Train without backprop

ajbrock (Andy Brock) June 25, 2017, 2:41pm 2

Not by default, but google shows several pytorch and luatorch projects:

1 Like

Explicitly passing gradients to torch.optim