[DOCS] Reload the library that could retain global gpu resources#8760
[DOCS] Reload the library that could retain global gpu resources#8760tqchen wants to merge 1 commit intoapache:mainfrom
Conversation
tqchen
requested review from
areusch,
comaniac,
jroesch,
junrushao,
merrymercy and
yzhliu
as code owners
|
background sphinx-gallery/sphinx-gallery#853 |
|
I didn't have any luck using just the module name nor explicitly removing all tensorflow* modules as they do for 'seaborn' (the latter triggers an internal tensorflow error). I've also tried launching the tensorflow tutorial as a sub-process but that didn't seem to actually separate out the module loads. |
| "min_reported_time": 60, | ||
| "expected_failing_examples": [], | ||
| "reset_modules": (force_gc, "matplotlib", "seaborn"), | ||
| "reset_modules": (force_gc, "matplotlib", "tensorflow", "torch", "onnx"), |
There was a problem hiding this comment.
i don't see torch or onnx here: https://github.com/sphinx-gallery/sphinx-gallery/blob/b41e328230f016b2089464b8f834fd76a6395eac/sphinx_gallery/scrapers.py#L561
what am i missing?
|
Good catch @areusch Seems indeed this would need some special support from sphinx gallery side. Perhaps we could reuse the same strategy here |
|
closing for now as it is more complicated than I initially thought |
4 participants
cc @areusch @mbs-octoml