nvalchemi.training.zero_gradients

nvalchemi.training.zero_gradients(opts)

Call zero_grad(set_to_none=True) on each optimizer.

Parameters:

opts (Iterable[torch.optim.Optimizer])

Return type:

None