bstriner
commented
7 years ago Great point on leave_inner! Github readme updated with some examples.
Progress bars will display correctly in normal for loops but putting them within other TQDM loops means that you can get an ETA for the whole process and nicely nested loops. TQDM cuts out all of that work regarding timing your code and estimating the ETA.
If you're training several models for many epochs in a notebook, probably best to use tqdm_notebook(leave=True) for your outer loop, and TQDMNotebookCallback(leave_inner=False, leave_outer=False) for training.
That should give you a total of 3 progress bars at any time: each model, each epoch, each batch.
Updates in the master branch on github. Haven't pushed to PyPI yet.
- Default is now leave_inner=False for the notebook. Readme shows an example with false and true.
- Additional options show_inner and show_outer can be used to suppress one or the other bars entirely.
- Updated readme that using TQDMCallback withing tqdm for loops isn't required but adds in some nice looking nested progress bars.
Cheers, Ben
stared
There is no need for extra loop, i.e.
for i in range(10)- epochs work as expected.(A side remark - I would default for one progress bar, as below (important for many-epoch training). Or at least - explain that it is an option.)