Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Pkg 1550 1.17.0 plus linux cuda #4

Merged
merged 9 commits into from
Mar 21, 2024
Merged

Pkg 1550 1.17.0 plus linux cuda #4

merged 9 commits into from
Mar 21, 2024

Conversation

danpetry
Copy link

@danpetry danpetry commented Feb 23, 2024 •
edited
Loading

Destination channel: defaults

Links

Explanation of changes:

  • Update to support CUDA 11.8 for the ticket; CPU versions updated alongside
  • Tests enabled and various fixes related to that
  • NB that support for CentOS 7 was dropped upstream (see changelog). Tests being run verifies that it's still ok.
  • CUDA version is building successfully on a build machine, log here
  • see commit messages for details for changes.

@danpetry danpetry force-pushed the PKG-1550-1.17.0-plus-linux-cuda branch from a44b2c5 to 62fe94f Compare February 23, 2024 22:26
@danpetry danpetry closed this Feb 23, 2024
@danpetry danpetry reopened this Feb 23, 2024
@danpetry danpetry closed this Feb 23, 2024
@danpetry danpetry reopened this Feb 23, 2024
@danpetry danpetry closed this Feb 27, 2024
@danpetry danpetry reopened this Feb 27, 2024
@danpetry danpetry changed the title Pkg 1550 1.17.0 plus linux cuda Pkg 1550 1.17.0 plus linux cuda skip CI Feb 27, 2024
@danpetry danpetry changed the title Pkg 1550 1.17.0 plus linux cuda skip CI Pkg 1550 1.17.0 plus linux cuda Feb 28, 2024
@danpetry danpetry changed the title Pkg 1550 1.17.0 plus linux cuda Pkg 1550 1.17.0 plus linux cuda skip CI Mar 4, 2024
@danpetry danpetry changed the title Pkg 1550 1.17.0 plus linux cuda skip CI Pkg 1550 1.17.0 plus linux cuda Mar 5, 2024
@danpetry danpetry changed the title Pkg 1550 1.17.0 plus linux cuda Pkg 1550 1.17.0 plus linux cuda skip CI Mar 5, 2024
@danpetry danpetry changed the title Pkg 1550 1.17.0 plus linux cuda skip CI Pkg 1550 1.17.0 plus linux cuda Mar 6, 2024
@danpetry danpetry closed this Mar 6, 2024
@danpetry danpetry reopened this Mar 6, 2024
@danpetry danpetry changed the title Pkg 1550 1.17.0 plus linux cuda Pkg 1550 1.17.0 plus linux cuda skip CI Mar 8, 2024
@danpetry danpetry changed the title Pkg 1550 1.17.0 plus linux cuda skip CI Pkg 1550 1.17.0 plus linux cuda Mar 14, 2024
@danpetry danpetry force-pushed the PKG-1550-1.17.0-plus-linux-cuda branch from 5082c18 to 9d6dc46 Compare March 14, 2024 23:05
@anaconda-pkg-build
Copy link

Linter check found the following problems: ERROR conda.cli.main_run:execute(124): `conda run conda-lint /tmp/abs_6avv63ksfn/clone` failed. (See above for error) The following problems have been found:

===== WARNINGS =====

  • clone/recipe/meta.yaml:45: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:43: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:47: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:44: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:56: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:48: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:58: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.

===== ERRORS =====

  • clone/recipe/build.sh:68: pip_install_args: pip install should be run with --no-deps and --no-build-isolation.
    ===== Final Report: =====
    1 Error and 7 Warnings were found

@danpetry danpetry marked this pull request as ready for review March 15, 2024 20:26
danpetry added 3 commits March 15, 2024 16:21
@danpetry
Update version and sha to 1.17.1
48343ff
@danpetry
Adjust patches
5b6c9c9 
- Remove upstreamed patch 16337.patch
- Remove patch that removes now-passing test
- Rebase other patches
- move patches to patches/ directory
- Use windows line endings for Windows patch
@danpetry
Build for CUDA execution provider for linux-64
41d1d29 
- Tell nvcc to use conda-provided compiler
- Provide CUDA variant in cbc.yaml
- Enable cuda profiling
- Add required CUDA runtime packages
- Specify cuda architecture explicitly
- Handle wheel names for CPU vs CUDA variants
@danpetry danpetry force-pushed the PKG-1550-1.17.0-plus-linux-cuda branch from 9d6dc46 to 4d1d232 Compare March 15, 2024 21:22
@anaconda-pkg-build
Copy link

Linter check found the following problems: ERROR conda.cli.main_run:execute(124): `conda run conda-lint /tmp/abs_1bjkqg4vm6/clone` failed. (See above for error) The following problems have been found:

===== WARNINGS =====

  • clone/recipe/meta.yaml:53: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:43: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:45: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:56: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:54: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
    ===== Final Report: =====
    0 Errors and 5 Warnings were found

danpetry added 4 commits March 15, 2024 16:32
@danpetry
Enable unit testing
c73ec54 
- skip CUDA test that's unsupported by g3x GPU
- skip custom op test which is segfaulting on linux cuda (both C++ and python API variants)
- skip onnx dependent tests; including onnx in the host environment leads to build errors

The segfault goes away when our libstdc++ is replaced by one from the system.
Work slated to correct the issue in that library, and we'll release this one.
See issue mentioned in patch for further details.
@danpetry
Use ninja for Windows builds
b9a4e0b 
Fixes required, because the default is VS2022 and adds an extra folder depth
compared with ninja:
 - Fix cwd name on windows; upstream adds the config to the pathname twice
 - Fix the xcopy path for the same reason
@danpetry
Upgrade osx sdk to provide std::filesystem::path and std::filesystem:…
0399a7d 
…:hash_value
@danpetry
Use current usual pip install args
fb54843
@danpetry danpetry force-pushed the PKG-1550-1.17.0-plus-linux-cuda branch from 4d1d232 to 6b10a93 Compare March 15, 2024 21:34
@anaconda-pkg-build
Copy link

Linter check found the following problems: ERROR conda.cli.main_run:execute(124): `conda run conda-lint /tmp/abs_2amjc1pg01/clone` failed. (See above for error) The following problems have been found:

===== WARNINGS =====

  • clone/recipe/meta.yaml:54: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:53: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:43: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:56: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:45: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
    ===== Final Report: =====
    0 Errors and 5 Warnings were found

JeanChristopheMorinPerso
JeanChristopheMorinPerso reviewed Mar 18, 2024
View reviewed changes
@danpetry
Copy link
Author

Thanks, due to the long build times I'll wait until I get another review before making changes.

Are these the only changes you want before an approval? if not, would be cool to get any other changes before I implement these, for the same reason, if you will be making more comments that is.

@danpetry
Copy link
Author

CUDA build takes 24 hours

JeanChristopheMorinPerso
JeanChristopheMorinPerso approved these changes Mar 18, 2024
View reviewed changes
Copy link

@JeanChristopheMorinPerso JeanChristopheMorinPerso left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM, on the condition that the comments are addressed before merging.

cbouss
cbouss approved these changes Mar 19, 2024
View reviewed changes
Copy link

@cbouss cbouss left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM, there is a lot of vendored stuff in the logs but that can be managed later.

@JeanChristopheMorinPerso
Copy link

LGTM, there is a lot of vendored stuff in the logs but that can be managed later.

I saw that too and came to the same conclusion as you.

@danpetry
Copy link
Author

Re protobuf, at least, I decided not to: conda-forge#57 (comment)

danpetry added 2 commits March 19, 2024 11:52
@danpetry
Refine dependencies:
9d34323 
- libdate and optional-lite aren't needed
- zlib and pybind11 don't have any upstream requirements, nor run_exports, so we just pin to latest for reproducibility
- clarifies that gmock is a test requirement (because of this, it doesn't need to be pinned)
@danpetry
Remove useless copy of wheels
db061be
@danpetry danpetry force-pushed the PKG-1550-1.17.0-plus-linux-cuda branch from 6b10a93 to db061be Compare March 19, 2024 16:53
@anaconda-pkg-build
Copy link

Linter check found the following problems: ERROR conda.cli.main_run:execute(124): `conda run conda-lint /tmp/abs_42xe8xyupk/clone` failed. (See above for error) The following problems have been found:

===== WARNINGS =====

  • clone/recipe/meta.yaml:43: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:53: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:45: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:54: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:56: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
    ===== Final Report: =====
    0 Errors and 5 Warnings were found

@danpetry danpetry merged commit 2115354 into main Mar 21, 2024
8 checks passed
@danpetry danpetry deleted the PKG-1550-1.17.0-plus-linux-cuda branch March 21, 2024 23:16
@anaconda-pkg-build
Copy link

Linter check found the following problems: ERROR conda.cli.main_run:execute(124): `conda run conda-lint /tmp/abs_8cqxowj8kc/clone` failed. (See above for error) The following problems have been found:

===== WARNINGS =====

  • clone/recipe/meta.yaml:54: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:53: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:45: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:43: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
  • clone/recipe/meta.yaml:56: host_section_needs_exact_pinnings: Linked libraries host should have exact version pinnings.
    ===== Final Report: =====
    0 Errors and 5 Warnings were found

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@cbouss cbouss cbouss approved these changes

@JeanChristopheMorinPerso JeanChristopheMorinPerso JeanChristopheMorinPerso approved these changes

@rafaelmartins rafaelmartins Awaiting requested review from rafaelmartins

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@danpetry @anaconda-pkg-build @JeanChristopheMorinPerso @cbouss