search

cfahlgren1 / webllm-playground

Run LLMs in the Browser with MLC / WebLLM ✨
https://huggingface.co/spaces/cfahlgren1/webllm-playground
91 stars 11 forks source link

Blazing fast with cool UI #3

Open Bhargav-Ravinuthala opened 1 month ago

Bhargav-Ravinuthala commented 1 month ago

Kudos for the wonderful work... This looks amazing!

As a newbie I wonder some of the models like Qwen or llava supports images as inputs. How useful it is developing such feature for this application to accept images as input along with the prompt?

cfahlgren1 commented 1 month ago

WebLLM has an example with Phi-3.5-Vision here.

Currently the playground is filtering out vision models here https://github.com/cfahlgren1/webllm-playground/blob/main/src/utils/llm.ts#L97-L99, but this would be a very neat feature as part of the demo to include a couple images and enable vision models. Feel free to take a stab at it if interested!