Closed dkrako closed 3 months ago
I think the problem is that we extend the reserved RAM usage at readthedocs, and #655 probably increased memory usage with the partition_by()
call.
The strange thing is, that we only use the ToyDataset
in the tutorial notebooks which doesn't have any groupby at all.
But then, probably there are trial columns added during dataset intialization so we get a single partition when calling partition_by()
in that case. #631 could help us to make a step forward to add a way to disable adding the parsed filename level information to a GazeDataFrame
.
The other strange thing is that this really is not a lot of data in the ToyDataset example..
Maybe the issue is more in this part:
/home/docs/checkouts/readthedocs.org/user_builds/pymovements/envs/latest/lib/python3.9/site-packages/nbsphinx/__init__.py:1058: RuntimeWarning: You are using an unsupported version of pandoc (2.5).
Your version must be at least (2.14.2) but less than (4.0.0).
Refer to https://pandoc.org/installing.html.
Continuing with doubts...
nbconvert.utils.pandoc.check_pandoc_version()
readthedocs has pandoc 2.5 installed here, and nbconvert
expects 2.14, so that could be an issue.
EDIT: Probably not, this line is also included in the log for builds that passed a month ago.
It's always the 10 minute tutorial when the kernel dies, let's try to find the cell(s) that lead to the dead kernel.
Current Behavior
Readthedocs fails during building the documentation.
Expected Behavior
No fail.
Minimum acceptance criteria
Failure Information (for bugs)
The commit this started to appear is the following: feat: trialize gaze.detect() (#655)
As the kernel is also dying during calculating the event properties, I'm worried that my implementation caused a bug