-
-
Notifications
You must be signed in to change notification settings - Fork 46
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.
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
PyTorch 2.5.0 #277
base: main
Are you sure you want to change the base?
PyTorch 2.5.0 #277
Conversation
@conda-forge-admin, please rerender |
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 ( |
Hi! This is the friendly automated conda-forge-webservice. I tried to rerender for you, but it looks like there was nothing to do. This message was generated by GitHub actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/11393752191. |
ee95e8c
to
5559a7e
Compare
You might need to update my patch to help find numpy, my suggestion didn't seem like it went through upstream.... |
0dffeb2
to
cd1d246
Compare
Submitted pytorch/pytorch#138287 for the nvtx patch |
Woohoo, it passed the cmake configuration stage and is now in building stage. By the way, I don't have permissions to run several CI jobs. Could I get those permissions or would someone like to take over this PR or create a new one? |
just use azure for now, that CI gets cloggled anyway, then you wait for ever ;) |
if you can get to the 6 hour timeout, we can switch back to the larger runner |
@conda-forge-admin, please rerender |
1 similar comment
@conda-forge-admin, please rerender |
@@ -20,6 +20,6 @@ github_actions: | |||
os_version: | |||
linux_64: cos7 | |||
provider: | |||
linux_64: github_actions | |||
linux_64: azure |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
there is a 1 minute timeout stated at the top of the file too.
@conda-forge-admin, please rerender |
There are other things you have to patch out.
|
recipe/meta.yaml
Outdated
@@ -303,6 +301,7 @@ outputs: | |||
requires: | |||
- {{ compiler('c') }} | |||
- {{ compiler('cxx') }} | |||
- boto3 # [osx and x86] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
the condition of osx and x86
seems dubious. its either always required (should be in the run section), or just required for testing (should be enabled for all architectures).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hmm, it's not in the root requirements.txt but in some CI requirements. Let me remove the condition for now.
recipe/meta.yaml
Outdated
@@ -277,7 +277,7 @@ outputs: | |||
# other requirements | |||
- python | |||
- typing_extensions >=4.8.0 | |||
- sympy >=1.13.1 | |||
- sympy 1.13.1 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
sorry for the rapid fire:
is there a known bug that makes sympy 1.13.3 incompatible with pytorch 2.4 or 2.5? otherwise i would rather unpin.
This kind of pinning
is great for tight distributions, but it goes against the sharing and interoperability of cocnda-forge.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
unpin = add a patch to make the pip check
more forgiving.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It seems there ware test failures on windows and mac with 1.13.2: pytorch/pytorch#133235
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not sure if the issue has been resolved in version 1.13.2. Let's patch to allow version 1.13.2 (or > 1.13.2) and see if the failures disappear (uncertain about how to test it yet though..)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@jeongseok-meta it may be appropriate to sometimes add !=1.13.2
and adding the PR you have as a reference.
@atalman are you able to point us to the exact tests that were failing so we can test with 1.13.3?
213cb75
to
1b40828
Compare
`simpy` was pinned to version 1.13.1 due to test failures with version 1.13.2 on Windows and mac, as reported in #133235. Now that a newer version, 1.13.3, has been released, this PR aims to verify if the test failure has been resolved and also allow building with newer versions for packaging purposes (e.g., conda-forge/pytorch-cpu-feedstock#277 (comment)). Pull Request resolved: #138338 Approved by: https://github.com/Skylion007, https://github.com/malfet Co-authored-by: Nikita Shulga <[email protected]>
Co-authored-by: Mark Harfouche <[email protected]>
Co-authored-by: Mark Harfouche <[email protected]>
7f59bbc
to
eff02b1
Compare
@conda-forge-admin, please rerender |
Thank you @jeongseok-meta this is amazing. it may be time to re-enable the ci-runners and i can make a dummy commit ;) |
Now I'm finally able to get a powerful linux machine and do local testing. So far, linux_64_* builds work fine, but I encountered other issues with linux_aarch6_*. Let me try to fix them and share the results (I might need help). |
(sorry for mixing different accounts) Here are the build logs: log_files.zip It seems it fails to build with error
|
It seems that maybe cuda 11.8 isn't working either on the linux64_. I'm not too interested in that in 2024, we can likely drop it. If you want to drop it, please say so in the first post so we can get other's feedback. The only
For linuxaarch it seems like you are missing
but compilation seems to be working!!!! We assume you can "emulate" aarch which requires a host package (outside of docker) like |
Sounds good! Made a comment: #270 (comment)
Rebuilding with the package installed... ⏳ Thanks! |
Don’t you want ppc as well? You might as well try. You are on a roll |
New build log with Only one failed:
Yes, but I'd like to keep this change minimal and address the rest in follow-up PRs. |
…nda-forge-pinning 2024.10.25.18.47.03
434eb7f
to
a1f33e6
Compare
I am not sure if we want to drop 11.8 + mkl (11.8 + generic works though) or fix it in this PR. Please let me know how you would like to proceed! |
Checklist
0
(if the version changed)conda-smithy
(Use the phrase@conda-forge-admin, please rerender
in a comment in this PR for automated rerendering)