Update tokenizer.py to decode list of tensors as well as single tensor

turboderp / exllamav2

A fast inference library for running LLMs locally on modern consumer-class GPUs

MIT License

3.23k stars 238 forks source link

Update tokenizer.py to decode list of tensors as well as single tensor #296

Closed ba2512005 closed 5 months ago

ba2512005 commented 5 months ago

This is an enhancement to fix functionality of Lollms-webui, specifically the lord of internet and wikipedia personalities, as well as anything else that uses workflows or returns list of tensors to be decoded.

This change identifies the type, and decodes based on the type. If a list is detected it calls itself recursively and decodes each item in the list until completion.

https://github.com/ParisNeo/lollms-webui/issues/480

ba2512005 commented 5 months ago

@turboderp please review and approve