So much thanks !!
I found use wiht learn.no_bar() is enough to work with %%capture
and if I don’t want information print by fit completely, I can do with learn.no_bar(), learn.no_logging(), which I won’t see anything completely.
I found there are three related functions, and below is a small note.
| no_bar | no_mbar | no_logging | |
|---|---|---|---|
| main bar | n | n | y |
| progress bar | n | y | y |
| sheet | n | n | n |
| L | y | n | n |
“progress bar” is phase (e.g. lr find) running.
“main bar” is progress of all epochs.
“sheet” is the table that shows val_loss and other things
“L” is a L object print by fit that records information recorded by sheet originally.