LarsKue/lightning-trainable

Copying HParams returns dict

Closed this issue · 1 comments

Copying a previously created hparams instance creates a dict instead.
Minimal working example:

import lightning_trainable as lt
hparams = dict(
  batch_size = 1024,
  max_epochs = 20,
  n_dims = 2,
  optimizer = {
    "name": "Adam",
    "lr": 3.e-3,
  },
  gradient_clip = 10.0,
)
hparams = lt.TrainableHParams(hparams)
hparams2 = hparams.copy()
print(type(hparams2)

I am not sure if this is intended. If it is, it wasn't obvious to me that I should expect this reaction.

Not intentional. Fixed in the next minor version.