What is the advantage of fine-tuning based on coder-instruct model rather than instruct?
#4
by
sangmini
- opened
Hello!
I'm curious about why models are trained on top of Qwen/Qwen2.5-Coder-7B-Instruct rather than the standard Qwen/Qwen2.5-7B-Instruct. I'd like to understand if there are references or evidence suggesting that coder-instruct models provide better performance as base models for fine-tuning.
Thanks
Thank you for your question! Based on our experiments, we found that fine-tuning on the Qwen/Qwen2.5-Coder-7B-Instruct base model yields better performance for function calling tasks compared to using the standard Qwen2.5-7B-Instruct model. This aligns with observations mentioned in XLAM-related research papers.