When continuing to train a model loaded from a checkpoint it will train for {max_epochs - "already trained epochs"}.
Example:
Loaded model already trained 200 epochs.
max_epochs = 400
=> training will be 200 epochs
The epoch counter on the left shows total trained epochs. The counter on the right shows trained epochs in the current run, but the total on the right is max_epochs although training will only be half of that.
When the training is finished the progressbar is only at the halfway line.
(Sorry, different example.)
When continuing to train a model loaded from a checkpoint it will train for {max_epochs - "already trained epochs"}.
Example:
The epoch counter on the left shows total trained epochs. The counter on the right shows trained epochs in the current run, but the total on the right is max_epochs although training will only be half of that.
When the training is finished the progressbar is only at the halfway line. (Sorry, different example.)