NER model training: relation between E and #

[DavidGerva]

I'm training a NER model to recognise very specific entities in a domain and I cannot understand what is the meaning of some output of the train procedure:

E are the epochs but what about #?
Is # the number of data processed in the given epoch?

And, related to this, in the spacy config.cfg file I can set both the params patience and max_steps and the only documentation I could find is this one: max_steps and/or patience are used to determine when the training should stop.
I did some tests and patience has never been considered, always on only max_steps.
Which is the difference between the two?

[polm]

Please see #7731 and #7465.

Basically # is iterations/steps and E is epochs.

max_steps is an absolute limit, but if you go patience steps without improvement training will terminate.

[DavidGerva]

Thank you again @polm, much clearer now!