Loss.. should be specified as either training loss or validation loss

[FlogramMatt]

System Info

Doesn't apply.. using runpod

Who can help?

No response

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

As you fine tune a new model, it shows a loss curve and numbers as it runs.. is that training loss or validation loss? Should be labelled

If training loss.. how can I view validation loss? I set 'val size', it should exist somewhere

Expected behavior

Should be labeled more clearly

[Rajatavaa]

Could you please elaborate more or provide a script to replicate this?

[FlogramMatt]

I did figure it out, it is training loss.

I mean when fine tuning the model, you have a 'loss' graph in bottom right and same for outputs in terminal and it's not obvious that it's training loss, not validation loss. Would be nice if labelled better.

[Rajatavaa]

I mean if your code doesn't specify to plot a graph then it will not show any graph.
And if you are plotting the graph then you can label it correctly