pythonPackages.deeplabcut: init at 2.1.6#61253
Conversation
|
It would probably be more appropriate to separate wxPython 4 changes into a separate PR. (but I'm not sure) |
|
@deliciouslytyped hmm not sure either, I couldn't find relevant guidance in the contribution docs. I put as one pull request as DeepLabCut needs WxPython4 to work, and separating could lead to a broken package if merged before WxPython, but my NixOS / open source experience is limited so happy for advice! |
|
Thanks, @risicle, made the suggested change! |
|
@tbenst Could you separate wxPython fix into a separate PR? We generally prefer orthogonal changes to be in separate commits. The changes appear to be significant enough to be reviewed separately, so doing it in two separate PR's should help. |
|
@tbenst On that note, since you're worried about breakage on separation you could probably wait for the wxPython4 to get merged and then do the deeplabcut. |
|
@veprbl @deliciouslytyped sure can do. will have to wait a bit due to looming deadlines |
|
updated now that wxPython4 is merged. Blocked until #71282 Also need to fix the # confirm that cudatoolkits are sync'd across dependencies
assert !(openMPISupport && cudaSupport) || matchesCudatoolkit openmpi;
assert !cudaSupport || matchesCudatoolkit magma;
# confirm that mkl is sync'd across dependencies
assert !mklSupport || mkl != null;
assert !(mklSupport && cudaSupport) || matchesMkl magma;
assert !mklSupport || (numpy.blasImplementation == "mkl" && numpy.blas == mkl); |
|
Ok, I fixed the imgaug issues. I think this is finally ready to merge! Verified working if cherry-picked onto 20.03. Note that numpy is currently broken on master. But fixing that is outside the scope of this PR (I hope...). Edit: forgot to mention, I figured out the hang. It's caused by an unknown bug in OpenBLAS. Hang does not occur with MKL. It's occurs in a call to Edit2: found a small issue, need to add wrapGapps, will push... |
2b35fc4 to
794626b
Compare
|
ok, all the gapps & numpy issues are resolved, but now boto3 is broken on master. |
This is fixed in |
|
I marked this as stale due to inactivity. → More info |
SuperSandro2000
left a comment
There was a problem hiding this comment.
Please rebase this and familiarize yourself with the updated python docs.
If this PR gets no reaction in the next months, I am suggesting to close it.
|
|
||
| script = writeText "deeplabcut" | ||
| '' | ||
| #!/usr/bin/env python |
There was a problem hiding this comment.
Please substitute this right away.
| postInstall = '' | ||
| ${old.postInstall} |
There was a problem hiding this comment.
| postInstall = '' | |
| ${old.postInstall} | |
| postInstall = old.postInstall + '' |
| postInstall = '' | ||
| ${old.postInstall} | ||
| mkdir -p $out/bin | ||
| cp ${script} $out/bin/deeplabcut |
There was a problem hiding this comment.
| cp ${script} $out/bin/deeplabcut | |
| ln -s ${script} $out/bin/deeplabcut |
| meta = with stdenvNoCC.lib; { | ||
| description = "Pretrained weights for ResNet"; | ||
| longDescription = '' | ||
| Deep residual networks, or ResNets for short, provided the breakthrough idea of identity mappings in order to enable training of very deep convolutional neural networks on ImageNet. See: Deep Residual Learning for Image Recognition by Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun, Dec 2015. |
There was a problem hiding this comment.
| Deep residual networks, or ResNets for short, provided the breakthrough idea of identity mappings in order to enable training of very deep convolutional neural networks on ImageNet. See: Deep Residual Learning for Image Recognition by Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun, Dec 2015. | |
| Deep residual networks, or ResNets for short, provided the breakthrough idea of identity mappings in order to enable training of very deep convolutional neural networks on ImageNet. | |
| See: Deep Residual Learning for Image Recognition by Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun, Dec 2015. |
| }); | ||
| }; | ||
|
|
||
| python = python37.override { |
There was a problem hiding this comment.
We no longer support 3.7 and version numbers should not be hardcoded like this
| ''; | ||
|
|
||
| checkPhase = '' | ||
| for f in tests/test*.py; do echo ========== $f ==========; python $f; done |
There was a problem hiding this comment.
| for f in tests/test*.py; do echo ========== $f ==========; python $f; done | |
| for f in tests/test*.py; do echo ========== $f ==========; ${python.interpreter} $f; done |
| # checkPhase = '' | ||
| # pytest | ||
| # ''; |
Motivation for this change
add deeplabcut, and fix derivation for wxPython40, which is broken and unlisted: #54235. Now working thanks to help from Phoenix team wxWidgets/Phoenix#1162 and @deliciouslytyped
Things done
sandboxinnix.confon non-NixOS)nix-shell -p nix-review --run "nix-review wip"./result/bin/)nix path-info -Sbefore and after)Resolves: #54235