AVX512 support#439
Conversation
|
Hi! This is the friendly automated conda-forge-linting service. I just wanted to let you know that I linted all conda-recipes in your PR ( I do have some suggestions for making it better though... For recipe/meta.yaml:
This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/16729133202. Examine the logs at this URL for more detail. |
|
@hmaarrfk can we merge this as well before I make the macOS packages? |
|
Yes ok with me. Please merge when you feel confident. |
|
Hi! This is the friendly automated conda-forge-linting service. I was trying to look for recipes to lint for you, but it appears we have a merge conflict. Please try to merge or rebase with the base branch to resolve this conflict. Please ping the 'conda-forge/core' team (using the |
|
Hi! This is the friendly automated conda-forge-linting service. I just wanted to let you know that I linted all conda-recipes in your PR ( I do have some suggestions for making it better though... For recipe/meta.yaml:
This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/22456640361. Examine the logs at this URL for more detail. |
…6.02.19.14.04.59 Other tools: - conda-build 25.9.0 - rattler-build 0.47.1 - rattler-build-conda-compat 1.4.5
| - cat test_tensorflow.py # [unix] | ||
| - run python test_tensorflow.py # [build_platform == target_platform] | ||
| - tf_upgrade_v2 --help # [build_platform == target_platform] | ||
| - run tf_upgrade_v2 --help # [build_platform == target_platform] |
There was a problem hiding this comment.
This didn't fix the test:
/home/conda/feedstock_root/build_artifacts/tensorflow-split_1771543504012/test_tmp/run_test.sh: line 10: run: command not found
There was a problem hiding this comment.
While you're debugging, please restrict testing here to one (preferably non-CUDA) build.
| - 12.6 | ||
| - None | ||
| - 12.8 | ||
| - 12.9 |
There was a problem hiding this comment.
this migrator should not be deleted, we still need 12.8 here (until 2.20 I guess?); it already has use_local: true so you can simply move 12.9 before 12.8 in the ordering here. and a rerender should do the correct thing (== stay on 12.8)
|
As an overall point; not sure if we're doing AVX512 builds anywhere else yet? Are there some benchmarks that we can reference for how much of a win this is over AVX2? Makes a big difference if it's 5% or 50%. |
This is not a difficult demand. I'm cancelling the builds. |
|
I skipped level=1, but since level=3 and level=4 failed and cuda vs non-cuda failed with different errors, I wanted to check all 4 builds. |
|
We have a pending pytorch merge that fixes several issues, which is IMO pretty self-evidently of higher priority than debugging work, especially for features rather than fixes. The PR here would block the server for 8-12h as-is, which I think is not reasonable under the circumstances. Though you should be able to run them in parallel if you're willing to try the cirrus setup from conda-forge/pytorch-cpu-feedstock#484 |
|
I had restarted all builds here as soon as runners became available again ~36h ago, but the interface with cirun is pretty fragile these days, and only one job started. The only recovery is to cancel and restart, but since the GH UI has no third way between "restart all failed jobs" and "restart one job", the former would have meant rerunning the just-failed job again. In any case, all the last of the 4 jobs whose output you wanted to see has now been started, 3 are done already. |
|
Moved some fixes from this PR to #477 |
Checklist
0(if the version changed)conda-smithy(Use the phrase@conda-forge-admin, please rerenderin a comment in this PR for automated rerendering)