Closed tonyyuandao closed 4 years ago
adding num_workers = 0
to dls fixed the issue. Pytorch multiprocess don't work on windows.
dls = ImageDataLoaders.from_name_func( path, get_image_files(path), valid_pct=0.2, seed=42, label_func=is_cat, item_tfms=Resize(224),num_workers = 0)
Win10, fresh start, just installed latest anaconda (python3.8), latest pytorch, and pip install fastai. (Failed at conda install, saying something like conflict with cuda-driver 10.2, so I tried pip and successed)
Then I made through to the code. "Running Your First Notebook"
Result:
Is this a pytorch problem? I tried
torch.cuda.is_available()
, it returned True.