TypeError: can't pickle weakref objects

[zzw-zwzhang]

I fixed it by adding .state_dict():
torch.save(ema.state_dict(), os.path.join(opt.output_dir, now + 'ema.pth'))
torch.save(ema2.state_dict(), os.path.join(opt.output_dir, now + 'ema2.pth'))

[tomguluson92]

ME TOO! This problem still exists though changed to DP from the original DDP!

-> torch.save(ema, os.path.join(opt.output_dir, now + 'ema.pth'))
(Pdb) n
TypeError: cannot pickle 'weakref' object

[xvdp]

change that line adding .state_dict(), same w all other lines where modules are saved
e.g. torch.save(ema.state_dict(), ...

[marcoamonteiro]

This may be related to different library versions since it works on our machines. Thanks for posting a solution for others who run into this issue.

[resnant]

This saved me. Thanks!

[zzhuolun]

change that line adding .state_dict(), same w all other lines where modules are saved e.g. torch.save(ema.state_dict(), ...

Thanks!

And in case someone else meets error while updating ema parameters at ema.update(generator_ddp.parameters()), e.g.

ValueError: dictionary update sequence element #0 has length 256; 2 is required

don't forget to change the ema loading method to:

ema = ExponentialMovingAverage(generator.parameters(), decay=0.999)
ema.load_state_dict(torch.load(os.path.join(opt.load_dir, "ema.pth"), map_location=device))

[ifti1394]

1 pickle.dump(model,open('model_70_version_01.pkl','wb'))

Error is showing : cannot pickle 'weakref' object