Fix source version check

[msaroufim]

@crcrpar can you try this instead?

Fixes #679 (comment)

This just basically takes advantage of the fact that every PyTorch version has the format X.Y.Z and then some stuff

# Test cases
import torch
from torchao.utils import torch_version_at_least

torch.__version__ = "2.5.0a0+git9f17037"
print(f"{torch.__version__}: {torch_version_at_least('2.5.0')}")  # Should be True
print(f"{torch.__version__}: {torch_version_at_least('2.4.0')}")  # Should be True


torch.__version__ = "2.5.0.dev20240708+cu121"
print(f"{torch.__version__}: {torch_version_at_least('2.5.0')}")  # Should be True
print(f"{torch.__version__}: {torch_version_at_least('2.4.0')}")  # Should be True


torch.__version__ = "2.5.0"
print(f"{torch.__version__}: {torch_version_at_least('2.4.0')}")  # Should be True
print(f"{torch.__version__}: {torch_version_at_least('2.5.0')}")  # Should be True

torch.__version__ = "2.4.0"
print(f"{torch.__version__}: {torch_version_at_least('2.4.0')}")  # Should be True
print(f"{torch.__version__}: {torch_version_at_least('2.5.0')}")  # Should be False

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/684

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit b943bc1 with merge base 5998389 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[jerryzh168]

@crcrpar can you try this instead?

Fixes #679 (comment)

This just basically takes advantage of the fact that every PyTorch version has the format X.Y.Z and then some stuff

# Test cases
import torch
from torchao.utils import torch_version_at_least

torch.__version__ = "2.5.0a0+git9f17037"
print(f"{torch.__version__}: {torch_version_at_least('2.5.0')}")  # Should be True
print(f"{torch.__version__}: {torch_version_at_least('2.4.0')}")  # Should be True


torch.__version__ = "2.5.0.dev20240708+cu121"
print(f"{torch.__version__}: {torch_version_at_least('2.5.0')}")  # Should be True
print(f"{torch.__version__}: {torch_version_at_least('2.4.0')}")  # Should be True


torch.__version__ = "2.5.0"
print(f"{torch.__version__}: {torch_version_at_least('2.4.0')}")  # Should be True
print(f"{torch.__version__}: {torch_version_at_least('2.5.0')}")  # Should be True

torch.__version__ = "2.4.0"
print(f"{torch.__version__}: {torch_version_at_least('2.4.0')}")  # Should be True
print(f"{torch.__version__}: {torch_version_at_least('2.5.0')}")  # Should be False

maybe add these examples to tests?

[crcrpar]

I'm confused, as I thought this should be False instead.
shouldn't 2.5.0a0 and 2.5.0.dev be <2.5.0 but >2.4?
I might misunderstand torch_version_at_least usage though

[msaroufim]

Strictly speaking this is an internal only utility that we leverage in case torch makes certain breaking changes or adds new features upstream. So I can't think of many cases where we'd do one thing with 2.5 nightlies vs 2.5 official releases.

[crcrpar]

understood.

[felipemello1]

@msaroufim , when do you guys plan to make a new release? Is it monthly?

[msaroufim]

We do monthly, next one branch cut is Sep 5 and release a few days later