Msabih

@ptrblck @mcarilli I am getting the same error but strangely this error always comes after 1 training epoch. First epoch is always fine. Also this goes away if I use...

@ptrblck I figured out the issue for my case. The problem occurs if someone is training and validating in one epoch and model.eval() is called in validation, for the next...

@mcarilli I got the error regardless of the opt_level on both O1 and O2. As i said, the error was due to me not calling train() after eval() for the...

@rbaghdadi @IHIaadj Is there lets say a very rough tentative time for this integration like April or May or June ?