I implemented a subclass of torch.autograd.Function class and its forward and backward function. I use Pycharm IDE. I add a breakpoint in backward function and wanted to debug the script , then I found the program did not stop at the breakpoint . I do not know why. It works well in other part of the script,even in forward function.
I thought it is the pycharm’s problem ,but when I run the script in another computer with pycharm ,the problem disappear . I mean it stop just the breakpoint.
The problem drives me mad. so if anyone how to solve the problem ,please help me.
I firstly used the pycharm 2017.1.2 and pytorch 0.1.11_5, then I ran the code in the computer with pycharm 2016.2.3 and pytorch 0.1.9.
The debug program could stop at the breakpoint when I ran code in the latter machines in which both software version are older. That confused me.