Yiximail

### Describe the bug I encountered an issue when attempting to use the `llama-65b-4bit-128g` model. The output results I generated were mostly composed of empty spaces or symbols.  I...

llama3 instruct models need multiple `eos_token_id` to make the output stop correctly

20

# Quick fix for llama3 doesn't stop correctly change `tokenizer_config.json` from `"eos_token": "",` to `"eos_token": "",` it should works If it's just a tokenizer bug, this enhancement issue doesn't need...

I tried to write a demo using the latest version of `Electron` and moved the recognize logic from the `index.html` to `main.js`.

The feature same as the impersonate feature on webui. Add an `impersonate` param to active this feature in `/v1/chat/completions`. Updated: Prepend `user_input` in impersonate reply for consistency It will always...