Evaluate model (i.e. get loss of test set w.r.t to trained model)

[pltrdy]

Hi,

I've been looking for a way to calculate loss on a kind of test set on a trained model.

The idea is to evaluate if adding general data is better for my model or not i.e. I have two training dataset A and B and a test-set T, I would like to train a model based on A, and on A+B and see if this data augmentation improves the model.

[Celebio]

Hi @pltrdy ,
We are planning to add the loss as a result of the test command soon.

Regards,
Onur

[pltrdy]

Hi @Celebio, thanks for your answer,

It seems that test is related to classification, right?

I'm talking about the embeddings only, i.e. fastTest training provides the training loss, I'd like to have a validation/test loss as well. Is it what you're talking about?

[iedmrc]

Any update on this?