Luca Bonfiglioli

bruh

Thank you, I will definitely try that next week. Another thing that might give some clue on the reported bug: for now I got it "fixed" by setting a warmup...

> Yeah, we noticed that as well with autofocus initially enabled, sometimes it needs more than one `set` command to get to take effect. It's some bug on the FW...

Why would there be a nn.Module among the submodules after calling `torch.jit.script`? Anyway, I am currently trying to simplify the code while still reproducing the bug to provide you with...

@narendasan Yes I think you understood correctly. The main reason why I don't script the whole thing before compiling with TRT is that to make it compile with TRT it...

please

I keep getting the exact same error with pytorch==1.13.1+cu117. This happens randomly roughly once every 7-8 minutes of training.

Yep, that's pretty much how I solved it too. But having to manually check each time a Compose is called should not be client code responsibility in my opinion. Also,...

I think this issue should at least get some consideration: this is a serious design flaw that hinders the possibility of using `rich` package to create any reasonably robust library/application....

I am switching back to TQDM, uglier but much more reliable. No one needs fancy graphics if the underlying software is rotting. Clearly `rich` is something one should avoid.