Apologies if I am misunderstanding your suggestion but I think I have tried that.
For the record, I am using the emacs ipython notebook, which makes fastprogress think I am in a jupyter notebook.
I can verify that it works when it it is in scope
But based on the fastai source code, I don’t currently see a way to override the progress_bar objects that are imported from core.py
Arf, that’s because fastai uses its own copy of progress_bar/master_bar. You should add:
fastai.basic_train.master_bar, fastai.basic_train.progress_bar = master_bar, progress_bar
after your force_console_behavior().
This is a bit hacky for now, but ultimately, I hope to have fastprogress recognize emacs or spyder from jupyter notebook.
Thanks so much! I am totally fine with the hack for now. It is really great to have the freedom to switch to console mode. I will definitely be using fastprogress in other projects now! Let me know if you ever need a test subject for future emacs ipython notebook functionality.
I have the latest version of FastAI Dev (pulled around 10 mins ago) and as per @sgugger instructions I have the appropriate jupyter lab extension:
jupyter labextension install @jupyter-widgets/jupyterlab-manager
However, despite this, I am not getting a progress bar during training:
learn.lr_find()
VBox(children=(HBox(children=(IntProgress(value=0, max=1), HTML(value='0.00% [0/1 00:00<00:00]'))), HTML(value…
I’m running this in JupyterLab. Please note that this used to work earlier.
I have no idea why it stopped working, but it’s probably ipywidget’s fault somehow. Note that you can force the console behavior to see something.
I deleted the environment and started afresh. Restarted and cleared out all the variables in the notebook. Its working now. Weird!
Hi Everyone,
any hints how to eliminate the progress bar - or at least the deletion of a Notebook’s cell for following calls?
learn.lr_find()
learn.fit_one_cycle()
Thanks,
Andras
This helped to get the Widgets (Like file upload) working for Jupyter lab. Thanks.