This looks good. Thank you @wintermute. I have a question. Is there a way to run each line of the code where the cursor is present? I don’t see it mentioned anywhere in the documentation. It’s cumbersome to create the #%% cells to run a new block of code.
Is there a way to execute line by line? I don’t think so.
I organised my code code into 4 or 5 cells - this is what the structure of my final dogs vs cats looks like
#%%
imports
#%%
collection of helper functions - this is where most of the logic is kept
#%%
# orchestrate training
execute training epochs
save weights to disk
#%%
# orchestrate predictions
load weights from disk
execute predictions
save predictions to disk
#%%
# orchestrate csv writing
load predictions from disk
write kaggle submission csv
As I refined each ‘orchestration’ section of code, I factored it into functions and put it into the 2nd ‘helper function’ block.
Logic that I’m still refining is kept at the orchestration level so I only need to repeatedly execute a single cell as I work on and test it.
There is also Hydrogen for GitHub’s Atom editor.
I tried installing to play around with but had some problems getting the linter package to work - YMMV
One advantage over VS Code is it looks like this can connect to a remote Jupyter instance.
I was having trouble using the ctrl+t command. I understand that we should generate our tags which I did using the plugin “jaydenlin.ctags-support” and also changed the workspace settings to include the path. But I am not able to get it work. Can you please help me?