A webgpu implementation of phi-3 mini-4k based on Xenova transformers.js
9.0K
Based on Xenova/experimental-phi3-webgpu , this is a docker standalone version of phi-3-mini-4k quantized 4bits that can run offline with webgpu compatible browser (Chrome, Brave...)
The container is AMD64 but works with Apple silicon.
#Run it
docker -p 8000:8000 rdmediation/phi3webgpu
| Machine | token/s |
|---|---|
| Macbook Intel i5 | 3 |
| Macbook Apple Silicon | 32 |
Content type
Image
Digest
sha256:c151ca8e1…
Size
3.9 GB
Last updated
about 2 years ago
docker pull rdmediation/phi3webgpu