Repository navigation
Doc build CI might update too frequently #351
Description
Since we merged #325 yesterday, the latest doc has been rebuilt several times. Taking the latest commit 8e4e4f2 as example, we see the gh-pages branch was updated frequently for two reasons:
- The Jupyter notebook hash is different in every run
- This might have to do with Building
cuda.bindingsdocs requires the presence of a physical GPU & valid driver #327
- This might have to do with Building
- The
environment.pickledoctrees generated by Sphinx are checked in- Q: Can we remove them from the artifacts?
Activity
leofang commentedon Jan 4, 2025
Related discussion: materialsproject/atomate2#345. Looks like the answer is yes.
Previously @trxcllnt and I had some conversation about binary blobs being checked in could cause
git checkoutslowness (it was for a different context). Perhaps Paul might have some tips for how to properly address this? 😉leofang commentedon Jan 4, 2025
On the pickle files: It seems to be my overlook in #183 because previously @vzhurba01 seems to be careful enough to not check them in? @vzhurba01 did your purge the pickle files before pushing?
leofang commentedon Jan 12, 2025
@vzhurba01 could you check if changing this from
forcetocachewould be enough to stop updating the hash?cuda-python/cuda_bindings/docs/source/conf.py
Line 35 in 57a710a
Otherwise, we should set it to
offor find another way (ex: just don't check in the Jupyter notebook to thegh-pagesbranch).xref: https://myst-nb.readthedocs.io/en/v0.10.1/use/execute.html#triggering-notebook-execution
vzhurba01 commentedon Jan 30, 2025
With the PR ready, posted past discussions on the questions raised here:
I simply copied everything from the
htmlfolder intodoc. Perhaps the.doctreesdidn't get picked up from my*wildcard.To resolve #327 I'm setting it to
offwith the why described in #424 (comment). The proposed PR applies the remaining changes needed to resolve this.leofang commentedon Feb 4, 2025
Looks like we fixed it!