Notebook copying to html error

Troubleshooting
1

Hi,

I did manage to get Guild to run my notebook, but I did get a couple warnings and 1 error:

2 Warnings

1). “WARNING: Skipping potential source code file ./tcn_kerasAPI_gpu_data_generator.ipynb because it’s too big. To control which files are copied, define ‘sourcecode’ for the operation in a Guild file.”

I’m just wondering why I was even seeing that warning when my actual run command was:

guild run tcn_kerasAPI_gpu_guild_ai_data_generator.ipynb

So you see I wasn’t even running the notebook that it says is “too big”. That’s a different notebook file entirely. Is there some reason Guild would be “copying over” a notebook file other than the one I’m running?

2). I’m not sure why I got this warning:

[NbConvertApp] WARNING | Config option kernel_spec_manager_class not recognized by NbConvertApp.

===================

1 Error

Tho’ I did manage to get a successful run and saw expected output images (in Tensorboard) for my notebook, at the end of the run, when Guild was trying to convert the notebook to html via “nbconvert”, I did see this error:

“AttributeError: module ‘jinja2’ has no attribute ‘Markup’”

I guess it’s not critical, but of course it’s nicer for a run not to end on an error of course :wink: :).

2

Guild copies all your sourcecode in order to run your notebook command.

So if you haven’t specified what sourcecode to include or exclude guild will automatically copy notebooks and python files. You can read more here: Guild File Reference

I am unsure about the kernel_spec_manager_class. Is this a class you have developed?

Regarding the error: that is a dependency problem. Try upgrading nbconvertor guildai.

3

@mabrandt_nasa that Guild is complaining about the size of a file it needs to run the notebook is a bug! Sorry about that.

I’ll open an issue so we can track that.

As for the warning and error messages, this seems like a potential library version conflict. I know jinja2 rev’d its APIs and I recall seeing that error come up before. I’ll take a look.

Thanks for the report!