Hi,So happy to have discovered this place
I quantized the whisper model according to the following command, but an error will be reported when importing the following model into unity.
optimum-cli export onnx --model ./whisper-tiny/ --task automatic-speech-recognition whisper-tiny-onnx
optimum-cli onnxruntime quantize --arm64 --onnx_model whisper-tiny-onnx -o qu_whisper-tiny-onnx
Models before and after quantification will be provided later:)
Here are some links that may be helpful:
optimum
Sentis does not support importing quantized models. You should use the Quantization API in version 1.4.0-pre.3 to quantize models in Sentis.
https://docs.unity3d.com/Packages/com.unity.sentis@1.4/manual/quantize-a-model.html
Hello, I used the above API to quantify the whisper-base model, but it failed. Normal onnx before quantification can be used for inference. What is the problem and how should I deal with it?
Below is my code and related error reports
Here is the model I used
optimum-cli export onnx --model ./whisper-tiny/ --task automatic-speech-recognition whisper-tiny-onnx
optimum-cli onnxruntime quantize --arm64 --onnx_model whisper-tiny-onnx -o qu_whisper-tiny-onnx
Interesting, we’ll take a look.
Thanks for the repro case
Hello, I saw that Sentis 1.5 has been updated. I would like to ask if the above problem has been fixed?
