Description
Describe the workflow you want to enable
When verbose>0, the fitting is shown as a progress bar including the average time/pace and last fit attempted time.
Describe your proposed solution
Implement tqdm-like or keras-like (familiar to epoch tracking) progress bar such that the feature can be enabled through an argument of verbose. This proposal is under the assumption that the iteration is accessible like for loop.
Describe alternatives you've considered, if relevant
No response
Additional context
I've been using GridSearchCV and brute-forcing parameters (regarding algorithms, domain-knowledge agnostic) to find the best model parameters while leaving the machine unattended (disconnect from the cloud). But sometimes, the kernel might freeze without any indication. Even with verbose tracking, it's hard to skim through the log, especially among hundreds of fits to deduce that the process has been interrupted.
The issue expanded from a simple inquiry of #6021 .