[BUG] in trainer.py

[ClemensSchwarke]

I don't think you can set the gradients to zero right before stepping the optimizer.

byol-pytorch/byol_pytorch/trainer.py

Lines 127 to 128 in 3cb1a5e

	self.optimizer.zero_grad()
	self.optimizer.step()

[lucidrains]

@ClemensSchwarke 🤦 thanks!