Reza2kn's picture
Use real FP16 ONNX WebGPU artifact
a7fb037 verified
|
Raw
History Blame Contribute Delete
642 Bytes
# Deploy
Create a new Hugging Face Space with SDK `gradio`, then upload this folder:
```bash
huggingface-cli repo create visualears-fastconformer-fa-webgpu --type space --space_sdk gradio
git clone https://huggingface.co/spaces/Reza2kn/visualears-fastconformer-fa-webgpu
rsync -av --delete visualears-webgpu-space/ visualears-fastconformer-fa-webgpu/
cd visualears-fastconformer-fa-webgpu
git add .
git commit -m "Add VisualEars WebGPU ASR demo"
git push
```
The app downloads the ONNX model from:
`Reza2kn/visualears-fastconformer-fa-full-ab-onnx-fp16`
The first browser load fetches about 461 MB. The browser should cache later loads.