Model save

README.md

@@ -0,0 +1,67 @@
+---
+library_name: transformers
+model_name: Qwen2.5-3B-Open-R1-GRPO-Self-TQA
+tags:
+- generated_from_trainer
+- trl
+- grpo
+licence: license
+---
+
+# Model Card for Qwen2.5-3B-Open-R1-GRPO-Self-TQA
+
+This model is a fine-tuned version of [None](https://huggingface.co/None).
+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="qingyangzhang/Qwen2.5-3B-Open-R1-GRPO-Self-TQA", 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/zqyoung1127-tianjin-university/huggingface/runs/yloacszm) 
+
+
+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: 0.14.0
+- Transformers: 4.48.3
+- Pytorch: 2.5.1+cu124
+- Datasets: 3.1.0
+- 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": 0.0001855157711543143,
+    "train_runtime": 422.7212,
+    "train_samples": 817,
+    "train_samples_per_second": 1.933,
+    "train_steps_per_second": 0.059
+}

generation_config.json

@@ -0,0 +1,14 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "repetition_penalty": 1.05,
+  "temperature": 0.7,
+  "top_k": 20,
+  "top_p": 0.8,
+  "transformers_version": "4.48.3"
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "total_flos": 0.0,
+    "train_loss": 0.0001855157711543143,
+    "train_runtime": 422.7212,
+    "train_samples": 817,
+    "train_samples_per_second": 1.933,
+    "train_steps_per_second": 0.059
+}

trainer_state.json

@@ -0,0 +1,342 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.975609756097561,
+  "eval_steps": 100,
+  "global_step": 25,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "completion_length": 22.076171875,
+      "epoch": 0.03902439024390244,
+      "grad_norm": 0.697234570980072,
+      "kl": 0.0,
+      "learning_rate": 6.666666666666666e-07,
+      "loss": 0.0,
+      "reward": 0.3046875,
+      "reward_std": 0.3262644410133362,
+      "rewards/semantic_entropy": 0.3046875,
+      "step": 1
+    },
+    {
+      "completion_length": 26.513671875,
+      "epoch": 0.07804878048780488,
+      "grad_norm": 0.44604623317718506,
+      "kl": 0.0,
+      "learning_rate": 1.3333333333333332e-06,
+      "loss": 0.0,
+      "reward": 0.40625,
+      "reward_std": 0.42861204594373703,
+      "rewards/semantic_entropy": 0.40625,
+      "step": 2
+    },
+    {
+      "completion_length": 22.556640625,
+      "epoch": 0.11707317073170732,
+      "grad_norm": 1.134758472442627,
+      "kl": 0.001371145248413086,
+      "learning_rate": 2e-06,
+      "loss": 0.0001,
+      "reward": 0.30078125,
+      "reward_std": 0.30227910727262497,
+      "rewards/semantic_entropy": 0.30078125,
+      "step": 3
+    },
+    {
+      "completion_length": 25.80078125,
+      "epoch": 0.15609756097560976,
+      "grad_norm": 0.8177841305732727,
+      "kl": 0.0005015134811401367,
+      "learning_rate": 1.9898214418809326e-06,
+      "loss": 0.0,
+      "reward": 0.302734375,
+      "reward_std": 0.3620207440108061,
+      "rewards/semantic_entropy": 0.302734375,
+      "step": 4
+    },
+    {
+      "completion_length": 23.333984375,
+      "epoch": 0.1951219512195122,
+      "grad_norm": 1.1548794507980347,
+      "kl": 0.0008560121059417725,
+      "learning_rate": 1.9594929736144973e-06,
+      "loss": 0.0,
+      "reward": 0.279296875,
+      "reward_std": 0.26689392141997814,
+      "rewards/semantic_entropy": 0.279296875,
+      "step": 5
+    },
+    {
+      "completion_length": 28.08984375,
+      "epoch": 0.23414634146341465,
+      "grad_norm": 2.489468812942505,
+      "kl": 0.008920669555664062,
+      "learning_rate": 1.9096319953545185e-06,
+      "loss": 0.0004,
+      "reward": 0.390625,
+      "reward_std": 0.3596664369106293,
+      "rewards/semantic_entropy": 0.390625,
+      "step": 6
+    },
+    {
+      "completion_length": 30.310546875,
+      "epoch": 0.2731707317073171,
+      "grad_norm": 0.4479486048221588,
+      "kl": 0.008846282958984375,
+      "learning_rate": 1.8412535328311812e-06,
+      "loss": 0.0004,
+      "reward": 0.4140625,
+      "reward_std": 0.41363558918237686,
+      "rewards/semantic_entropy": 0.4140625,
+      "step": 7
+    },
+    {
+      "completion_length": 30.3671875,
+      "epoch": 0.3121951219512195,
+      "grad_norm": 0.38852936029434204,
+      "kl": 0.0185089111328125,
+      "learning_rate": 1.7557495743542582e-06,
+      "loss": 0.0007,
+      "reward": 0.380859375,
+      "reward_std": 0.3279299959540367,
+      "rewards/semantic_entropy": 0.380859375,
+      "step": 8
+    },
+    {
+      "completion_length": 32.099609375,
+      "epoch": 0.35121951219512193,
+      "grad_norm": 0.4612264037132263,
+      "kl": 0.03018951416015625,
+      "learning_rate": 1.6548607339452852e-06,
+      "loss": 0.0012,
+      "reward": 0.3671875,
+      "reward_std": 0.40921078994870186,
+      "rewards/semantic_entropy": 0.3671875,
+      "step": 9
+    },
+    {
+      "completion_length": 34.060546875,
+      "epoch": 0.3902439024390244,
+      "grad_norm": 0.4247353971004486,
+      "kl": 0.029693603515625,
+      "learning_rate": 1.5406408174555977e-06,
+      "loss": 0.0012,
+      "reward": 0.408203125,
+      "reward_std": 0.4044051952660084,
+      "rewards/semantic_entropy": 0.408203125,
+      "step": 10
+    },
+    {
+      "completion_length": 32.00390625,
+      "epoch": 0.4292682926829268,
+      "grad_norm": 0.41794028878211975,
+      "kl": 0.0347137451171875,
+      "learning_rate": 1.4154150130018865e-06,
+      "loss": 0.0014,
+      "reward": 0.337890625,
+      "reward_std": 0.34299428947269917,
+      "rewards/semantic_entropy": 0.337890625,
+      "step": 11
+    },
+    {
+      "completion_length": 38.1015625,
+      "epoch": 0.4682926829268293,
+      "grad_norm": 0.32552075386047363,
+      "kl": 0.0347442626953125,
+      "learning_rate": 1.2817325568414297e-06,
+      "loss": 0.0014,
+      "reward": 0.400390625,
+      "reward_std": 0.4432142451405525,
+      "rewards/semantic_entropy": 0.400390625,
+      "step": 12
+    },
+    {
+      "completion_length": 38.935546875,
+      "epoch": 0.5073170731707317,
+      "grad_norm": 0.3407377600669861,
+      "kl": 0.045013427734375,
+      "learning_rate": 1.1423148382732853e-06,
+      "loss": 0.0018,
+      "reward": 0.41015625,
+      "reward_std": 0.4198991097509861,
+      "rewards/semantic_entropy": 0.41015625,
+      "step": 13
+    },
+    {
+      "completion_length": 33.248046875,
+      "epoch": 0.5463414634146342,
+      "grad_norm": 0.7667419910430908,
+      "kl": 0.03955078125,
+      "learning_rate": 1e-06,
+      "loss": 0.0016,
+      "reward": 0.3515625,
+      "reward_std": 0.39246053621172905,
+      "rewards/semantic_entropy": 0.3515625,
+      "step": 14
+    },
+    {
+      "completion_length": 34.533203125,
+      "epoch": 0.5853658536585366,
+      "grad_norm": 0.6600065231323242,
+      "kl": 0.045623779296875,
+      "learning_rate": 8.576851617267149e-07,
+      "loss": 0.0018,
+      "reward": 0.388671875,
+      "reward_std": 0.3779917135834694,
+      "rewards/semantic_entropy": 0.388671875,
+      "step": 15
+    },
+    {
+      "completion_length": 31.439453125,
+      "epoch": 0.624390243902439,
+      "grad_norm": 0.5913363695144653,
+      "kl": 0.03285980224609375,
+      "learning_rate": 7.182674431585702e-07,
+      "loss": 0.0013,
+      "reward": 0.41796875,
+      "reward_std": 0.3863699361681938,
+      "rewards/semantic_entropy": 0.41796875,
+      "step": 16
+    },
+    {
+      "completion_length": 31.44140625,
+      "epoch": 0.6634146341463415,
+      "grad_norm": 0.342813640832901,
+      "kl": 0.030029296875,
+      "learning_rate": 5.845849869981136e-07,
+      "loss": 0.0012,
+      "reward": 0.365234375,
+      "reward_std": 0.3471006993204355,
+      "rewards/semantic_entropy": 0.365234375,
+      "step": 17
+    },
+    {
+      "completion_length": 30.06640625,
+      "epoch": 0.7024390243902439,
+      "grad_norm": 0.5251995325088501,
+      "kl": 0.0410003662109375,
+      "learning_rate": 4.5935918254440274e-07,
+      "loss": 0.0016,
+      "reward": 0.33203125,
+      "reward_std": 0.3720228523015976,
+      "rewards/semantic_entropy": 0.33203125,
+      "step": 18
+    },
+    {
+      "completion_length": 36.03125,
+      "epoch": 0.7414634146341463,
+      "grad_norm": 0.44380083680152893,
+      "kl": 0.02307891845703125,
+      "learning_rate": 3.45139266054715e-07,
+      "loss": 0.0009,
+      "reward": 0.369140625,
+      "reward_std": 0.40155548974871635,
+      "rewards/semantic_entropy": 0.369140625,
+      "step": 19
+    },
+    {
+      "completion_length": 29.765625,
+      "epoch": 0.7804878048780488,
+      "grad_norm": 0.5805388689041138,
+      "kl": 0.03347015380859375,
+      "learning_rate": 2.4425042564574185e-07,
+      "loss": 0.0013,
+      "reward": 0.376953125,
+      "reward_std": 0.37975445948541164,
+      "rewards/semantic_entropy": 0.376953125,
+      "step": 20
+    },
+    {
+      "completion_length": 33.17578125,
+      "epoch": 0.8195121951219512,
+      "grad_norm": 0.4464017450809479,
+      "kl": 0.011322021484375,
+      "learning_rate": 1.5874646716881868e-07,
+      "loss": 0.0005,
+      "reward": 0.3671875,
+      "reward_std": 0.40596737898886204,
+      "rewards/semantic_entropy": 0.3671875,
+      "step": 21
+    },
+    {
+      "completion_length": 26.4453125,
+      "epoch": 0.8585365853658536,
+      "grad_norm": 0.5445716977119446,
+      "kl": 0.028438568115234375,
+      "learning_rate": 9.036800464548156e-08,
+      "loss": 0.0011,
+      "reward": 0.4375,
+      "reward_std": 0.39070899225771427,
+      "rewards/semantic_entropy": 0.4375,
+      "step": 22
+    },
+    {
+      "completion_length": 28.7578125,
+      "epoch": 0.8975609756097561,
+      "grad_norm": 0.49012911319732666,
+      "kl": 0.018596649169921875,
+      "learning_rate": 4.050702638550274e-08,
+      "loss": 0.0007,
+      "reward": 0.39453125,
+      "reward_std": 0.3807212747633457,
+      "rewards/semantic_entropy": 0.39453125,
+      "step": 23
+    },
+    {
+      "completion_length": 26.87109375,
+      "epoch": 0.9365853658536586,
+      "grad_norm": 0.47687703371047974,
+      "kl": 0.03133201599121094,
+      "learning_rate": 1.0178558119067315e-08,
+      "loss": 0.0013,
+      "reward": 0.30859375,
+      "reward_std": 0.3204674646258354,
+      "rewards/semantic_entropy": 0.30859375,
+      "step": 24
+    },
+    {
+      "completion_length": 28.25390625,
+      "epoch": 0.975609756097561,
+      "grad_norm": 0.623159646987915,
+      "kl": 0.02616119384765625,
+      "learning_rate": 0.0,
+      "loss": 0.001,
+      "reward": 0.375,
+      "reward_std": 0.3317893110215664,
+      "rewards/semantic_entropy": 0.375,
+      "step": 25
+    },
+    {
+      "epoch": 0.975609756097561,
+      "step": 25,
+      "total_flos": 0.0,
+      "train_loss": 0.0001855157711543143,
+      "train_runtime": 422.7212,
+      "train_samples_per_second": 1.933,
+      "train_steps_per_second": 0.059
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 25,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "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": 1,
+  "trial_name": null,
+  "trial_params": null
+}