How to anneal LR of optimizer?

1

Currently it seems that you’d input the LR from the start with optim.SGD for example. How do you anneal it?

2

Please refer to the imagenet example.

4 Likes
3

Or you could just recreate the optimizer

1 Like
4

If you recreate the optimizer, would it lose its weights, it seems to be? What do you mean by recreate?