search

UnicomAI / Unichat-llama3-Chinese

Apache License 2.0
340 stars 34 forks source link

回答的答案中夹杂很多重复网址信息 #10

Open Belonger opened 4 months ago

Belonger commented 4 months ago

1

Belonger commented 4 months ago

这个是什么原因造成的？

zhqfdn commented 4 months ago

而且太多对话还到不了点子上

UnicomAI commented 4 months ago

这个是什么原因造成的？

这个不是llama3模板，自己的模板，看下用对没

UnicomAI commented 4 months ago

而且太多对话还到不了点子上

只是做了微调，后面会用中文做二次预训练

Belonger commented 4 months ago

@UnicomAI 我用的是llama3的 prompt template 1

Belonger commented 4 months ago

难道是我用错了吗？

UnicomAI commented 4 months ago

难道是我用错了吗？

跑下主页代码，打印prompt看下，template在tokenizer_config.json

Belonger commented 4 months ago

@UnicomAI 2

Belonger commented 4 months ago

@UnicomAI 可以加个联系方式吗，

Belonger commented 4 months ago

@UnicomAI 3

UnicomAI commented 4 months ago

@UnicomAI 可以加个联系方式吗，

qq群：635964480

Belonger commented 4 months ago

感谢博主，已解决该问题，如果用的llama factory工具，把_register_template( name="llama3-unichat", format_user=StringFormatter(slots=["Human:{{content}}\nAssistant:"]), format_assistant=StringFormatter(slots=["{{content}}<|end_of_text|>"]), format_system=StringFormatter(slots=["<|begin_of_text|>{{content}}"]), default_system="A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions.\n" )注册进template.py就可以， 4