QwenLM / Qwen2.5-Coder

Qwen2.5-Coder is the code version of Qwen2.5, the large language model series developed by Qwen team, Alibaba Cloud.
819 stars 74 forks source link

关于补全模型微调 #82

Closed Soulscb closed 3 months ago

Soulscb commented 4 months ago

你好,我们想微调base模型的补全能力,但是发现你们的chat模型有单独的格式,请问需要遵从你们的微调格式吗?

cyente commented 3 months ago

建议暂时不要使用微调格式

80

Soulscb commented 3 months ago

感谢您的回复,请问数据拼接过程中,bos和eos的拼接是怎么做的呢?微调FIM的格式是否是 bos + input + output + eos

cyente commented 3 months ago

https://github.com/QwenLM/CodeQwen1.5/blob/main/examples/CodeQwen1.5-base-fim.py

fim的格式参考此处example