Model save

README.md

@@ -0,0 +1,68 @@
+---
+base_model: meta-llama/Llama-3.1-8B-Instruct
+library_name: transformers
+model_name: new-Llama3.1-8b-instruct-RL
+tags:
+- generated_from_trainer
+- trl
+- grpo
+licence: license
+---
+
+# Model Card for new-Llama3.1-8b-instruct-RL
+
+This model is a fine-tuned version of [meta-llama/Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct).
+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="sravanthib/new-Llama3.1-8b-instruct-RL", 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/golden-goose/huggingface/runs/k9bt3x5h) 
+
+
+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": 0.009057209056256146,
+    "train_runtime": 7830.5569,
+    "train_samples": 7500,
+    "train_samples_per_second": 0.958,
+    "train_steps_per_second": 0.007
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "total_flos": 0.0,
+    "train_loss": 0.009057209056256146,
+    "train_runtime": 7830.5569,
+    "train_samples": 7500,
+    "train_samples_per_second": 0.958,
+    "train_steps_per_second": 0.007
+}

trainer_state.json

@@ -0,0 +1,217 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9893390191897654,
+  "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": 559.5167694091797,
+      "epoch": 0.017057569296375266,
+      "grad_norm": 0.012841759249567986,
+      "kl": 0.0,
+      "learning_rate": 5e-07,
+      "loss": 0.0087,
+      "reward": 0.11495536100119352,
+      "reward_std": 0.195150856859982,
+      "rewards/accuracy_reward": 0.11495536100119352,
+      "rewards/format_reward": 0.0,
+      "step": 1
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 555.0773105621338,
+      "epoch": 0.08528784648187633,
+      "grad_norm": 0.014003963209688663,
+      "kl": 0.0001754462718963623,
+      "learning_rate": 2.5e-06,
+      "loss": 0.0047,
+      "reward": 0.11049107764847577,
+      "reward_std": 0.19013801333494484,
+      "rewards/accuracy_reward": 0.11049107764847577,
+      "rewards/format_reward": 0.0,
+      "step": 5
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 538.5312759399415,
+      "epoch": 0.17057569296375266,
+      "grad_norm": 0.01436670869588852,
+      "kl": 0.00021240711212158203,
+      "learning_rate": 2.956412726139078e-06,
+      "loss": 0.0048,
+      "reward": 0.13459821995347737,
+      "reward_std": 0.20578723903745413,
+      "rewards/accuracy_reward": 0.13459821995347737,
+      "rewards/format_reward": 0.0,
+      "step": 10
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 534.3341758728027,
+      "epoch": 0.255863539445629,
+      "grad_norm": 0.014759319834411144,
+      "kl": 0.00021679401397705077,
+      "learning_rate": 2.7836719084521715e-06,
+      "loss": 0.0129,
+      "reward": 0.11629464784637093,
+      "reward_std": 0.19606330525130033,
+      "rewards/accuracy_reward": 0.11629464784637093,
+      "rewards/format_reward": 0.0,
+      "step": 15
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 532.0986877441406,
+      "epoch": 0.3411513859275053,
+      "grad_norm": 0.01396097894757986,
+      "kl": 0.00022270679473876954,
+      "learning_rate": 2.4946839873611927e-06,
+      "loss": 0.0112,
+      "reward": 0.1265625054948032,
+      "reward_std": 0.2010766550898552,
+      "rewards/accuracy_reward": 0.1265625054948032,
+      "rewards/format_reward": 0.0,
+      "step": 20
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 551.1631965637207,
+      "epoch": 0.42643923240938164,
+      "grad_norm": 0.014589853584766388,
+      "kl": 0.00024003982543945311,
+      "learning_rate": 2.1156192081791355e-06,
+      "loss": 0.0043,
+      "reward": 0.1200892923399806,
+      "reward_std": 0.19121555145829916,
+      "rewards/accuracy_reward": 0.1200892923399806,
+      "rewards/format_reward": 0.0,
+      "step": 25
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 567.554711151123,
+      "epoch": 0.511727078891258,
+      "grad_norm": 0.013035917654633522,
+      "kl": 0.00024027824401855468,
+      "learning_rate": 1.6808050203829845e-06,
+      "loss": 0.0111,
+      "reward": 0.11250000512227416,
+      "reward_std": 0.19112712442874907,
+      "rewards/accuracy_reward": 0.11250000512227416,
+      "rewards/format_reward": 0.0,
+      "step": 30
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 555.3761436462403,
+      "epoch": 0.5970149253731343,
+      "grad_norm": 0.014955726452171803,
+      "kl": 0.0002460002899169922,
+      "learning_rate": 1.2296174432791415e-06,
+      "loss": 0.0133,
+      "reward": 0.11696429131552577,
+      "reward_std": 0.1858937400393188,
+      "rewards/accuracy_reward": 0.11696429131552577,
+      "rewards/format_reward": 0.0,
+      "step": 35
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 547.1129707336426,
+      "epoch": 0.6823027718550106,
+      "grad_norm": 0.014341042377054691,
+      "kl": 0.00026259422302246096,
+      "learning_rate": 8.029152419343472e-07,
+      "loss": 0.0045,
+      "reward": 0.1279017912223935,
+      "reward_std": 0.19265865609049798,
+      "rewards/accuracy_reward": 0.1279017912223935,
+      "rewards/format_reward": 0.0,
+      "step": 40
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 559.8629776000977,
+      "epoch": 0.767590618336887,
+      "grad_norm": 0.014500826597213745,
+      "kl": 0.0002582550048828125,
+      "learning_rate": 4.3933982822017883e-07,
+      "loss": 0.0107,
+      "reward": 0.12209822023287416,
+      "reward_std": 0.19157440066337586,
+      "rewards/accuracy_reward": 0.12209822023287416,
+      "rewards/format_reward": 0.0,
+      "step": 45
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 566.6609634399414,
+      "epoch": 0.8528784648187633,
+      "grad_norm": 0.014964189380407333,
+      "kl": 0.00025937557220458987,
+      "learning_rate": 1.718159615201853e-07,
+      "loss": 0.0148,
+      "reward": 0.12544643357396126,
+      "reward_std": 0.20125691760331393,
+      "rewards/accuracy_reward": 0.12544643357396126,
+      "rewards/format_reward": 0.0,
+      "step": 50
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 555.6741317749023,
+      "epoch": 0.9381663113006397,
+      "grad_norm": 0.012106225825846195,
+      "kl": 0.0002725839614868164,
+      "learning_rate": 2.4570139579284723e-08,
+      "loss": 0.0091,
+      "reward": 0.12388393525034189,
+      "reward_std": 0.1997589457780123,
+      "rewards/accuracy_reward": 0.12388393525034189,
+      "rewards/format_reward": 0.0,
+      "step": 55
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 555.804059346517,
+      "epoch": 0.9893390191897654,
+      "kl": 0.0002773602803548177,
+      "reward": 0.13988095987588167,
+      "reward_std": 0.21711470931768417,
+      "rewards/accuracy_reward": 0.13988095987588167,
+      "rewards/format_reward": 0.0,
+      "step": 58,
+      "total_flos": 0.0,
+      "train_loss": 0.009057209056256146,
+      "train_runtime": 7830.5569,
+      "train_samples_per_second": 0.958,
+      "train_steps_per_second": 0.007
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 58,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 5,
+  "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": 16,
+  "trial_name": null,
+  "trial_params": null
+}