facebookresearch/madgrad

Issues

• How about CIFAR100?

  #14 opened a year ago by twmht
  6

• Turn off optimization with lr==0.0

  #16 opened a year ago by AngledLuffa
  3

• Optimizers from 1.1 incompatible with 1.2

  #15 opened a year ago by AngledLuffa
  0

• why not AdamW style weight decay

  #1 opened 3 years ago by nestordemeure
  16

• Does it work for Keras?

  #8 opened 3 years ago by ritulahoti
  1

• Regret analysis

  #7 opened 3 years ago by samanemami
  1

• Compatible with Keras?

  #6 opened 3 years ago by andreaskeviin
  1

• Feature request

  #2 opened 3 years ago by Atul997
  2

• Speech

  #3 opened 3 years ago by deepconsc
  4

• Bug with state initialization

  #4 opened 3 years ago by tmabraham
  3