Hi!
I suppose I am not alone in trying to use fastai as part of a project that runs in a remote server (not in Jupyter, not in a local machine).
This makes it very hard to access graphic data. I believe an easy and reasonable way to fix this would be to return the printed objects in the plot functions. Standard users would keep on working as usual, while others could now take these objects and export them or explore them in any other way.
Right now I am having problems with
learn.recorder.plot() (matplotlib plot)
data_lm.show_batch() (HTML object)
Any thoughts on this?
Edit:
For plots my current solution is to use -X when connecting with ssh, then
matplotlib.use('tkagg')
self.learner.recorder.plot(skip_end=15)
plt.show()
(Don’t forget to import matplotlib and matplotlib.pyplot as plt in that script.)