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/jmiy00ow) 
+
+
+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": 0.2185068104801507,
+    "train_runtime": 17682.4744,
+    "train_samples": 7500,
+    "train_samples_per_second": 0.848,
+    "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": 0.2185068104801507,
+    "train_runtime": 17682.4744,
+    "train_samples": 7500,
+    "train_samples_per_second": 0.848,
+    "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": 0.28463842001337397,
+      "kl": 0.0,
+      "learning_rate": 5e-07,
+      "loss": 0.134,
+      "reward": 0.5082886815071106,
+      "reward_std": 0.15963527746498585,
+      "rewards/accuracy_reward": 0.745535746216774,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.23724707029759884,
+      "step": 1
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 974.6809825897217,
+      "epoch": 0.1702127659574468,
+      "grad_norm": 0.23728338609141092,
+      "kl": 7.671117782592773e-05,
+      "learning_rate": 2.5e-06,
+      "loss": 0.132,
+      "reward": 0.4996245112270117,
+      "reward_std": 0.1835802630521357,
+      "rewards/accuracy_reward": 0.737583739683032,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.23795922426506877,
+      "step": 5
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 867.940104675293,
+      "epoch": 0.3404255319148936,
+      "grad_norm": 0.6283062851573521,
+      "kl": 0.008861899375915527,
+      "learning_rate": 2.956412726139078e-06,
+      "loss": 0.1317,
+      "reward": 0.5490380935370922,
+      "reward_std": 0.1904804239049554,
+      "rewards/accuracy_reward": 0.7609375298023224,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.21189943961799146,
+      "step": 10
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 730.2546096801758,
+      "epoch": 0.5106382978723404,
+      "grad_norm": 1.310897088360317,
+      "kl": 0.26617431640625,
+      "learning_rate": 2.7836719084521715e-06,
+      "loss": 0.1354,
+      "reward": 0.47696074694395063,
+      "reward_std": 0.23283899687230586,
+      "rewards/accuracy_reward": 0.6552455656230449,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.17828481681644917,
+      "step": 15
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 575.1072868347168,
+      "epoch": 0.6808510638297872,
+      "grad_norm": 2.0875900059869346,
+      "kl": 0.91943359375,
+      "learning_rate": 2.4946839873611927e-06,
+      "loss": 0.178,
+      "reward": 0.44084297828376295,
+      "reward_std": 0.2517234329134226,
+      "rewards/accuracy_reward": 0.5812500230967999,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.14040705244988202,
+      "step": 20
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 543.0458961486817,
+      "epoch": 0.851063829787234,
+      "grad_norm": 8.217171705965015,
+      "kl": 1.1748046875,
+      "learning_rate": 2.1156192081791355e-06,
+      "loss": 0.2204,
+      "reward": 0.4586034990847111,
+      "reward_std": 0.23402524627745153,
+      "rewards/accuracy_reward": 0.5911830633878707,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.13257956448942423,
+      "step": 25
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 521.1100387573242,
+      "epoch": 1.0340425531914894,
+      "grad_norm": 6.762082627347706,
+      "kl": 0.7322998046875,
+      "learning_rate": 1.6808050203829845e-06,
+      "loss": 0.1845,
+      "reward": 0.5005072712898254,
+      "reward_std": 0.23026448152959347,
+      "rewards/accuracy_reward": 0.627120566368103,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.1266132965683937,
+      "step": 30
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 494.15995559692385,
+      "epoch": 1.2042553191489362,
+      "grad_norm": 3.1612480284068982,
+      "kl": 1.55126953125,
+      "learning_rate": 1.2296174432791415e-06,
+      "loss": 0.2725,
+      "reward": 0.48895372450351715,
+      "reward_std": 0.24462027363479139,
+      "rewards/accuracy_reward": 0.6095982402563095,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.12064452040940524,
+      "step": 35
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 483.87390594482423,
+      "epoch": 1.374468085106383,
+      "grad_norm": 4.294437421320427,
+      "kl": 1.28154296875,
+      "learning_rate": 8.029152419343472e-07,
+      "loss": 0.2563,
+      "reward": 0.47126407250761987,
+      "reward_std": 0.2647111125290394,
+      "rewards/accuracy_reward": 0.5893973462283612,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.1181332778185606,
+      "step": 40
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 481.56776733398436,
+      "epoch": 1.5446808510638297,
+      "grad_norm": 4.491684991262943,
+      "kl": 1.42666015625,
+      "learning_rate": 4.3933982822017883e-07,
+      "loss": 0.2752,
+      "reward": 0.4363360181450844,
+      "reward_std": 0.2855977095663548,
+      "rewards/accuracy_reward": 0.5539062716066837,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.11757025569677353,
+      "step": 45
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 484.0088371276855,
+      "epoch": 1.7148936170212767,
+      "grad_norm": 2.1484729806652747,
+      "kl": 1.5796875,
+      "learning_rate": 1.718159615201853e-07,
+      "loss": 0.2947,
+      "reward": 0.4461195237934589,
+      "reward_std": 0.28383450768887997,
+      "rewards/accuracy_reward": 0.564285734295845,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.11816622000187635,
+      "step": 50
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 489.8710029602051,
+      "epoch": 1.8851063829787233,
+      "grad_norm": 1.8313195035052112,
+      "kl": 1.49287109375,
+      "learning_rate": 2.4570139579284723e-08,
+      "loss": 0.2968,
+      "reward": 0.44390707574784755,
+      "reward_std": 0.281308988109231,
+      "rewards/accuracy_reward": 0.5635044880211353,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.11959741283208132,
+      "step": 55
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 490.9578692118327,
+      "epoch": 1.9872340425531916,
+      "kl": 1.4850260416666667,
+      "reward": 0.4616190791130066,
+      "reward_std": 0.27286908278862637,
+      "rewards/accuracy_reward": 0.5811012176175913,
+      "rewards/format_reward": 0.0,
+      "rewards/len_penalty": -0.1194821372628212,
+      "step": 58,
+      "total_flos": 0.0,
+      "train_loss": 0.2185068104801507,
+      "train_runtime": 17682.4744,
+      "train_samples_per_second": 0.848,
+      "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
+}