Kark07/dapo_phi3.5_code

1---2base_model: microsoft/Phi-3.5-mini-instruct3library_name: peft4model_name: dapo_phi3.5_code5tags:6- base_model:adapter:microsoft/Phi-3.5-mini-instruct7- grpo8- lora9- transformers10- trl11licence: license12pipeline_tag: text-generation13---14 15# Model Card for dapo_phi3.5_code16 17This model is a fine-tuned version of [microsoft/Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct).18It has been trained using [TRL](https://github.com/huggingface/trl).19 20## Quick start21 22```python23from transformers import pipeline24 25question = "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?"26generator = pipeline("text-generation", model="None", device="cuda")27output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]28print(output["generated_text"])29```30 31## Training procedure32 33 34 35 36This 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).37 38### Framework versions39 40- PEFT 0.18.141- TRL: 0.24.042- Transformers: 4.57.643- Pytorch: 2.9.144- Datasets: 4.3.045- Tokenizers: 0.22.246 47## Citations48 49Cite GRPO as:50 51```bibtex52@article{shao2024deepseekmath,53    title        = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},54    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},55    year         = 2024,56    eprint       = {arXiv:2402.03300},57}58 59```60 61Cite TRL as:62    63```bibtex64@misc{vonwerra2022trl,65	title        = {{TRL: Transformer Reinforcement Learning}},66	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 Gallou{\'e}dec},67	year         = 2020,68	journal      = {GitHub repository},69	publisher    = {GitHub},70	howpublished = {\url{https://github.com/huggingface/trl}}71}72```