Skip to content

Conversation

mohit-surana
Copy link

Original issue at: #390

Optimizers in the autograd package do not have well defined exit criteria and thus users are forced to set a given number of iterations upfront.

The changes in the PR improve upon this by allowing the user to return a (has) converged value from the optional callback function sent to the optimizer.

@mohit-surana
Copy link
Author

@dougalm @duvenaud @mattjj
Kindly review. Thanks!

@asross
Copy link

asross commented Oct 16, 2018

👍 This would be pretty helpful!

@maedoc
Copy link

maedoc commented Feb 2, 2019

I was going to make a PR with exactly this functionality (well, with raise StopIteration instead of a return True), but it's already done.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants