Problem

TimesFMFinetuner currently trains silently. Long runs give no per-batch feedback.

Proposal

Add an optional tqdm progress bar around the train/val DataLoader loops (rank 0 only).
Introduce a single flag in FinetuningConfig:

progress_bar: bool = True

Default remains True (can be set to False to keep old behavior).

Implementation details

• Wrap loops in _train_epoch and _validate with tqdm when config.progress_bar and rank == 0.
• Show running average loss via set_postfix every log_every_n_steps.

Backward compatibility

Setting progress_bar=False yields the exact current behavior (no bars). Otherwise, there’s only additional stdout output.

Testing

• Tiny smoke test to ensure training runs with progress_bar=True without crashing.

Happy to submit a PR with the change. Let me know if you prefer a different flag name or default!