LarsKue/lightning-trainable

Issues

• Make toy models deterministic

  #27 opened 7 months ago by thelostscout
  1

• omegaconf

  #26 opened a year ago by ThisIsForReview
  1

• No last.ckpt

  #25 opened a year ago by ThisIsForReview
  1

• Cannot run lightning-trainable example

  #24 opened a year ago by ThisIsForReview
  6

• Continued training `model.fit()` creates spurious validation entries in external log

  #10 opened a year ago by fdraxler
  5

• How to validate without logger?

  #12 opened a year ago by fdraxler
  5

• HParams validation batch has `self.training = False`

  #20 opened a year ago by fdraxler
  1

• load_from_checkpoint loads hparams as dict instead of class HParams

  #18 opened a year ago by thelostscout
  6

• Copying HParams returns dict

  #13 opened a year ago by thelostscout
  1

• Progressbar confusing when continuing training of a loaded model.

  #8 opened 2 years ago by HeiseMax
  1

• on_train_start does not transfer batch to self.device

  #16 opened a year ago by LarsKue
  0

• HParams with dicts (i.e. almost all) store dicts as special AttributeDict type in `hparams.yaml`

  #21 opened a year ago by fdraxler
  3

• No support for cpu only

  #22 opened a year ago by thelostscout
  2

• Allow single-layer (linear) `FullyConnectedNetwork`

  #23 opened a year ago by LarsKue
  1

• Use default engine for resolving class specs for loggers as well

  #11 opened a year ago by fdraxler
  3

• Cannot assign new keys to hparams of type dict

  #17 opened a year ago by thelostscout
  0

• pip doesn't install the dependency tomli

  #7 opened 2 years ago by Fjallripa
  1