search

huggingface / optimum-quanto

A pytorch quantization backend for optimum
Apache License 2.0
645 stars 36 forks source link

Got stuck when train resnet50 with QAT #183

Closed catsled closed 1 month ago

catsled commented 2 months ago

When I quantize a ResNet50 model and train it, it may get stuck at the first backward pass; it's normal when I train a non-quantized model..

GPU:A800 data:Cifar10

image

image

how should i do?

dacorvo commented 2 months ago

It is strange that you get integer output. Did you take a look at the MNIST example ? If not, maybe compare with your script and see if something is different.

catsled commented 2 months ago

It is strange that you get integer output. Did you take a look at the MNIST example ? If not, maybe compare with your script and see if something is different.

ok : )

catsled commented 2 months ago

got stuck when calibration(i did't add it before ) image image

dacorvo commented 2 months ago

How long does it take to run evaluate on the float model ?

catsled commented 2 months ago

How long does it take to run evaluate on the float model ?

about 8s

github-actions[bot] commented 1 month ago

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 5 days.

github-actions[bot] commented 1 month ago

This issue was closed because it has been stalled for 5 days with no activity.