Model save

README.md

@@ -0,0 +1,68 @@
+---
+base_model: Qwen/Qwen2.5-Math-7B
+library_name: transformers
+model_name: Qwen-2.5-Math-7B-test1
+tags:
+- generated_from_trainer
+- trl
+- grpo
+licence: license
+---
+
+# Model Card for Qwen-2.5-Math-7B-test1
+
+This model is a fine-tuned version of [Qwen/Qwen2.5-Math-7B](https://huggingface.co/Qwen/Qwen2.5-Math-7B).
+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="chenggong1995/Qwen-2.5-Math-7B-test1", 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/gongc1995-city-university-of-hong-kong/huggingface/runs/o0kq9n2f) 
+
+
+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: 3.3.2
+- Tokenizers: 0.21.0
+
+## 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": -0.0030980921536684037,
+    "train_runtime": 7058.3489,
+    "train_samples": 100,
+    "train_samples_per_second": 0.028,
+    "train_steps_per_second": 0.014
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "max_new_tokens": 2048,
+  "transformers_version": "4.49.0"
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "total_flos": 0.0,
+    "train_loss": -0.0030980921536684037,
+    "train_runtime": 7058.3489,
+    "train_samples": 100,
+    "train_samples_per_second": 0.028,
+    "train_steps_per_second": 0.014
+}

trainer_state.json

@@ -0,0 +1,389 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0,
+  "eval_steps": 49,
+  "global_step": 100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 379.71429443359375,
+      "epoch": 0.02,
+      "grad_norm": 2.3801450729370117,
+      "kl": 0.0,
+      "learning_rate": 5e-08,
+      "loss": -0.0295,
+      "reward": 0.8928571939468384,
+      "reward_std": 0.26375192403793335,
+      "rewards/accuracy_reward": 0.8571429252624512,
+      "rewards/format_reward": 0.0,
+      "rewards/tag_count_reward": 0.0357142873108387,
+      "step": 1
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 602.6250305175781,
+      "epoch": 0.1,
+      "grad_norm": 1.251340627670288,
+      "kl": 0.0002231597900390625,
+      "learning_rate": 2.5e-07,
+      "loss": -0.004,
+      "reward": 0.5044643133878708,
+      "reward_std": 0.4692006930708885,
+      "rewards/accuracy_reward": 0.5000000298023224,
+      "rewards/format_reward": 0.0,
+      "rewards/tag_count_reward": 0.004464285913854837,
+      "step": 5
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 700.3143188476563,
+      "epoch": 0.2,
+      "grad_norm": 3.892239809036255,
+      "kl": 0.00029048919677734377,
+      "learning_rate": 5e-07,
+      "loss": -0.0521,
+      "reward": 0.5571428656578064,
+      "reward_std": 0.30040043592453003,
+      "rewards/accuracy_reward": 0.5285714507102967,
+      "rewards/format_reward": 0.0,
+      "rewards/tag_count_reward": 0.02857142984867096,
+      "step": 10
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 622.2857299804688,
+      "epoch": 0.3,
+      "grad_norm": 1.9563440084457397,
+      "kl": 0.000347137451171875,
+      "learning_rate": 4.96201938253052e-07,
+      "loss": -0.0156,
+      "reward": 0.5642857372760772,
+      "reward_std": 0.4562141001224518,
+      "rewards/accuracy_reward": 0.5000000238418579,
+      "rewards/format_reward": 0.01428571492433548,
+      "rewards/tag_count_reward": 0.05000000074505806,
+      "step": 15
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 623.8857543945312,
+      "epoch": 0.4,
+      "grad_norm": 3.7430624961853027,
+      "kl": 0.0006534576416015625,
+      "learning_rate": 4.849231551964771e-07,
+      "loss": -0.0311,
+      "reward": 0.5000000238418579,
+      "reward_std": 0.39751608967781066,
+      "rewards/accuracy_reward": 0.4428571581840515,
+      "rewards/format_reward": 0.01428571492433548,
+      "rewards/tag_count_reward": 0.04285714328289032,
+      "step": 20
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 519.84287109375,
+      "epoch": 0.5,
+      "grad_norm": 2.7288520336151123,
+      "kl": 0.007835006713867188,
+      "learning_rate": 4.6650635094610966e-07,
+      "loss": 0.083,
+      "reward": 0.6428571701049804,
+      "reward_std": 0.4448284685611725,
+      "rewards/accuracy_reward": 0.5714285880327225,
+      "rewards/format_reward": 0.02857142984867096,
+      "rewards/tag_count_reward": 0.04285714626312256,
+      "step": 25
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 572.8857360839844,
+      "epoch": 0.6,
+      "grad_norm": 8.5879487991333,
+      "kl": 0.0009708404541015625,
+      "learning_rate": 4.415111107797445e-07,
+      "loss": -0.056,
+      "reward": 0.8035714507102967,
+      "reward_std": 0.42971263229846957,
+      "rewards/accuracy_reward": 0.7142857491970063,
+      "rewards/format_reward": 0.01428571492433548,
+      "rewards/tag_count_reward": 0.07500000335276127,
+      "step": 30
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 606.6429016113282,
+      "epoch": 0.7,
+      "grad_norm": 1.614426612854004,
+      "kl": 0.00061492919921875,
+      "learning_rate": 4.106969024216348e-07,
+      "loss": 0.0218,
+      "reward": 0.7428571701049804,
+      "reward_std": 0.49451608657836915,
+      "rewards/accuracy_reward": 0.6571428775787354,
+      "rewards/format_reward": 0.02857142984867096,
+      "rewards/tag_count_reward": 0.05714286118745804,
+      "step": 35
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 659.6000305175781,
+      "epoch": 0.8,
+      "grad_norm": 0.00019049388356506824,
+      "kl": 0.008877372741699219,
+      "learning_rate": 3.75e-07,
+      "loss": 0.0187,
+      "reward": 0.45714287757873534,
+      "reward_std": 0.3365379124879837,
+      "rewards/accuracy_reward": 0.4285714507102966,
+      "rewards/format_reward": 0.0,
+      "rewards/tag_count_reward": 0.02857142984867096,
+      "step": 40
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 681.0714599609375,
+      "epoch": 0.9,
+      "grad_norm": 3.8056037425994873,
+      "kl": 0.0019466400146484375,
+      "learning_rate": 3.355050358314172e-07,
+      "loss": 0.0472,
+      "reward": 0.5178571701049804,
+      "reward_std": 0.3290700912475586,
+      "rewards/accuracy_reward": 0.4571428894996643,
+      "rewards/format_reward": 0.0,
+      "rewards/tag_count_reward": 0.06071428954601288,
+      "step": 45
+    },
+    {
+      "epoch": 0.98,
+      "eval_clip_ratio": 0.0,
+      "eval_completion_length": 594.2248850097657,
+      "eval_kl": 0.003310443878173828,
+      "eval_loss": -0.02919997088611126,
+      "eval_reward": 0.534714311145246,
+      "eval_reward_std": 0.3895806306153536,
+      "eval_rewards/accuracy_reward": 0.4940000242590904,
+      "eval_rewards/format_reward": 0.006000000268220901,
+      "eval_rewards/tag_count_reward": 0.03471428740769625,
+      "eval_runtime": 2236.5694,
+      "eval_samples_per_second": 0.224,
+      "eval_steps_per_second": 0.016,
+      "step": 49
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 594.1571655273438,
+      "epoch": 1.0,
+      "grad_norm": 0.0004400305333547294,
+      "kl": 0.001186370849609375,
+      "learning_rate": 2.934120444167326e-07,
+      "loss": -0.0472,
+      "reward": 0.4107142984867096,
+      "reward_std": 0.2270268350839615,
+      "rewards/accuracy_reward": 0.40000001192092893,
+      "rewards/format_reward": 0.0,
+      "rewards/tag_count_reward": 0.01071428656578064,
+      "step": 50
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 699.0428955078125,
+      "epoch": 1.1,
+      "grad_norm": 0.6008018851280212,
+      "kl": 0.0012775421142578124,
+      "learning_rate": 2.5e-07,
+      "loss": -0.0518,
+      "reward": 0.5821428835391999,
+      "reward_std": 0.44157396256923676,
+      "rewards/accuracy_reward": 0.5428571760654449,
+      "rewards/format_reward": 0.01428571492433548,
+      "rewards/tag_count_reward": 0.02500000111758709,
+      "step": 55
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 676.3571655273438,
+      "epoch": 1.2,
+      "grad_norm": 0.7341341376304626,
+      "kl": 0.0003894805908203125,
+      "learning_rate": 2.065879555832674e-07,
+      "loss": -0.0145,
+      "reward": 0.6964286088943481,
+      "reward_std": 0.32321630120277406,
+      "rewards/accuracy_reward": 0.685714328289032,
+      "rewards/format_reward": 0.0,
+      "rewards/tag_count_reward": 0.010714286193251609,
+      "step": 60
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 484.5143188476562,
+      "epoch": 1.3,
+      "grad_norm": 2.903294086456299,
+      "kl": 0.003038787841796875,
+      "learning_rate": 1.6449496416858282e-07,
+      "loss": -0.0505,
+      "reward": 0.6428571611642837,
+      "reward_std": 0.3603375285863876,
+      "rewards/accuracy_reward": 0.5857143193483353,
+      "rewards/format_reward": 0.02857142984867096,
+      "rewards/tag_count_reward": 0.02857142984867096,
+      "step": 65
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 664.8000244140625,
+      "epoch": 1.4,
+      "grad_norm": 10.34648609161377,
+      "kl": 0.0013919830322265624,
+      "learning_rate": 1.2500000000000005e-07,
+      "loss": 0.0629,
+      "reward": 0.4928571581840515,
+      "reward_std": 0.5296794831752777,
+      "rewards/accuracy_reward": 0.45714287757873534,
+      "rewards/format_reward": 0.01428571492433548,
+      "rewards/tag_count_reward": 0.021428572386503218,
+      "step": 70
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 597.857177734375,
+      "epoch": 1.5,
+      "grad_norm": 1.6099767684936523,
+      "kl": 0.0036956787109375,
+      "learning_rate": 8.930309757836516e-08,
+      "loss": 0.0399,
+      "reward": 0.4571428716182709,
+      "reward_std": 0.3776773989200592,
+      "rewards/accuracy_reward": 0.4571428716182709,
+      "rewards/format_reward": 0.0,
+      "rewards/tag_count_reward": 0.0,
+      "step": 75
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 570.3000366210938,
+      "epoch": 1.6,
+      "grad_norm": 0.7647049427032471,
+      "kl": 0.0009387969970703125,
+      "learning_rate": 5.848888922025552e-08,
+      "loss": -0.0569,
+      "reward": 0.6071428894996643,
+      "reward_std": 0.3395166456699371,
+      "rewards/accuracy_reward": 0.5714285969734192,
+      "rewards/format_reward": 0.0,
+      "rewards/tag_count_reward": 0.035714288055896756,
+      "step": 80
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 619.4714538574219,
+      "epoch": 1.7,
+      "grad_norm": 1.8258100748062134,
+      "kl": 0.002353668212890625,
+      "learning_rate": 3.349364905389032e-08,
+      "loss": 0.0276,
+      "reward": 0.44642859399318696,
+      "reward_std": 0.27503437697887423,
+      "rewards/accuracy_reward": 0.4428571656346321,
+      "rewards/format_reward": 0.0,
+      "rewards/tag_count_reward": 0.00357142873108387,
+      "step": 85
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 535.4000244140625,
+      "epoch": 1.8,
+      "grad_norm": 3.521362066268921,
+      "kl": 0.0031871795654296875,
+      "learning_rate": 1.507684480352292e-08,
+      "loss": 0.0045,
+      "reward": 0.5464286029338836,
+      "reward_std": 0.48063153624534605,
+      "rewards/accuracy_reward": 0.5000000238418579,
+      "rewards/format_reward": 0.01428571492433548,
+      "rewards/tag_count_reward": 0.03214285895228386,
+      "step": 90
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 741.9714477539062,
+      "epoch": 1.9,
+      "grad_norm": 2.457162857055664,
+      "kl": 0.0006832122802734375,
+      "learning_rate": 3.798061746947995e-09,
+      "loss": 0.0415,
+      "reward": 0.5500000238418579,
+      "reward_std": 0.369550421833992,
+      "rewards/accuracy_reward": 0.5285714507102967,
+      "rewards/format_reward": 0.0,
+      "rewards/tag_count_reward": 0.02142857201397419,
+      "step": 95
+    },
+    {
+      "epoch": 1.96,
+      "eval_clip_ratio": 0.0,
+      "eval_completion_length": 588.6673402709961,
+      "eval_kl": 0.004097911834716797,
+      "eval_loss": -0.02776484563946724,
+      "eval_reward": 0.5458571693897247,
+      "eval_reward_std": 0.39641910094022753,
+      "eval_rewards/accuracy_reward": 0.5062857375442982,
+      "eval_rewards/format_reward": 0.006285714566707611,
+      "eval_rewards/tag_count_reward": 0.03328571600466967,
+      "eval_runtime": 2226.3251,
+      "eval_samples_per_second": 0.225,
+      "eval_steps_per_second": 0.016,
+      "step": 98
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 761.4000244140625,
+      "epoch": 2.0,
+      "grad_norm": 0.7996540069580078,
+      "kl": 0.0015918731689453125,
+      "learning_rate": 0.0,
+      "loss": -0.0241,
+      "reward": 0.6500000298023224,
+      "reward_std": 0.4324204444885254,
+      "rewards/accuracy_reward": 0.5571428954601287,
+      "rewards/format_reward": 0.042857144773006436,
+      "rewards/tag_count_reward": 0.05000000223517418,
+      "step": 100
+    },
+    {
+      "epoch": 2.0,
+      "step": 100,
+      "total_flos": 0.0,
+      "train_loss": -0.0030980921536684037,
+      "train_runtime": 7058.3489,
+      "train_samples_per_second": 0.028,
+      "train_steps_per_second": 0.014
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 100,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 500,
+  "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": 2,
+  "trial_name": null,
+  "trial_params": null
+}