search

dbolya / yolact

A simple, fully convolutional model for real-time instance segmentation.
MIT License
5k stars 1.33k forks source link

Quantization: Train on cpu for post training quantization #484

Open sabine1993 opened 4 years ago

sabine1993 commented 4 years ago

Hey everyone, I want to train the model on CPU to use the pytorch post dynamic quantization afterwards. and then run the eval.py on cuda to improve the inference.

I am having trouble running the model only on cpu.

Does any one have suggestions how I can run it only on the cpu? And did anyone quantizise the model already?

Thanks a lot for your help

abhigoku10 commented 4 years ago

@sabine1993 had to modify the code to run on cpu , did not quantize the model , do let us knw if you find any progress