You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Feb 7, 2024. It is now read-only.
I've investigated this issue and there's no way around it without change PyTorch itself.
A python program may do non_torchy_tensor.set_(tochy_tensor). If the torchy tensor isn't materialized, we won't store non_torchy_tensor as a shared tensor on the trace. So when the trace is flushed this tensor hasn't the metadata updated.
Sign up for freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
It required this patch:
But ideally it wouldn't be needed, as
shallow_copy_from
would be called between Torchy objects. So why pickle used different objects?This is the backtrace, while executing a TorchVision model:
The text was updated successfully, but these errors were encountered: