search

microsoft / onnxruntime

ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
https://onnxruntime.ai
MIT License
14.05k stars 2.83k forks source link

Example script for quantization aware training #7996

Open rohanshingade opened 3 years ago

rohanshingade commented 3 years ago

Is your feature request related to a problem? Please describe. I'm trying to perform quantization. I converted the model using quantize_qat, but not able to figure out how to do training using the quantized model. Any example script related to this will be very helpful.

System information

Describe alternatives you've considered I checked the web for resources similar to this, how to perform quantization aware training using onnxruntime. didn't get anything relevant.

yufenglee commented 3 years ago

@rohanshingade , onnxruntime doesn't support retraining now. You need to use the original framework to do the QAT retraining.

rohanshingade commented 3 years ago

@yufenglee that means i need to do QAT in pytorch and then convert to onnx model, instead of QAT using onnxruntime. Is this what you are suggesting?

stale[bot] commented 2 years ago

This issue has been automatically marked as stale due to inactivity and will be closed in 7 days if no further activity occurs. If further support is needed, please provide an update and/or more details.