Model save

README.md

@@ -0,0 +1,68 @@
+---
+base_model: zijianh/DeepSeek-R1-Distill-Qwen-7B-RL-length-penalty-low-new
+library_name: transformers
+model_name: DeepSeek-R1-Distill-Qwen-7B-RL-length-penalty-low-high-0_1-new
+tags:
+- generated_from_trainer
+- trl
+- grpo
+licence: license
+---
+
+# Model Card for DeepSeek-R1-Distill-Qwen-7B-RL-length-penalty-low-high-0_1-new
+
+This model is a fine-tuned version of [zijianh/DeepSeek-R1-Distill-Qwen-7B-RL-length-penalty-low-new](https://huggingface.co/zijianh/DeepSeek-R1-Distill-Qwen-7B-RL-length-penalty-low-new).
+It has been trained using [TRL](https://github.com/huggingface/trl).
+
+## Quick start
+
+```python
+from transformers import pipeline
+
+question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
+generator = pipeline("text-generation", model="zijianh/DeepSeek-R1-Distill-Qwen-7B-RL-length-penalty-low-high-0_1-new", device="cuda")
+output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
+print(output["generated_text"])
+```
+
+## Training procedure
+
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/sota_mavens-university-of-michigan/huggingface/runs/nbrpsw2q) 
+
+
+This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
+
+### Framework versions
+
+- TRL: 0.16.0.dev0
+- Transformers: 4.49.0
+- Pytorch: 2.5.1
+- Datasets: 3.4.1
+- Tokenizers: 0.21.1
+
+## Citations
+
+Cite GRPO as:
+
+```bibtex
+@article{zhihong2024deepseekmath,
+    title        = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
+    author       = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo},
+    year         = 2024,
+    eprint       = {arXiv:2402.03300},
+}
+
+```
+
+Cite TRL as:
+    
+```bibtex
+@misc{vonwerra2022trl,
+	title        = {{TRL: Transformer Reinforcement Learning}},
+	author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
+	year         = 2020,
+	journal      = {GitHub repository},
+	publisher    = {GitHub},
+	howpublished = {\url{https://github.com/huggingface/trl}}
+}
+```

all_results.json

@@ -0,0 +1,8 @@
+{
+    "total_flos": 0.0,
+    "train_loss": 0.1866766851523827,
+    "train_runtime": 17842.1504,
+    "train_samples": 7500,
+    "train_samples_per_second": 0.841,
+    "train_steps_per_second": 0.003
+}

generation_config.json

@@ -0,0 +1,9 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151646,
+  "do_sample": true,
+  "eos_token_id": 151643,
+  "temperature": 0.6,
+  "top_p": 0.95,
+  "transformers_version": "4.49.0"
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "total_flos": 0.0,
+    "train_loss": 0.1866766851523827,
+    "train_runtime": 17842.1504,
+    "train_samples": 7500,
+    "train_samples_per_second": 0.841,
+    "train_steps_per_second": 0.003
+}

trainer_state.json

@@ -0,0 +1,230 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9872340425531916,
+  "eval_steps": 100,
+  "global_step": 58,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 971.7639999389648,
+      "epoch": 0.03404255319148936,
+      "grad_norm": 0.2631084113360146,
+      "kl": 0.0,
+      "learning_rate": 5e-07,
+      "loss": 0.1238,
+      "reward": 0.698086328804493,
+      "reward_std": 0.13120541395619512,
+      "rewards/accuracy_reward": 0.745535746216774,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.04744941322132945,
+      "step": 1
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 979.8407287597656,
+      "epoch": 0.1702127659574468,
+      "grad_norm": 0.22684407077545118,
+      "kl": 7.544457912445068e-05,
+      "learning_rate": 2.5e-06,
+      "loss": 0.1155,
+      "reward": 0.6869497802108526,
+      "reward_std": 0.16120319138281047,
+      "rewards/accuracy_reward": 0.734793558716774,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.047843783744610846,
+      "step": 5
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 872.467677307129,
+      "epoch": 0.3404255319148936,
+      "grad_norm": 0.4684158119972997,
+      "kl": 0.00807795524597168,
+      "learning_rate": 2.956412726139078e-06,
+      "loss": 0.1146,
+      "reward": 0.7163276463747025,
+      "reward_std": 0.16980856116861104,
+      "rewards/accuracy_reward": 0.7589286044239998,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.04260096037760377,
+      "step": 10
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 751.1273788452148,
+      "epoch": 0.5106382978723404,
+      "grad_norm": 1.2177198858526401,
+      "kl": 0.34739990234375,
+      "learning_rate": 2.7836719084521715e-06,
+      "loss": 0.1264,
+      "reward": 0.6330783508718014,
+      "reward_std": 0.20811637695878743,
+      "rewards/accuracy_reward": 0.6697544969618321,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.03667614138685167,
+      "step": 15
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 613.7534912109375,
+      "epoch": 0.6808510638297872,
+      "grad_norm": 1.3980210820254566,
+      "kl": 0.744873046875,
+      "learning_rate": 2.4946839873611927e-06,
+      "loss": 0.1578,
+      "reward": 0.5512815967202187,
+      "reward_std": 0.23243679329752923,
+      "rewards/accuracy_reward": 0.581250024586916,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.0299684323836118,
+      "step": 20
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 582.3771484375,
+      "epoch": 0.851063829787234,
+      "grad_norm": 1.8890671871166043,
+      "kl": 0.9321044921875,
+      "learning_rate": 2.1156192081791355e-06,
+      "loss": 0.1965,
+      "reward": 0.56687613427639,
+      "reward_std": 0.21850477196276188,
+      "rewards/accuracy_reward": 0.5953125260770321,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.028436384443193675,
+      "step": 25
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 557.1671844482422,
+      "epoch": 1.0340425531914894,
+      "grad_norm": 6.969148820152461,
+      "kl": 0.90537109375,
+      "learning_rate": 1.6808050203829845e-06,
+      "loss": 0.1776,
+      "reward": 0.609768246114254,
+      "reward_std": 0.20966911502182484,
+      "rewards/accuracy_reward": 0.6368303805589676,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.027062145154923202,
+      "step": 30
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 529.0748001098633,
+      "epoch": 1.2042553191489362,
+      "grad_norm": 1.6903572527831305,
+      "kl": 0.88388671875,
+      "learning_rate": 1.2296174432791415e-06,
+      "loss": 0.1976,
+      "reward": 0.6035189718008042,
+      "reward_std": 0.20989050157368183,
+      "rewards/accuracy_reward": 0.6293527103960515,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.025833730399608613,
+      "step": 35
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 518.8400924682617,
+      "epoch": 1.374468085106383,
+      "grad_norm": 1.140228131246192,
+      "kl": 1.342578125,
+      "learning_rate": 8.029152419343472e-07,
+      "loss": 0.2382,
+      "reward": 0.5803579963743687,
+      "reward_std": 0.23556350730359554,
+      "rewards/accuracy_reward": 0.6056919865310192,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.025333988945931196,
+      "step": 40
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 512.1945533752441,
+      "epoch": 1.5446808510638297,
+      "grad_norm": 1.4935623924313746,
+      "kl": 0.83896484375,
+      "learning_rate": 4.3933982822017883e-07,
+      "loss": 0.2043,
+      "reward": 0.5334726713597775,
+      "reward_std": 0.25458414256572726,
+      "rewards/accuracy_reward": 0.5584821730852128,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.025009499955922367,
+      "step": 45
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 510.34209899902345,
+      "epoch": 1.7148936170212767,
+      "grad_norm": 2.5529208106611776,
+      "kl": 1.512890625,
+      "learning_rate": 1.718159615201853e-07,
+      "loss": 0.2564,
+      "reward": 0.5088979430496693,
+      "reward_std": 0.26402820013463496,
+      "rewards/accuracy_reward": 0.5338169865310192,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.02491904767230153,
+      "step": 50
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 509.2735771179199,
+      "epoch": 1.8851063829787233,
+      "grad_norm": 1.2580199216017853,
+      "kl": 1.3421875,
+      "learning_rate": 2.4570139579284723e-08,
+      "loss": 0.2429,
+      "reward": 0.5227894008159637,
+      "reward_std": 0.24820317029953004,
+      "rewards/accuracy_reward": 0.5476562783122063,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.02486687391065061,
+      "step": 55
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 513.0390764872233,
+      "epoch": 1.9872340425531916,
+      "kl": 1.3177083333333333,
+      "reward": 0.5240900913874308,
+      "reward_std": 0.256128067150712,
+      "rewards/accuracy_reward": 0.54892115543286,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.024831065131972235,
+      "step": 58,
+      "total_flos": 0.0,
+      "train_loss": 0.1866766851523827,
+      "train_runtime": 17842.1504,
+      "train_samples_per_second": 0.841,
+      "train_steps_per_second": 0.003
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 58,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 10,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}