Spaces:

---

Duplicated from  autotrain-projects/autotrain-advanced

Hankbeasley

/

polycrest-coder

like 0

Sleeping

I would like to attempt to train the Qwen32B-r1 distilled model for better coding. Using a rewards training. I am going to attempt to get the Qwen32B model to out preform Deep Seek R1 on code test via reinforcement learning. I have gathered hundreds of accept/reject samples via automated testing of the coding exercises at https://github.com/exercism/ . I feel like I will need around 150GB vram in order to complete the training.