Epochs vs Training Steps in Model Configuration

[ml-ts03]

Hello, I am working on single-step forecasting with some of the models in this library, performing optimization with variable batch_size as one of the considered hyperparameters.

I have some doubts about model configuration in regards to the "max_steps" parameter (for both Models, and AutoModels) since the documentation mentions it as the "maximum number of training steps" (which I'd assume refers to weight updates after each batch).

However the output (PyTorch Lightning progress bar) returns "Epoch x: 100%|██████████|: ..." going up until Epoch of "max_steps" -1 (e.g. if I set max_steps=300, it will go up to "Epoch 299", assuming there's no early stopping).

Having this in mind, does "max_steps" define the maximum number of epochs (running through every batch for each epoch) or does it actually define the maximum number of training steps (meaning weight updates/batches processed) for the entirety of the fitting process?

Thank you in advance!