validation loss

#3
by jacek2024 - opened

I see that results from this model are worse than from original llama 3, I wonder why validation loss is not changing during your training, is this correct behaviour?

Cognitive Computations org

It is not worse than original llama 3, we do not train the instruct version, we train the base model.

Crystalcareai changed discussion status to closed

Sign up or log in to comment