Unpin submodlib-py and remove linux only install constraint

[eshwarprasadS]

Since there is a new 0.0.2 release tag from submodlib-py, remove requirement versioning pin and remove linux-only requirement (as 0.0.2 supports MacOS builds)

[bbrowning]

I'm wondering if at this point we remove the version constrain entirely? Or something like<= 1.0.0? That way we can pull in 0.0.1, 0.0.2, etc without needing to explicitly bump, since we work with both on Linux and Mac will automatically pick up 0.0.2.

[eshwarprasadS]

I'm wondering if at this point we remove the version constrain entirely? Or something like<= 1.0.0? That way we can pull in 0.0.1, 0.0.2, etc without needing to explicitly bump, since we work with both on Linux and Mac will automatically pick up 0.0.2.

This sounds fine as well. Only reason I am wary of uncapped versioning for the dependency is due to its potential instability, specifically for this dependency. Not strictly opposed to this, just leaned towards the safer side, since we have had previous issues with this library and its stability. Open to removing the cap entirely for now, given this, WDYT?

[bbrowning]

I think removing the cap entirely is reasonable for now, and we can address if some incompatible releases come out while we're still supporting this code.

[dhellmann]

Please do not pin to an exact version like this. If we had to deliver a fix in submodlib-py downstream we would not be able to ship the incompatible versions of the library. There are some guidelines for managing versions documented in https://github.com/instructlab/dev-docs/blob/main/docs/dependency-management.md.

I think in this case a version range like <0.1.0 or even <1.0.0 would make the most sense.

If you want to pin to a specific version for your CI jobs, use a constraints file separate from the requirements file.

[courtneypacheco]

@eshwarprasadS I see the small E2E job is failing. In our core repo, we had to temporarily disable this job: instructlab/instructlab#3290

See linked issue for more information about why we temporarily disabled it: instructlab/instructlab#3289

So we can ignore the small E2E job failure here once @dhellmann's review comments are addressed.

[eshwarprasadS]

@eshwarprasadS I see the small E2E job is failing. In our core repo, we had to temporarily disable this job: instructlab/instructlab#3290

See linked issue for more information about why we temporarily disabled it: instructlab/instructlab#3289

So we can ignore the small E2E job failure here once @dhellmann's review comments are addressed.

would we want to get another PR in for the e2e job failure, or can we ignore and merge for now, if the comments here are resolved?

[courtneypacheco]

I suggest we merge this PR (with the small E2E failure) if it looks good to everyone, and comment out the small E2E job in another PR. This way, we don’t block this PR’s progress.

@bbrowning can you review this PR once more?

[eshwarprasadS]

Wonderful, thanks everyone, for all the help in getting this in. Will put up another PR for disabling the small E2E job (for the time being).

[bbrowning]

@Mergifyio backport release-v0.8

[mergify]

backport release-v0.8

✅ Backports have been created

Details

• #604 Unpin submodlib-py and remove linux only install constraint (backport #603) has been created for branch release-v0.8