Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Web] Can't use Conv3DNaive on webgpu #22974

Open
pdufour opened this issue Nov 29, 2024 · 1 comment
Open

[Web] Can't use Conv3DNaive on webgpu #22974

pdufour opened this issue Nov 29, 2024 · 1 comment
Assignees
Labels
ep:WebGPU ort-web webgpu provider platform:web issues related to ONNX Runtime web; typically submitted using template

Comments

@pdufour
Copy link

pdufour commented Nov 29, 2024

Describe the issue

When there is a Conv3d op in an q4f16 quantized onnx model, it throws the following error:

An uncaught WebGPU validation error was raised: Error while parsing WGSL: :51:15 error: return statement type must match its function return type, returned 'f16', expected 'f32'
              return get_xByIndices(aIndices);
              ^^^^^^


 - While validating [ShaderModuleDescriptor ""Conv3DNaive""]
 - While calling [Device].CreateShaderModule([ShaderModuleDescriptor ""Conv3DNaive""]).

I think it's this op here specifically that is causing the error:
Image

This does not happen on WASM, but only on the webgpu exec provider.

To reproduce

Visit https://huggingface.co/spaces/pdufour/Qwen2VL_TransformersJS_Demo (space I setup to use this model)
Select q4f16 model to load
Select example image and type in text and hit enter

Actual results
See error mentioned above

Expected results
Should process query

ONNX Runtime Version or Commit ID

1.20.1

Execution Provider

'webgpu' (WebGPU)

@pdufour pdufour added the platform:web issues related to ONNX Runtime web; typically submitted using template label Nov 29, 2024
@github-actions github-actions bot added the ep:WebGPU ort-web webgpu provider label Nov 29, 2024
@fs-eire
Copy link
Contributor

fs-eire commented Dec 2, 2024

Need to implement f16 support for Conv3D.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
ep:WebGPU ort-web webgpu provider platform:web issues related to ONNX Runtime web; typically submitted using template
Projects
None yet
Development

No branches or pull requests

2 participants