I'm unable to start an inference session in an audio worklet context. I'm not sure what all of the obstacles to getting onnxruntime working in that context but I think the current build system may assume there is a global object called "self" that doesn't exist in that context. There are also other unusual characteristics of that context, some of which are documented here.
To reproduce
Here is a repo with my best attempt to run onnxruntime-web in an audio worklet. I end up with this error in the console
Error: no available backend found. ERR: [wasm] TypeError: o.cpus is not a function worklet.js line 19 > eval:85:11
Describe the issue
I'm unable to start an inference session in an audio worklet context. I'm not sure what all of the obstacles to getting onnxruntime working in that context but I think the current build system may assume there is a global object called "self" that doesn't exist in that context. There are also other unusual characteristics of that context, some of which are documented here.
To reproduce
Here is a repo with my best attempt to run onnxruntime-web in an audio worklet. I end up with this error in the console
Urgency
No response
ONNX Runtime Installation
Released Package
ONNX Runtime Version or Commit ID
"onnxruntime-web": "^1.12.1"
Execution Provider
WASM