Oleksiy Ostapenko
Oleksiy Ostapenko
I think the model published on hf was trained with `train_on_inputs` set to `False`.
For what I have seen, two data loaders are practical **a)** for batch learning regimes to make sure that in each batch one has labeled and unlabeled samples and **b)**...
@jlamypoirier can I merge this one? I effects many files, but its mostly only SSM related changes + minor things related to lr schedule being passed as a string
@jlamypoirier apologies for delayed reply, yes, it should be ready. Just need to run local tests and verify everything is ok, will merge after.
Yes @tscholak, addressed. Using metrics dict instead.
@jlamypoirier yes, will address your comments today. Sorry, it was deprioritised in favour of mamba.
@jlamypoirier I think I addressed all the comments.