Adding callback 'on_training_failed'?

[maxpv]

At the early stages of the training of the model, I find sometimes having troubles with generators, which causes an error while training -and stops it.

I was wondering why can't we have a callback in the vein of on_training_fails, for example, if we insert at this line:

except Exception as e:
   callbacks.on_training_fails(e)

or

except Exception as e:
   callbacks.on_training_fails(e, epoch, generator_output)