feat: Support `top_p` and `presence_pennalty` configure in the ggml plugin
alabulei1 opened this issue · comments
alabulei1 commented
Summary
This feature is requested from a user.
The popular web chatbot ui framework like ChatGPTNextWeb and Lobe Hub supports top_p
and presence_pennalty
. Because LlamaEdge doesn't support these two parameters, so the api server created by LlamaEdge can't be used in above two framework.
I think it'd better if we can support these two parameters
Details
Example:
-d '{
"messages": [{"role": "user", "content": "李白 诗仙代表作"}],
"model": "Mixtral-8x7B-Instruct-v0.1",
"temperature": 0.7,
"stream": true,
"presence_penalty": 0,
"frequency_penalty": 0,
"top_p": 1
}'
Appendix
No response