PyTorch custom operators for graph operations

[valsdav]

The PR add to the externals 3 small libraries of pytorch operators for optimized graph clustering and sparse tensors operations:

• Pytorch Sparse https://github.com/rusty1s/pytorch_sparse
• Pytorch Cluster https://github.com/rusty1s/pytorch_cluster
• Pytorch Scatter https://github.com/rusty1s/pytorch_scatter

These libraries are used in graph neural networks to perform clustering operations, i.e. in the DRN model cms-sw/cmssw#37134

A pytorch-custom-ops tool is defined to link all of them.
Only the CPU ops are compiled for the moment. Kernels for CUDA are also available.

PR validation

I have linked successfully a CMSSW test, loading a DRN TorchScript model. The PyTorch engine finds the operators and does not complain. I will create a PR to CMSSW with tests in the next days.

[cmsbuild]

A new Pull Request was created by @valsdav for branch IB/CMSSW_14_1_X/master.

@iarspider, @smuzaffar, @aandvalenzuela, @cmsbuild can you please review it and eventually sign? Thanks.
@antoniovilela, @sextonkennedy, @rappoccio you are the release manager for this.
cms-bot commands are listed here

[cmsbuild]

cms-bot internal usage

[smuzaffar]

@valsdav , can you please apply the following changes

• Update the comment # Make sure the default c++sdt stand is c++11 to # Make sure the default c++sdt stand is c++14 (as the grep command after this comment is looking for 14
• Remove the empty %post sections
• pytorch-custom-ops.xml tool file is not going to be part of cmssw as there is no pytorch-custom-ops.spec. I would suggest to create a meta package pytorch-custom-ops.spec which does nothing in %prep, %build, %install sections but it depends on other pytorch-* tools e.g something like [a]
• Add pytorch-custom-ops dependency in cmssw-tool-conf.spec .

[a]

### RPM external pytorch-custom-ops 1.0
Source: none
Requires: pytorch-scatter
Requires: pytorch-cluster
Requires: pytorch-sparse
%prep
%build
%install

[cmsbuild]

Pull request #9094 was updated.

[smuzaffar]

please test

[smuzaffar]

@valsdav , can you please update all the newly added xml files to have PYTORCH_ instead of TORCH_ e.g. replace TORCH_CLUSTER_BASE with PYTORCH_CLUSTER_BASE every where. The name of this variable should match the name of tool.

[valsdav]

fixed in 492ec08

[cmsbuild]

Pull request #9094 was updated.

[smuzaffar]

please test

[smuzaffar]

please test for el8_aarch64_gcc12

[smuzaffar]

please test for el8_ppc64le_gcc12

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-f196af/38419/summary.html
COMMIT: 492ec08
CMSSW: CMSSW_14_1_X_2024-03-25-2300/el8_ppc64le_gcc12
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmsdist/9094/38419/install.sh to create a dev area with all the needed externals and cmssw changes.

The following merge commits were also included on top of IB + this PR after doing git cms-merge-topic:

• @cms-sw cmssw-osenv: Fix for KRB5 file #9100

You can see more details here:
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-f196af/38419/git-recent-commits.json
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-f196af/38419/git-merge-result

[cmsbuild]

-1

Failed Tests: RelVals
Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-f196af/38418/summary.html
COMMIT: 492ec08
CMSSW: CMSSW_14_1_X_2024-03-25-2300/el8_aarch64_gcc12
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmsdist/9094/38418/install.sh to create a dev area with all the needed externals and cmssw changes.

The following merge commits were also included on top of IB + this PR after doing git cms-merge-topic:

• @cms-sw cmssw-osenv: Fix for KRB5 file #9100

You can see more details here:
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-f196af/38418/git-recent-commits.json
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-f196af/38418/git-merge-result

RelVals

The relvals timed out after 4 hours.

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-f196af/38403/summary.html
COMMIT: 492ec08
CMSSW: CMSSW_14_1_X_2024-03-25-1100/el8_amd64_gcc12
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmsdist/9094/38403/install.sh to create a dev area with all the needed externals and cmssw changes.

The following merge commits were also included on top of IB + this PR after doing git cms-merge-topic:

• @cms-sw Update Geant4 to version 11.2.1 #9086
• @jmduarte AXOL1TL v3.0.2 #9091

You can see more details here:
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-f196af/38403/git-recent-commits.json
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-f196af/38403/git-merge-result

Comparison Summary

Summary:

• You potentially added 38 lines to the logs
• ROOTFileChecks: Some differences in event products or their sizes found
• Reco comparison results: 51890 differences found in the comparisons
• DQMHistoTests: Total files compared: 48
• DQMHistoTests: Total histograms compared: 3297537
• DQMHistoTests: Total failures: 305653
• DQMHistoTests: Total nulls: 255
• DQMHistoTests: Total successes: 2991609
• DQMHistoTests: Total skipped: 20
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.686 KiB( 47 files compared)
• DQMHistoSizes: changed ( 10224.0 ): 0.596 KiB SiStrip/MechanicalView
• DQMHistoSizes: changed ( 13034.0 ): 0.357 KiB SiStrip/MechanicalView
• DQMHistoSizes: changed ( 250202.181 ): 0.234 KiB SiStrip/MechanicalView
• DQMHistoSizes: changed ( 25202.0 ): -0.225 KiB SiStrip/MechanicalView
• DQMHistoSizes: changed ( 7.3 ): -0.276 KiB SiStrip/MechanicalView
• Checked 202 log files, 165 edm output root files, 48 DQM output files
• TriggerResults: found differences in 20 / 46 workflows

[smuzaffar]

+externals

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next IB/CMSSW_14_1_X/master IBs (tests are also fine). This pull request will now be reviewed by the release team before it's merged. @sextonkennedy, @rappoccio, @antoniovilela (and backports should be raised in the release meeting by the corresponding L2)