Model save

README.md

@@ -0,0 +1,68 @@
+---
+base_model: sai-lakkshmii/Finetuned-llama-8b_SFT_step5000
+library_name: transformers
+model_name: Final-try-Llama3.1-8b-instruct-RL
+tags:
+- generated_from_trainer
+- trl
+- grpo
+licence: license
+---
+
+# Model Card for Final-try-Llama3.1-8b-instruct-RL
+
+This model is a fine-tuned version of [sai-lakkshmii/Finetuned-llama-8b_SFT_step5000](https://huggingface.co/sai-lakkshmii/Finetuned-llama-8b_SFT_step5000).
+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/Final-try-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/qvf6i4dv) 
+
+
+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": -0.0886706564703892,
+    "train_runtime": 9867.8585,
+    "train_samples": 7500,
+    "train_samples_per_second": 0.76,
+    "train_steps_per_second": 0.006
+}

generation_config.json

@@ -0,0 +1,12 @@
+{
+  "bos_token_id": 128000,
+  "do_sample": true,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "temperature": 0.6,
+  "top_p": 0.9,
+  "transformers_version": "4.49.0"
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "total_flos": 0.0,
+    "train_loss": -0.0886706564703892,
+    "train_runtime": 9867.8585,
+    "train_samples": 7500,
+    "train_samples_per_second": 0.76,
+    "train_steps_per_second": 0.006
+}

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": 465.1015853881836,
+      "epoch": 0.017057569296375266,
+      "grad_norm": 0.46851029992103577,
+      "kl": 0.0,
+      "learning_rate": 5e-07,
+      "loss": 0.0311,
+      "reward": 0.3024553693830967,
+      "reward_std": 0.2792186979204416,
+      "rewards/accuracy_reward": 0.3024553693830967,
+      "rewards/format_reward": 0.0,
+      "step": 1
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 479.50532627105713,
+      "epoch": 0.08528784648187633,
+      "grad_norm": 4.368417263031006,
+      "kl": 0.011774897575378418,
+      "learning_rate": 2.5e-06,
+      "loss": 0.0178,
+      "reward": 0.3108259071595967,
+      "reward_std": 0.282580086030066,
+      "rewards/accuracy_reward": 0.3108259071595967,
+      "rewards/format_reward": 0.0,
+      "step": 5
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 493.8828353881836,
+      "epoch": 0.17057569296375266,
+      "grad_norm": 0.43728089332580566,
+      "kl": 0.0563232421875,
+      "learning_rate": 2.956412726139078e-06,
+      "loss": 0.0367,
+      "reward": 0.5104910925030708,
+      "reward_std": 0.28240326046943665,
+      "rewards/accuracy_reward": 0.5104910925030708,
+      "rewards/format_reward": 0.0,
+      "step": 10
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 539.2189956665039,
+      "epoch": 0.255863539445629,
+      "grad_norm": 3.359680652618408,
+      "kl": 0.243646240234375,
+      "learning_rate": 2.7836719084521715e-06,
+      "loss": 0.0724,
+      "reward": 0.44196430668234826,
+      "reward_std": 0.2963893756270409,
+      "rewards/accuracy_reward": 0.44196430668234826,
+      "rewards/format_reward": 0.0,
+      "step": 15
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 636.4513717651367,
+      "epoch": 0.3411513859275053,
+      "grad_norm": 2.16961669921875,
+      "kl": 0.4969482421875,
+      "learning_rate": 2.4946839873611927e-06,
+      "loss": 0.1423,
+      "reward": 0.31763394474983214,
+      "reward_std": 0.28798535354435445,
+      "rewards/accuracy_reward": 0.31763394474983214,
+      "rewards/format_reward": 0.0,
+      "step": 20
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 533.4765838623047,
+      "epoch": 0.42643923240938164,
+      "grad_norm": 8.851544380187988,
+      "kl": 0.75478515625,
+      "learning_rate": 2.1156192081791355e-06,
+      "loss": 0.0974,
+      "reward": 0.3165178719907999,
+      "reward_std": 0.2697611689567566,
+      "rewards/accuracy_reward": 0.3165178719907999,
+      "rewards/format_reward": 0.0,
+      "step": 25
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 486.86542587280275,
+      "epoch": 0.511727078891258,
+      "grad_norm": 1.5197241306304932,
+      "kl": 0.635107421875,
+      "learning_rate": 1.6808050203829845e-06,
+      "loss": -0.0795,
+      "reward": 0.37879465594887735,
+      "reward_std": 0.28313881531357765,
+      "rewards/accuracy_reward": 0.37879465594887735,
+      "rewards/format_reward": 0.0,
+      "step": 30
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 651.0556098937989,
+      "epoch": 0.5970149253731343,
+      "grad_norm": 251.43299865722656,
+      "kl": 1.340673828125,
+      "learning_rate": 1.2296174432791415e-06,
+      "loss": -0.1861,
+      "reward": 0.35892858989536763,
+      "reward_std": 0.2909108128398657,
+      "rewards/accuracy_reward": 0.35892858989536763,
+      "rewards/format_reward": 0.0,
+      "step": 35
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 531.6004737854004,
+      "epoch": 0.6823027718550106,
+      "grad_norm": 0.8638615012168884,
+      "kl": 1.24169921875,
+      "learning_rate": 8.029152419343472e-07,
+      "loss": -0.2823,
+      "reward": 0.2919642984867096,
+      "reward_std": 0.30123363584280016,
+      "rewards/accuracy_reward": 0.2919642984867096,
+      "rewards/format_reward": 0.0,
+      "step": 40
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 535.4821655273438,
+      "epoch": 0.767590618336887,
+      "grad_norm": 0.8260723948478699,
+      "kl": 1.6283203125,
+      "learning_rate": 4.3933982822017883e-07,
+      "loss": -0.2886,
+      "reward": 0.2575892962515354,
+      "reward_std": 0.28938031755387783,
+      "rewards/accuracy_reward": 0.2575892962515354,
+      "rewards/format_reward": 0.0,
+      "step": 45
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 574.5600708007812,
+      "epoch": 0.8528784648187633,
+      "grad_norm": 0.8665443658828735,
+      "kl": 1.6455078125,
+      "learning_rate": 1.718159615201853e-07,
+      "loss": -0.2452,
+      "reward": 0.2620535843074322,
+      "reward_std": 0.2727984722703695,
+      "rewards/accuracy_reward": 0.2620535843074322,
+      "rewards/format_reward": 0.0,
+      "step": 50
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 633.8752510070801,
+      "epoch": 0.9381663113006397,
+      "grad_norm": 0.7306084036827087,
+      "kl": 1.2185546875,
+      "learning_rate": 2.4570139579284723e-08,
+      "loss": -0.1955,
+      "reward": 0.2723214395344257,
+      "reward_std": 0.2689169988036156,
+      "rewards/accuracy_reward": 0.2723214395344257,
+      "rewards/format_reward": 0.0,
+      "step": 55
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 655.2425740559896,
+      "epoch": 0.9893390191897654,
+      "kl": 1.1414388020833333,
+      "reward": 0.2812500136593978,
+      "reward_std": 0.2797961321969827,
+      "rewards/accuracy_reward": 0.2812500136593978,
+      "rewards/format_reward": 0.0,
+      "step": 58,
+      "total_flos": 0.0,
+      "train_loss": -0.0886706564703892,
+      "train_runtime": 9867.8585,
+      "train_samples_per_second": 0.76,
+      "train_steps_per_second": 0.006
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 58,
+  "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": 16,
+  "trial_name": null,
+  "trial_params": null
+}