[v3.x] Cannot load whisper-v3-large-turbo

[liuhuapiaoyuan]

System Info

3.x

Environment/Platform

• Website/web-app
• Browser extension
• Server-side (e.g., Node.js, Deno, Bun)
• Desktop app (e.g., Electron)
• Other (e.g., VSCode extension)

Description

Can't create a session. ERROR_CODE: 1, ERROR_MESSAGE: Deserialize tensor onnx::MatMul_4069 failed.Failed to load external data file ""encoder_model.onnx_data"", error: Module.MountedFiles is not available.

Reproduction

pipeline(
'automatic-speech-recognition',
'onnx-community/whisper-large-v3-turbo',
{ progress_callback,device:"webgpu" },
)

[xenova]

Hi there 👋 There are 2 ways to fix this:

1. Use a dtype which is < 2GB:

   pipeline(
       'automatic-speech-recognition',
       'onnx-community/whisper-large-v3-turbo',
       {
           progress_callback,
           device:"webgpu",
           dtype: {
               encoder_model: "fp16",
               decoder_model_merged: "q4",
           }
       },
   )

2. Use external data format for the fp32 encoder (which is > 2GB):

   pipeline(
       'automatic-speech-recognition',
       'onnx-community/whisper-large-v3-turbo',
       {
           progress_callback,
           device:"webgpu",
           use_external_data_format: {
               encoder_model: true,
               decoder_model_merged: false,
           }
       },
   )

---

Let me know if that helps!