Hi all, I run into a problem only recenly
after creating the model
learn = tabular_learner(data, layers=[200, 100], ps=0.5, metrics=[accuracy], y_range=(0, 1.2), wd=1e-1).to_fp16()
learn.lr_find()
learn.recorder.plot()
learn.fit_one_cycle(10)
The accuracy and loss are always worse, however if I run it without lr_find(), the accuracy becomes much better.
learn = tabular_learner(data, layers=[200, 100], ps=0.5, metrics=[accuracy], y_range=(0, 1.2), wd=1e-1).to_fp16()
learn.fit_one_cycle(10)
I think most of the notebook in the lesson can run it direcly after lr_find() and recorder.plot(), is it my problem or my dataset is not doing well? Thanks a lot.