.0439,
+      "step": 58
+    },
+    {
+      "epoch": 0.7937219730941704,
+      "grad_norm": 2.608729600906372,
+      "learning_rate": 2.95e-06,
+      "loss": 0.057,
+      "step": 59
+    },
+    {
+      "epoch": 0.8071748878923767,
+      "grad_norm": 2.365234851837158,
+      "learning_rate": 3e-06,
+      "loss": 0.0547,
+      "step": 60
+    },
+    {
+      "epoch": 0.820627802690583,
+      "grad_norm": 0.787550687789917,
+      "learning_rate": 3.05e-06,
+      "loss": 0.0209,
+      "step": 61
+    },
+    {
+      "epoch": 0.8340807174887892,
+      "grad_norm": 0.7686442732810974,
+      "learning_rate": 3.1000000000000004e-06,
+      "loss": 0.0221,
+      "step": 62
+    },
+    {
+      "epoch": 0.8475336322869955,
+      "grad_norm": 1.2510555982589722,
+      "learning_rate": 3.1500000000000003e-06,
+      "loss": 0.0165,
+      "step": 63
+    },
+    {
+      "epoch": 0.8609865470852018,
+      "grad_norm": 0.8923770189285278,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 0.0187,
+      "step": 64
+    },
+    {
+      "epoch": 0.874439461883408,
+      "grad_norm": 0.8052615523338318,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.0266,
+      "step": 65
+    },
+    {
+      "epoch": 0.8878923766816144,
+      "grad_norm": 0.6710303425788879,
+      "learning_rate": 3.3000000000000006e-06,
+      "loss": 0.0154,
+      "step": 66
+    },
+    {
+      "epoch": 0.9013452914798207,
+      "grad_norm": 0.5213025212287903,
+      "learning_rate": 3.3500000000000005e-06,
+      "loss": 0.0085,
+      "step": 67
+    },
+    {
+      "epoch": 0.9147982062780269,
+      "grad_norm": 0.5758580565452576,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 0.0133,
+      "step": 68
+    },
+    {
+      "epoch": 0.9282511210762332,
+      "grad_norm": 0.6828752160072327,
+      "learning_rate": 3.45e-06,
+      "loss": 0.0186,
+      "step": 69
+    },
+    {
+      "epoch": 0.9417040358744395,
+      "grad_norm": 0.6814988255500793,
+      "learning_rate": 3.5e-06,
+      "loss": 0.0215,
+      "step": 70
+    },
+    {
+      "epoch": 0.9551569506726457,
+      "grad_norm": 0.718296229839325,
+      "learning_rate": 3.5500000000000003e-06,
+      "loss": 0.0204,
+      "step": 71
+    },
+    {
+      "epoch": 0.968609865470852,
+      "grad_norm": 0.7816944122314453,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 0.0184,
+      "step": 72
+    },
+    {
+      "epoch": 0.9820627802690582,
+      "grad_norm": 0.6058817505836487,
+      "learning_rate": 3.65e-06,
+      "loss": 0.0179,
+      "step": 73
+    },
+    {
+      "epoch": 0.9955156950672646,
+      "grad_norm": 1.0496101379394531,
+      "learning_rate": 3.7e-06,
+      "loss": 0.032,
+      "step": 74
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.0496101379394531,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.0137,
+      "step": 75
+    },
+    {
+      "epoch": 1.0134529147982063,
+      "grad_norm": 1.598720669746399,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.0211,
+      "step": 76
+    },
+    {
+      "epoch": 1.0269058295964126,
+      "grad_norm": 0.7792187333106995,
+      "learning_rate": 3.85e-06,
+      "loss": 0.0213,
+      "step": 77
+    },
+    {
+      "epoch": 1.0403587443946187,
+      "grad_norm": 0.7717252373695374,
+      "learning_rate": 3.900000000000001e-06,
+      "loss": 0.0151,
+      "step": 78
+    },
+    {
+      "epoch": 1.053811659192825,
+      "grad_norm": 0.39334648847579956,
+      "learning_rate": 3.95e-06,
+      "loss": 0.0186,
+      "step": 79
+    },
+    {
+      "epoch": 1.0672645739910314,
+      "grad_norm": 0.9775457382202148,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.0043,
+      "step": 80
+    },
+    {
+      "epoch": 1.0807174887892377,
+      "grad_norm": 1.2425150871276855,
+      "learning_rate": 4.05e-06,
+      "loss": 0.0179,
+      "step": 81
+    },
+    {
+      "epoch": 1.094170403587444,
+      "grad_norm": 0.3884654939174652,
+      "learning_rate": 4.1e-06,
+      "loss": 0.008,
+      "step": 82
+    },
+    {
+      "epoch": 1.1076233183856503,
+      "grad_norm": 0.4746466875076294,
+      "learning_rate": 4.15e-06,
+      "loss": 0.0144,
+      "step": 83
+    },
+    {
+      "epoch": 1.1210762331838564,
+      "grad_norm": 0.6812214255332947,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 0.0072,
+      "step": 84
+    },
+    {
+      "epoch": 1.1345291479820627,
+      "grad_norm": 0.5414469838142395,
+      "learning_rate": 4.25e-06,
+      "loss": 0.016,
+      "step": 85
+    },
+    {
+      "epoch": 1.147982062780269,
+      "grad_norm": 0.8709024786949158,
+      "learning_rate": 4.3e-06,
+      "loss": 0.0154,
+      "step": 86
+    },
+    {
+      "epoch": 1.1614349775784754,
+      "grad_norm": 0.3181096613407135,
+      "learning_rate": 4.350000000000001e-06,
+      "loss": 0.0039,
+      "step": 87
+    },
+    {
+      "epoch": 1.1748878923766817,
+      "grad_norm": 0.2581265866756439,
+      "learning_rate": 4.4e-06,
+      "loss": 0.0014,
+      "step": 88
+    },
+    {
+      "epoch": 1.188340807174888,
+      "grad_norm": 0.10870776325464249,
+      "learning_rate": 4.450000000000001e-06,
+      "loss": 0.0088,
+      "step": 89
+    },
+    {
+      "epoch": 1.201793721973094,
+      "grad_norm": 0.45314452052116394,
+      "learning_rate": 4.5e-06,
+      "loss": 0.0019,
+      "step": 90
+    },
+    {
+      "epoch": 1.2152466367713004,
+      "grad_norm": 0.7328381538391113,
+      "learning_rate": 4.5500000000000005e-06,
+      "loss": 0.0082,
+      "step": 91
+    },
+    {
+      "epoch": 1.2286995515695067,
+      "grad_norm": 0.6641069650650024,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 0.0075,
+      "step": 92
+    },
+    {
+      "epoch": 1.242152466367713,
+      "grad_norm": 0.27984222769737244,
+      "learning_rate": 4.65e-06,
+      "loss": 0.0009,
+      "step": 93
+    },
+    {
+      "epoch": 1.2556053811659194,
+      "grad_norm": 0.8341127634048462,
+      "learning_rate": 4.7e-06,
+      "loss": 0.0204,
+      "step": 94
+    },
+    {
+      "epoch": 1.2690582959641254,
+      "grad_norm": 1.0140557289123535,
+      "learning_rate": 4.75e-06,
+      "loss": 0.0228,
+      "step": 95
+    },
+    {
+      "epoch": 1.2825112107623318,
+      "grad_norm": 0.9439787268638611,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.027,
+      "step": 96
+    },
+    {
+      "epoch": 1.295964125560538,
+      "grad_norm": 1.3867762088775635,
+      "learning_rate": 4.85e-06,
+      "loss": 0.0223,
+      "step": 97
+    },
+    {
+      "epoch": 1.3094170403587444,
+      "grad_norm": 0.4747941493988037,
+      "learning_rate": 4.9000000000000005e-06,
+      "loss": 0.0048,
+      "step": 98
+    },
+    {
+      "epoch": 1.3228699551569507,
+      "grad_norm": 0.5673424601554871,
+      "learning_rate": 4.95e-06,
+      "loss": 0.0033,
+      "step": 99
+    },
+    {
+      "epoch": 1.336322869955157,
+      "grad_norm": 0.24510182440280914,
+      "learning_rate": 5e-06,
+      "loss": 0.0048,
+      "step": 100
+    },
+    {
+      "epoch": 1.3497757847533634,
+      "grad_norm": 0.6084151268005371,
+      "learning_rate": 4.99989574668946e-06,
+      "loss": 0.032,
+      "step": 101
+    },
+    {
+      "epoch": 1.3632286995515694,
+      "grad_norm": 0.606272280216217,
+      "learning_rate": 4.999582995452842e-06,
+      "loss": 0.0058,
+      "step": 102
+    },
+    {
+      "epoch": 1.3766816143497758,
+      "grad_norm": 0.2270481288433075,
+      "learning_rate": 4.999061772374426e-06,
+      "loss": 0.0043,
+      "step": 103
+    },
+    {
+      "epoch": 1.390134529147982,
+      "grad_norm": 0.2832431495189667,
+      "learning_rate": 4.998332120925598e-06,
+      "loss": 0.0079,
+      "step": 104
+    },
+    {
+      "epoch": 1.4035874439461884,
+      "grad_norm": 0.44565466046333313,
+      "learning_rate": 4.9973941019612235e-06,
+      "loss": 0.017,
+      "step": 105
+    },
+    {
+      "epoch": 1.4170403587443947,
+      "grad_norm": 0.3518397808074951,
+      "learning_rate": 4.996247793714565e-06,
+      "loss": 0.0057,
+      "step": 106
+    },
+    {
+      "epoch": 1.4304932735426008,
+      "grad_norm": 0.41977858543395996,
+      "learning_rate": 4.994893291790768e-06,
+      "loss": 0.0056,
+      "step": 107
+    },
+    {
+      "epoch": 1.4439461883408071,
+      "grad_norm": 0.22567600011825562,
+      "learning_rate": 4.993330709158879e-06,
+      "loss": 0.0067,
+      "step": 108
+    },
+    {
+      "epoch": 1.4573991031390134,
+      "grad_norm": 0.2184637039899826,
+      "learning_rate": 4.9915601761424304e-06,
+      "loss": 0.0043,
+      "step": 109
+    },
+    {
+      "epoch": 1.4708520179372198,
+      "grad_norm": 0.24834735691547394,
+      "learning_rate": 4.989581840408562e-06,
+      "loss": 0.0034,
+      "step": 110
+    },
+    {
+      "epoch": 1.484304932735426,
+      "grad_norm": 0.13165591657161713,
+      "learning_rate": 4.987395866955716e-06,
+      "loss": 0.0041,
+      "step": 111
+    },
+    {
+      "epoch": 1.4977578475336322,
+      "grad_norm": 0.27152982354164124,
+      "learning_rate": 4.9850024380998655e-06,
+      "loss": 0.0034,
+      "step": 112
+    },
+    {
+      "epoch": 1.5112107623318387,
+      "grad_norm": 0.2793160676956177,
+      "learning_rate": 4.982401753459317e-06,
+      "loss": 0.0049,
+      "step": 113
+    },
+    {
+      "epoch": 1.5246636771300448,
+      "grad_norm": 0.06674113124608994,
+      "learning_rate": 4.979594029938058e-06,
+      "loss": 0.0034,
+      "step": 114
+    },
+    {
+      "epoch": 1.5381165919282511,
+      "grad_norm": 0.2601087689399719,
+      "learning_rate": 4.976579501707665e-06,
+      "loss": 0.0025,
+      "step": 115
+    },
+    {
+      "epoch": 1.5515695067264574,
+      "grad_norm": 0.17075951397418976,
+      "learning_rate": 4.973358420187776e-06,
+      "loss": 0.0005,
+      "step": 116
+    },
+    {
+      "epoch": 1.5650224215246635,
+      "grad_norm": 0.11211276799440384,
+      "learning_rate": 4.969931054025122e-06,
+      "loss": 0.0059,
+      "step": 117
+    },
+    {
+      "epoch": 1.57847533632287,
+      "grad_norm": 0.5140838027000427,
+      "learning_rate": 4.966297689071117e-06,
+      "loss": 0.018,
+      "step": 118
+    },
+    {
+      "epoch": 1.5919282511210762,
+      "grad_norm": 0.5356995463371277,
+      "learning_rate": 4.962458628358021e-06,
+      "loss": 0.0007,
+      "step": 119
+    },
+    {
+      "epoch": 1.6053811659192825,
+      "grad_norm": 0.3967442214488983,
+      "learning_rate": 4.958414192073665e-06,
+      "loss": 0.014,
+      "step": 120
+    },
+    {
+      "epoch": 1.6188340807174888,
+      "grad_norm": 0.4567921757698059,
+      "learning_rate": 4.954164717534748e-06,
+      "loss": 0.0027,
+      "step": 121
+    },
+    {
+      "epoch": 1.6322869955156951,
+      "grad_norm": 0.4405613839626312,
+      "learning_rate": 4.949710559158699e-06,
+      "loss": 0.0135,
+      "step": 122
+    },
+    {
+      "epoch": 1.6457399103139014,
+      "grad_norm": 0.7164422869682312,
+      "learning_rate": 4.945052088434123e-06,
+      "loss": 0.0207,
+      "step": 123
+    },
+    {
+      "epoch": 1.6591928251121075,
+      "grad_norm": 0.47173142433166504,
+      "learning_rate": 4.940189693889819e-06,
+      "loss": 0.0019,
+      "step": 124
+    },
+    {
+      "epoch": 1.672645739910314,
+      "grad_norm": 0.4606887996196747,
+      "learning_rate": 4.9351237810623655e-06,
+      "loss": 0.0125,
+      "step": 125
+    },
+    {
+      "epoch": 1.6860986547085202,
+      "grad_norm": 0.7209401726722717,
+      "learning_rate": 4.929854772462312e-06,
+      "loss": 0.012,
+      "step": 126
+    },
+    {
+      "epoch": 1.6995515695067265,
+      "grad_norm": 1.7617985010147095,
+      "learning_rate": 4.924383107538929e-06,
+      "loss": 0.0079,
+      "step": 127
+    },
+    {
+      "epoch": 1.7130044843049328,
+      "grad_norm": 1.0861084461212158,
+      "learning_rate": 4.918709242643563e-06,
+      "loss": 0.0091,
+      "step": 128
+    },
+    {
+      "epoch": 1.726457399103139,
+      "grad_norm": 0.11263061314821243,
+      "learning_rate": 4.9128336509915746e-06,
+      "loss": 0.0006,
+      "step": 129
+    },
+    {
+      "epoch": 1.7399103139013454,
+      "grad_norm": 0.3411642909049988,
+      "learning_rate": 4.906756822622865e-06,
+      "loss": 0.0037,
+      "step": 130
+    },
+    {
+      "epoch": 1.7533632286995515,
+      "grad_norm": 0.1550491899251938,
+      "learning_rate": 4.900479264361017e-06,
+      "loss": 0.0056,
+      "step": 131
+    },
+    {
+      "epoch": 1.7668161434977578,
+      "grad_norm": 0.8448930382728577,
+      "learning_rate": 4.894001499771015e-06,
+      "loss": 0.0135,
+      "step": 132
+    },
+    {
+      "epoch": 1.7802690582959642,
+      "grad_norm": 0.5105615258216858,
+      "learning_rate": 4.887324069115582e-06,
+      "loss": 0.009,
+      "step": 133
+    },
+    {
+      "epoch": 1.7937219730941703,
+      "grad_norm": 0.5042226910591125,
+      "learning_rate": 4.880447529310118e-06,
+      "loss": 0.0022,
+      "step": 134
+    },
+    {
+      "epoch": 1.8071748878923768,
+      "grad_norm": 0.15797697007656097,
+      "learning_rate": 4.873372453876255e-06,
+      "loss": 0.0073,
+      "step": 135
+    },
+    {
+      "epoch": 1.8206278026905829,
+      "grad_norm": 0.47805944085121155,
+      "learning_rate": 4.866099432894023e-06,
+      "loss": 0.0084,
+      "step": 136
+    },
+    {
+      "epoch": 1.8340807174887892,
+      "grad_norm": 0.3370952904224396,
+      "learning_rate": 4.858629072952635e-06,
+      "loss": 0.0037,
+      "step": 137
+    },
+    {
+      "epoch": 1.8475336322869955,
+      "grad_norm": 0.30983835458755493,
+      "learning_rate": 4.850961997099892e-06,
+      "loss": 0.0024,
+      "step": 138
+    },
+    {
+      "epoch": 1.8609865470852018,
+      "grad_norm": 0.2800588011741638,
+      "learning_rate": 4.843098844790228e-06,
+      "loss": 0.0032,
+      "step": 139
+    },
+    {
+      "epoch": 1.8744394618834082,
+      "grad_norm": 0.2037343531847,
+      "learning_rate": 4.835040271831371e-06,
+      "loss": 0.0016,
+      "step": 140
+    },
+    {
+      "epoch": 1.8878923766816142,
+      "grad_norm": 0.45981553196907043,
+      "learning_rate": 4.826786950329646e-06,
+      "loss": 0.0035,
+      "step": 141
+    },
+    {
+      "epoch": 1.9013452914798208,
+      "grad_norm": 0.17092454433441162,
+      "learning_rate": 4.818339568633926e-06,
+      "loss": 0.0068,
+      "step": 142
+    },
+    {
+      "epoch": 1.9147982062780269,
+      "grad_norm": 0.5339077711105347,
+      "learning_rate": 4.809698831278217e-06,
+      "loss": 0.0026,
+      "step": 143
+    },
+    {
+      "epoch": 1.9282511210762332,
+      "grad_norm": 0.5208529829978943,
+      "learning_rate": 4.800865458922899e-06,
+      "loss": 0.0047,
+      "step": 144
+    },
+    {
+      "epoch": 1.9417040358744395,
+      "grad_norm": 0.7488933801651001,
+      "learning_rate": 4.79184018829462e-06,
+      "loss": 0.0096,
+      "step": 145
+    },
+    {
+      "epoch": 1.9551569506726456,
+      "grad_norm": 0.5727106332778931,
+      "learning_rate": 4.782623772124854e-06,
+      "loss": 0.001,
+      "step": 146
+    },
+    {
+      "epoch": 1.9686098654708521,
+      "grad_norm": 0.3813195526599884,
+      "learning_rate": 4.77321697908712e-06,
+      "loss": 0.0085,
+      "step": 147
+    },
+    {
+      "epoch": 1.9820627802690582,
+      "grad_norm": 0.5406109094619751,
+      "learning_rate": 4.763620593732867e-06,
+      "loss": 0.0017,
+      "step": 148
+    },
+    {
+      "epoch": 1.9955156950672646,
+      "grad_norm": 0.391985148191452,
+      "learning_rate": 4.7538354164260515e-06,
+      "loss": 0.0019,
+      "step": 149
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.391985148191452,
+      "learning_rate": 4.743862263276376e-06,
+      "loss": 0.0003,
+      "step": 150
+    },
+    {
+      "epoch": 2.013452914798206,
+      "grad_norm": 0.10133524239063263,
+      "learning_rate": 4.733701966071226e-06,
+      "loss": 0.0012,
+      "step": 151
+    },
+    {
+      "epoch": 2.0269058295964126,
+      "grad_norm": 0.2032414674758911,
+      "learning_rate": 4.723355372206297e-06,
+      "loss": 0.0001,
+      "step": 152
+    },
+    {
+      "epoch": 2.0403587443946187,
+      "grad_norm": 0.00796876847743988,
+      "learning_rate": 4.712823344614921e-06,
+      "loss": 0.0025,
+      "step": 153
+    },
+    {
+      "epoch": 2.0538116591928253,
+      "grad_norm": 0.36200040578842163,
+      "learning_rate": 4.702106761696091e-06,
+      "loss": 0.0003,
+      "step": 154
+    },
+    {
+      "epoch": 2.0672645739910314,
+      "grad_norm": 0.19646115601062775,
+      "learning_rate": 4.691206517241205e-06,
+      "loss": 0.0009,
+      "step": 155
+    },
+    {
+      "epoch": 2.0807174887892375,
+      "grad_norm": 0.06766581535339355,
+      "learning_rate": 4.68012352035952e-06,
+      "loss": 0.0001,
+      "step": 156
+    },
+    {
+      "epoch": 2.094170403587444,
+      "grad_norm": 0.00797713827341795,
+      "learning_rate": 4.668858695402326e-06,
+      "loss": 0.0001,
+      "step": 157
+    },
+    {
+      "epoch": 2.10762331838565,
+      "grad_norm": 0.020394539460539818,
+      "learning_rate": 4.657412981885862e-06,
+      "loss": 0.0002,
+      "step": 158
+    },
+    {
+      "epoch": 2.1210762331838566,
+      "grad_norm": 0.017584379762411118,
+      "learning_rate": 4.645787334412945e-06,
+      "loss": 0.0002,
+      "step": 159
+    },
+    {
+      "epoch": 2.1345291479820627,
+      "grad_norm": 0.07967082411050797,
+      "learning_rate": 4.633982722593367e-06,
+      "loss": 0.0003,
+      "step": 160
+    },
+    {
+      "epoch": 2.1479820627802693,
+      "grad_norm": 0.01606675237417221,
+      "learning_rate": 4.622000130963015e-06,
+      "loss": 0.0003,
+      "step": 161
+    },
+    {
+      "epoch": 2.1614349775784754,
+      "grad_norm": 0.06052660569548607,
+      "learning_rate": 4.6098405589017685e-06,
+      "loss": 0.0007,
+      "step": 162
+    },
+    {
+      "epoch": 2.1748878923766815,
+      "grad_norm": 0.07991409301757812,
+      "learning_rate": 4.597505020550138e-06,
+      "loss": 0.0002,
+      "step": 163
+    },
+    {
+      "epoch": 2.188340807174888,
+      "grad_norm": 0.04991272836923599,
+      "learning_rate": 4.584994544724695e-06,
+      "loss": 0.0001,
+      "step": 164
+    },
+    {
+      "epoch": 2.201793721973094,
+      "grad_norm": 0.027111921459436417,
+      "learning_rate": 4.572310174832255e-06,
+      "loss": 0.0001,
+      "step": 165
+    },
+    {
+      "epoch": 2.2152466367713006,
+      "grad_norm": 0.006566982250660658,
+      "learning_rate": 4.5594529687828615e-06,
+      "loss": 0.0,
+      "step": 166
+    },
+    {
+      "epoch": 2.2286995515695067,
+      "grad_norm": 0.002437079790979624,
+      "learning_rate": 4.546423998901549e-06,
+      "loss": 0.0004,
+      "step": 167
+    },
+    {
+      "epoch": 2.242152466367713,
+      "grad_norm": 0.08434150367975235,
+      "learning_rate": 4.533224351838914e-06,
+      "loss": 0.0001,
+      "step": 168
+    },
+    {
+      "epoch": 2.2556053811659194,
+      "grad_norm": 0.013094129040837288,
+      "learning_rate": 4.519855128480478e-06,
+      "loss": 0.0001,
+      "step": 169
+    },
+    {
+      "epoch": 2.2690582959641254,
+      "grad_norm": 0.002770340768620372,
+      "learning_rate": 4.5063174438548775e-06,
+      "loss": 0.0001,
+      "step": 170
+    },
+    {
+      "epoch": 2.282511210762332,
+      "grad_norm": 0.021480072289705276,
+      "learning_rate": 4.492612427040864e-06,
+      "loss": 0.0001,
+      "step": 171
+    },
+    {
+      "epoch": 2.295964125560538,
+      "grad_norm": 0.0028862387407571077,
+      "learning_rate": 4.478741221073136e-06,
+      "loss": 0.0001,
+      "step": 172
+    },
+    {
+      "epoch": 2.3094170403587446,
+      "grad_norm": 0.01969303749501705,
+      "learning_rate": 4.464704982847008e-06,
+      "loss": 0.0,
+      "step": 173
+    },
+    {
+      "epoch": 2.3228699551569507,
+      "grad_norm": 0.00244798487983644,
+      "learning_rate": 4.450504883021923e-06,
+      "loss": 0.0,
+      "step": 174
+    },
+    {
+      "epoch": 2.336322869955157,
+      "grad_norm": 0.0036638586316257715,
+      "learning_rate": 4.436142105923814e-06,
+      "loss": 0.0009,
+      "step": 175
+    },
+    {
+      "epoch": 2.3497757847533634,
+      "grad_norm": 0.4378701448440552,
+      "learning_rate": 4.4216178494463305e-06,
+      "loss": 0.0003,
+      "step": 176
+    },
+    {
+      "epoch": 2.3632286995515694,
+      "grad_norm": 0.004919757135212421,
+      "learning_rate": 4.406933324950929e-06,
+      "loss": 0.0,
+      "step": 177
+    },
+    {
+      "epoch": 2.376681614349776,
+      "grad_norm": 0.0061010573990643024,
+      "learning_rate": 4.392089757165841e-06,
+      "loss": 0.0001,
+      "step": 178
+    },
+    {
+      "epoch": 2.390134529147982,
+      "grad_norm": 0.00750540429726243,
+      "learning_rate": 4.377088384083935e-06,
+      "loss": 0.0004,
+      "step": 179
+    },
+    {
+      "epoch": 2.403587443946188,
+      "grad_norm": 0.18055735528469086,
+      "learning_rate": 4.361930456859455e-06,
+      "loss": 0.0001,
+      "step": 180
+    },
+    {
+      "epoch": 2.4170403587443947,
+      "grad_norm": 0.00540179992094636,
+      "learning_rate": 4.346617239703676e-06,
+      "loss": 0.0017,
+      "step": 181
+    },
+    {
+      "epoch": 2.430493273542601,
+      "grad_norm": 0.6051300764083862,
+      "learning_rate": 4.3311500097794655e-06,
+      "loss": 0.0002,
+      "step": 182
+    },
+    {
+      "epoch": 2.4439461883408073,
+      "grad_norm": 0.04465539753437042,
+      "learning_rate": 4.315530057094763e-06,
+      "loss": 0.0001,
+      "step": 183
+    },
+    {
+      "epoch": 2.4573991031390134,
+      "grad_norm": 0.009125534445047379,
+      "learning_rate": 4.2997586843949905e-06,
+      "loss": 0.0001,
+      "step": 184
+    },
+    {
+      "epoch": 2.4708520179372195,
+      "grad_norm": 0.009754106402397156,
+      "learning_rate": 4.2838372070544e-06,
+      "loss": 0.0001,
+      "step": 185
+    },
+    {
+      "epoch": 2.484304932735426,
+      "grad_norm": 0.03795509785413742,
+      "learning_rate": 4.267766952966369e-06,
+      "loss": 0.0006,
+      "step": 186
+    },
+    {
+      "epoch": 2.497757847533632,
+      "grad_norm": 0.19477951526641846,
+      "learning_rate": 4.25154926243265e-06,
+      "loss": 0.0013,
+      "step": 187
+    },
+    {
+      "epoch": 2.5112107623318387,
+      "grad_norm": 0.27783000469207764,
+      "learning_rate": 4.2351854880515856e-06,
+      "loss": 0.0001,
+      "step": 188
+    },
+    {
+      "epoch": 2.524663677130045,
+      "grad_norm": 0.030597640201449394,
+      "learning_rate": 4.218676994605295e-06,
+      "loss": 0.0,
+      "step": 189
+    },
+    {
+      "epoch": 2.538116591928251,
+      "grad_norm": 0.019212787970900536,
+      "learning_rate": 4.202025158945855e-06,
+      "loss": 0.0001,
+      "step": 190
+    },
+    {
+      "epoch": 2.5515695067264574,
+      "grad_norm": 0.017139675095677376,
+      "learning_rate": 4.185231369880461e-06,
+      "loss": 0.0001,
+      "step": 191
+    },
+    {
+      "epoch": 2.5650224215246635,
+      "grad_norm": 0.0013709627091884613,
+      "learning_rate": 4.168297028055599e-06,
+      "loss": 0.0002,
+      "step": 192
+    },
+    {
+      "epoch": 2.57847533632287,
+      "grad_norm": 0.07277967780828476,
+      "learning_rate": 4.151223545840225e-06,
+      "loss": 0.0,
+      "step": 193
+    },
+    {
+      "epoch": 2.591928251121076,
+      "grad_norm": 0.0014003290561959147,
+      "learning_rate": 4.134012347207974e-06,
+      "loss": 0.0001,
+      "step": 194
+    },
+    {
+      "epoch": 2.6053811659192823,
+      "grad_norm": 0.04370618611574173,
+      "learning_rate": 4.116664867618395e-06,
+      "loss": 0.0004,
+      "step": 195
+    },
+    {
+      "epoch": 2.618834080717489,
+      "grad_norm": 0.24697266519069672,
+      "learning_rate": 4.099182553897228e-06,
+      "loss": 0.0,
+      "step": 196
+    },
+    {
+      "epoch": 2.6322869955156953,
+      "grad_norm": 0.0013013904681429267,
+      "learning_rate": 4.081566864115741e-06,
+      "loss": 0.0,
+      "step": 197
+    },
+    {
+      "epoch": 2.6457399103139014,
+      "grad_norm": 0.001239327946677804,
+      "learning_rate": 4.063819267469114e-06,
+      "loss": 0.0,
+      "step": 198
+    },
+    {
+      "epoch": 2.6591928251121075,
+      "grad_norm": 0.007082940544933081,
+      "learning_rate": 4.04594124415391e-06,
+      "loss": 0.0,
+      "step": 199
+    },
+    {
+      "epoch": 2.672645739910314,
+      "grad_norm": 0.00556205864995718,
+      "learning_rate": 4.027934285244624e-06,
+      "loss": 0.0001,
+      "step": 200
+    },
+    {
+      "epoch": 2.68609865470852,
+      "grad_norm": 0.025378312915563583,
+      "learning_rate": 4.009799892569317e-06,
+      "loss": 0.0001,
+      "step": 201
+    },
+    {
+      "epoch": 2.6995515695067267,
+      "grad_norm": 0.006344004534184933,
+      "learning_rate": 3.991539578584368e-06,
+      "loss": 0.0,
+      "step": 202
+    },
+    {
+      "epoch": 2.713004484304933,
+      "grad_norm": 0.0016575426561757922,
+      "learning_rate": 3.973154866248323e-06,
+      "loss": 0.0,
+      "step": 203
+    },
+    {
+      "epoch": 2.726457399103139,
+      "grad_norm": 0.0020909749437123537,
+      "learning_rate": 3.9546472888948825e-06,
+      "loss": 0.0,
+      "step": 204
+    },
+    {
+      "epoch": 2.7399103139013454,
+      "grad_norm": 0.004855802282691002,
+      "learning_rate": 3.936018390105013e-06,
+      "loss": 0.0,
+      "step": 205
+    },
+    {
+      "epoch": 2.7533632286995515,
+      "grad_norm": 0.0082467095926404,
+      "learning_rate": 3.917269723578212e-06,
+      "loss": 0.0,
+      "step": 206
+    },
+    {
+      "epoch": 2.766816143497758,
+      "grad_norm": 0.0013344286708161235,
+      "learning_rate": 3.898402853002921e-06,
+      "loss": 0.0,
+      "step": 207
+    },
+    {
+      "epoch": 2.780269058295964,
+      "grad_norm": 0.0034061160404235125,
+      "learning_rate": 3.879419351926115e-06,
+      "loss": 0.0,
+      "step": 208
+    },
+    {
+      "epoch": 2.7937219730941703,
+      "grad_norm": 0.011912211775779724,
+      "learning_rate": 3.86032080362206e-06,
+      "loss": 0.0,
+      "step": 209
+    },
+    {
+      "epoch": 2.807174887892377,
+      "grad_norm": 0.00134057376999408,
+      "learning_rate": 3.841108800960264e-06,
+      "loss": 0.0,
+      "step": 210
+    },
+    {
+      "epoch": 2.820627802690583,
+      "grad_norm": 0.0016560767544433475,
+      "learning_rate": 3.8217849462726334e-06,
+      "loss": 0.0,
+      "step": 211
+    },
+    {
+      "epoch": 2.8340807174887894,
+      "grad_norm": 0.006285225041210651,
+      "learning_rate": 3.802350851219826e-06,
+      "loss": 0.0,
+      "step": 212
+    },
+    {
+      "epoch": 2.8475336322869955,
+      "grad_norm": 0.0018534021219238639,
+      "learning_rate": 3.7828081366568388e-06,
+      "loss": 0.0,
+      "step": 213
+    },
+    {
+      "epoch": 2.8609865470852016,
+      "grad_norm": 0.0041048116981983185,
+      "learning_rate": 3.763158432497824e-06,
+      "loss": 0.0,
+      "step": 214
+    },
+    {
+      "epoch": 2.874439461883408,
+      "grad_norm": 0.0023216214030981064,
+      "learning_rate": 3.743403377580149e-06,
+      "loss": 0.0,
+      "step": 215
+    },
+    {
+      "epoch": 2.8878923766816142,
+      "grad_norm": 0.000994804548099637,
+      "learning_rate": 3.723544619527714e-06,
+      "loss": 0.0,
+      "step": 216
+    },
+    {
+      "epoch": 2.901345291479821,
+      "grad_norm": 0.003168656025081873,
+      "learning_rate": 3.703583814613536e-06,
+      "loss": 0.0,
+      "step": 217
+    },
+    {
+      "epoch": 2.914798206278027,
+      "grad_norm": 0.0013909138506278396,
+      "learning_rate": 3.6835226276216087e-06,
+      "loss": 0.0,
+      "step": 218
+    },
+    {
+      "epoch": 2.928251121076233,
+      "grad_norm": 0.008506865240633488,
+      "learning_rate": 3.663362731708059e-06,
+      "loss": 0.0,
+      "step": 219
+    },
+    {
+      "epoch": 2.9417040358744395,
+      "grad_norm": 0.0017819993663579226,
+      "learning_rate": 3.6431058082615966e-06,
+      "loss": 0.0,
+      "step": 220
+    },
+    {
+      "epoch": 2.9551569506726456,
+      "grad_norm": 0.0014212332898750901,
+      "learning_rate": 3.6227535467632873e-06,
+      "loss": 0.0,
+      "step": 221
+    },
+    {
+      "epoch": 2.968609865470852,
+      "grad_norm": 0.003146026050671935,
+      "learning_rate": 3.6023076446456415e-06,
+      "loss": 0.0,
+      "step": 222
+    },
+    {
+      "epoch": 2.9820627802690582,
+      "grad_norm": 0.0012437553377822042,
+      "learning_rate": 3.581769807151044e-06,
+      "loss": 0.0,
+      "step": 223
+    },
+    {
+      "epoch": 2.9955156950672643,
+      "grad_norm": 0.017311880365014076,
+      "learning_rate": 3.561141747189538e-06,
+      "loss": 0.0001,
+      "step": 224
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.012313771061599255,
+      "learning_rate": 3.5404251851959537e-06,
+      "loss": 0.0,
+      "step": 225
+    },
+    {
+      "epoch": 3.013452914798206,
+      "grad_norm": 0.0043451120145618916,
+      "learning_rate": 3.519621848986428e-06,
+      "loss": 0.0,
+      "step": 226
+    },
+    {
+      "epoch": 3.0269058295964126,
+      "grad_norm": 0.002464739605784416,
+      "learning_rate": 3.498733473614298e-06,
+      "loss": 0.0,
+      "step": 227
+    },
+    {
+      "epoch": 3.0403587443946187,
+      "grad_norm": 0.0019529856508597732,
+      "learning_rate": 3.47776180122539e-06,
+      "loss": 0.0,
+      "step": 228
+    },
+    {
+      "epoch": 3.0538116591928253,
+      "grad_norm": 0.0013189928140491247,
+      "learning_rate": 3.4567085809127247e-06,
+      "loss": 0.0,
+      "step": 229
+    },
+    {
+      "epoch": 3.0672645739910314,
+      "grad_norm": 0.0021529668010771275,
+      "learning_rate": 3.435575568570633e-06,
+      "loss": 0.0,
+      "step": 230
+    },
+    {
+      "epoch": 3.0807174887892375,
+      "grad_norm": 0.007290184032171965,
+      "learning_rate": 3.4143645267483144e-06,
+      "loss": 0.0,
+      "step": 231
+    },
+    {
+      "epoch": 3.094170403587444,
+      "grad_norm": 0.001345694880001247,
+      "learning_rate": 3.393077224502832e-06,
+      "loss": 0.0,
+      "step": 232
+    },
+    {
+      "epoch": 3.10762331838565,
+      "grad_norm": 0.004343180451542139,
+      "learning_rate": 3.3717154372515716e-06,
+      "loss": 0.0,
+      "step": 233
+    },
+    {
+      "epoch": 3.1210762331838566,
+      "grad_norm": 0.0010816323338076472,
+      "learning_rate": 3.350280946624166e-06,
+      "loss": 0.0,
+      "step": 234
+    },
+    {
+      "epoch": 3.1345291479820627,
+      "grad_norm": 0.003842687699943781,
+      "learning_rate": 3.3287755403139007e-06,
+      "loss": 0.0,
+      "step": 235
+    },
+    {
+      "epoch": 3.1479820627802693,
+      "grad_norm": 0.0013102421071380377,
+      "learning_rate": 3.3072010119286156e-06,
+      "loss": 0.0,
+      "step": 236
+    },
+    {
+      "epoch": 3.1614349775784754,
+      "grad_norm": 0.0014620574656873941,
+      "learning_rate": 3.2855591608411203e-06,
+      "loss": 0.0,
+      "step": 237
+    },
+    {
+      "epoch": 3.1748878923766815,
+      "grad_norm": 0.0007613594643771648,
+      "learning_rate": 3.2638517920391095e-06,
+      "loss": 0.0,
+      "step": 238
+    },
+    {
+      "epoch": 3.188340807174888,
+      "grad_norm": 0.001927120960317552,
+      "learning_rate": 3.2420807159746333e-06,
+      "loss": 0.0,
+      "step": 239
+    },
+    {
+      "epoch": 3.201793721973094,
+      "grad_norm": 0.0047145108692348,
+      "learning_rate": 3.2202477484130947e-06,
+      "loss": 0.0,
+      "step": 240
+    },
+    {
+      "epoch": 3.2152466367713006,
+      "grad_norm": 0.001525243278592825,
+      "learning_rate": 3.1983547102818104e-06,
+      "loss": 0.0,
+      "step": 241
+    },
+    {
+      "epoch": 3.2286995515695067,
+      "grad_norm": 0.0008274471038021147,
+      "learning_rate": 3.1764034275181436e-06,
+      "loss": 0.0,
+      "step": 242
+    },
+    {
+      "epoch": 3.242152466367713,
+      "grad_norm": 0.0009357924573123455,
+      "learning_rate": 3.1543957309172136e-06,
+      "loss": 0.0,
+      "step": 243
+    },
+    {
+      "epoch": 3.2556053811659194,
+      "grad_norm": 0.0016225146828219295,
+      "learning_rate": 3.132333455979202e-06,
+      "loss": 0.0,
+      "step": 244
+    },
+    {
+      "epoch": 3.2690582959641254,
+      "grad_norm": 0.013506707735359669,
+      "learning_rate": 3.1102184427562696e-06,
+      "loss": 0.0001,
+      "step": 245
+    },
+    {
+      "epoch": 3.282511210762332,
+      "grad_norm": 0.0010258476249873638,
+      "learning_rate": 3.0880525356990898e-06,
+      "loss": 0.0,
+      "step": 246
+    },
+    {
+      "epoch": 3.295964125560538,
+      "grad_norm": 0.006462691817432642,
+      "learning_rate": 3.0658375835030148e-06,
+      "loss": 0.0,
+      "step": 247
+    },
+    {
+      "epoch": 3.3094170403587446,
+      "grad_norm": 0.0011441456153988838,
+      "learning_rate": 3.043575438953893e-06,
+      "loss": 0.0,
+      "step": 248
+    },
+    {
+      "epoch": 3.3228699551569507,
+      "grad_norm": 0.000751970277633518,
+      "learning_rate": 3.02126795877354e-06,
+      "loss": 0.0,
+      "step": 249
+    },
+    {
+      "epoch": 3.336322869955157,
+      "grad_norm": 0.006623424123972654,
+      "learning_rate": 2.9989170034648823e-06,
+      "loss": 0.0,
+      "step": 250
+    },
+    {
+      "epoch": 3.3497757847533634,
+      "grad_norm": 0.0021056546829640865,
+      "learning_rate": 2.9765244371567873e-06,
+      "loss": 0.0,
+      "step": 251
+    },
+    {
+      "epoch": 3.3632286995515694,
+      "grad_norm": 0.003343602642416954,
+      "learning_rate": 2.9540921274485913e-06,
+      "loss": 0.0,
+      "step": 252
+    },
+    {
+      "epoch": 3.376681614349776,
+      "grad_norm": 0.0032404386438429356,
+      "learning_rate": 2.9316219452543342e-06,
+      "loss": 0.0,
+      "step": 253
+    },
+    {
+      "epoch": 3.390134529147982,
+      "grad_norm": 0.02200383134186268,
+      "learning_rate": 2.9091157646467205e-06,
+      "loss": 0.0001,
+      "step": 254
+    },
+    {
+      "epoch": 3.403587443946188,
+      "grad_norm": 0.0015861240681260824,
+      "learning_rate": 2.886575462700821e-06,
+      "loss": 0.0,
+      "step": 255
+    },
+    {
+      "epoch": 3.4170403587443947,
+      "grad_norm": 0.0023104625288397074,
+      "learning_rate": 2.864002919337513e-06,
+      "loss": 0.0,
+      "step": 256
+    },
+    {
+      "epoch": 3.430493273542601,
+      "grad_norm": 0.0009963945485651493,
+      "learning_rate": 2.8414000171666952e-06,
+      "loss": 0.0,
+      "step": 257
+    },
+    {
+      "epoch": 3.4439461883408073,
+      "grad_norm": 0.006071928422898054,
+      "learning_rate": 2.81876864133027e-06,
+      "loss": 0.0,
+      "step": 258
+    },
+    {
+      "epoch": 3.4573991031390134,
+      "grad_norm": 0.003969075623899698,
+      "learning_rate": 2.7961106793449217e-06,
+      "loss": 0.0,
+      "step": 259
+    },
+    {
+      "epoch": 3.4708520179372195,
+      "grad_norm": 0.0010353871621191502,
+      "learning_rate": 2.773428020944687e-06,
+      "loss": 0.0,
+      "step": 260
+    },
+    {
+      "epoch": 3.484304932735426,
+      "grad_norm": 0.003665305208414793,
+      "learning_rate": 2.7507225579233487e-06,
+      "loss": 0.0,
+      "step": 261
+    },
+    {
+      "epoch": 3.497757847533632,
+      "grad_norm": 0.003311133710667491,
+      "learning_rate": 2.727996183976659e-06,
+      "loss": 0.0,
+      "step": 262
+    },
+    {
+      "epoch": 3.5112107623318387,
+      "grad_norm": 0.004625072702765465,
+      "learning_rate": 2.705250794544393e-06,
+      "loss": 0.0,
+      "step": 263
+    },
+    {
+      "epoch": 3.524663677130045,
+      "grad_norm": 0.0023010042496025562,
+      "learning_rate": 2.682488286652269e-06,
+      "loss": 0.0,
+      "step": 264
+    },
+    {
+      "epoch": 3.538116591928251,
+      "grad_norm": 0.0008537416579201818,
+      "learning_rate": 2.6597105587537307e-06,
+      "loss": 0.0,
+      "step": 265
+    },
+    {
+      "epoch": 3.5515695067264574,
+      "grad_norm": 0.0014414336765184999,
+      "learning_rate": 2.6369195105716087e-06,
+      "loss": 0.0,
+      "step": 266
+    },
+    {
+      "epoch": 3.5650224215246635,
+      "grad_norm": 0.003060834715142846,
+      "learning_rate": 2.614117042939685e-06,
+      "loss": 0.0,
+      "step": 267
+    },
+    {
+      "epoch": 3.57847533632287,
+      "grad_norm": 0.002594274003058672,
+      "learning_rate": 2.591305057644148e-06,
+      "loss": 0.0,
+      "step": 268
+    },
+    {
+      "epoch": 3.591928251121076,
+      "grad_norm": 0.0009433673694729805,
+      "learning_rate": 2.5684854572649876e-06,
+      "loss": 0.0,
+      "step": 269
+    },
+    {
+      "epoch": 3.6053811659192823,
+      "grad_norm": 0.000922717503271997,
+      "learning_rate": 2.5456601450173123e-06,
+      "loss": 0.0,
+      "step": 270
+    },
+    {
+      "epoch": 3.618834080717489,
+      "grad_norm": 0.0013827934162691236,
+      "learning_rate": 2.522831024592615e-06,
+      "loss": 0.0,
+      "step": 271
+    },
+    {
+      "epoch": 3.6322869955156953,
+      "grad_norm": 0.0007690931670367718,
+      "learning_rate": 2.5e-06,
+      "loss": 0.0,
+      "step": 272
+    },
+    {
+      "epoch": 3.6457399103139014,
+      "grad_norm": 0.002565343165770173,
+      "learning_rate": 2.4771689754073856e-06,
+      "loss": 0.0,
+      "step": 273
+    },
+    {
+      "epoch": 3.6591928251121075,
+      "grad_norm": 0.0013943302910774946,
+      "learning_rate": 2.454339854982688e-06,
+      "loss": 0.0,
+      "step": 274
+    },
+    {
+      "epoch": 3.672645739910314,
+      "grad_norm": 0.0006668591522611678,
+      "learning_rate": 2.4315145427350132e-06,
+      "loss": 0.0,
+      "step": 275
+    },
+    {
+      "epoch": 3.68609865470852,
+      "grad_norm": 0.0025589216966181993,
+      "learning_rate": 2.408694942355853e-06,
+      "loss": 0.0,
+      "step": 276
+    },
+    {
+      "epoch": 3.6995515695067267,
+      "grad_norm": 0.001420054235495627,
+      "learning_rate": 2.3858829570603157e-06,
+      "loss": 0.0,
+      "step": 277
+    },
+    {
+      "epoch": 3.713004484304933,
+      "grad_norm": 0.0014220779994502664,
+      "learning_rate": 2.363080489428391e-06,
+      "loss": 0.0,
+      "step": 278
+    },
+    {
+      "epoch": 3.726457399103139,
+      "grad_norm": 0.0009400748531334102,
+      "learning_rate": 2.3402894412462697e-06,
+      "loss": 0.0,
+      "step": 279
+    },
+    {
+      "epoch": 3.7399103139013454,
+      "grad_norm": 0.0006919830339029431,
+      "learning_rate": 2.317511713347731e-06,
+      "loss": 0.0,
+      "step": 280
+    },
+    {
+      "epoch": 3.7533632286995515,
+      "grad_norm": 0.0011200032895430923,
+      "learning_rate": 2.2947492054556075e-06,
+      "loss": 0.0,
+      "step": 281
+    },
+    {
+      "epoch": 3.766816143497758,
+      "grad_norm": 0.0008829529979266226,
+      "learning_rate": 2.272003816023341e-06,
+      "loss": 0.0,
+      "step": 282
+    },
+    {
+      "epoch": 3.780269058295964,
+      "grad_norm": 0.0009032113594003022,
+      "learning_rate": 2.2492774420766517e-06,
+      "loss": 0.0,
+      "step": 283
+    },
+    {
+      "epoch": 3.7937219730941703,
+      "grad_norm": 0.0012500348966568708,
+      "learning_rate": 2.2265719790553147e-06,
+      "loss": 0.0,
+      "step": 284
+    },
+    {
+      "epoch": 3.807174887892377,
+      "grad_norm": 0.0008029688615351915,
+      "learning_rate": 2.20388932065508e-06,
+      "loss": 0.0,
+      "step": 285
+    },
+    {
+      "epoch": 3.820627802690583,
+      "grad_norm": 0.0015973382396623492,
+      "learning_rate": 2.1812313586697307e-06,
+      "loss": 0.0,
+      "step": 286
+    },
+    {
+      "epoch": 3.8340807174887894,
+      "grad_norm": 0.0021750489249825478,
+      "learning_rate": 2.1585999828333065e-06,
+      "loss": 0.0,
+      "step": 287
+    },
+    {
+      "epoch": 3.8475336322869955,
+      "grad_norm": 0.004780885297805071,
+      "learning_rate": 2.1359970806624886e-06,
+      "loss": 0.0,
+      "step": 288
+    },
+    {
+      "epoch": 3.8609865470852016,
+      "grad_norm": 0.0007556549389846623,
+      "learning_rate": 2.11342453729918e-06,
+      "loss": 0.0,
+      "step": 289
+    },
+    {
+      "epoch": 3.874439461883408,
+      "grad_norm": 0.0006669393624179065,
+      "learning_rate": 2.0908842353532803e-06,
+      "loss": 0.0,
+      "step": 290
+    },
+    {
+      "epoch": 3.8878923766816142,
+      "grad_norm": 0.001046078628860414,
+      "learning_rate": 2.0683780547456666e-06,
+      "loss": 0.0,
+      "step": 291
+    },
+    {
+      "epoch": 3.901345291479821,
+      "grad_norm": 0.004534538835287094,
+      "learning_rate": 2.045907872551409e-06,
+      "loss": 0.0,
+      "step": 292
+    },
+    {
+      "epoch": 3.914798206278027,
+      "grad_norm": 0.0008524219738319516,
+      "learning_rate": 2.0234755628432135e-06,
+      "loss": 0.0,
+      "step": 293
+    },
+    {
+      "epoch": 3.928251121076233,
+      "grad_norm": 0.0006630662246607244,
+      "learning_rate": 2.0010829965351185e-06,
+      "loss": 0.0,
+      "step": 294
+    },
+    {
+      "epoch": 3.9417040358744395,
+      "grad_norm": 0.008820832706987858,
+      "learning_rate": 1.978732041226461e-06,
+      "loss": 0.0,
+      "step": 295
+    },
+    {
+      "epoch": 3.9551569506726456,
+      "grad_norm": 0.0007411285769194365,
+      "learning_rate": 1.956424561046108e-06,
+      "loss": 0.0,
+      "step": 296
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 444,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 6,
+  "save_steps": 74,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.586020432455926e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-296/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db2163a23b611a8c5fa6dfb9e81074ea1793ff3e8d206475ef1251ddbdea9baa
+size 7992

checkpoint-296/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-370/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-370/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": [
+    "up_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj",
+    "k_proj",
+    "v_proj",
+    "q_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-370/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fcb573e4403e9fb0209774f20ee5addb5dbadc561d765e8565956bd509fb02e
+size 1291899552

checkpoint-370/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-370/global_step367/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2db0297f9b120ccf0a7e044aed442728964f1c879339a6afcc1b7ceef4137744
+size 7751120668

checkpoint-370/global_step367/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd628e1955fdebcb0a983c0bba4fadbd074d5679557401ff96fe51b8d288b20a
+size 1396311192

checkpoint-370/latest

@@ -0,0 +1 @@
+global_step367

checkpoint-370/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acb2ceeccd437a922c89d6c21f143b48f366fac9cb8539e0239c6d444ab4dbfe
+size 14244

checkpoint-370/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba23d83f6ddb50bfb3a2cd224d2f9d1db11450518348141bc44f8108b3169607
+size 1064

checkpoint-370/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-370/tokenizer.json

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

checkpoint-370/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-370/trainer_state.json

@@ -0,0 +1,2623 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.941704035874439,
+  "eval_steps": 500,
+  "global_step": 370,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.013452914798206279,
+      "grad_norm": 95.4926986694336,
+      "learning_rate": 5.0000000000000004e-08,
+      "loss": 12.2856,
+      "step": 1
+    },
+    {
+      "epoch": 0.026905829596412557,
+      "grad_norm": 93.69285583496094,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 12.2383,
+      "step": 2
+    },
+    {
+      "epoch": 0.04035874439461883,
+      "grad_norm": 95.09840393066406,
+      "learning_rate": 1.5000000000000002e-07,
+      "loss": 12.1293,
+      "step": 3
+    },
+    {
+      "epoch": 0.053811659192825115,
+      "grad_norm": 95.04216766357422,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 12.1453,
+      "step": 4
+    },
+    {
+      "epoch": 0.06726457399103139,
+      "grad_norm": 93.44210052490234,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 12.165,
+      "step": 5
+    },
+    {
+      "epoch": 0.08071748878923767,
+      "grad_norm": 93.28514862060547,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 12.063,
+      "step": 6
+    },
+    {
+      "epoch": 0.09417040358744394,
+      "grad_norm": 95.3654556274414,
+      "learning_rate": 3.5000000000000004e-07,
+      "loss": 11.9676,
+      "step": 7
+    },
+    {
+      "epoch": 0.10762331838565023,
+      "grad_norm": 96.05154418945312,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 12.0911,
+      "step": 8
+    },
+    {
+      "epoch": 0.1210762331838565,
+      "grad_norm": 97.69881439208984,
+      "learning_rate": 4.5000000000000003e-07,
+      "loss": 12.0717,
+      "step": 9
+    },
+    {
+      "epoch": 0.13452914798206278,
+      "grad_norm": 95.55254364013672,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 11.9752,
+      "step": 10
+    },
+    {
+      "epoch": 0.14798206278026907,
+      "grad_norm": 95.9182357788086,
+      "learning_rate": 5.5e-07,
+      "loss": 11.9413,
+      "step": 11
+    },
+    {
+      "epoch": 0.16143497757847533,
+      "grad_norm": 95.40771484375,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 11.7523,
+      "step": 12
+    },
+    {
+      "epoch": 0.17488789237668162,
+      "grad_norm": 94.40055847167969,
+      "learning_rate": 6.5e-07,
+      "loss": 11.6384,
+      "step": 13
+    },
+    {
+      "epoch": 0.18834080717488788,
+      "grad_norm": 93.58352661132812,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 11.4293,
+      "step": 14
+    },
+    {
+      "epoch": 0.20179372197309417,
+      "grad_norm": 94.48737335205078,
+      "learning_rate": 7.5e-07,
+      "loss": 11.1445,
+      "step": 15
+    },
+    {
+      "epoch": 0.21524663677130046,
+      "grad_norm": 92.44265747070312,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 10.7705,
+      "step": 16
+    },
+    {
+      "epoch": 0.22869955156950672,
+      "grad_norm": 90.97422790527344,
+      "learning_rate": 8.500000000000001e-07,
+      "loss": 10.3754,
+      "step": 17
+    },
+    {
+      "epoch": 0.242152466367713,
+      "grad_norm": 88.54856872558594,
+      "learning_rate": 9.000000000000001e-07,
+      "loss": 10.0019,
+      "step": 18
+    },
+    {
+      "epoch": 0.2556053811659193,
+      "grad_norm": 88.39138793945312,
+      "learning_rate": 9.500000000000001e-07,
+      "loss": 9.5093,
+      "step": 19
+    },
+    {
+      "epoch": 0.26905829596412556,
+      "grad_norm": 86.55109405517578,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 9.2342,
+      "step": 20
+    },
+    {
+      "epoch": 0.2825112107623318,
+      "grad_norm": 80.62335205078125,
+      "learning_rate": 1.0500000000000001e-06,
+      "loss": 8.597,
+      "step": 21
+    },
+    {
+      "epoch": 0.29596412556053814,
+      "grad_norm": 73.67768859863281,
+      "learning_rate": 1.1e-06,
+      "loss": 8.1122,
+      "step": 22
+    },
+    {
+      "epoch": 0.3094170403587444,
+      "grad_norm": 64.57353210449219,
+      "learning_rate": 1.1500000000000002e-06,
+      "loss": 7.6455,
+      "step": 23
+    },
+    {
+      "epoch": 0.32286995515695066,
+      "grad_norm": 55.2818603515625,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 7.2493,
+      "step": 24
+    },
+    {
+      "epoch": 0.336322869955157,
+      "grad_norm": 48.274452209472656,
+      "learning_rate": 1.25e-06,
+      "loss": 7.0377,
+      "step": 25
+    },
+    {
+      "epoch": 0.34977578475336324,
+      "grad_norm": 42.7370491027832,
+      "learning_rate": 1.3e-06,
+      "loss": 6.5782,
+      "step": 26
+    },
+    {
+      "epoch": 0.3632286995515695,
+      "grad_norm": 39.297462463378906,
+      "learning_rate": 1.3500000000000002e-06,
+      "loss": 6.2558,
+      "step": 27
+    },
+    {
+      "epoch": 0.37668161434977576,
+      "grad_norm": 37.91667938232422,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 5.9809,
+      "step": 28
+    },
+    {
+      "epoch": 0.3901345291479821,
+      "grad_norm": 37.87322998046875,
+      "learning_rate": 1.45e-06,
+      "loss": 5.7268,
+      "step": 29
+    },
+    {
+      "epoch": 0.40358744394618834,
+      "grad_norm": 36.48906707763672,
+      "learning_rate": 1.5e-06,
+      "loss": 5.449,
+      "step": 30
+    },
+    {
+      "epoch": 0.4170403587443946,
+      "grad_norm": 36.38510513305664,
+      "learning_rate": 1.5500000000000002e-06,
+      "loss": 5.1884,
+      "step": 31
+    },
+    {
+      "epoch": 0.4304932735426009,
+      "grad_norm": 35.656829833984375,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 4.899,
+      "step": 32
+    },
+    {
+      "epoch": 0.4439461883408072,
+      "grad_norm": 34.09960174560547,
+      "learning_rate": 1.6500000000000003e-06,
+      "loss": 4.5842,
+      "step": 33
+    },
+    {
+      "epoch": 0.45739910313901344,
+      "grad_norm": 32.74240493774414,
+      "learning_rate": 1.7000000000000002e-06,
+      "loss": 4.3009,
+      "step": 34
+    },
+    {
+      "epoch": 0.47085201793721976,
+      "grad_norm": 31.867507934570312,
+      "learning_rate": 1.75e-06,
+      "loss": 3.9865,
+      "step": 35
+    },
+    {
+      "epoch": 0.484304932735426,
+      "grad_norm": 30.741374969482422,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 3.6916,
+      "step": 36
+    },
+    {
+      "epoch": 0.4977578475336323,
+      "grad_norm": 27.8775577545166,
+      "learning_rate": 1.85e-06,
+      "loss": 3.3719,
+      "step": 37
+    },
+    {
+      "epoch": 0.5112107623318386,
+      "grad_norm": 25.97083282470703,
+      "learning_rate": 1.9000000000000002e-06,
+      "loss": 3.0907,
+      "step": 38
+    },
+    {
+      "epoch": 0.5246636771300448,
+      "grad_norm": 23.62006950378418,
+      "learning_rate": 1.9500000000000004e-06,
+      "loss": 2.8336,
+      "step": 39
+    },
+    {
+      "epoch": 0.5381165919282511,
+      "grad_norm": 23.80520248413086,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 2.5717,
+      "step": 40
+    },
+    {
+      "epoch": 0.5515695067264574,
+      "grad_norm": 25.32924461364746,
+      "learning_rate": 2.05e-06,
+      "loss": 2.3658,
+      "step": 41
+    },
+    {
+      "epoch": 0.5650224215246636,
+      "grad_norm": 26.20570182800293,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 2.2443,
+      "step": 42
+    },
+    {
+      "epoch": 0.57847533632287,
+      "grad_norm": 24.581693649291992,
+      "learning_rate": 2.15e-06,
+      "loss": 1.926,
+      "step": 43
+    },
+    {
+      "epoch": 0.5919282511210763,
+      "grad_norm": 24.414310455322266,
+      "learning_rate": 2.2e-06,
+      "loss": 1.7034,
+      "step": 44
+    },
+    {
+      "epoch": 0.6053811659192825,
+      "grad_norm": 22.691083908081055,
+      "learning_rate": 2.25e-06,
+      "loss": 1.4857,
+      "step": 45
+    },
+    {
+      "epoch": 0.6188340807174888,
+      "grad_norm": 20.669803619384766,
+      "learning_rate": 2.3000000000000004e-06,
+      "loss": 1.2415,
+      "step": 46
+    },
+    {
+      "epoch": 0.6322869955156951,
+      "grad_norm": 20.149641036987305,
+      "learning_rate": 2.35e-06,
+      "loss": 0.997,
+      "step": 47
+    },
+    {
+      "epoch": 0.6457399103139013,
+      "grad_norm": 18.632596969604492,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 0.7552,
+      "step": 48
+    },
+    {
+      "epoch": 0.6591928251121076,
+      "grad_norm": 16.93793296813965,
+      "learning_rate": 2.4500000000000003e-06,
+      "loss": 0.5883,
+      "step": 49
+    },
+    {
+      "epoch": 0.672645739910314,
+      "grad_norm": 14.432519912719727,
+      "learning_rate": 2.5e-06,
+      "loss": 0.4382,
+      "step": 50
+    },
+    {
+      "epoch": 0.6860986547085202,
+      "grad_norm": 11.829660415649414,
+      "learning_rate": 2.55e-06,
+      "loss": 0.2983,
+      "step": 51
+    },
+    {
+      "epoch": 0.6995515695067265,
+      "grad_norm": 8.680500030517578,
+      "learning_rate": 2.6e-06,
+      "loss": 0.1988,
+      "step": 52
+    },
+    {
+      "epoch": 0.7130044843049327,
+      "grad_norm": 6.53156852722168,
+      "learning_rate": 2.6500000000000005e-06,
+      "loss": 0.1589,
+      "step": 53
+    },
+    {
+      "epoch": 0.726457399103139,
+      "grad_norm": 2.9756624698638916,
+      "learning_rate": 2.7000000000000004e-06,
+      "loss": 0.0686,
+      "step": 54
+    },
+    {
+      "epoch": 0.7399103139013453,
+      "grad_norm": 5.545580863952637,
+      "learning_rate": 2.7500000000000004e-06,
+      "loss": 0.0865,
+      "step": 55
+    },
+    {
+      "epoch": 0.7533632286995515,
+      "grad_norm": 4.045405387878418,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 0.0949,
+      "step": 56
+    },
+    {
+      "epoch": 0.7668161434977578,
+      "grad_norm": 1.6688120365142822,
+      "learning_rate": 2.85e-06,
+      "loss": 0.0396,
+      "step": 57
+    },
+    {
+      "epoch": 0.7802690582959642,
+      "grad_norm": 2.4520657062530518,
+      "learning_rate": 2.9e-06,
+      "loss": 0.0439,
+      "step": 58
+    },
+    {
+      "epoch": 0.7937219730941704,
+      "grad_norm": 2.608729600906372,
+      "learning_rate": 2.95e-06,
+      "loss": 0.057,
+      "step": 59
+    },
+    {
+      "epoch": 0.8071748878923767,
+      "grad_norm": 2.365234851837158,
+      "learning_rate": 3e-06,
+      "loss": 0.0547,
+      "step": 60
+    },
+    {
+      "epoch": 0.820627802690583,
+      "grad_norm": 0.787550687789917,
+      "learning_rate": 3.05e-06,
+      "loss": 0.0209,
+      "step": 61
+    },
+    {
+      "epoch": 0.8340807174887892,
+      "grad_norm": 0.7686442732810974,
+      "learning_rate": 3.1000000000000004e-06,
+      "loss": 0.0221,
+      "step": 62
+    },
+    {
+      "epoch": 0.8475336322869955,
+      "grad_norm": 1.2510555982589722,
+      "learning_rate": 3.1500000000000003e-06,
+      "loss": 0.0165,
+      "step": 63
+    },
+    {
+      "epoch": 0.8609865470852018,
+      "grad_norm": 0.8923770189285278,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 0.0187,
+      "step": 64
+    },
+    {
+      "epoch": 0.874439461883408,
+      "grad_norm": 0.8052615523338318,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.0266,
+      "step": 65
+    },
+    {
+      "epoch": 0.8878923766816144,
+      "grad_norm": 0.6710303425788879,
+      "learning_rate": 3.3000000000000006e-06,
+      "loss": 0.0154,
+      "step": 66
+    },
+    {
+      "epoch": 0.9013452914798207,
+      "grad_norm": 0.5213025212287903,
+      "learning_rate": 3.3500000000000005e-06,
+      "loss": 0.0085,
+      "step": 67
+    },
+    {
+      "epoch": 0.9147982062780269,
+      "grad_norm": 0.5758580565452576,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 0.0133,
+      "step": 68
+    },
+    {
+      "epoch": 0.9282511210762332,
+      "grad_norm": 0.6828752160072327,
+      "learning_rate": 3.45e-06,
+      "loss": 0.0186,
+      "step": 69
+    },
+    {
+      "epoch": 0.9417040358744395,
+      "grad_norm": 0.6814988255500793,
+      "learning_rate": 3.5e-06,
+      "loss": 0.0215,
+      "step": 70
+    },
+    {
+      "epoch": 0.9551569506726457,
+      "grad_norm": 0.718296229839325,
+      "learning_rate": 3.5500000000000003e-06,
+      "loss": 0.0204,
+      "step": 71
+    },
+    {
+      "epoch": 0.968609865470852,
+      "grad_norm": 0.7816944122314453,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 0.0184,
+      "step": 72
+    },
+    {
+      "epoch": 0.9820627802690582,
+      "grad_norm": 0.6058817505836487,
+      "learning_rate": 3.65e-06,
+      "loss": 0.0179,
+      "step": 73
+    },
+    {
+      "epoch": 0.9955156950672646,
+      "grad_norm": 1.0496101379394531,
+      "learning_rate": 3.7e-06,
+      "loss": 0.032,
+      "step": 74
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.0496101379394531,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.0137,
+      "step": 75
+    },
+    {
+      "epoch": 1.0134529147982063,
+      "grad_norm": 1.598720669746399,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.0211,
+      "step": 76
+    },
+    {
+      "epoch": 1.0269058295964126,
+      "grad_norm": 0.7792187333106995,
+      "learning_rate": 3.85e-06,
+      "loss": 0.0213,
+      "step": 77
+    },
+    {
+      "epoch": 1.0403587443946187,
+      "grad_norm": 0.7717252373695374,
+      "learning_rate": 3.900000000000001e-06,
+      "loss": 0.0151,
+      "step": 78
+    },
+    {
+      "epoch": 1.053811659192825,
+      "grad_norm": 0.39334648847579956,
+      "learning_rate": 3.95e-06,
+      "loss": 0.0186,
+      "step": 79
+    },
+    {
+      "epoch": 1.0672645739910314,
+      "grad_norm": 0.9775457382202148,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.0043,
+      "step": 80
+    },
+    {
+      "epoch": 1.0807174887892377,
+      "grad_norm": 1.2425150871276855,
+      "learning_rate": 4.05e-06,
+      "loss": 0.0179,
+      "step": 81
+    },
+    {
+      "epoch": 1.094170403587444,
+      "grad_norm": 0.3884654939174652,
+      "learning_rate": 4.1e-06,
+      "loss": 0.008,
+      "step": 82
+    },
+    {
+      "epoch": 1.1076233183856503,
+      "grad_norm": 0.4746466875076294,
+      "learning_rate": 4.15e-06,
+      "loss": 0.0144,
+      "step": 83
+    },
+    {
+      "epoch": 1.1210762331838564,
+      "grad_norm": 0.6812214255332947,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 0.0072,
+      "step": 84
+    },
+    {
+      "epoch": 1.1345291479820627,
+      "grad_norm": 0.5414469838142395,
+      "learning_rate": 4.25e-06,
+      "loss": 0.016,
+      "step": 85
+    },
+    {
+      "epoch": 1.147982062780269,
+      "grad_norm": 0.8709024786949158,
+      "learning_rate": 4.3e-06,
+      "loss": 0.0154,
+      "step": 86
+    },
+    {
+      "epoch": 1.1614349775784754,
+      "grad_norm": 0.3181096613407135,
+      "learning_rate": 4.350000000000001e-06,
+      "loss": 0.0039,
+      "step": 87
+    },
+    {
+      "epoch": 1.1748878923766817,
+      "grad_norm": 0.2581265866756439,
+      "learning_rate": 4.4e-06,
+      "loss": 0.0014,
+      "step": 88
+    },
+    {
+      "epoch": 1.188340807174888,
+      "grad_norm": 0.10870776325464249,
+      "learning_rate": 4.450000000000001e-06,
+      "loss": 0.0088,
+      "step": 89
+    },
+    {
+      "epoch": 1.201793721973094,
+      "grad_norm": 0.45314452052116394,
+      "learning_rate": 4.5e-06,
+      "loss": 0.0019,
+      "step": 90
+    },
+    {
+      "epoch": 1.2152466367713004,
+      "grad_norm": 0.7328381538391113,
+      "learning_rate": 4.5500000000000005e-06,
+      "loss": 0.0082,
+      "step": 91
+    },
+    {
+      "epoch": 1.2286995515695067,
+      "grad_norm": 0.6641069650650024,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 0.0075,
+      "step": 92
+    },
+    {
+      "epoch": 1.242152466367713,
+      "grad_norm": 0.27984222769737244,
+      "learning_rate": 4.65e-06,
+      "loss": 0.0009,
+      "step": 93
+    },
+    {
+      "epoch": 1.2556053811659194,
+      "grad_norm": 0.8341127634048462,
+      "learning_rate": 4.7e-06,
+      "loss": 0.0204,
+      "step": 94
+    },
+    {
+      "epoch": 1.2690582959641254,
+      "grad_norm": 1.0140557289123535,
+      "learning_rate": 4.75e-06,
+      "loss": 0.0228,
+      "step": 95
+    },
+    {
+      "epoch": 1.2825112107623318,
+      "grad_norm": 0.9439787268638611,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.027,
+      "step": 96
+    },
+    {
+      "epoch": 1.295964125560538,
+      "grad_norm": 1.3867762088775635,
+      "learning_rate": 4.85e-06,
+      "loss": 0.0223,
+      "step": 97
+    },
+    {
+      "epoch": 1.3094170403587444,
+      "grad_norm": 0.4747941493988037,
+      "learning_rate": 4.9000000000000005e-06,
+      "loss": 0.0048,
+      "step": 98
+    },
+    {
+      "epoch": 1.3228699551569507,
+      "grad_norm": 0.5673424601554871,
+      "learning_rate": 4.95e-06,
+      "loss": 0.0033,
+      "step": 99
+    },
+    {
+      "epoch": 1.336322869955157,
+      "grad_norm": 0.24510182440280914,
+      "learning_rate": 5e-06,
+      "loss": 0.0048,
+      "step": 100
+    },
+    {
+      "epoch": 1.3497757847533634,
+      "grad_norm": 0.6084151268005371,
+      "learning_rate": 4.99989574668946e-06,
+      "loss": 0.032,
+      "step": 101
+    },
+    {
+      "epoch": 1.3632286995515694,
+      "grad_norm": 0.606272280216217,
+      "learning_rate": 4.999582995452842e-06,
+      "loss": 0.0058,
+      "step": 102
+    },
+    {
+      "epoch": 1.3766816143497758,
+      "grad_norm": 0.2270481288433075,
+      "learning_rate": 4.999061772374426e-06,
+      "loss": 0.0043,
+      "step": 103
+    },
+    {
+      "epoch": 1.390134529147982,
+      "grad_norm": 0.2832431495189667,
+      "learning_rate": 4.998332120925598e-06,
+      "loss": 0.0079,
+      "step": 104
+    },
+    {
+      "epoch": 1.4035874439461884,
+      "grad_norm": 0.44565466046333313,
+      "learning_rate": 4.9973941019612235e-06,
+      "loss": 0.017,
+      "step": 105
+    },
+    {
+      "epoch": 1.4170403587443947,
+      "grad_norm": 0.3518397808074951,
+      "learning_rate": 4.996247793714565e-06,
+      "loss": 0.0057,
+      "step": 106
+    },
+    {
+      "epoch": 1.4304932735426008,
+      "grad_norm": 0.41977858543395996,
+      "learning_rate": 4.994893291790768e-06,
+      "loss": 0.0056,
+      "step": 107
+    },
+    {
+      "epoch": 1.4439461883408071,
+      "grad_norm": 0.22567600011825562,
+      "learning_rate": 4.993330709158879e-06,
+      "loss": 0.0067,
+      "step": 108
+    },
+    {
+      "epoch": 1.4573991031390134,
+      "grad_norm": 0.2184637039899826,
+      "learning_rate": 4.9915601761424304e-06,
+      "loss": 0.0043,
+      "step": 109
+    },
+    {
+      "epoch": 1.4708520179372198,
+      "grad_norm": 0.24834735691547394,
+      "learning_rate": 4.989581840408562e-06,
+      "loss": 0.0034,
+      "step": 110
+    },
+    {
+      "epoch": 1.484304932735426,
+      "grad_norm": 0.13165591657161713,
+      "learning_rate": 4.987395866955716e-06,
+      "loss": 0.0041,
+      "step": 111
+    },
+    {
+      "epoch": 1.4977578475336322,
+      "grad_norm": 0.27152982354164124,
+      "learning_rate": 4.9850024380998655e-06,
+      "loss": 0.0034,
+      "step": 112
+    },
+    {
+      "epoch": 1.5112107623318387,
+      "grad_norm": 0.2793160676956177,
+      "learning_rate": 4.982401753459317e-06,
+      "loss": 0.0049,
+      "step": 113
+    },
+    {
+      "epoch": 1.5246636771300448,
+      "grad_norm": 0.06674113124608994,
+      "learning_rate": 4.979594029938058e-06,
+      "loss": 0.0034,
+      "step": 114
+    },
+    {
+      "epoch": 1.5381165919282511,
+      "grad_norm": 0.2601087689399719,
+      "learning_rate": 4.976579501707665e-06,
+      "loss": 0.0025,
+      "step": 115
+    },
+    {
+      "epoch": 1.5515695067264574,
+      "grad_norm": 0.17075951397418976,
+      "learning_rate": 4.973358420187776e-06,
+      "loss": 0.0005,
+      "step": 116
+    },
+    {
+      "epoch": 1.5650224215246635,
+      "grad_norm": 0.11211276799440384,
+      "learning_rate": 4.969931054025122e-06,
+      "loss": 0.0059,
+      "step": 117
+    },
+    {
+      "epoch": 1.57847533632287,
+      "grad_norm": 0.5140838027000427,
+      "learning_rate": 4.966297689071117e-06,
+      "loss": 0.018,
+      "step": 118
+    },
+    {
+      "epoch": 1.5919282511210762,
+      "grad_norm": 0.5356995463371277,
+      "learning_rate": 4.962458628358021e-06,
+      "loss": 0.0007,
+      "step": 119
+    },
+    {
+      "epoch": 1.6053811659192825,
+      "grad_norm": 0.3967442214488983,
+      "learning_rate": 4.958414192073665e-06,
+      "loss": 0.014,
+      "step": 120
+    },
+    {
+      "epoch": 1.6188340807174888,
+      "grad_norm": 0.4567921757698059,
+      "learning_rate": 4.954164717534748e-06,
+      "loss": 0.0027,
+      "step": 121
+    },
+    {
+      "epoch": 1.6322869955156951,
+      "grad_norm": 0.4405613839626312,
+      "learning_rate": 4.949710559158699e-06,
+      "loss": 0.0135,
+      "step": 122
+    },
+    {
+      "epoch": 1.6457399103139014,
+      "grad_norm": 0.7164422869682312,
+      "learning_rate": 4.945052088434123e-06,
+      "loss": 0.0207,
+      "step": 123
+    },
+    {
+      "epoch": 1.6591928251121075,
+      "grad_norm": 0.47173142433166504,
+      "learning_rate": 4.940189693889819e-06,
+      "loss": 0.0019,
+      "step": 124
+    },
+    {
+      "epoch": 1.672645739910314,
+      "grad_norm": 0.4606887996196747,
+      "learning_rate": 4.9351237810623655e-06,
+      "loss": 0.0125,
+      "step": 125
+    },
+    {
+      "epoch": 1.6860986547085202,
+      "grad_norm": 0.7209401726722717,
+      "learning_rate": 4.929854772462312e-06,
+      "loss": 0.012,
+      "step": 126
+    },
+    {
+      "epoch": 1.6995515695067265,
+      "grad_norm": 1.7617985010147095,
+      "learning_rate": 4.924383107538929e-06,
+      "loss": 0.0079,
+      "step": 127
+    },
+    {
+      "epoch": 1.7130044843049328,
+      "grad_norm": 1.0861084461212158,
+      "learning_rate": 4.918709242643563e-06,
+      "loss": 0.0091,
+      "step": 128
+    },
+    {
+      "epoch": 1.726457399103139,
+      "grad_norm": 0.11263061314821243,
+      "learning_rate": 4.9128336509915746e-06,
+      "loss": 0.0006,
+      "step": 129
+    },
+    {
+      "epoch": 1.7399103139013454,
+      "grad_norm": 0.3411642909049988,
+      "learning_rate": 4.906756822622865e-06,
+      "loss": 0.0037,
+      "step": 130
+    },
+    {
+      "epoch": 1.7533632286995515,
+      "grad_norm": 0.1550491899251938,
+      "learning_rate": 4.900479264361017e-06,
+      "loss": 0.0056,
+      "step": 131
+    },
+    {
+      "epoch": 1.7668161434977578,
+      "grad_norm": 0.8448930382728577,
+      "learning_rate": 4.894001499771015e-06,
+      "loss": 0.0135,
+      "step": 132
+    },
+    {
+      "epoch": 1.7802690582959642,
+      "grad_norm": 0.5105615258216858,
+      "learning_rate": 4.887324069115582e-06,
+      "loss": 0.009,
+      "step": 133
+    },
+    {
+      "epoch": 1.7937219730941703,
+      "grad_norm": 0.5042226910591125,
+      "learning_rate": 4.880447529310118e-06,
+      "loss": 0.0022,
+      "step": 134
+    },
+    {
+      "epoch": 1.8071748878923768,
+      "grad_norm": 0.15797697007656097,
+      "learning_rate": 4.873372453876255e-06,
+      "loss": 0.0073,
+      "step": 135
+    },
+    {
+      "epoch": 1.8206278026905829,
+      "grad_norm": 0.47805944085121155,
+      "learning_rate": 4.866099432894023e-06,
+      "loss": 0.0084,
+      "step": 136
+    },
+    {
+      "epoch": 1.8340807174887892,
+      "grad_norm": 0.3370952904224396,
+      "learning_rate": 4.858629072952635e-06,
+      "loss": 0.0037,
+      "step": 137
+    },
+    {
+      "epoch": 1.8475336322869955,
+      "grad_norm": 0.30983835458755493,
+      "learning_rate": 4.850961997099892e-06,
+      "loss": 0.0024,
+      "step": 138
+    },
+    {
+      "epoch": 1.8609865470852018,
+      "grad_norm": 0.2800588011741638,
+      "learning_rate": 4.843098844790228e-06,
+      "loss": 0.0032,
+      "step": 139
+    },
+    {
+      "epoch": 1.8744394618834082,
+      "grad_norm": 0.2037343531847,
+      "learning_rate": 4.835040271831371e-06,
+      "loss": 0.0016,
+      "step": 140
+    },
+    {
+      "epoch": 1.8878923766816142,
+      "grad_norm": 0.45981553196907043,
+      "learning_rate": 4.826786950329646e-06,
+      "loss": 0.0035,
+      "step": 141
+    },
+    {
+      "epoch": 1.9013452914798208,
+      "grad_norm": 0.17092454433441162,
+      "learning_rate": 4.818339568633926e-06,
+      "loss": 0.0068,
+      "step": 142
+    },
+    {
+      "epoch": 1.9147982062780269,
+      "grad_norm": 0.5339077711105347,
+      "learning_rate": 4.809698831278217e-06,
+      "loss": 0.0026,
+      "step": 143
+    },
+    {
+      "epoch": 1.9282511210762332,
+      "grad_norm": 0.5208529829978943,
+      "learning_rate": 4.800865458922899e-06,
+      "loss": 0.0047,
+      "step": 144
+    },
+    {
+      "epoch": 1.9417040358744395,
+      "grad_norm": 0.7488933801651001,
+      "learning_rate": 4.79184018829462e-06,
+      "loss": 0.0096,
+      "step": 145
+    },
+    {
+      "epoch": 1.9551569506726456,
+      "grad_norm": 0.5727106332778931,
+      "learning_rate": 4.782623772124854e-06,
+      "loss": 0.001,
+      "step": 146
+    },
+    {
+      "epoch": 1.9686098654708521,
+      "grad_norm": 0.3813195526599884,
+      "learning_rate": 4.77321697908712e-06,
+      "loss": 0.0085,
+      "step": 147
+    },
+    {
+      "epoch": 1.9820627802690582,
+      "grad_norm": 0.5406109094619751,
+      "learning_rate": 4.763620593732867e-06,
+      "loss": 0.0017,
+      "step": 148
+    },
+    {
+      "epoch": 1.9955156950672646,
+      "grad_norm": 0.391985148191452,
+      "learning_rate": 4.7538354164260515e-06,
+      "loss": 0.0019,
+      "step": 149
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.391985148191452,
+      "learning_rate": 4.743862263276376e-06,
+      "loss": 0.0003,
+      "step": 150
+    },
+    {
+      "epoch": 2.013452914798206,
+      "grad_norm": 0.10133524239063263,
+      "learning_rate": 4.733701966071226e-06,
+      "loss": 0.0012,
+      "step": 151
+    },
+    {
+      "epoch": 2.0269058295964126,
+      "grad_norm": 0.2032414674758911,
+      "learning_rate": 4.723355372206297e-06,
+      "loss": 0.0001,
+      "step": 152
+    },
+    {
+      "epoch": 2.0403587443946187,
+      "grad_norm": 0.00796876847743988,
+      "learning_rate": 4.712823344614921e-06,
+      "loss": 0.0025,
+      "step": 153
+    },
+    {
+      "epoch": 2.0538116591928253,
+      "grad_norm": 0.36200040578842163,
+      "learning_rate": 4.702106761696091e-06,
+      "loss": 0.0003,
+      "step": 154
+    },
+    {
+      "epoch": 2.0672645739910314,
+      "grad_norm": 0.19646115601062775,
+      "learning_rate": 4.691206517241205e-06,
+      "loss": 0.0009,
+      "step": 155
+    },
+    {
+      "epoch": 2.0807174887892375,
+      "grad_norm": 0.06766581535339355,
+      "learning_rate": 4.68012352035952e-06,
+      "loss": 0.0001,
+      "step": 156
+    },
+    {
+      "epoch": 2.094170403587444,
+      "grad_norm": 0.00797713827341795,
+      "learning_rate": 4.668858695402326e-06,
+      "loss": 0.0001,
+      "step": 157
+    },
+    {
+      "epoch": 2.10762331838565,
+      "grad_norm": 0.020394539460539818,
+      "learning_rate": 4.657412981885862e-06,
+      "loss": 0.0002,
+      "step": 158
+    },
+    {
+      "epoch": 2.1210762331838566,
+      "grad_norm": 0.017584379762411118,
+      "learning_rate": 4.645787334412945e-06,
+      "loss": 0.0002,
+      "step": 159
+    },
+    {
+      "epoch": 2.1345291479820627,
+      "grad_norm": 0.07967082411050797,
+      "learning_rate": 4.633982722593367e-06,
+      "loss": 0.0003,
+      "step": 160
+    },
+    {
+      "epoch": 2.1479820627802693,
+      "grad_norm": 0.01606675237417221,
+      "learning_rate": 4.622000130963015e-06,
+      "loss": 0.0003,
+      "step": 161
+    },
+    {
+      "epoch": 2.1614349775784754,
+      "grad_norm": 0.06052660569548607,
+      "learning_rate": 4.6098405589017685e-06,
+      "loss": 0.0007,
+      "step": 162
+    },
+    {
+      "epoch": 2.1748878923766815,
+      "grad_norm": 0.07991409301757812,
+      "learning_rate": 4.597505020550138e-06,
+      "loss": 0.0002,
+      "step": 163
+    },
+    {
+      "epoch": 2.188340807174888,
+      "grad_norm": 0.04991272836923599,
+      "learning_rate": 4.584994544724695e-06,
+      "loss": 0.0001,
+      "step": 164
+    },
+    {
+      "epoch": 2.201793721973094,
+      "grad_norm": 0.027111921459436417,
+      "learning_rate": 4.572310174832255e-06,
+      "loss": 0.0001,
+      "step": 165
+    },
+    {
+      "epoch": 2.2152466367713006,
+      "grad_norm": 0.006566982250660658,
+      "learning_rate": 4.5594529687828615e-06,
+      "loss": 0.0,
+      "step": 166
+    },
+    {
+      "epoch": 2.2286995515695067,
+      "grad_norm": 0.002437079790979624,
+      "learning_rate": 4.546423998901549e-06,
+      "loss": 0.0004,
+      "step": 167
+    },
+    {
+      "epoch": 2.242152466367713,
+      "grad_norm": 0.08434150367975235,
+      "learning_rate": 4.533224351838914e-06,
+      "loss": 0.0001,
+      "step": 168
+    },
+    {
+      "epoch": 2.2556053811659194,
+      "grad_norm": 0.013094129040837288,
+      "learning_rate": 4.519855128480478e-06,
+      "loss": 0.0001,
+      "step": 169
+    },
+    {
+      "epoch": 2.2690582959641254,
+      "grad_norm": 0.002770340768620372,
+      "learning_rate": 4.5063174438548775e-06,
+      "loss": 0.0001,
+      "step": 170
+    },
+    {
+      "epoch": 2.282511210762332,
+      "grad_norm": 0.021480072289705276,
+      "learning_rate": 4.492612427040864e-06,
+      "loss": 0.0001,
+      "step": 171
+    },
+    {
+      "epoch": 2.295964125560538,
+      "grad_norm": 0.0028862387407571077,
+      "learning_rate": 4.478741221073136e-06,
+      "loss": 0.0001,
+      "step": 172
+    },
+    {
+      "epoch": 2.3094170403587446,
+      "grad_norm": 0.01969303749501705,
+      "learning_rate": 4.464704982847008e-06,
+      "loss": 0.0,
+      "step": 173
+    },
+    {
+      "epoch": 2.3228699551569507,
+      "grad_norm": 0.00244798487983644,
+      "learning_rate": 4.450504883021923e-06,
+      "loss": 0.0,
+      "step": 174
+    },
+    {
+      "epoch": 2.336322869955157,
+      "grad_norm": 0.0036638586316257715,
+      "learning_rate": 4.436142105923814e-06,
+      "loss": 0.0009,
+      "step": 175
+    },
+    {
+      "epoch": 2.3497757847533634,
+      "grad_norm": 0.4378701448440552,
+      "learning_rate": 4.4216178494463305e-06,
+      "loss": 0.0003,
+      "step": 176
+    },
+    {
+      "epoch": 2.3632286995515694,
+      "grad_norm": 0.004919757135212421,
+      "learning_rate": 4.406933324950929e-06,
+      "loss": 0.0,
+      "step": 177
+    },
+    {
+      "epoch": 2.376681614349776,
+      "grad_norm": 0.0061010573990643024,
+      "learning_rate": 4.392089757165841e-06,
+      "loss": 0.0001,
+      "step": 178
+    },
+    {
+      "epoch": 2.390134529147982,
+      "grad_norm": 0.00750540429726243,
+      "learning_rate": 4.377088384083935e-06,
+      "loss": 0.0004,
+      "step": 179
+    },
+    {
+      "epoch": 2.403587443946188,
+      "grad_norm": 0.18055735528469086,
+      "learning_rate": 4.361930456859455e-06,
+      "loss": 0.0001,
+      "step": 180
+    },
+    {
+      "epoch": 2.4170403587443947,
+      "grad_norm": 0.00540179992094636,
+      "learning_rate": 4.346617239703676e-06,
+      "loss": 0.0017,
+      "step": 181
+    },
+    {
+      "epoch": 2.430493273542601,
+      "grad_norm": 0.6051300764083862,
+      "learning_rate": 4.3311500097794655e-06,
+      "loss": 0.0002,
+      "step": 182
+    },
+    {
+      "epoch": 2.4439461883408073,
+      "grad_norm": 0.04465539753437042,
+      "learning_rate": 4.315530057094763e-06,
+      "loss": 0.0001,
+      "step": 183
+    },
+    {
+      "epoch": 2.4573991031390134,
+      "grad_norm": 0.009125534445047379,
+      "learning_rate": 4.2997586843949905e-06,
+      "loss": 0.0001,
+      "step": 184
+    },
+    {
+      "epoch": 2.4708520179372195,
+      "grad_norm": 0.009754106402397156,
+      "learning_rate": 4.2838372070544e-06,
+      "loss": 0.0001,
+      "step": 185
+    },
+    {
+      "epoch": 2.484304932735426,
+      "grad_norm": 0.03795509785413742,
+      "learning_rate": 4.267766952966369e-06,
+      "loss": 0.0006,
+      "step": 186
+    },
+    {
+      "epoch": 2.497757847533632,
+      "grad_norm": 0.19477951526641846,
+      "learning_rate": 4.25154926243265e-06,
+      "loss": 0.0013,
+      "step": 187
+    },
+    {
+      "epoch": 2.5112107623318387,
+      "grad_norm": 0.27783000469207764,
+      "learning_rate": 4.2351854880515856e-06,
+      "loss": 0.0001,
+      "step": 188
+    },
+    {
+      "epoch": 2.524663677130045,
+      "grad_norm": 0.030597640201449394,
+      "learning_rate": 4.218676994605295e-06,
+      "loss": 0.0,
+      "step": 189
+    },
+    {
+      "epoch": 2.538116591928251,
+      "grad_norm": 0.019212787970900536,
+      "learning_rate": 4.202025158945855e-06,
+      "loss": 0.0001,
+      "step": 190
+    },
+    {
+      "epoch": 2.5515695067264574,
+      "grad_norm": 0.017139675095677376,
+      "learning_rate": 4.185231369880461e-06,
+      "loss": 0.0001,
+      "step": 191
+    },
+    {
+      "epoch": 2.5650224215246635,
+      "grad_norm": 0.0013709627091884613,
+      "learning_rate": 4.168297028055599e-06,
+      "loss": 0.0002,
+      "step": 192
+    },
+    {
+      "epoch": 2.57847533632287,
+      "grad_norm": 0.07277967780828476,
+      "learning_rate": 4.151223545840225e-06,
+      "loss": 0.0,
+      "step": 193
+    },
+    {
+      "epoch": 2.591928251121076,
+      "grad_norm": 0.0014003290561959147,
+      "learning_rate": 4.134012347207974e-06,
+      "loss": 0.0001,
+      "step": 194
+    },
+    {
+      "epoch": 2.6053811659192823,
+      "grad_norm": 0.04370618611574173,
+      "learning_rate": 4.116664867618395e-06,
+      "loss": 0.0004,
+      "step": 195
+    },
+    {
+      "epoch": 2.618834080717489,
+      "grad_norm": 0.24697266519069672,
+      "learning_rate": 4.099182553897228e-06,
+      "loss": 0.0,
+      "step": 196
+    },
+    {
+      "epoch": 2.6322869955156953,
+      "grad_norm": 0.0013013904681429267,
+      "learning_rate": 4.081566864115741e-06,
+      "loss": 0.0,
+      "step": 197
+    },
+    {
+      "epoch": 2.6457399103139014,
+      "grad_norm": 0.001239327946677804,
+      "learning_rate": 4.063819267469114e-06,
+      "loss": 0.0,
+      "step": 198
+    },
+    {
+      "epoch": 2.6591928251121075,
+      "grad_norm": 0.007082940544933081,
+      "learning_rate": 4.04594124415391e-06,
+      "loss": 0.0,
+      "step": 199
+    },
+    {
+      "epoch": 2.672645739910314,
+      "grad_norm": 0.00556205864995718,
+      "learning_rate": 4.027934285244624e-06,
+      "loss": 0.0001,
+      "step": 200
+    },
+    {
+      "epoch": 2.68609865470852,
+      "grad_norm": 0.025378312915563583,
+      "learning_rate": 4.009799892569317e-06,
+      "loss": 0.0001,
+      "step": 201
+    },
+    {
+      "epoch": 2.6995515695067267,
+      "grad_norm": 0.006344004534184933,
+      "learning_rate": 3.991539578584368e-06,
+      "loss": 0.0,
+      "step": 202
+    },
+    {
+      "epoch": 2.713004484304933,
+      "grad_norm": 0.0016575426561757922,
+      "learning_rate": 3.973154866248323e-06,
+      "loss": 0.0,
+      "step": 203
+    },
+    {
+      "epoch": 2.726457399103139,
+      "grad_norm": 0.0020909749437123537,
+      "learning_rate": 3.9546472888948825e-06,
+      "loss": 0.0,
+      "step": 204
+    },
+    {
+      "epoch": 2.7399103139013454,
+      "grad_norm": 0.004855802282691002,
+      "learning_rate": 3.936018390105013e-06,
+      "loss": 0.0,
+      "step": 205
+    },
+    {
+      "epoch": 2.7533632286995515,
+      "grad_norm": 0.0082467095926404,
+      "learning_rate": 3.917269723578212e-06,
+      "loss": 0.0,
+      "step": 206
+    },
+    {
+      "epoch": 2.766816143497758,
+      "grad_norm": 0.0013344286708161235,
+      "learning_rate": 3.898402853002921e-06,
+      "loss": 0.0,
+      "step": 207
+    },
+    {
+      "epoch": 2.780269058295964,
+      "grad_norm": 0.0034061160404235125,
+      "learning_rate": 3.879419351926115e-06,
+      "loss": 0.0,
+      "step": 208
+    },
+    {
+      "epoch": 2.7937219730941703,
+      "grad_norm": 0.011912211775779724,
+      "learning_rate": 3.86032080362206e-06,
+      "loss": 0.0,
+      "step": 209
+    },
+    {
+      "epoch": 2.807174887892377,
+      "grad_norm": 0.00134057376999408,
+      "learning_rate": 3.841108800960264e-06,
+      "loss": 0.0,
+      "step": 210
+    },
+    {
+      "epoch": 2.820627802690583,
+      "grad_norm": 0.0016560767544433475,
+      "learning_rate": 3.8217849462726334e-06,
+      "loss": 0.0,
+      "step": 211
+    },
+    {
+      "epoch": 2.8340807174887894,
+      "grad_norm": 0.006285225041210651,
+      "learning_rate": 3.802350851219826e-06,
+      "loss": 0.0,
+      "step": 212
+    },
+    {
+      "epoch": 2.8475336322869955,
+      "grad_norm": 0.0018534021219238639,
+      "learning_rate": 3.7828081366568388e-06,
+      "loss": 0.0,
+      "step": 213
+    },
+    {
+      "epoch": 2.8609865470852016,
+      "grad_norm": 0.0041048116981983185,
+      "learning_rate": 3.763158432497824e-06,
+      "loss": 0.0,
+      "step": 214
+    },
+    {
+      "epoch": 2.874439461883408,
+      "grad_norm": 0.0023216214030981064,
+      "learning_rate": 3.743403377580149e-06,
+      "loss": 0.0,
+      "step": 215
+    },
+    {
+      "epoch": 2.8878923766816142,
+      "grad_norm": 0.000994804548099637,
+      "learning_rate": 3.723544619527714e-06,
+      "loss": 0.0,
+      "step": 216
+    },
+    {
+      "epoch": 2.901345291479821,
+      "grad_norm": 0.003168656025081873,
+      "learning_rate": 3.703583814613536e-06,
+      "loss": 0.0,
+      "step": 217
+    },
+    {
+      "epoch": 2.914798206278027,
+      "grad_norm": 0.0013909138506278396,
+      "learning_rate": 3.6835226276216087e-06,
+      "loss": 0.0,
+      "step": 218
+    },
+    {
+      "epoch": 2.928251121076233,
+      "grad_norm": 0.008506865240633488,
+      "learning_rate": 3.663362731708059e-06,
+      "loss": 0.0,
+      "step": 219
+    },
+    {
+      "epoch": 2.9417040358744395,
+      "grad_norm": 0.0017819993663579226,
+      "learning_rate": 3.6431058082615966e-06,
+      "loss": 0.0,
+      "step": 220
+    },
+    {
+      "epoch": 2.9551569506726456,
+      "grad_norm": 0.0014212332898750901,
+      "learning_rate": 3.6227535467632873e-06,
+      "loss": 0.0,
+      "step": 221
+    },
+    {
+      "epoch": 2.968609865470852,
+      "grad_norm": 0.003146026050671935,
+      "learning_rate": 3.6023076446456415e-06,
+      "loss": 0.0,
+      "step": 222
+    },
+    {
+      "epoch": 2.9820627802690582,
+      "grad_norm": 0.0012437553377822042,
+      "learning_rate": 3.581769807151044e-06,
+      "loss": 0.0,
+      "step": 223
+    },
+    {
+      "epoch": 2.9955156950672643,
+      "grad_norm": 0.017311880365014076,
+      "learning_rate": 3.561141747189538e-06,
+      "loss": 0.0001,
+      "step": 224
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.012313771061599255,
+      "learning_rate": 3.5404251851959537e-06,
+      "loss": 0.0,
+      "step": 225
+    },
+    {
+      "epoch": 3.013452914798206,
+      "grad_norm": 0.0043451120145618916,
+      "learning_rate": 3.519621848986428e-06,
+      "loss": 0.0,
+      "step": 226
+    },
+    {
+      "epoch": 3.0269058295964126,
+      "grad_norm": 0.002464739605784416,
+      "learning_rate": 3.498733473614298e-06,
+      "loss": 0.0,
+      "step": 227
+    },
+    {
+      "epoch": 3.0403587443946187,
+      "grad_norm": 0.0019529856508597732,
+      "learning_rate": 3.47776180122539e-06,
+      "loss": 0.0,
+      "step": 228
+    },
+    {
+      "epoch": 3.0538116591928253,
+      "grad_norm": 0.0013189928140491247,
+      "learning_rate": 3.4567085809127247e-06,
+      "loss": 0.0,
+      "step": 229
+    },
+    {
+      "epoch": 3.0672645739910314,
+      "grad_norm": 0.0021529668010771275,
+      "learning_rate": 3.435575568570633e-06,
+      "loss": 0.0,
+      "step": 230
+    },
+    {
+      "epoch": 3.0807174887892375,
+      "grad_norm": 0.007290184032171965,
+      "learning_rate": 3.4143645267483144e-06,
+      "loss": 0.0,
+      "step": 231
+    },
+    {
+      "epoch": 3.094170403587444,
+      "grad_norm": 0.001345694880001247,
+      "learning_rate": 3.393077224502832e-06,
+      "loss": 0.0,
+      "step": 232
+    },
+    {
+      "epoch": 3.10762331838565,
+      "grad_norm": 0.004343180451542139,
+      "learning_rate": 3.3717154372515716e-06,
+      "loss": 0.0,
+      "step": 233
+    },
+    {
+      "epoch": 3.1210762331838566,
+      "grad_norm": 0.0010816323338076472,
+      "learning_rate": 3.350280946624166e-06,
+      "loss": 0.0,
+      "step": 234
+    },
+    {
+      "epoch": 3.1345291479820627,
+      "grad_norm": 0.003842687699943781,
+      "learning_rate": 3.3287755403139007e-06,
+      "loss": 0.0,
+      "step": 235
+    },
+    {
+      "epoch": 3.1479820627802693,
+      "grad_norm": 0.0013102421071380377,
+      "learning_rate": 3.3072010119286156e-06,
+      "loss": 0.0,
+      "step": 236
+    },
+    {
+      "epoch": 3.1614349775784754,
+      "grad_norm": 0.0014620574656873941,
+      "learning_rate": 3.2855591608411203e-06,
+      "loss": 0.0,
+      "step": 237
+    },
+    {
+      "epoch": 3.1748878923766815,
+      "grad_norm": 0.0007613594643771648,
+      "learning_rate": 3.2638517920391095e-06,
+      "loss": 0.0,
+      "step": 238
+    },
+    {
+      "epoch": 3.188340807174888,
+      "grad_norm": 0.001927120960317552,
+      "learning_rate": 3.2420807159746333e-06,
+      "loss": 0.0,
+      "step": 239
+    },
+    {
+      "epoch": 3.201793721973094,
+      "grad_norm": 0.0047145108692348,
+      "learning_rate": 3.2202477484130947e-06,
+      "loss": 0.0,
+      "step": 240
+    },
+    {
+      "epoch": 3.2152466367713006,
+      "grad_norm": 0.001525243278592825,
+      "learning_rate": 3.1983547102818104e-06,
+      "loss": 0.0,
+      "step": 241
+    },
+    {
+      "epoch": 3.2286995515695067,
+      "grad_norm": 0.0008274471038021147,
+      "learning_rate": 3.1764034275181436e-06,
+      "loss": 0.0,
+      "step": 242
+    },
+    {
+      "epoch": 3.242152466367713,
+      "grad_norm": 0.0009357924573123455,
+      "learning_rate": 3.1543957309172136e-06,
+      "loss": 0.0,
+      "step": 243
+    },
+    {
+      "epoch": 3.2556053811659194,
+      "grad_norm": 0.0016225146828219295,
+      "learning_rate": 3.132333455979202e-06,
+      "loss": 0.0,
+      "step": 244
+    },
+    {
+      "epoch": 3.2690582959641254,
+      "grad_norm": 0.013506707735359669,
+      "learning_rate": 3.1102184427562696e-06,
+      "loss": 0.0001,
+      "step": 245
+    },
+    {
+      "epoch": 3.282511210762332,
+      "grad_norm": 0.0010258476249873638,
+      "learning_rate": 3.0880525356990898e-06,
+      "loss": 0.0,
+      "step": 246
+    },
+    {
+      "epoch": 3.295964125560538,
+      "grad_norm": 0.006462691817432642,
+      "learning_rate": 3.0658375835030148e-06,
+      "loss": 0.0,
+      "step": 247
+    },
+    {
+      "epoch": 3.3094170403587446,
+      "grad_norm": 0.0011441456153988838,
+      "learning_rate": 3.043575438953893e-06,
+      "loss": 0.0,
+      "step": 248
+    },
+    {
+      "epoch": 3.3228699551569507,
+      "grad_norm": 0.000751970277633518,
+      "learning_rate": 3.02126795877354e-06,
+      "loss": 0.0,
+      "step": 249
+    },
+    {
+      "epoch": 3.336322869955157,
+      "grad_norm": 0.006623424123972654,
+      "learning_rate": 2.9989170034648823e-06,
+      "loss": 0.0,
+      "step": 250
+    },
+    {
+      "epoch": 3.3497757847533634,
+      "grad_norm": 0.0021056546829640865,
+      "learning_rate": 2.9765244371567873e-06,
+      "loss": 0.0,
+      "step": 251
+    },
+    {
+      "epoch": 3.3632286995515694,
+      "grad_norm": 0.003343602642416954,
+      "learning_rate": 2.9540921274485913e-06,
+      "loss": 0.0,
+      "step": 252
+    },
+    {
+      "epoch": 3.376681614349776,
+      "grad_norm": 0.0032404386438429356,
+      "learning_rate": 2.9316219452543342e-06,
+      "loss": 0.0,
+      "step": 253
+    },
+    {
+      "epoch": 3.390134529147982,
+      "grad_norm": 0.02200383134186268,
+      "learning_rate": 2.9091157646467205e-06,
+      "loss": 0.0001,
+      "step": 254
+    },
+    {
+      "epoch": 3.403587443946188,
+      "grad_norm": 0.0015861240681260824,
+      "learning_rate": 2.886575462700821e-06,
+      "loss": 0.0,
+      "step": 255
+    },
+    {
+      "epoch": 3.4170403587443947,
+      "grad_norm": 0.0023104625288397074,
+      "learning_rate": 2.864002919337513e-06,
+      "loss": 0.0,
+      "step": 256
+    },
+    {
+      "epoch": 3.430493273542601,
+      "grad_norm": 0.0009963945485651493,
+      "learning_rate": 2.8414000171666952e-06,
+      "loss": 0.0,
+      "step": 257
+    },
+    {
+      "epoch": 3.4439461883408073,
+      "grad_norm": 0.006071928422898054,
+      "learning_rate": 2.81876864133027e-06,
+      "loss": 0.0,
+      "step": 258
+    },
+    {
+      "epoch": 3.4573991031390134,
+      "grad_norm": 0.003969075623899698,
+      "learning_rate": 2.7961106793449217e-06,
+      "loss": 0.0,
+      "step": 259
+    },
+    {
+      "epoch": 3.4708520179372195,
+      "grad_norm": 0.0010353871621191502,
+      "learning_rate": 2.773428020944687e-06,
+      "loss": 0.0,
+      "step": 260
+    },
+    {
+      "epoch": 3.484304932735426,
+      "grad_norm": 0.003665305208414793,
+      "learning_rate": 2.7507225579233487e-06,
+      "loss": 0.0,
+      "step": 261
+    },
+    {
+      "epoch": 3.497757847533632,
+      "grad_norm": 0.003311133710667491,
+      "learning_rate": 2.727996183976659e-06,
+      "loss": 0.0,
+      "step": 262
+    },
+    {
+      "epoch": 3.5112107623318387,
+      "grad_norm": 0.004625072702765465,
+      "learning_rate": 2.705250794544393e-06,
+      "loss": 0.0,
+      "step": 263
+    },
+    {
+      "epoch": 3.524663677130045,
+      "grad_norm": 0.0023010042496025562,
+      "learning_rate": 2.682488286652269e-06,
+      "loss": 0.0,
+      "step": 264
+    },
+    {
+      "epoch": 3.538116591928251,
+      "grad_norm": 0.0008537416579201818,
+      "learning_rate": 2.6597105587537307e-06,
+      "loss": 0.0,
+      "step": 265
+    },
+    {
+      "epoch": 3.5515695067264574,
+      "grad_norm": 0.0014414336765184999,
+      "learning_rate": 2.6369195105716087e-06,
+      "loss": 0.0,
+      "step": 266
+    },
+    {
+      "epoch": 3.5650224215246635,
+      "grad_norm": 0.003060834715142846,
+      "learning_rate": 2.614117042939685e-06,
+      "loss": 0.0,
+      "step": 267
+    },
+    {
+      "epoch": 3.57847533632287,
+      "grad_norm": 0.002594274003058672,
+      "learning_rate": 2.591305057644148e-06,
+      "loss": 0.0,
+      "step": 268
+    },
+    {
+      "epoch": 3.591928251121076,
+      "grad_norm": 0.0009433673694729805,
+      "learning_rate": 2.5684854572649876e-06,
+      "loss": 0.0,
+      "step": 269
+    },
+    {
+      "epoch": 3.6053811659192823,
+      "grad_norm": 0.000922717503271997,
+      "learning_rate": 2.5456601450173123e-06,
+      "loss": 0.0,
+      "step": 270
+    },
+    {
+      "epoch": 3.618834080717489,
+      "grad_norm": 0.0013827934162691236,
+      "learning_rate": 2.522831024592615e-06,
+      "loss": 0.0,
+      "step": 271
+    },
+    {
+      "epoch": 3.6322869955156953,
+      "grad_norm": 0.0007690931670367718,
+      "learning_rate": 2.5e-06,
+      "loss": 0.0,
+      "step": 272
+    },
+    {
+      "epoch": 3.6457399103139014,
+      "grad_norm": 0.002565343165770173,
+      "learning_rate": 2.4771689754073856e-06,
+      "loss": 0.0,
+      "step": 273
+    },
+    {
+      "epoch": 3.6591928251121075,
+      "grad_norm": 0.0013943302910774946,
+      "learning_rate": 2.454339854982688e-06,
+      "loss": 0.0,
+      "step": 274
+    },
+    {
+      "epoch": 3.672645739910314,
+      "grad_norm": 0.0006668591522611678,
+      "learning_rate": 2.4315145427350132e-06,
+      "loss": 0.0,
+      "step": 275
+    },
+    {
+      "epoch": 3.68609865470852,
+      "grad_norm": 0.0025589216966181993,
+      "learning_rate": 2.408694942355853e-06,
+      "loss": 0.0,
+      "step": 276
+    },
+    {
+      "epoch": 3.6995515695067267,
+      "grad_norm": 0.001420054235495627,
+      "learning_rate": 2.3858829570603157e-06,
+      "loss": 0.0,
+      "step": 277
+    },
+    {
+      "epoch": 3.713004484304933,
+      "grad_norm": 0.0014220779994502664,
+      "learning_rate": 2.363080489428391e-06,
+      "loss": 0.0,
+      "step": 278
+    },
+    {
+      "epoch": 3.726457399103139,
+      "grad_norm": 0.0009400748531334102,
+      "learning_rate": 2.3402894412462697e-06,
+      "loss": 0.0,
+      "step": 279
+    },
+    {
+      "epoch": 3.7399103139013454,
+      "grad_norm": 0.0006919830339029431,
+      "learning_rate": 2.317511713347731e-06,
+      "loss": 0.0,
+      "step": 280
+    },
+    {
+      "epoch": 3.7533632286995515,
+      "grad_norm": 0.0011200032895430923,
+      "learning_rate": 2.2947492054556075e-06,
+      "loss": 0.0,
+      "step": 281
+    },
+    {
+      "epoch": 3.766816143497758,
+      "grad_norm": 0.0008829529979266226,
+      "learning_rate": 2.272003816023341e-06,
+      "loss": 0.0,
+      "step": 282
+    },
+    {
+      "epoch": 3.780269058295964,
+      "grad_norm": 0.0009032113594003022,
+      "learning_rate": 2.2492774420766517e-06,
+      "loss": 0.0,
+      "step": 283
+    },
+    {
+      "epoch": 3.7937219730941703,
+      "grad_norm": 0.0012500348966568708,
+      "learning_rate": 2.2265719790553147e-06,
+      "loss": 0.0,
+      "step": 284
+    },
+    {
+      "epoch": 3.807174887892377,
+      "grad_norm": 0.0008029688615351915,
+      "learning_rate": 2.20388932065508e-06,
+      "loss": 0.0,
+      "step": 285
+    },
+    {
+      "epoch": 3.820627802690583,
+      "grad_norm": 0.0015973382396623492,
+      "learning_rate": 2.1812313586697307e-06,
+      "loss": 0.0,
+      "step": 286
+    },
+    {
+      "epoch": 3.8340807174887894,
+      "grad_norm": 0.0021750489249825478,
+      "learning_rate": 2.1585999828333065e-06,
+      "loss": 0.0,
+      "step": 287
+    },
+    {
+      "epoch": 3.8475336322869955,
+      "grad_norm": 0.004780885297805071,
+      "learning_rate": 2.1359970806624886e-06,
+      "loss": 0.0,
+      "step": 288
+    },
+    {
+      "epoch": 3.8609865470852016,
+      "grad_norm": 0.0007556549389846623,
+      "learning_rate": 2.11342453729918e-06,
+      "loss": 0.0,
+      "step": 289
+    },
+    {
+      "epoch": 3.874439461883408,
+      "grad_norm": 0.0006669393624179065,
+      "learning_rate": 2.0908842353532803e-06,
+      "loss": 0.0,
+      "step": 290
+    },
+    {
+      "epoch": 3.8878923766816142,
+      "grad_norm": 0.001046078628860414,
+      "learning_rate": 2.0683780547456666e-06,
+      "loss": 0.0,
+      "step": 291
+    },
+    {
+      "epoch": 3.901345291479821,
+      "grad_norm": 0.004534538835287094,
+      "learning_rate": 2.045907872551409e-06,
+      "loss": 0.0,
+      "step": 292
+    },
+    {
+      "epoch": 3.914798206278027,
+      "grad_norm": 0.0008524219738319516,
+      "learning_rate": 2.0234755628432135e-06,
+      "loss": 0.0,
+      "step": 293
+    },
+    {
+      "epoch": 3.928251121076233,
+      "grad_norm": 0.0006630662246607244,
+      "learning_rate": 2.0010829965351185e-06,
+      "loss": 0.0,
+      "step": 294
+    },
+    {
+      "epoch": 3.9417040358744395,
+      "grad_norm": 0.008820832706987858,
+      "learning_rate": 1.978732041226461e-06,
+      "loss": 0.0,
+      "step": 295
+    },
+    {
+      "epoch": 3.9551569506726456,
+      "grad_norm": 0.0007411285769194365,
+      "learning_rate": 1.956424561046108e-06,
+      "loss": 0.0,
+      "step": 296
+    },
+    {
+      "epoch": 3.968609865470852,
+      "grad_norm": 0.004659746307879686,
+      "learning_rate": 1.934162416496986e-06,
+      "loss": 0.0,
+      "step": 297
+    },
+    {
+      "epoch": 3.9820627802690582,
+      "grad_norm": 0.004637689795345068,
+      "learning_rate": 1.911947464300911e-06,
+      "loss": 0.0,
+      "step": 298
+    },
+    {
+      "epoch": 3.9955156950672643,
+      "grad_norm": 0.000994283240288496,
+      "learning_rate": 1.8897815572437304e-06,
+      "loss": 0.0,
+      "step": 299
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 0.000994283240288496,
+      "learning_rate": 1.8676665440207982e-06,
+      "loss": 0.0,
+      "step": 300
+    },
+    {
+      "epoch": 4.013452914798206,
+      "grad_norm": 0.0013737012632191181,
+      "learning_rate": 1.845604269082787e-06,
+      "loss": 0.0,
+      "step": 301
+    },
+    {
+      "epoch": 4.026905829596412,
+      "grad_norm": 0.0019984773825854063,
+      "learning_rate": 1.8235965724818563e-06,
+      "loss": 0.0,
+      "step": 302
+    },
+    {
+      "epoch": 4.040358744394619,
+      "grad_norm": 0.001266778097487986,
+      "learning_rate": 1.8016452897181902e-06,
+      "loss": 0.0,
+      "step": 303
+    },
+    {
+      "epoch": 4.053811659192825,
+      "grad_norm": 0.0029514820780605078,
+      "learning_rate": 1.7797522515869062e-06,
+      "loss": 0.0,
+      "step": 304
+    },
+    {
+      "epoch": 4.067264573991031,
+      "grad_norm": 0.00543377036228776,
+      "learning_rate": 1.7579192840253678e-06,
+      "loss": 0.0,
+      "step": 305
+    },
+    {
+      "epoch": 4.0807174887892375,
+      "grad_norm": 0.0013654204085469246,
+      "learning_rate": 1.7361482079608916e-06,
+      "loss": 0.0,
+      "step": 306
+    },
+    {
+      "epoch": 4.0941704035874436,
+      "grad_norm": 0.0031392457894980907,
+      "learning_rate": 1.7144408391588812e-06,
+      "loss": 0.0,
+      "step": 307
+    },
+    {
+      "epoch": 4.1076233183856505,
+      "grad_norm": 0.008237143978476524,
+      "learning_rate": 1.6927989880713852e-06,
+      "loss": 0.0,
+      "step": 308
+    },
+    {
+      "epoch": 4.121076233183857,
+      "grad_norm": 0.003675549989566207,
+      "learning_rate": 1.6712244596861008e-06,
+      "loss": 0.0,
+      "step": 309
+    },
+    {
+      "epoch": 4.134529147982063,
+      "grad_norm": 0.0007359271985478699,
+      "learning_rate": 1.6497190533758351e-06,
+      "loss": 0.0,
+      "step": 310
+    },
+    {
+      "epoch": 4.147982062780269,
+      "grad_norm": 0.006465969141572714,
+      "learning_rate": 1.628284562748429e-06,
+      "loss": 0.0,
+      "step": 311
+    },
+    {
+      "epoch": 4.161434977578475,
+      "grad_norm": 0.0018335963832214475,
+      "learning_rate": 1.6069227754971683e-06,
+      "loss": 0.0,
+      "step": 312
+    },
+    {
+      "epoch": 4.174887892376682,
+      "grad_norm": 0.0007329874206334352,
+      "learning_rate": 1.5856354732516865e-06,
+      "loss": 0.0,
+      "step": 313
+    },
+    {
+      "epoch": 4.188340807174888,
+      "grad_norm": 0.000715755217242986,
+      "learning_rate": 1.5644244314293672e-06,
+      "loss": 0.0,
+      "step": 314
+    },
+    {
+      "epoch": 4.201793721973094,
+      "grad_norm": 0.001220604288391769,
+      "learning_rate": 1.5432914190872757e-06,
+      "loss": 0.0,
+      "step": 315
+    },
+    {
+      "epoch": 4.2152466367713,
+      "grad_norm": 0.002464067656546831,
+      "learning_rate": 1.5222381987746104e-06,
+      "loss": 0.0,
+      "step": 316
+    },
+    {
+      "epoch": 4.228699551569507,
+      "grad_norm": 0.002008558949455619,
+      "learning_rate": 1.5012665263857024e-06,
+      "loss": 0.0,
+      "step": 317
+    },
+    {
+      "epoch": 4.242152466367713,
+      "grad_norm": 0.0008084883447736502,
+      "learning_rate": 1.4803781510135723e-06,
+      "loss": 0.0,
+      "step": 318
+    },
+    {
+      "epoch": 4.255605381165919,
+      "grad_norm": 0.0006517472211271524,
+      "learning_rate": 1.4595748148040467e-06,
+      "loss": 0.0,
+      "step": 319
+    },
+    {
+      "epoch": 4.2690582959641254,
+      "grad_norm": 0.004320419393479824,
+      "learning_rate": 1.4388582528104628e-06,
+      "loss": 0.0,
+      "step": 320
+    },
+    {
+      "epoch": 4.2825112107623315,
+      "grad_norm": 0.0016021894989535213,
+      "learning_rate": 1.4182301928489556e-06,
+      "loss": 0.0,
+      "step": 321
+    },
+    {
+      "epoch": 4.2959641255605385,
+      "grad_norm": 0.002214021747931838,
+      "learning_rate": 1.3976923553543587e-06,
+      "loss": 0.0,
+      "step": 322
+    },
+    {
+      "epoch": 4.309417040358745,
+      "grad_norm": 0.000659990357235074,
+      "learning_rate": 1.3772464532367124e-06,
+      "loss": 0.0,
+      "step": 323
+    },
+    {
+      "epoch": 4.322869955156951,
+      "grad_norm": 0.001591618754900992,
+      "learning_rate": 1.3568941917384038e-06,
+      "loss": 0.0,
+      "step": 324
+    },
+    {
+      "epoch": 4.336322869955157,
+      "grad_norm": 0.005078968591988087,
+      "learning_rate": 1.3366372682919415e-06,
+      "loss": 0.0,
+      "step": 325
+    },
+    {
+      "epoch": 4.349775784753363,
+      "grad_norm": 0.001620083930902183,
+      "learning_rate": 1.316477372378392e-06,
+      "loss": 0.0,
+      "step": 326
+    },
+    {
+      "epoch": 4.36322869955157,
+      "grad_norm": 0.0006833957741037011,
+      "learning_rate": 1.2964161853864655e-06,
+      "loss": 0.0,
+      "step": 327
+    },
+    {
+      "epoch": 4.376681614349776,
+      "grad_norm": 0.001587000209838152,
+      "learning_rate": 1.2764553804722868e-06,
+      "loss": 0.0,
+      "step": 328
+    },
+    {
+      "epoch": 4.390134529147982,
+      "grad_norm": 0.0026333138812333345,
+      "learning_rate": 1.256596622419852e-06,
+      "loss": 0.0,
+      "step": 329
+    },
+    {
+      "epoch": 4.403587443946188,
+      "grad_norm": 0.0006281823152676225,
+      "learning_rate": 1.236841567502177e-06,
+      "loss": 0.0,
+      "step": 330
+    },
+    {
+      "epoch": 4.417040358744394,
+      "grad_norm": 0.0014232223620638251,
+      "learning_rate": 1.2171918633431623e-06,
+      "loss": 0.0,
+      "step": 331
+    },
+    {
+      "epoch": 4.430493273542601,
+      "grad_norm": 0.0007285006577149034,
+      "learning_rate": 1.1976491487801747e-06,
+      "loss": 0.0,
+      "step": 332
+    },
+    {
+      "epoch": 4.443946188340807,
+      "grad_norm": 0.0009421680588275194,
+      "learning_rate": 1.1782150537273665e-06,
+      "loss": 0.0,
+      "step": 333
+    },
+    {
+      "epoch": 4.457399103139013,
+      "grad_norm": 0.002453515538945794,
+      "learning_rate": 1.1588911990397364e-06,
+      "loss": 0.0,
+      "step": 334
+    },
+    {
+      "epoch": 4.4708520179372195,
+      "grad_norm": 0.003878404852002859,
+      "learning_rate": 1.139679196377941e-06,
+      "loss": 0.0,
+      "step": 335
+    },
+    {
+      "epoch": 4.484304932735426,
+      "grad_norm": 0.0005806125700473785,
+      "learning_rate": 1.1205806480738852e-06,
+      "loss": 0.0,
+      "step": 336
+    },
+    {
+      "epoch": 4.497757847533633,
+      "grad_norm": 0.0008392960298806429,
+      "learning_rate": 1.1015971469970795e-06,
+      "loss": 0.0,
+      "step": 337
+    },
+    {
+      "epoch": 4.511210762331839,
+      "grad_norm": 0.003920732066035271,
+      "learning_rate": 1.0827302764217887e-06,
+      "loss": 0.0,
+      "step": 338
+    },
+    {
+      "epoch": 4.524663677130045,
+      "grad_norm": 0.0006277119391597807,
+      "learning_rate": 1.063981609894987e-06,
+      "loss": 0.0,
+      "step": 339
+    },
+    {
+      "epoch": 4.538116591928251,
+      "grad_norm": 0.0008221868192777038,
+      "learning_rate": 1.0453527111051183e-06,
+      "loss": 0.0,
+      "step": 340
+    },
+    {
+      "epoch": 4.551569506726457,
+      "grad_norm": 0.002272391924634576,
+      "learning_rate": 1.0268451337516774e-06,
+      "loss": 0.0,
+      "step": 341
+    },
+    {
+      "epoch": 4.565022421524664,
+      "grad_norm": 0.0025312723591923714,
+      "learning_rate": 1.0084604214156324e-06,
+      "loss": 0.0,
+      "step": 342
+    },
+    {
+      "epoch": 4.57847533632287,
+      "grad_norm": 0.0016034027794376016,
+      "learning_rate": 9.902001074306836e-07,
+      "loss": 0.0,
+      "step": 343
+    },
+    {
+      "epoch": 4.591928251121076,
+      "grad_norm": 0.00481824204325676,
+      "learning_rate": 9.72065714755377e-07,
+      "loss": 0.0,
+      "step": 344
+    },
+    {
+      "epoch": 4.605381165919282,
+      "grad_norm": 0.000697854848112911,
+      "learning_rate": 9.5405875584609e-07,
+      "loss": 0.0,
+      "step": 345
+    },
+    {
+      "epoch": 4.618834080717489,
+      "grad_norm": 0.0007224497967399657,
+      "learning_rate": 9.361807325308861e-07,
+      "loss": 0.0,
+      "step": 346
+    },
+    {
+      "epoch": 4.632286995515695,
+      "grad_norm": 0.0008536895620636642,
+      "learning_rate": 9.184331358842594e-07,
+      "loss": 0.0,
+      "step": 347
+    },
+    {
+      "epoch": 4.645739910313901,
+      "grad_norm": 0.0019302460132166743,
+      "learning_rate": 9.008174461027724e-07,
+      "loss": 0.0,
+      "step": 348
+    },
+    {
+      "epoch": 4.6591928251121075,
+      "grad_norm": 0.0005941809504292905,
+      "learning_rate": 8.833351323816064e-07,
+      "loss": 0.0,
+      "step": 349
+    },
+    {
+      "epoch": 4.672645739910314,
+      "grad_norm": 0.0042343768291175365,
+      "learning_rate": 8.659876527920277e-07,
+      "loss": 0.0,
+      "step": 350
+    },
+    {
+      "epoch": 4.68609865470852,
+      "grad_norm": 0.002343270927667618,
+      "learning_rate": 8.487764541597765e-07,
+      "loss": 0.0,
+      "step": 351
+    },
+    {
+      "epoch": 4.699551569506727,
+      "grad_norm": 0.0008750990382395685,
+      "learning_rate": 8.317029719444017e-07,
+      "loss": 0.0,
+      "step": 352
+    },
+    {
+      "epoch": 4.713004484304933,
+      "grad_norm": 0.0015886047622188926,
+      "learning_rate": 8.147686301195384e-07,
+      "loss": 0.0,
+      "step": 353
+    },
+    {
+      "epoch": 4.726457399103139,
+      "grad_norm": 0.0005803987733088434,
+      "learning_rate": 7.979748410541452e-07,
+      "loss": 0.0,
+      "step": 354
+    },
+    {
+      "epoch": 4.739910313901345,
+      "grad_norm": 0.0011800318025052547,
+      "learning_rate": 7.813230053947054e-07,
+      "loss": 0.0,
+      "step": 355
+    },
+    {
+      "epoch": 4.753363228699552,
+      "grad_norm": 0.00053420290350914,
+      "learning_rate": 7.648145119484152e-07,
+      "loss": 0.0,
+      "step": 356
+    },
+    {
+      "epoch": 4.766816143497758,
+      "grad_norm": 0.0007576547213830054,
+      "learning_rate": 7.484507375673505e-07,
+      "loss": 0.0,
+      "step": 357
+    },
+    {
+      "epoch": 4.780269058295964,
+      "grad_norm": 0.0006002706359140575,
+      "learning_rate": 7.322330470336314e-07,
+      "loss": 0.0,
+      "step": 358
+    },
+    {
+      "epoch": 4.79372197309417,
+      "grad_norm": 0.001380656729452312,
+      "learning_rate": 7.161627929456005e-07,
+      "loss": 0.0,
+      "step": 359
+    },
+    {
+      "epoch": 4.807174887892376,
+      "grad_norm": 0.000663610000628978,
+      "learning_rate": 7.002413156050109e-07,
+      "loss": 0.0,
+      "step": 360
+    },
+    {
+      "epoch": 4.820627802690583,
+      "grad_norm": 0.002041428815573454,
+      "learning_rate": 6.844699429052377e-07,
+      "loss": 0.0,
+      "step": 361
+    },
+    {
+      "epoch": 4.834080717488789,
+      "grad_norm": 0.0009557300363667309,
+      "learning_rate": 6.688499902205345e-07,
+      "loss": 0.0,
+      "step": 362
+    },
+    {
+      "epoch": 4.8475336322869955,
+      "grad_norm": 0.0006233525928109884,
+      "learning_rate": 6.533827602963244e-07,
+      "loss": 0.0,
+      "step": 363
+    },
+    {
+      "epoch": 4.860986547085202,
+      "grad_norm": 0.0005438902880996466,
+      "learning_rate": 6.380695431405453e-07,
+      "loss": 0.0,
+      "step": 364
+    },
+    {
+      "epoch": 4.874439461883408,
+      "grad_norm": 0.003903225064277649,
+      "learning_rate": 6.229116159160653e-07,
+      "loss": 0.0,
+      "step": 365
+    },
+    {
+      "epoch": 4.887892376681615,
+      "grad_norm": 0.0008262547198683023,
+      "learning_rate": 6.079102428341588e-07,
+      "loss": 0.0,
+      "step": 366
+    },
+    {
+      "epoch": 4.901345291479821,
+      "grad_norm": 0.0006261409143917263,
+      "learning_rate": 5.930666750490724e-07,
+      "loss": 0.0,
+      "step": 367
+    },
+    {
+      "epoch": 4.914798206278027,
+      "grad_norm": 0.000995171838440001,
+      "learning_rate": 5.783821505536696e-07,
+      "loss": 0.0,
+      "step": 368
+    },
+    {
+      "epoch": 4.928251121076233,
+      "grad_norm": 0.0018100445158779621,
+      "learning_rate": 5.63857894076186e-07,
+      "loss": 0.0,
+      "step": 369
+    },
+    {
+      "epoch": 4.941704035874439,
+      "grad_norm": 0.001178378821350634,
+      "learning_rate": 5.494951169780777e-07,
+      "loss": 0.0,
+      "step": 370
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 444,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 6,
+  "save_steps": 74,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.226925182892114e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-370/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db2163a23b611a8c5fa6dfb9e81074ea1793ff3e8d206475ef1251ddbdea9baa
+size 7992

checkpoint-370/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)