Model save

README.md

@@ -0,0 +1,68 @@
+---
+base_model: GSAI-ML/LLaDA-8B-Instruct
+library_name: transformers
+model_name: LLaDA-8B-Instruct-DLPO
+tags:
+- generated_from_trainer
+- trl
+- dlpo
+licence: license
+---
+
+# Model Card for LLaDA-8B-Instruct-DLPO
+
+This model is a fine-tuned version of [GSAI-ML/LLaDA-8B-Instruct](https://huggingface.co/GSAI-ML/LLaDA-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="howey/LLaDA-8B-Instruct-DLPO", 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/hhy-tue/huggingface/runs/auoqt4fz) 
+
+
+This model was trained with DLPO, 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 DLPO 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": -0.004999998807907104,
+    "train_runtime": 1556.2446,
+    "train_samples": 93733,
+    "train_samples_per_second": 0.074,
+    "train_steps_per_second": 0.016
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 126080,
+  "eos_token_id": 126081,
+  "transformers_version": "4.49.0"
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "total_flos": 0.0,
+    "train_loss": -0.004999998807907104,
+    "train_runtime": 1556.2446,
+    "train_samples": 93733,
+    "train_samples_per_second": 0.074,
+    "train_steps_per_second": 0.016
+}

trainer_state.json

@@ -0,0 +1,189 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.869565217391305,
+  "eval_steps": 500,
+  "global_step": 25,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.75,
+      "epoch": 0.17391304347826086,
+      "grad_norm": 2422.725830078125,
+      "learning_rate": 6.666666666666666e-07,
+      "loss": 0.0,
+      "reward": 0.25,
+      "rewards/accuracy_reward": 0.25,
+      "step": 1
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 376.25,
+      "epoch": 0.34782608695652173,
+      "grad_norm": 321.9919738769531,
+      "learning_rate": 1.3333333333333332e-06,
+      "loss": -0.25,
+      "reward": 0.25,
+      "rewards/accuracy_reward": 0.25,
+      "step": 2
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 347.125,
+      "epoch": 0.6956521739130435,
+      "grad_norm": 527.0032958984375,
+      "learning_rate": 1.9898214418809326e-06,
+      "loss": -0.125,
+      "reward": 0.125,
+      "rewards/accuracy_reward": 0.125,
+      "step": 4
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 307.0,
+      "epoch": 1.1739130434782608,
+      "grad_norm": 0.0,
+      "learning_rate": 1.9096319953545185e-06,
+      "loss": 0.0625,
+      "reward": 0.0,
+      "rewards/accuracy_reward": 0.0,
+      "step": 6
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 383.625,
+      "epoch": 1.5217391304347827,
+      "grad_norm": 8402.4208984375,
+      "learning_rate": 1.7557495743542582e-06,
+      "loss": 0.125,
+      "reward": 0.0,
+      "rewards/accuracy_reward": 0.0,
+      "step": 8
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 270.125,
+      "epoch": 1.8695652173913042,
+      "grad_norm": 255.75051879882812,
+      "learning_rate": 1.5406408174555977e-06,
+      "loss": 0.0,
+      "reward": 0.125,
+      "rewards/accuracy_reward": 0.125,
+      "step": 10
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 250.875,
+      "epoch": 2.3478260869565215,
+      "grad_norm": 0.0,
+      "learning_rate": 1.2817325568414297e-06,
+      "loss": 0.0,
+      "reward": 0.0,
+      "rewards/accuracy_reward": 0.0,
+      "step": 12
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 270.0,
+      "epoch": 2.6956521739130435,
+      "grad_norm": 0.0,
+      "learning_rate": 1e-06,
+      "loss": 0.125,
+      "reward": 0.125,
+      "rewards/accuracy_reward": 0.125,
+      "step": 14
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 157.25,
+      "epoch": 3.1739130434782608,
+      "grad_norm": 0.0,
+      "learning_rate": 7.182674431585702e-07,
+      "loss": 0.0,
+      "reward": 0.0,
+      "rewards/accuracy_reward": 0.0,
+      "step": 16
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 232.75,
+      "epoch": 3.5217391304347827,
+      "grad_norm": 0.0,
+      "learning_rate": 4.5935918254440274e-07,
+      "loss": 0.0,
+      "reward": 0.125,
+      "rewards/accuracy_reward": 0.125,
+      "step": 18
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 253.5,
+      "epoch": 3.869565217391304,
+      "grad_norm": 1907.75390625,
+      "learning_rate": 2.4425042564574185e-07,
+      "loss": -0.1875,
+      "reward": 0.125,
+      "rewards/accuracy_reward": 0.125,
+      "step": 20
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 327.5,
+      "epoch": 4.3478260869565215,
+      "grad_norm": 6329.58837890625,
+      "learning_rate": 9.036800464548156e-08,
+      "loss": 0.0625,
+      "reward": 0.125,
+      "rewards/accuracy_reward": 0.125,
+      "step": 22
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 242.5,
+      "epoch": 4.695652173913043,
+      "grad_norm": 0.0,
+      "learning_rate": 1.0178558119067315e-08,
+      "loss": -0.125,
+      "reward": 0.125,
+      "rewards/accuracy_reward": 0.125,
+      "step": 24
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 277.0,
+      "epoch": 4.869565217391305,
+      "reward": 0.0,
+      "rewards/accuracy_reward": 0.0,
+      "step": 25,
+      "total_flos": 0.0,
+      "train_loss": -0.004999998807907104,
+      "train_runtime": 1556.2446,
+      "train_samples_per_second": 0.074,
+      "train_steps_per_second": 0.016
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 25,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "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": 1,
+  "trial_name": null,
+  "trial_params": null
+}