skottmckay
commented
1 year ago onnxruntime-gpu is really onnxruntime-cuda.
The CoreML execution provider is the best option but depends on the model as to which operators can use CoreML. Are you able to share the model?
Alternatively if you set log severity to verbose and look for 'Node placements' in the output you can see how many nodes are able to run on CoreML. If ORT has to switch between the CPU execution provider and CoreML execution provider that will have a significant performance penalty.
DeepFace-user
vladiliescu
Describe the issue
onnxruntime-gpu only in the case of window cuda, why not Mac m1, not support, only install onnxruntime, onnx model rendering performance is very slow, very slow; Can't understand, is Mac m1 currently not supported?
To reproduce
Reasoning is slow, slow to death
Urgency
No response
Platform
Mac
OS Version
13.4.1
ONNX Runtime Installation
Built from Source
ONNX Runtime Version or Commit ID
onnxruntime-1.15.1
ONNX Runtime API
Python
Architecture
ARM64
Execution Provider
CoreML
Execution Provider Library Version
No response
Model File
No response
Is this a quantized model?
Yes