Michael Lazos
Michael Lazos
Kicked off a run here: https://github.com/pytorch/pytorch/actions/runs/8977631712
Couldn't this change be a one line change to `generic_jump` that just checks if the VariableTracker is a UserDefinedObjectVariable and then you jump? Why are the other code changes necessary?...
In this case the model is calling `torch.jit.isinstance` which handles isinstance typing containers properly. Currently looking to inline that.
I opted to just graph break on it, supporting it was taking more time than it was worth when the graph break at least eliminates the error and allows compilation...
@pytorchbot merge -f "Unrelated test failure"
> jolly good show, I was wondering about what this was used for lol Yee, it's already managed properly in restore, so it doesn't need to be saved basically nodes...
@stas00 Thanks for the suggestions, I think these all seem super useful and are doable. You're correct that this is not for everything, this is initially for the PyTorch 2.0...
> Are there any configs that can be removed because of this change? e.g. `log_level` or `output_code`? Yeah good catch, I will remove these
> This looks super useful to me, thank you for doing it! A very minor suggestion (🚲🏠) is to use + for increasing the log level instead of >. I...
cc @stas00 @albanD @ezyang @williamwen42 I have a [design doc](https://docs.google.com/document/d/e/2PACX-1vTVFQSOoIGNscL72Xa9_R92ahNAjVZ_5GGFyxn32DNiSMi2QTD9mHVur6dEv1aYmg4XHqy6xdO4MPnu/pub) with more detail