Puffin-Phi-v2 running on the browser with Wasm!!
#4
by
radames
- opened
Hi
@teknium
,
We have been working on a ML framework in Rust called Candle.
@lmz
, has implemented the Phi architecture in the framework. With Puffin-Phi-v2 quantized model, you can run it completely offline on your browser using Wasm. Candle makes it easy target the build to Wasm.
Here is the demo:
https://huggingface.co/spaces/radames/Candle-Phi-1.5-Wasm
The responses are remarkably better when using Puffin-Phi-v2 compared to the base Phi! (Tested with Candle Wasm quantized)
Thank you for setting that up!