Tuner class resetting the enable_progress_bar flag to True. I can’t retrace where it does that in the source code, but even when enable_progress_bar is set to False in Trainer, I still see it in the output log.
lr_trainer = L.Trainer(default_root_dir = config['model_save_path'], logger=comet_logger, accelerator='gpu', log_every_n_steps = 1, enable_progress_bar=False, max_epochs=-1)
# Create a Tuner
tuner = L.pytorch.tuner.Tuner(lr_trainer)
# Run learning rate finder
lr_finder = tuner.lr_find(m, datamodule=data_module, min_lr=1e-6, max_lr=1e-1, num_training=5000)
I’m using Lightning 2.1.2 and running the code on a SLURM server.