Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,20 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-100/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-1000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-1100/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-1200/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-1300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-1400/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-1404/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-200/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-400/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-600/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-700/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-800/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/checkpoint-900/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+2/tokenizer.json filter=lfs diff=lfs merge=lfs -text

2/README.md

@@ -0,0 +1,62 @@
+---
+library_name: peft
+license: other
+base_model: /root/autodl-tmp/Qwen3-8B
+tags:
+- llama-factory
+- lora
+- generated_from_trainer
+model-index:
+- name: '2'
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# 2
+
+This model is a fine-tuned version of [/root/autodl-tmp/Qwen3-8B](https://huggingface.co//root/autodl-tmp/Qwen3-8B) on the LAW and the LAWlegal datasets.
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 0.0001
+- train_batch_size: 1
+- eval_batch_size: 8
+- seed: 42
+- distributed_type: multi-GPU
+- num_devices: 2
+- gradient_accumulation_steps: 8
+- total_train_batch_size: 16
+- total_eval_batch_size: 16
+- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- lr_scheduler_type: cosine
+- num_epochs: 1.0
+- mixed_precision_training: Native AMP
+
+### Training results
+
+
+
+### Framework versions
+
+- PEFT 0.15.1
+- Transformers 4.51.3
+- Pytorch 2.3.0+cu121
+- Datasets 3.5.0
+- Tokenizers 0.21.0

2/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "/root/autodl-tmp/Qwen3-8B",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 8,
+  "lora_bias": false,
+  "lora_dropout": 0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 4,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": true
+}

2/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a8b269cca579b1edb239928e0d508ce1b9a87a02089570c091b785832df8bcd
+size 21889736

2/added_tokens.json

@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|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
+}

2/all_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 0.9995550413811516,
+    "num_input_tokens_seen": 29121176,
+    "total_flos": 1.3243117909594604e+18,
+    "train_loss": 0.19406711987280778,
+    "train_runtime": 19252.8015,
+    "train_samples_per_second": 1.167,
+    "train_steps_per_second": 0.073
+}

2/checkpoint-100/README.md

@@ -0,0 +1,202 @@
+---
+base_model: /root/autodl-tmp/Qwen3-8B
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.15.1

2/checkpoint-100/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "/root/autodl-tmp/Qwen3-8B",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 8,
+  "lora_bias": false,
+  "lora_dropout": 0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 4,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": true
+}

2/checkpoint-100/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fe6ad31d909f41448594ba6946f7e00f439e6275f85fd4aee80fdd273d4e5c3
+size 21889736

2/checkpoint-100/added_tokens.json

@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|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
+}

2/checkpoint-100/global_step100/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d40b8c4949d723df5ed2c9ea72d6290c4f8464a72c50f66a9caa526b3e3051b9
+size 133228077

2/checkpoint-100/global_step100/zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7037c0521ea824ca44d0543c3100740c3b3aaf4190b5e3df1524b69b9b1d90cb
+size 65501576

2/checkpoint-100/global_step100/zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ce27b3ccfdf5ca6a8cd2398eca54fc0e59b06459f6c4970ff4d8137dd6c22d3
+size 65501704

2/checkpoint-100/latest

@@ -0,0 +1 @@
+global_step100

2/checkpoint-100/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8d6a959372d5e0c2ea025dd26c9d0ad2046fce19352056cae8074dcbd0a6fd4
+size 14512

2/checkpoint-100/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f68a37892a1b445d21bb35cc10bf7a058a6f9ec8c363f5ed156ff4f49d90fb6
+size 14512

2/checkpoint-100/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b89769ae34d6c4b2010219db05cd08bd2e022e9ebb9d70a0d02c0d5adec44d84
+size 1064

2/checkpoint-100/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
+  }
+}

2/checkpoint-100/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4
+size 11422654

2/checkpoint-100/tokenizer_config.json

@@ -0,0 +1,241 @@
+{
+  "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
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "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\\n' }}\n    {%- endif %}\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    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set content = message.content %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in message.content %}\n                {%- set content = message.content.split('</think>')[-1].lstrip('\\n') %}\n                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first 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    {%- if enable_thinking is defined and enable_thinking is false %}\n        {{- '<think>\\n\\n</think>\\n\\n' }}\n    {%- endif %}\n{%- endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

2/checkpoint-100/trainer_state.json

@@ -0,0 +1,194 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.07119337901575154,
+  "eval_steps": 500,
+  "global_step": 100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003559668950787577,
+      "grad_norm": 1.069462776184082,
+      "learning_rate": 9.999799726899262e-05,
+      "loss": 1.4623,
+      "num_input_tokens_seen": 97000,
+      "step": 5
+    },
+    {
+      "epoch": 0.007119337901575154,
+      "grad_norm": 0.6442700624465942,
+      "learning_rate": 9.998986144924251e-05,
+      "loss": 1.0085,
+      "num_input_tokens_seen": 185728,
+      "step": 10
+    },
+    {
+      "epoch": 0.01067900685236273,
+      "grad_norm": 0.7749446630477905,
+      "learning_rate": 9.997546838764065e-05,
+      "loss": 0.8166,
+      "num_input_tokens_seen": 274048,
+      "step": 15
+    },
+    {
+      "epoch": 0.014238675803150308,
+      "grad_norm": 0.6821078658103943,
+      "learning_rate": 9.995481988576968e-05,
+      "loss": 0.6584,
+      "num_input_tokens_seen": 395024,
+      "step": 20
+    },
+    {
+      "epoch": 0.017798344753937885,
+      "grad_norm": 0.6945608258247375,
+      "learning_rate": 9.992791852820709e-05,
+      "loss": 0.4809,
+      "num_input_tokens_seen": 484056,
+      "step": 25
+    },
+    {
+      "epoch": 0.02135801370472546,
+      "grad_norm": 0.30774927139282227,
+      "learning_rate": 9.989476768220168e-05,
+      "loss": 0.3719,
+      "num_input_tokens_seen": 588744,
+      "step": 30
+    },
+    {
+      "epoch": 0.024917682655513036,
+      "grad_norm": 0.3996995985507965,
+      "learning_rate": 9.985537149725205e-05,
+      "loss": 0.3598,
+      "num_input_tokens_seen": 687824,
+      "step": 35
+    },
+    {
+      "epoch": 0.028477351606300615,
+      "grad_norm": 0.26959237456321716,
+      "learning_rate": 9.980973490458728e-05,
+      "loss": 0.3405,
+      "num_input_tokens_seen": 810104,
+      "step": 40
+    },
+    {
+      "epoch": 0.03203702055708819,
+      "grad_norm": 0.26429229974746704,
+      "learning_rate": 9.97578636165496e-05,
+      "loss": 0.3165,
+      "num_input_tokens_seen": 910048,
+      "step": 45
+    },
+    {
+      "epoch": 0.03559668950787577,
+      "grad_norm": 0.32172468304634094,
+      "learning_rate": 9.969976412587944e-05,
+      "loss": 0.292,
+      "num_input_tokens_seen": 1022144,
+      "step": 50
+    },
+    {
+      "epoch": 0.039156358458663346,
+      "grad_norm": 0.28077825903892517,
+      "learning_rate": 9.96354437049027e-05,
+      "loss": 0.2923,
+      "num_input_tokens_seen": 1126152,
+      "step": 55
+    },
+    {
+      "epoch": 0.04271602740945092,
+      "grad_norm": 0.2883082926273346,
+      "learning_rate": 9.956491040462052e-05,
+      "loss": 0.2807,
+      "num_input_tokens_seen": 1252376,
+      "step": 60
+    },
+    {
+      "epoch": 0.0462756963602385,
+      "grad_norm": 0.314687043428421,
+      "learning_rate": 9.948817305370143e-05,
+      "loss": 0.2779,
+      "num_input_tokens_seen": 1353176,
+      "step": 65
+    },
+    {
+      "epoch": 0.04983536531102607,
+      "grad_norm": 0.38759657740592957,
+      "learning_rate": 9.94052412573764e-05,
+      "loss": 0.2632,
+      "num_input_tokens_seen": 1458928,
+      "step": 70
+    },
+    {
+      "epoch": 0.05339503426181365,
+      "grad_norm": 0.3042629063129425,
+      "learning_rate": 9.931612539623643e-05,
+      "loss": 0.28,
+      "num_input_tokens_seen": 1550808,
+      "step": 75
+    },
+    {
+      "epoch": 0.05695470321260123,
+      "grad_norm": 0.2713088393211365,
+      "learning_rate": 9.922083662493329e-05,
+      "loss": 0.2927,
+      "num_input_tokens_seen": 1658104,
+      "step": 80
+    },
+    {
+      "epoch": 0.060514372163388806,
+      "grad_norm": 0.2769795060157776,
+      "learning_rate": 9.911938687078324e-05,
+      "loss": 0.2728,
+      "num_input_tokens_seen": 1763840,
+      "step": 85
+    },
+    {
+      "epoch": 0.06407404111417637,
+      "grad_norm": 0.3135468363761902,
+      "learning_rate": 9.901178883227414e-05,
+      "loss": 0.2562,
+      "num_input_tokens_seen": 1885016,
+      "step": 90
+    },
+    {
+      "epoch": 0.06763371006496396,
+      "grad_norm": 0.3024243116378784,
+      "learning_rate": 9.889805597747588e-05,
+      "loss": 0.2564,
+      "num_input_tokens_seen": 2001056,
+      "step": 95
+    },
+    {
+      "epoch": 0.07119337901575154,
+      "grad_norm": 0.29715123772621155,
+      "learning_rate": 9.877820254235471e-05,
+      "loss": 0.2392,
+      "num_input_tokens_seen": 2076528,
+      "step": 100
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1404,
+  "num_input_tokens_seen": 2076528,
+  "num_train_epochs": 1,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.443198752731955e+16,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

2/checkpoint-100/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53d6f2b6ddc39413c7f5106d4dfc20a60221e892597125a4051acc98e75cf41e
+size 7352

2/checkpoint-100/zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/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 . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+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, weights_only=False)
+
+        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 tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # 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}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] 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")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+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 = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * 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
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        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}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        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 to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=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
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    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.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    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")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               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_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``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
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+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_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    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_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

2/checkpoint-1000/README.md

@@ -0,0 +1,202 @@
+---
+base_model: /root/autodl-tmp/Qwen3-8B
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.15.1

2/checkpoint-1000/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "/root/autodl-tmp/Qwen3-8B",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 8,
+  "lora_bias": false,
+  "lora_dropout": 0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 4,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": true
+}

2/checkpoint-1000/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ea33f60b95b36233813970b1cad9aebe8d9f0bf1379369bdac3623d2f62af7f
+size 21889736

2/checkpoint-1000/added_tokens.json

@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|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
+}

2/checkpoint-1000/global_step1000/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e82d63a880cfb762ee5a1095a9b9c8394f33ab337ec8d6e8ca21d5c4e95d4be
+size 133228077

2/checkpoint-1000/global_step1000/zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ca6ecd594b87e44bb2fe093cb7a59f8eab0f8bd785157765d6e294cb1c95159
+size 65501576

2/checkpoint-1000/global_step1000/zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39cae1e945fe56433b3281f0469ce50930c8a438658f65ebde8e8e499ccbc08c
+size 65501704

2/checkpoint-1000/latest

@@ -0,0 +1 @@
+global_step1000

2/checkpoint-1000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8d6a959372d5e0c2ea025dd26c9d0ad2046fce19352056cae8074dcbd0a6fd4
+size 14512

2/checkpoint-1000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f68a37892a1b445d21bb35cc10bf7a058a6f9ec8c363f5ed156ff4f49d90fb6
+size 14512

2/checkpoint-1000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ea813e47c515744a6308c8143c3b1d70b704d9e8500020af4d6e28d64e8ebe0
+size 1064

2/checkpoint-1000/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
+  }
+}

2/checkpoint-1000/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4
+size 11422654

2/checkpoint-1000/tokenizer_config.json

@@ -0,0 +1,241 @@
+{
+  "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
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "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\\n' }}\n    {%- endif %}\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    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set content = message.content %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in message.content %}\n                {%- set content = message.content.split('</think>')[-1].lstrip('\\n') %}\n                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first 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    {%- if enable_thinking is defined and enable_thinking is false %}\n        {{- '<think>\\n\\n</think>\\n\\n' }}\n    {%- endif %}\n{%- endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

2/checkpoint-1000/trainer_state.json

@@ -0,0 +1,1634 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.7119337901575153,
+  "eval_steps": 500,
+  "global_step": 1000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003559668950787577,
+      "grad_norm": 1.069462776184082,
+      "learning_rate": 9.999799726899262e-05,
+      "loss": 1.4623,
+      "num_input_tokens_seen": 97000,
+      "step": 5
+    },
+    {
+      "epoch": 0.007119337901575154,
+      "grad_norm": 0.6442700624465942,
+      "learning_rate": 9.998986144924251e-05,
+      "loss": 1.0085,
+      "num_input_tokens_seen": 185728,
+      "step": 10
+    },
+    {
+      "epoch": 0.01067900685236273,
+      "grad_norm": 0.7749446630477905,
+      "learning_rate": 9.997546838764065e-05,
+      "loss": 0.8166,
+      "num_input_tokens_seen": 274048,
+      "step": 15
+    },
+    {
+      "epoch": 0.014238675803150308,
+      "grad_norm": 0.6821078658103943,
+      "learning_rate": 9.995481988576968e-05,
+      "loss": 0.6584,
+      "num_input_tokens_seen": 395024,
+      "step": 20
+    },
+    {
+      "epoch": 0.017798344753937885,
+      "grad_norm": 0.6945608258247375,
+      "learning_rate": 9.992791852820709e-05,
+      "loss": 0.4809,
+      "num_input_tokens_seen": 484056,
+      "step": 25
+    },
+    {
+      "epoch": 0.02135801370472546,
+      "grad_norm": 0.30774927139282227,
+      "learning_rate": 9.989476768220168e-05,
+      "loss": 0.3719,
+      "num_input_tokens_seen": 588744,
+      "step": 30
+    },
+    {
+      "epoch": 0.024917682655513036,
+      "grad_norm": 0.3996995985507965,
+      "learning_rate": 9.985537149725205e-05,
+      "loss": 0.3598,
+      "num_input_tokens_seen": 687824,
+      "step": 35
+    },
+    {
+      "epoch": 0.028477351606300615,
+      "grad_norm": 0.26959237456321716,
+      "learning_rate": 9.980973490458728e-05,
+      "loss": 0.3405,
+      "num_input_tokens_seen": 810104,
+      "step": 40
+    },
+    {
+      "epoch": 0.03203702055708819,
+      "grad_norm": 0.26429229974746704,
+      "learning_rate": 9.97578636165496e-05,
+      "loss": 0.3165,
+      "num_input_tokens_seen": 910048,
+      "step": 45
+    },
+    {
+      "epoch": 0.03559668950787577,
+      "grad_norm": 0.32172468304634094,
+      "learning_rate": 9.969976412587944e-05,
+      "loss": 0.292,
+      "num_input_tokens_seen": 1022144,
+      "step": 50
+    },
+    {
+      "epoch": 0.039156358458663346,
+      "grad_norm": 0.28077825903892517,
+      "learning_rate": 9.96354437049027e-05,
+      "loss": 0.2923,
+      "num_input_tokens_seen": 1126152,
+      "step": 55
+    },
+    {
+      "epoch": 0.04271602740945092,
+      "grad_norm": 0.2883082926273346,
+      "learning_rate": 9.956491040462052e-05,
+      "loss": 0.2807,
+      "num_input_tokens_seen": 1252376,
+      "step": 60
+    },
+    {
+      "epoch": 0.0462756963602385,
+      "grad_norm": 0.314687043428421,
+      "learning_rate": 9.948817305370143e-05,
+      "loss": 0.2779,
+      "num_input_tokens_seen": 1353176,
+      "step": 65
+    },
+    {
+      "epoch": 0.04983536531102607,
+      "grad_norm": 0.38759657740592957,
+      "learning_rate": 9.94052412573764e-05,
+      "loss": 0.2632,
+      "num_input_tokens_seen": 1458928,
+      "step": 70
+    },
+    {
+      "epoch": 0.05339503426181365,
+      "grad_norm": 0.3042629063129425,
+      "learning_rate": 9.931612539623643e-05,
+      "loss": 0.28,
+      "num_input_tokens_seen": 1550808,
+      "step": 75
+    },
+    {
+      "epoch": 0.05695470321260123,
+      "grad_norm": 0.2713088393211365,
+      "learning_rate": 9.922083662493329e-05,
+      "loss": 0.2927,
+      "num_input_tokens_seen": 1658104,
+      "step": 80
+    },
+    {
+      "epoch": 0.060514372163388806,
+      "grad_norm": 0.2769795060157776,
+      "learning_rate": 9.911938687078324e-05,
+      "loss": 0.2728,
+      "num_input_tokens_seen": 1763840,
+      "step": 85
+    },
+    {
+      "epoch": 0.06407404111417637,
+      "grad_norm": 0.3135468363761902,
+      "learning_rate": 9.901178883227414e-05,
+      "loss": 0.2562,
+      "num_input_tokens_seen": 1885016,
+      "step": 90
+    },
+    {
+      "epoch": 0.06763371006496396,
+      "grad_norm": 0.3024243116378784,
+      "learning_rate": 9.889805597747588e-05,
+      "loss": 0.2564,
+      "num_input_tokens_seen": 2001056,
+      "step": 95
+    },
+    {
+      "epoch": 0.07119337901575154,
+      "grad_norm": 0.29715123772621155,
+      "learning_rate": 9.877820254235471e-05,
+      "loss": 0.2392,
+      "num_input_tokens_seen": 2076528,
+      "step": 100
+    },
+    {
+      "epoch": 0.07475304796653912,
+      "grad_norm": 0.3076399564743042,
+      "learning_rate": 9.865224352899119e-05,
+      "loss": 0.264,
+      "num_input_tokens_seen": 2210064,
+      "step": 105
+    },
+    {
+      "epoch": 0.07831271691732669,
+      "grad_norm": 0.2700069546699524,
+      "learning_rate": 9.852019470370253e-05,
+      "loss": 0.2483,
+      "num_input_tokens_seen": 2306936,
+      "step": 110
+    },
+    {
+      "epoch": 0.08187238586811427,
+      "grad_norm": 0.34573429822921753,
+      "learning_rate": 9.838207259506891e-05,
+      "loss": 0.2462,
+      "num_input_tokens_seen": 2413016,
+      "step": 115
+    },
+    {
+      "epoch": 0.08543205481890184,
+      "grad_norm": 0.29131779074668884,
+      "learning_rate": 9.82378944918648e-05,
+      "loss": 0.2403,
+      "num_input_tokens_seen": 2543504,
+      "step": 120
+    },
+    {
+      "epoch": 0.08899172376968942,
+      "grad_norm": 0.341266930103302,
+      "learning_rate": 9.80876784408948e-05,
+      "loss": 0.2276,
+      "num_input_tokens_seen": 2640240,
+      "step": 125
+    },
+    {
+      "epoch": 0.092551392720477,
+      "grad_norm": 0.31063246726989746,
+      "learning_rate": 9.793144324473473e-05,
+      "loss": 0.221,
+      "num_input_tokens_seen": 2743912,
+      "step": 130
+    },
+    {
+      "epoch": 0.09611106167126457,
+      "grad_norm": 0.3486584722995758,
+      "learning_rate": 9.776920845937816e-05,
+      "loss": 0.2547,
+      "num_input_tokens_seen": 2840632,
+      "step": 135
+    },
+    {
+      "epoch": 0.09967073062205214,
+      "grad_norm": 0.3306490182876587,
+      "learning_rate": 9.760099439178852e-05,
+      "loss": 0.2314,
+      "num_input_tokens_seen": 2928208,
+      "step": 140
+    },
+    {
+      "epoch": 0.10323039957283972,
+      "grad_norm": 0.37140771746635437,
+      "learning_rate": 9.742682209735727e-05,
+      "loss": 0.221,
+      "num_input_tokens_seen": 3033456,
+      "step": 145
+    },
+    {
+      "epoch": 0.1067900685236273,
+      "grad_norm": 0.31432077288627625,
+      "learning_rate": 9.724671337726854e-05,
+      "loss": 0.2158,
+      "num_input_tokens_seen": 3139640,
+      "step": 150
+    },
+    {
+      "epoch": 0.11034973747441489,
+      "grad_norm": 0.3638117015361786,
+      "learning_rate": 9.706069077577001e-05,
+      "loss": 0.2277,
+      "num_input_tokens_seen": 3240696,
+      "step": 155
+    },
+    {
+      "epoch": 0.11390940642520246,
+      "grad_norm": 0.2973380982875824,
+      "learning_rate": 9.686877757735127e-05,
+      "loss": 0.2287,
+      "num_input_tokens_seen": 3341736,
+      "step": 160
+    },
+    {
+      "epoch": 0.11746907537599004,
+      "grad_norm": 0.318136066198349,
+      "learning_rate": 9.66709978038292e-05,
+      "loss": 0.2201,
+      "num_input_tokens_seen": 3424968,
+      "step": 165
+    },
+    {
+      "epoch": 0.12102874432677761,
+      "grad_norm": 0.397605299949646,
+      "learning_rate": 9.646737621134112e-05,
+      "loss": 0.2175,
+      "num_input_tokens_seen": 3534184,
+      "step": 170
+    },
+    {
+      "epoch": 0.12458841327756519,
+      "grad_norm": 0.37626489996910095,
+      "learning_rate": 9.625793828724618e-05,
+      "loss": 0.2359,
+      "num_input_tokens_seen": 3631120,
+      "step": 175
+    },
+    {
+      "epoch": 0.12814808222835275,
+      "grad_norm": 0.441562294960022,
+      "learning_rate": 9.604271024693495e-05,
+      "loss": 0.2284,
+      "num_input_tokens_seen": 3736360,
+      "step": 180
+    },
+    {
+      "epoch": 0.13170775117914035,
+      "grad_norm": 0.3418191969394684,
+      "learning_rate": 9.582171903054816e-05,
+      "loss": 0.2125,
+      "num_input_tokens_seen": 3830560,
+      "step": 185
+    },
+    {
+      "epoch": 0.13526742012992793,
+      "grad_norm": 0.40202295780181885,
+      "learning_rate": 9.559499229960451e-05,
+      "loss": 0.2133,
+      "num_input_tokens_seen": 3920536,
+      "step": 190
+    },
+    {
+      "epoch": 0.1388270890807155,
+      "grad_norm": 0.29538485407829285,
+      "learning_rate": 9.536255843353832e-05,
+      "loss": 0.1935,
+      "num_input_tokens_seen": 4021840,
+      "step": 195
+    },
+    {
+      "epoch": 0.14238675803150308,
+      "grad_norm": 0.3438551425933838,
+      "learning_rate": 9.512444652614728e-05,
+      "loss": 0.2249,
+      "num_input_tokens_seen": 4123792,
+      "step": 200
+    },
+    {
+      "epoch": 0.14594642698229066,
+      "grad_norm": 0.31994742155075073,
+      "learning_rate": 9.48806863819507e-05,
+      "loss": 0.2117,
+      "num_input_tokens_seen": 4227968,
+      "step": 205
+    },
+    {
+      "epoch": 0.14950609593307823,
+      "grad_norm": 0.2823421359062195,
+      "learning_rate": 9.463130851245898e-05,
+      "loss": 0.2326,
+      "num_input_tokens_seen": 4337096,
+      "step": 210
+    },
+    {
+      "epoch": 0.1530657648838658,
+      "grad_norm": 0.36814814805984497,
+      "learning_rate": 9.437634413235436e-05,
+      "loss": 0.2211,
+      "num_input_tokens_seen": 4442592,
+      "step": 215
+    },
+    {
+      "epoch": 0.15662543383465338,
+      "grad_norm": 0.31317004561424255,
+      "learning_rate": 9.41158251555839e-05,
+      "loss": 0.1706,
+      "num_input_tokens_seen": 4540672,
+      "step": 220
+    },
+    {
+      "epoch": 0.16018510278544096,
+      "grad_norm": 0.34285885095596313,
+      "learning_rate": 9.384978419136468e-05,
+      "loss": 0.2111,
+      "num_input_tokens_seen": 4646952,
+      "step": 225
+    },
+    {
+      "epoch": 0.16374477173622853,
+      "grad_norm": 0.30709314346313477,
+      "learning_rate": 9.357825454010213e-05,
+      "loss": 0.2043,
+      "num_input_tokens_seen": 4748104,
+      "step": 230
+    },
+    {
+      "epoch": 0.1673044406870161,
+      "grad_norm": 0.331566721200943,
+      "learning_rate": 9.330127018922194e-05,
+      "loss": 0.2049,
+      "num_input_tokens_seen": 4857120,
+      "step": 235
+    },
+    {
+      "epoch": 0.17086410963780368,
+      "grad_norm": 0.34622254967689514,
+      "learning_rate": 9.301886580891562e-05,
+      "loss": 0.2129,
+      "num_input_tokens_seen": 4967288,
+      "step": 240
+    },
+    {
+      "epoch": 0.17442377858859126,
+      "grad_norm": 0.3246297240257263,
+      "learning_rate": 9.273107674780102e-05,
+      "loss": 0.1882,
+      "num_input_tokens_seen": 5077952,
+      "step": 245
+    },
+    {
+      "epoch": 0.17798344753937884,
+      "grad_norm": 0.3121411204338074,
+      "learning_rate": 9.243793902849763e-05,
+      "loss": 0.1998,
+      "num_input_tokens_seen": 5170960,
+      "step": 250
+    },
+    {
+      "epoch": 0.1815431164901664,
+      "grad_norm": 0.39862558245658875,
+      "learning_rate": 9.213948934311767e-05,
+      "loss": 0.1945,
+      "num_input_tokens_seen": 5281608,
+      "step": 255
+    },
+    {
+      "epoch": 0.185102785440954,
+      "grad_norm": 0.30746859312057495,
+      "learning_rate": 9.183576504867327e-05,
+      "loss": 0.1983,
+      "num_input_tokens_seen": 5398464,
+      "step": 260
+    },
+    {
+      "epoch": 0.18866245439174156,
+      "grad_norm": 0.3460409343242645,
+      "learning_rate": 9.152680416240059e-05,
+      "loss": 0.211,
+      "num_input_tokens_seen": 5503696,
+      "step": 265
+    },
+    {
+      "epoch": 0.19222212334252914,
+      "grad_norm": 0.42770832777023315,
+      "learning_rate": 9.121264535700107e-05,
+      "loss": 0.1943,
+      "num_input_tokens_seen": 5619400,
+      "step": 270
+    },
+    {
+      "epoch": 0.19578179229331671,
+      "grad_norm": 0.30582112073898315,
+      "learning_rate": 9.089332795580086e-05,
+      "loss": 0.186,
+      "num_input_tokens_seen": 5719208,
+      "step": 275
+    },
+    {
+      "epoch": 0.1993414612441043,
+      "grad_norm": 0.38887447118759155,
+      "learning_rate": 9.056889192782866e-05,
+      "loss": 0.2011,
+      "num_input_tokens_seen": 5814880,
+      "step": 280
+    },
+    {
+      "epoch": 0.20290113019489187,
+      "grad_norm": 0.4006999731063843,
+      "learning_rate": 9.023937788281278e-05,
+      "loss": 0.1941,
+      "num_input_tokens_seen": 5910072,
+      "step": 285
+    },
+    {
+      "epoch": 0.20646079914567944,
+      "grad_norm": 0.373165100812912,
+      "learning_rate": 8.990482706609805e-05,
+      "loss": 0.1893,
+      "num_input_tokens_seen": 6022608,
+      "step": 290
+    },
+    {
+      "epoch": 0.21002046809646702,
+      "grad_norm": 0.3396100699901581,
+      "learning_rate": 8.95652813534831e-05,
+      "loss": 0.1945,
+      "num_input_tokens_seen": 6135608,
+      "step": 295
+    },
+    {
+      "epoch": 0.2135801370472546,
+      "grad_norm": 0.4196370542049408,
+      "learning_rate": 8.922078324597879e-05,
+      "loss": 0.2082,
+      "num_input_tokens_seen": 6241088,
+      "step": 300
+    },
+    {
+      "epoch": 0.2171398059980422,
+      "grad_norm": 0.39760822057724,
+      "learning_rate": 8.88713758644883e-05,
+      "loss": 0.1875,
+      "num_input_tokens_seen": 6353520,
+      "step": 305
+    },
+    {
+      "epoch": 0.22069947494882977,
+      "grad_norm": 0.30599263310432434,
+      "learning_rate": 8.851710294440973e-05,
+      "loss": 0.1876,
+      "num_input_tokens_seen": 6454480,
+      "step": 310
+    },
+    {
+      "epoch": 0.22425914389961735,
+      "grad_norm": 0.4323524832725525,
+      "learning_rate": 8.815800883016168e-05,
+      "loss": 0.2053,
+      "num_input_tokens_seen": 6563816,
+      "step": 315
+    },
+    {
+      "epoch": 0.22781881285040492,
+      "grad_norm": 0.3267993927001953,
+      "learning_rate": 8.779413846963267e-05,
+      "loss": 0.1994,
+      "num_input_tokens_seen": 6658448,
+      "step": 320
+    },
+    {
+      "epoch": 0.2313784818011925,
+      "grad_norm": 0.372569739818573,
+      "learning_rate": 8.742553740855506e-05,
+      "loss": 0.1768,
+      "num_input_tokens_seen": 6772528,
+      "step": 325
+    },
+    {
+      "epoch": 0.23493815075198007,
+      "grad_norm": 0.3330164849758148,
+      "learning_rate": 8.705225178480398e-05,
+      "loss": 0.1916,
+      "num_input_tokens_seen": 6875864,
+      "step": 330
+    },
+    {
+      "epoch": 0.23849781970276765,
+      "grad_norm": 0.4003218710422516,
+      "learning_rate": 8.66743283226223e-05,
+      "loss": 0.2065,
+      "num_input_tokens_seen": 6972880,
+      "step": 335
+    },
+    {
+      "epoch": 0.24205748865355523,
+      "grad_norm": 0.3475089371204376,
+      "learning_rate": 8.629181432677213e-05,
+      "loss": 0.2037,
+      "num_input_tokens_seen": 7091480,
+      "step": 340
+    },
+    {
+      "epoch": 0.2456171576043428,
+      "grad_norm": 0.29793596267700195,
+      "learning_rate": 8.59047576766137e-05,
+      "loss": 0.1915,
+      "num_input_tokens_seen": 7198744,
+      "step": 345
+    },
+    {
+      "epoch": 0.24917682655513038,
+      "grad_norm": 0.39443179965019226,
+      "learning_rate": 8.551320682011228e-05,
+      "loss": 0.1884,
+      "num_input_tokens_seen": 7306808,
+      "step": 350
+    },
+    {
+      "epoch": 0.2527364955059179,
+      "grad_norm": 0.37115100026130676,
+      "learning_rate": 8.511721076777389e-05,
+      "loss": 0.1932,
+      "num_input_tokens_seen": 7414544,
+      "step": 355
+    },
+    {
+      "epoch": 0.2562961644567055,
+      "grad_norm": 0.3560400903224945,
+      "learning_rate": 8.471681908651067e-05,
+      "loss": 0.1879,
+      "num_input_tokens_seen": 7526376,
+      "step": 360
+    },
+    {
+      "epoch": 0.25985583340749313,
+      "grad_norm": 0.36505258083343506,
+      "learning_rate": 8.43120818934367e-05,
+      "loss": 0.1882,
+      "num_input_tokens_seen": 7622440,
+      "step": 365
+    },
+    {
+      "epoch": 0.2634155023582807,
+      "grad_norm": 0.3832385241985321,
+      "learning_rate": 8.390304984959454e-05,
+      "loss": 0.1955,
+      "num_input_tokens_seen": 7729176,
+      "step": 370
+    },
+    {
+      "epoch": 0.2669751713090683,
+      "grad_norm": 0.32355767488479614,
+      "learning_rate": 8.348977415361434e-05,
+      "loss": 0.1977,
+      "num_input_tokens_seen": 7846344,
+      "step": 375
+    },
+    {
+      "epoch": 0.27053484025985586,
+      "grad_norm": 0.31665217876434326,
+      "learning_rate": 8.3072306535305e-05,
+      "loss": 0.1944,
+      "num_input_tokens_seen": 7945632,
+      "step": 380
+    },
+    {
+      "epoch": 0.27409450921064343,
+      "grad_norm": 0.3798852562904358,
+      "learning_rate": 8.265069924917925e-05,
+      "loss": 0.1926,
+      "num_input_tokens_seen": 8048344,
+      "step": 385
+    },
+    {
+      "epoch": 0.277654178161431,
+      "grad_norm": 0.3776335120201111,
+      "learning_rate": 8.222500506791304e-05,
+      "loss": 0.2007,
+      "num_input_tokens_seen": 8154184,
+      "step": 390
+    },
+    {
+      "epoch": 0.2812138471122186,
+      "grad_norm": 0.3042481243610382,
+      "learning_rate": 8.179527727573975e-05,
+      "loss": 0.1763,
+      "num_input_tokens_seen": 8268128,
+      "step": 395
+    },
+    {
+      "epoch": 0.28477351606300616,
+      "grad_norm": 0.3481425940990448,
+      "learning_rate": 8.136156966178081e-05,
+      "loss": 0.1897,
+      "num_input_tokens_seen": 8366648,
+      "step": 400
+    },
+    {
+      "epoch": 0.28833318501379374,
+      "grad_norm": 0.345595121383667,
+      "learning_rate": 8.092393651331275e-05,
+      "loss": 0.1998,
+      "num_input_tokens_seen": 8473872,
+      "step": 405
+    },
+    {
+      "epoch": 0.2918928539645813,
+      "grad_norm": 0.30839142203330994,
+      "learning_rate": 8.048243260897217e-05,
+      "loss": 0.1844,
+      "num_input_tokens_seen": 8575560,
+      "step": 410
+    },
+    {
+      "epoch": 0.2954525229153689,
+      "grad_norm": 0.3298119008541107,
+      "learning_rate": 8.003711321189895e-05,
+      "loss": 0.1816,
+      "num_input_tokens_seen": 8692528,
+      "step": 415
+    },
+    {
+      "epoch": 0.29901219186615646,
+      "grad_norm": 0.3186253607273102,
+      "learning_rate": 7.95880340628191e-05,
+      "loss": 0.1894,
+      "num_input_tokens_seen": 8792544,
+      "step": 420
+    },
+    {
+      "epoch": 0.30257186081694404,
+      "grad_norm": 0.3161162734031677,
+      "learning_rate": 7.913525137306756e-05,
+      "loss": 0.1859,
+      "num_input_tokens_seen": 8911488,
+      "step": 425
+    },
+    {
+      "epoch": 0.3061315297677316,
+      "grad_norm": 0.36929574608802795,
+      "learning_rate": 7.86788218175523e-05,
+      "loss": 0.1867,
+      "num_input_tokens_seen": 9015008,
+      "step": 430
+    },
+    {
+      "epoch": 0.3096911987185192,
+      "grad_norm": 0.5688495635986328,
+      "learning_rate": 7.821880252766025e-05,
+      "loss": 0.2022,
+      "num_input_tokens_seen": 9115512,
+      "step": 435
+    },
+    {
+      "epoch": 0.31325086766930677,
+      "grad_norm": 0.3410090506076813,
+      "learning_rate": 7.775525108410615e-05,
+      "loss": 0.1674,
+      "num_input_tokens_seen": 9216312,
+      "step": 440
+    },
+    {
+      "epoch": 0.31681053662009434,
+      "grad_norm": 0.41320088505744934,
+      "learning_rate": 7.728822550972523e-05,
+      "loss": 0.1834,
+      "num_input_tokens_seen": 9319432,
+      "step": 445
+    },
+    {
+      "epoch": 0.3203702055708819,
+      "grad_norm": 0.3464307487010956,
+      "learning_rate": 7.681778426221042e-05,
+      "loss": 0.1671,
+      "num_input_tokens_seen": 9426264,
+      "step": 450
+    },
+    {
+      "epoch": 0.3239298745216695,
+      "grad_norm": 0.42251551151275635,
+      "learning_rate": 7.634398622679517e-05,
+      "loss": 0.1873,
+      "num_input_tokens_seen": 9536096,
+      "step": 455
+    },
+    {
+      "epoch": 0.32748954347245707,
+      "grad_norm": 0.41400083899497986,
+      "learning_rate": 7.586689070888284e-05,
+      "loss": 0.1915,
+      "num_input_tokens_seen": 9648984,
+      "step": 460
+    },
+    {
+      "epoch": 0.33104921242324464,
+      "grad_norm": 0.35248515009880066,
+      "learning_rate": 7.53865574266234e-05,
+      "loss": 0.1822,
+      "num_input_tokens_seen": 9772136,
+      "step": 465
+    },
+    {
+      "epoch": 0.3346088813740322,
+      "grad_norm": 0.43567249178886414,
+      "learning_rate": 7.490304650343841e-05,
+      "loss": 0.185,
+      "num_input_tokens_seen": 9873504,
+      "step": 470
+    },
+    {
+      "epoch": 0.3381685503248198,
+      "grad_norm": 0.4128562808036804,
+      "learning_rate": 7.441641846049556e-05,
+      "loss": 0.1794,
+      "num_input_tokens_seen": 9994952,
+      "step": 475
+    },
+    {
+      "epoch": 0.34172821927560737,
+      "grad_norm": 0.45978060364723206,
+      "learning_rate": 7.3926734209133e-05,
+      "loss": 0.1694,
+      "num_input_tokens_seen": 10096568,
+      "step": 480
+    },
+    {
+      "epoch": 0.34528788822639495,
+      "grad_norm": 0.42638832330703735,
+      "learning_rate": 7.343405504323519e-05,
+      "loss": 0.172,
+      "num_input_tokens_seen": 10193976,
+      "step": 485
+    },
+    {
+      "epoch": 0.3488475571771825,
+      "grad_norm": 0.31733188033103943,
+      "learning_rate": 7.293844263156072e-05,
+      "loss": 0.1722,
+      "num_input_tokens_seen": 10314512,
+      "step": 490
+    },
+    {
+      "epoch": 0.3524072261279701,
+      "grad_norm": 0.4016326665878296,
+      "learning_rate": 7.243995901002312e-05,
+      "loss": 0.1932,
+      "num_input_tokens_seen": 10420280,
+      "step": 495
+    },
+    {
+      "epoch": 0.3559668950787577,
+      "grad_norm": 0.30861759185791016,
+      "learning_rate": 7.193866657392597e-05,
+      "loss": 0.1778,
+      "num_input_tokens_seen": 10512840,
+      "step": 500
+    },
+    {
+      "epoch": 0.35952656402954525,
+      "grad_norm": 0.358877032995224,
+      "learning_rate": 7.143462807015271e-05,
+      "loss": 0.1595,
+      "num_input_tokens_seen": 10601008,
+      "step": 505
+    },
+    {
+      "epoch": 0.3630862329803328,
+      "grad_norm": 0.34301072359085083,
+      "learning_rate": 7.092790658931273e-05,
+      "loss": 0.163,
+      "num_input_tokens_seen": 10686288,
+      "step": 510
+    },
+    {
+      "epoch": 0.3666459019311204,
+      "grad_norm": 0.32205110788345337,
+      "learning_rate": 7.041856555784421e-05,
+      "loss": 0.181,
+      "num_input_tokens_seen": 10792520,
+      "step": 515
+    },
+    {
+      "epoch": 0.370205570881908,
+      "grad_norm": 0.44235751032829285,
+      "learning_rate": 6.990666873007505e-05,
+      "loss": 0.1762,
+      "num_input_tokens_seen": 10885304,
+      "step": 520
+    },
+    {
+      "epoch": 0.37376523983269555,
+      "grad_norm": 0.36452656984329224,
+      "learning_rate": 6.939228018024275e-05,
+      "loss": 0.1647,
+      "num_input_tokens_seen": 10993776,
+      "step": 525
+    },
+    {
+      "epoch": 0.3773249087834831,
+      "grad_norm": 0.5452892184257507,
+      "learning_rate": 6.887546429447419e-05,
+      "loss": 0.1669,
+      "num_input_tokens_seen": 11110832,
+      "step": 530
+    },
+    {
+      "epoch": 0.3808845777342707,
+      "grad_norm": 0.517644464969635,
+      "learning_rate": 6.835628576272638e-05,
+      "loss": 0.1873,
+      "num_input_tokens_seen": 11201704,
+      "step": 535
+    },
+    {
+      "epoch": 0.3844442466850583,
+      "grad_norm": 0.3548191785812378,
+      "learning_rate": 6.783480957068934e-05,
+      "loss": 0.1711,
+      "num_input_tokens_seen": 11293424,
+      "step": 540
+    },
+    {
+      "epoch": 0.38800391563584585,
+      "grad_norm": 0.3615559935569763,
+      "learning_rate": 6.731110099165164e-05,
+      "loss": 0.1764,
+      "num_input_tokens_seen": 11394280,
+      "step": 545
+    },
+    {
+      "epoch": 0.39156358458663343,
+      "grad_norm": 0.4043658375740051,
+      "learning_rate": 6.678522557833024e-05,
+      "loss": 0.184,
+      "num_input_tokens_seen": 11477712,
+      "step": 550
+    },
+    {
+      "epoch": 0.395123253537421,
+      "grad_norm": 0.3802790343761444,
+      "learning_rate": 6.625724915466526e-05,
+      "loss": 0.1681,
+      "num_input_tokens_seen": 11593608,
+      "step": 555
+    },
+    {
+      "epoch": 0.3986829224882086,
+      "grad_norm": 0.5214163661003113,
+      "learning_rate": 6.572723780758069e-05,
+      "loss": 0.1764,
+      "num_input_tokens_seen": 11706168,
+      "step": 560
+    },
+    {
+      "epoch": 0.40224259143899616,
+      "grad_norm": 0.4910968840122223,
+      "learning_rate": 6.519525787871235e-05,
+      "loss": 0.1841,
+      "num_input_tokens_seen": 11836904,
+      "step": 565
+    },
+    {
+      "epoch": 0.40580226038978373,
+      "grad_norm": 0.508994996547699,
+      "learning_rate": 6.466137595610388e-05,
+      "loss": 0.1755,
+      "num_input_tokens_seen": 11952440,
+      "step": 570
+    },
+    {
+      "epoch": 0.4093619293405713,
+      "grad_norm": 0.3509501516819,
+      "learning_rate": 6.412565886587185e-05,
+      "loss": 0.1661,
+      "num_input_tokens_seen": 12050640,
+      "step": 575
+    },
+    {
+      "epoch": 0.4129215982913589,
+      "grad_norm": 0.43870943784713745,
+      "learning_rate": 6.358817366384122e-05,
+      "loss": 0.1835,
+      "num_input_tokens_seen": 12159120,
+      "step": 580
+    },
+    {
+      "epoch": 0.41648126724214646,
+      "grad_norm": 0.45170462131500244,
+      "learning_rate": 6.304898762715186e-05,
+      "loss": 0.1668,
+      "num_input_tokens_seen": 12264824,
+      "step": 585
+    },
+    {
+      "epoch": 0.42004093619293403,
+      "grad_norm": 0.3374248445034027,
+      "learning_rate": 6.250816824583747e-05,
+      "loss": 0.1731,
+      "num_input_tokens_seen": 12382032,
+      "step": 590
+    },
+    {
+      "epoch": 0.4236006051437216,
+      "grad_norm": 0.3665522336959839,
+      "learning_rate": 6.19657832143779e-05,
+      "loss": 0.1558,
+      "num_input_tokens_seen": 12494104,
+      "step": 595
+    },
+    {
+      "epoch": 0.4271602740945092,
+      "grad_norm": 0.32165008783340454,
+      "learning_rate": 6.142190042322569e-05,
+      "loss": 0.1763,
+      "num_input_tokens_seen": 12609472,
+      "step": 600
+    },
+    {
+      "epoch": 0.43071994304529676,
+      "grad_norm": 0.47890809178352356,
+      "learning_rate": 6.087658795030837e-05,
+      "loss": 0.1795,
+      "num_input_tokens_seen": 12723800,
+      "step": 605
+    },
+    {
+      "epoch": 0.4342796119960844,
+      "grad_norm": 0.3581695556640625,
+      "learning_rate": 6.032991405250702e-05,
+      "loss": 0.1658,
+      "num_input_tokens_seen": 12812152,
+      "step": 610
+    },
+    {
+      "epoch": 0.43783928094687197,
+      "grad_norm": 0.3896292746067047,
+      "learning_rate": 5.9781947157112536e-05,
+      "loss": 0.1781,
+      "num_input_tokens_seen": 12913336,
+      "step": 615
+    },
+    {
+      "epoch": 0.44139894989765954,
+      "grad_norm": 0.3902624845504761,
+      "learning_rate": 5.9232755853260635e-05,
+      "loss": 0.1712,
+      "num_input_tokens_seen": 13000856,
+      "step": 620
+    },
+    {
+      "epoch": 0.4449586188484471,
+      "grad_norm": 0.5472461581230164,
+      "learning_rate": 5.868240888334653e-05,
+      "loss": 0.1803,
+      "num_input_tokens_seen": 13110056,
+      "step": 625
+    },
+    {
+      "epoch": 0.4485182877992347,
+      "grad_norm": 0.364450067281723,
+      "learning_rate": 5.813097513442035e-05,
+      "loss": 0.1697,
+      "num_input_tokens_seen": 13199584,
+      "step": 630
+    },
+    {
+      "epoch": 0.45207795675002227,
+      "grad_norm": 0.4172821342945099,
+      "learning_rate": 5.757852362956463e-05,
+      "loss": 0.1652,
+      "num_input_tokens_seen": 13300400,
+      "step": 635
+    },
+    {
+      "epoch": 0.45563762570080985,
+      "grad_norm": 0.5252318382263184,
+      "learning_rate": 5.702512351925464e-05,
+      "loss": 0.1757,
+      "num_input_tokens_seen": 13414232,
+      "step": 640
+    },
+    {
+      "epoch": 0.4591972946515974,
+      "grad_norm": 0.3265618681907654,
+      "learning_rate": 5.6470844072702764e-05,
+      "loss": 0.1671,
+      "num_input_tokens_seen": 13508520,
+      "step": 645
+    },
+    {
+      "epoch": 0.462756963602385,
+      "grad_norm": 0.3361479938030243,
+      "learning_rate": 5.591575466918816e-05,
+      "loss": 0.1757,
+      "num_input_tokens_seen": 13630000,
+      "step": 650
+    },
+    {
+      "epoch": 0.46631663255317257,
+      "grad_norm": 0.39177238941192627,
+      "learning_rate": 5.5359924789372396e-05,
+      "loss": 0.1772,
+      "num_input_tokens_seen": 13746784,
+      "step": 655
+    },
+    {
+      "epoch": 0.46987630150396015,
+      "grad_norm": 0.3369304835796356,
+      "learning_rate": 5.480342400660268e-05,
+      "loss": 0.1642,
+      "num_input_tokens_seen": 13847368,
+      "step": 660
+    },
+    {
+      "epoch": 0.4734359704547477,
+      "grad_norm": 0.3846571445465088,
+      "learning_rate": 5.424632197820324e-05,
+      "loss": 0.1684,
+      "num_input_tokens_seen": 13947616,
+      "step": 665
+    },
+    {
+      "epoch": 0.4769956394055353,
+      "grad_norm": 0.4429912269115448,
+      "learning_rate": 5.368868843675642e-05,
+      "loss": 0.188,
+      "num_input_tokens_seen": 14033504,
+      "step": 670
+    },
+    {
+      "epoch": 0.4805553083563229,
+      "grad_norm": 0.4972701370716095,
+      "learning_rate": 5.31305931813741e-05,
+      "loss": 0.164,
+      "num_input_tokens_seen": 14128176,
+      "step": 675
+    },
+    {
+      "epoch": 0.48411497730711045,
+      "grad_norm": 0.38014113903045654,
+      "learning_rate": 5.2572106068961026e-05,
+      "loss": 0.1778,
+      "num_input_tokens_seen": 14226296,
+      "step": 680
+    },
+    {
+      "epoch": 0.487674646257898,
+      "grad_norm": 0.4327758550643921,
+      "learning_rate": 5.201329700547076e-05,
+      "loss": 0.1822,
+      "num_input_tokens_seen": 14341248,
+      "step": 685
+    },
+    {
+      "epoch": 0.4912343152086856,
+      "grad_norm": 0.3764859735965729,
+      "learning_rate": 5.145423593715557e-05,
+      "loss": 0.1544,
+      "num_input_tokens_seen": 14454128,
+      "step": 690
+    },
+    {
+      "epoch": 0.4947939841594732,
+      "grad_norm": 0.4072837829589844,
+      "learning_rate": 5.089499284181122e-05,
+      "loss": 0.1521,
+      "num_input_tokens_seen": 14558352,
+      "step": 695
+    },
+    {
+      "epoch": 0.49835365311026075,
+      "grad_norm": 0.45584630966186523,
+      "learning_rate": 5.0335637720017817e-05,
+      "loss": 0.1549,
+      "num_input_tokens_seen": 14632632,
+      "step": 700
+    },
+    {
+      "epoch": 0.5019133220610483,
+      "grad_norm": 0.4504990577697754,
+      "learning_rate": 4.977624058637783e-05,
+      "loss": 0.1645,
+      "num_input_tokens_seen": 14733712,
+      "step": 705
+    },
+    {
+      "epoch": 0.5054729910118358,
+      "grad_norm": 0.3279663026332855,
+      "learning_rate": 4.921687146075244e-05,
+      "loss": 0.1786,
+      "num_input_tokens_seen": 14856480,
+      "step": 710
+    },
+    {
+      "epoch": 0.5090326599626235,
+      "grad_norm": 0.43772807717323303,
+      "learning_rate": 4.865760035949695e-05,
+      "loss": 0.1683,
+      "num_input_tokens_seen": 14963032,
+      "step": 715
+    },
+    {
+      "epoch": 0.512592328913411,
+      "grad_norm": 0.41034719347953796,
+      "learning_rate": 4.809849728669702e-05,
+      "loss": 0.1737,
+      "num_input_tokens_seen": 15056864,
+      "step": 720
+    },
+    {
+      "epoch": 0.5161519978641986,
+      "grad_norm": 0.35806897282600403,
+      "learning_rate": 4.7539632225406095e-05,
+      "loss": 0.1635,
+      "num_input_tokens_seen": 15169176,
+      "step": 725
+    },
+    {
+      "epoch": 0.5197116668149863,
+      "grad_norm": 0.410163015127182,
+      "learning_rate": 4.6981075128885693e-05,
+      "loss": 0.1682,
+      "num_input_tokens_seen": 15264616,
+      "step": 730
+    },
+    {
+      "epoch": 0.5232713357657738,
+      "grad_norm": 0.4012819826602936,
+      "learning_rate": 4.642289591184934e-05,
+      "loss": 0.1834,
+      "num_input_tokens_seen": 15375512,
+      "step": 735
+    },
+    {
+      "epoch": 0.5268310047165614,
+      "grad_norm": 0.35860753059387207,
+      "learning_rate": 4.586516444171122e-05,
+      "loss": 0.1708,
+      "num_input_tokens_seen": 15468760,
+      "step": 740
+    },
+    {
+      "epoch": 0.5303906736673489,
+      "grad_norm": 0.4758777916431427,
+      "learning_rate": 4.530795052984104e-05,
+      "loss": 0.1738,
+      "num_input_tokens_seen": 15570704,
+      "step": 745
+    },
+    {
+      "epoch": 0.5339503426181366,
+      "grad_norm": 0.37693336606025696,
+      "learning_rate": 4.475132392282556e-05,
+      "loss": 0.1647,
+      "num_input_tokens_seen": 15668584,
+      "step": 750
+    },
+    {
+      "epoch": 0.5375100115689241,
+      "grad_norm": 0.34583452343940735,
+      "learning_rate": 4.4195354293738484e-05,
+      "loss": 0.1557,
+      "num_input_tokens_seen": 15746296,
+      "step": 755
+    },
+    {
+      "epoch": 0.5410696805197117,
+      "grad_norm": 0.36316967010498047,
+      "learning_rate": 4.364011123341947e-05,
+      "loss": 0.1651,
+      "num_input_tokens_seen": 15868760,
+      "step": 760
+    },
+    {
+      "epoch": 0.5446293494704992,
+      "grad_norm": 0.46182507276535034,
+      "learning_rate": 4.308566424176336e-05,
+      "loss": 0.1691,
+      "num_input_tokens_seen": 15961016,
+      "step": 765
+    },
+    {
+      "epoch": 0.5481890184212869,
+      "grad_norm": 0.4366365075111389,
+      "learning_rate": 4.253208271902091e-05,
+      "loss": 0.1762,
+      "num_input_tokens_seen": 16068296,
+      "step": 770
+    },
+    {
+      "epoch": 0.5517486873720744,
+      "grad_norm": 0.3218953311443329,
+      "learning_rate": 4.197943595711198e-05,
+      "loss": 0.1603,
+      "num_input_tokens_seen": 16189096,
+      "step": 775
+    },
+    {
+      "epoch": 0.555308356322862,
+      "grad_norm": 0.48252201080322266,
+      "learning_rate": 4.142779313095223e-05,
+      "loss": 0.1627,
+      "num_input_tokens_seen": 16303344,
+      "step": 780
+    },
+    {
+      "epoch": 0.5588680252736495,
+      "grad_norm": 0.5327666401863098,
+      "learning_rate": 4.087722328979438e-05,
+      "loss": 0.1752,
+      "num_input_tokens_seen": 16413888,
+      "step": 785
+    },
+    {
+      "epoch": 0.5624276942244372,
+      "grad_norm": 0.4146188795566559,
+      "learning_rate": 4.032779534858544e-05,
+      "loss": 0.156,
+      "num_input_tokens_seen": 16520296,
+      "step": 790
+    },
+    {
+      "epoch": 0.5659873631752247,
+      "grad_norm": 0.4379608631134033,
+      "learning_rate": 3.9779578079340554e-05,
+      "loss": 0.176,
+      "num_input_tokens_seen": 16622024,
+      "step": 795
+    },
+    {
+      "epoch": 0.5695470321260123,
+      "grad_norm": 0.6401222944259644,
+      "learning_rate": 3.9232640102534786e-05,
+      "loss": 0.1524,
+      "num_input_tokens_seen": 16734184,
+      "step": 800
+    },
+    {
+      "epoch": 0.5731067010767998,
+      "grad_norm": 0.48411625623703003,
+      "learning_rate": 3.86870498785139e-05,
+      "loss": 0.171,
+      "num_input_tokens_seen": 16828056,
+      "step": 805
+    },
+    {
+      "epoch": 0.5766663700275875,
+      "grad_norm": 0.3455513119697571,
+      "learning_rate": 3.814287569892512e-05,
+      "loss": 0.1599,
+      "num_input_tokens_seen": 16937648,
+      "step": 810
+    },
+    {
+      "epoch": 0.580226038978375,
+      "grad_norm": 0.35831761360168457,
+      "learning_rate": 3.760018567816908e-05,
+      "loss": 0.1694,
+      "num_input_tokens_seen": 17043496,
+      "step": 815
+    },
+    {
+      "epoch": 0.5837857079291626,
+      "grad_norm": 0.4048117399215698,
+      "learning_rate": 3.705904774487396e-05,
+      "loss": 0.157,
+      "num_input_tokens_seen": 17141592,
+      "step": 820
+    },
+    {
+      "epoch": 0.5873453768799501,
+      "grad_norm": 0.4098561704158783,
+      "learning_rate": 3.651952963339282e-05,
+      "loss": 0.1624,
+      "num_input_tokens_seen": 17260592,
+      "step": 825
+    },
+    {
+      "epoch": 0.5909050458307378,
+      "grad_norm": 0.5280525088310242,
+      "learning_rate": 3.598169887532521e-05,
+      "loss": 0.1637,
+      "num_input_tokens_seen": 17352008,
+      "step": 830
+    },
+    {
+      "epoch": 0.5944647147815253,
+      "grad_norm": 0.3961341083049774,
+      "learning_rate": 3.5445622791064356e-05,
+      "loss": 0.1622,
+      "num_input_tokens_seen": 17484016,
+      "step": 835
+    },
+    {
+      "epoch": 0.5980243837323129,
+      "grad_norm": 0.3392258584499359,
+      "learning_rate": 3.491136848137053e-05,
+      "loss": 0.1614,
+      "num_input_tokens_seen": 17569392,
+      "step": 840
+    },
+    {
+      "epoch": 0.6015840526831004,
+      "grad_norm": 0.41437968611717224,
+      "learning_rate": 3.4379002818972124e-05,
+      "loss": 0.1607,
+      "num_input_tokens_seen": 17660056,
+      "step": 845
+    },
+    {
+      "epoch": 0.6051437216338881,
+      "grad_norm": 0.3347319960594177,
+      "learning_rate": 3.384859244019511e-05,
+      "loss": 0.1549,
+      "num_input_tokens_seen": 17753416,
+      "step": 850
+    },
+    {
+      "epoch": 0.6087033905846756,
+      "grad_norm": 0.41625234484672546,
+      "learning_rate": 3.3320203736622184e-05,
+      "loss": 0.1621,
+      "num_input_tokens_seen": 17850256,
+      "step": 855
+    },
+    {
+      "epoch": 0.6122630595354632,
+      "grad_norm": 0.3651588261127472,
+      "learning_rate": 3.2793902846782534e-05,
+      "loss": 0.1651,
+      "num_input_tokens_seen": 17947344,
+      "step": 860
+    },
+    {
+      "epoch": 0.6158227284862507,
+      "grad_norm": 0.4354760944843292,
+      "learning_rate": 3.226975564787322e-05,
+      "loss": 0.1718,
+      "num_input_tokens_seen": 18052208,
+      "step": 865
+    },
+    {
+      "epoch": 0.6193823974370384,
+      "grad_norm": 0.41476985812187195,
+      "learning_rate": 3.174782774751338e-05,
+      "loss": 0.1567,
+      "num_input_tokens_seen": 18180416,
+      "step": 870
+    },
+    {
+      "epoch": 0.6229420663878259,
+      "grad_norm": 0.42029616236686707,
+      "learning_rate": 3.122818447553201e-05,
+      "loss": 0.1649,
+      "num_input_tokens_seen": 18285576,
+      "step": 875
+    },
+    {
+      "epoch": 0.6265017353386135,
+      "grad_norm": 0.3612583577632904,
+      "learning_rate": 3.071089087579074e-05,
+      "loss": 0.1486,
+      "num_input_tokens_seen": 18382008,
+      "step": 880
+    },
+    {
+      "epoch": 0.630061404289401,
+      "grad_norm": 0.3782946467399597,
+      "learning_rate": 3.019601169804216e-05,
+      "loss": 0.1498,
+      "num_input_tokens_seen": 18463520,
+      "step": 885
+    },
+    {
+      "epoch": 0.6336210732401887,
+      "grad_norm": 0.5428029894828796,
+      "learning_rate": 2.9683611389825167e-05,
+      "loss": 0.1605,
+      "num_input_tokens_seen": 18573728,
+      "step": 890
+    },
+    {
+      "epoch": 0.6371807421909762,
+      "grad_norm": 0.3950955271720886,
+      "learning_rate": 2.917375408839803e-05,
+      "loss": 0.1647,
+      "num_input_tokens_seen": 18670936,
+      "step": 895
+    },
+    {
+      "epoch": 0.6407404111417638,
+      "grad_norm": 0.515082597732544,
+      "learning_rate": 2.8666503612710226e-05,
+      "loss": 0.1499,
+      "num_input_tokens_seen": 18762568,
+      "step": 900
+    },
+    {
+      "epoch": 0.6443000800925514,
+      "grad_norm": 0.40842828154563904,
+      "learning_rate": 2.8161923455414367e-05,
+      "loss": 0.1505,
+      "num_input_tokens_seen": 18865968,
+      "step": 905
+    },
+    {
+      "epoch": 0.647859749043339,
+      "grad_norm": 0.3697022795677185,
+      "learning_rate": 2.7660076774918708e-05,
+      "loss": 0.1681,
+      "num_input_tokens_seen": 18981472,
+      "step": 910
+    },
+    {
+      "epoch": 0.6514194179941265,
+      "grad_norm": 0.5153518319129944,
+      "learning_rate": 2.7161026387481636e-05,
+      "loss": 0.1698,
+      "num_input_tokens_seen": 19079608,
+      "step": 915
+    },
+    {
+      "epoch": 0.6549790869449141,
+      "grad_norm": 0.6159951686859131,
+      "learning_rate": 2.666483475934885e-05,
+      "loss": 0.1589,
+      "num_input_tokens_seen": 19168792,
+      "step": 920
+    },
+    {
+      "epoch": 0.6585387558957017,
+      "grad_norm": 0.4294576644897461,
+      "learning_rate": 2.6171563998934605e-05,
+      "loss": 0.1779,
+      "num_input_tokens_seen": 19292088,
+      "step": 925
+    },
+    {
+      "epoch": 0.6620984248464893,
+      "grad_norm": 0.4489574730396271,
+      "learning_rate": 2.5681275849047482e-05,
+      "loss": 0.1797,
+      "num_input_tokens_seen": 19393184,
+      "step": 930
+    },
+    {
+      "epoch": 0.6656580937972768,
+      "grad_norm": 0.38424602150917053,
+      "learning_rate": 2.5194031679162067e-05,
+      "loss": 0.1585,
+      "num_input_tokens_seen": 19486688,
+      "step": 935
+    },
+    {
+      "epoch": 0.6692177627480644,
+      "grad_norm": 0.3428622782230377,
+      "learning_rate": 2.4709892477737262e-05,
+      "loss": 0.1527,
+      "num_input_tokens_seen": 19583928,
+      "step": 940
+    },
+    {
+      "epoch": 0.672777431698852,
+      "grad_norm": 0.3572031855583191,
+      "learning_rate": 2.422891884458241e-05,
+      "loss": 0.163,
+      "num_input_tokens_seen": 19682760,
+      "step": 945
+    },
+    {
+      "epoch": 0.6763371006496396,
+      "grad_norm": 0.4243921637535095,
+      "learning_rate": 2.3751170983272e-05,
+      "loss": 0.1659,
+      "num_input_tokens_seen": 19784712,
+      "step": 950
+    },
+    {
+      "epoch": 0.6798967696004271,
+      "grad_norm": 0.4836697578430176,
+      "learning_rate": 2.3276708693609943e-05,
+      "loss": 0.1662,
+      "num_input_tokens_seen": 19890408,
+      "step": 955
+    },
+    {
+      "epoch": 0.6834564385512147,
+      "grad_norm": 0.3940310478210449,
+      "learning_rate": 2.2805591364144447e-05,
+      "loss": 0.1689,
+      "num_input_tokens_seen": 19988040,
+      "step": 960
+    },
+    {
+      "epoch": 0.6870161075020023,
+      "grad_norm": 0.41803377866744995,
+      "learning_rate": 2.233787796473432e-05,
+      "loss": 0.1497,
+      "num_input_tokens_seen": 20074552,
+      "step": 965
+    },
+    {
+      "epoch": 0.6905757764527899,
+      "grad_norm": 0.37162527441978455,
+      "learning_rate": 2.187362703916766e-05,
+      "loss": 0.1631,
+      "num_input_tokens_seen": 20181064,
+      "step": 970
+    },
+    {
+      "epoch": 0.6941354454035775,
+      "grad_norm": 0.37610870599746704,
+      "learning_rate": 2.141289669783401e-05,
+      "loss": 0.1752,
+      "num_input_tokens_seen": 20293072,
+      "step": 975
+    },
+    {
+      "epoch": 0.697695114354365,
+      "grad_norm": 0.39441388845443726,
+      "learning_rate": 2.0955744610450618e-05,
+      "loss": 0.1575,
+      "num_input_tokens_seen": 20394384,
+      "step": 980
+    },
+    {
+      "epoch": 0.7012547833051527,
+      "grad_norm": 0.33397364616394043,
+      "learning_rate": 2.050222799884387e-05,
+      "loss": 0.1389,
+      "num_input_tokens_seen": 20497680,
+      "step": 985
+    },
+    {
+      "epoch": 0.7048144522559402,
+      "grad_norm": 0.39059650897979736,
+      "learning_rate": 2.0052403629786858e-05,
+      "loss": 0.1625,
+      "num_input_tokens_seen": 20597280,
+      "step": 990
+    },
+    {
+      "epoch": 0.7083741212067278,
+      "grad_norm": 0.4324747323989868,
+      "learning_rate": 1.9606327807893902e-05,
+      "loss": 0.1546,
+      "num_input_tokens_seen": 20715592,
+      "step": 995
+    },
+    {
+      "epoch": 0.7119337901575153,
+      "grad_norm": 0.4057559072971344,
+      "learning_rate": 1.9164056368572846e-05,
+      "loss": 0.1759,
+      "num_input_tokens_seen": 20830456,
+      "step": 1000
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1404,
+  "num_input_tokens_seen": 20830456,
+  "num_train_epochs": 1,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.472838083134095e+17,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

2/checkpoint-1000/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53d6f2b6ddc39413c7f5106d4dfc20a60221e892597125a4051acc98e75cf41e
+size 7352

2/checkpoint-1000/zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/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 . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+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, weights_only=False)
+
+        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 tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # 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}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] 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")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+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 = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * 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
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        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}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        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 to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=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
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    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.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    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")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               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_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``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
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+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_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    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_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

2/checkpoint-1100/README.md

@@ -0,0 +1,202 @@
+---
+base_model: /root/autodl-tmp/Qwen3-8B
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.15.1

2/checkpoint-1100/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "/root/autodl-tmp/Qwen3-8B",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 8,
+  "lora_bias": false,
+  "lora_dropout": 0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 4,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": true
+}

2/checkpoint-1100/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc980566540beab038e36e4b044f63f9bb8b9ea3480ead65503fde29989ae98a
+size 21889736

2/checkpoint-1100/added_tokens.json

@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|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
+}

2/checkpoint-1100/global_step1100/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:031470bac929ee8b41ace3e23b641c253aba531aeaaf043b805f5dd87fd6b04c
+size 133228077

2/checkpoint-1100/global_step1100/zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4587341834363525c009e1d93ddc604d47814005495ace04a593afa44560f06b
+size 65501576