README.md

metadata

library_name: transformers
datasets:
  - PowerInfer/QWQ-LONGCOT-500K
  - PowerInfer/LONGCOT-Refine-500K
base_model:
  - Qwen/Qwen2.5-7B-Instruct
license: apache-2.0

Qwen2.5-7B-Instruct-QwQ

A QwQ style model trained from Qwen/Qwen2.5-7B-Instruct

• 1.7 epoch on PowerInfer/QWQ-LONGCOT-500K and PowerInfer/LONGCOT-Refine-500K
• recommended parameters: temperature=0.7 top_p=0.8 repetition_penalty=1.1 max_tokens=16384
• AIME24: 35.33% (average of 5 runs)