Nicolas Hug

Also @ichamatidis : > `torch.hub.load('pytorch/vision:v1.9.0','resnet50', pretrained=True)` This should be either `vision:v0.9.0` or `vision:v0.10.0`, as torchvision's version 1.9 doesn't exist.

Thank you for the report @gurdeepmaurya . As you noted, this is largely due to an unsolved issue on the pytorch Dataloader's side and there isn't much that can be...

Hi @mortal-Zero , thanks for the proposal. Trying to figure out the scope here, can you share more details about where in your training pipeline you're using these functions? E.g....

Hi @vritansh, thanks for the feature request. Sure, we can pass the `print-freq` arg to `evaluate()` as well. Feel free to submit a PR. > finding it difficult to see...

Hi @8uurg , I guess these come from diverging version of libjpeg[-turbo]. Could you please share the PIL version, as well as the output of `ldd _imaging.so` (from https://stackoverflow.com/a/24397115) Could...

Thanks for the output > libjpeg.so.9 => /lib/python3.10/site-packages/PIL/../../../libjpeg.so.9 (0x00007f1b34cad000) I think that's it: PIL is relying on `libjpeg` while torchvision is relying on `libjpeg-turbo`. They're both jpeg-compliant and, from past...

Hi @davideboschetto , looking at https://github.com/pytorch/vision/pull/2113 and the other conversations linked in this PR, it seems that 1/9 was chosen mostly for its empirical performance

Thanks a lot for this great investigation Philip. @lezcano I tend to have a different intuition from yours: if `resize` is much faster than `compiled(resize)`, then perhaps the speed-up gained...

@cleebp you'll need pytorch 2.0 if you're using torchvision 0.15 - you can refer to our compatibility table [here](https://github.com/pytorch/vision#installation)

Hi @ganessh22 , Using cuda tensors within a multi-processing context is not really supported unfortuntately. See e.g. https://pytorch.org/docs/stable/data.html#multi-process-data-loading and the other resources linked from there > It is generally not...