Training cell not printing

Hi there,
I am using pycharm, connected to paperspace, to fine-tune a pre-trained model as per Lesson 6 of the 2022 course.

Everything seems to work correctly but for the fact that I can’t seem to get pycharm to print-out the table that shows the error rate and other metrics per epoch:

The same thing happens with vscode. I believe it has to do with the fact that I am connected to a remote server. Everything else prints out correctly - I can preview images, see the suggested LR graph, etc.

When I copy the code into paperspace and run it in one of their notebooks, I get the expected loading bars and table:

I’m wondering if anyone knows a way to make the learn.fine_tune “louder” - to force it to print more explicitly.

My code is identical (direct download) to the “road to the top pt 1” notebook: First Steps: Road to the Top, Part 1 | Kaggle

Further information: The model does train – When I include cbs = CSVLogger() in my vision learner declaration, it produces a credible-looking file:

Jetbrains IDEs don’t have good output cell support for Jupyter Notebooks, and therefore don’t support the output created by fastai. Nothing can be done except waiting until Jetbrains improves the output cell. I do want to use Jetbrains Dataspell, but Jupyter support is like 90% of the way there, and the remaining 10% makes it either difficult or annoying to use.

VS Code does have proper output cell support. I remember having the same issue as you: VS Code wasn’t initially producing the output fastai was making. I think it’s because I may have accidently clicked cancel to a prompt VS Code produced that said it needed to install additional stuff to have output cells work properly.

I reinstalled VS Code after, and then the fastai output was properly displayed. So perhaps try reinstalling it. Though when uninstalling it, make sure you remove all files (e.g., cache files, library files, and whatnot), not only the application file. This will allow for a fresh install.

This worked for me:

1 Like