Add mamba-ssm

[weiji14]

Mamba state-space model! Repo at https://github.com/state-spaces/mamba

Checklist

• Title of this PR is meaningful: e.g. "Adding my_nifty_package", not "updated meta.yaml".
• License file is packaged (see here for an example).
• Source is from official source.
• Package does not vendor other packages. (If a package uses the source of another package, they should be separate packages or the licenses of all packages need to be packaged).
• If static libraries are linked in, the license of the static library is packaged.
• Package does not ship static libraries. If static libraries are needed, follow CFEP-18.
• Build number is 0.
• A tarball (url) rather than a repo (e.g. git_url) is used in your recipe (see here for more details).
• GitHub users listed in the maintainer section have posted a comment confirming they are willing to be listed there.
• When in trouble, please check our knowledge base documentation before pinging a team.

[github-actions]

Hi! This is the staged-recipes linter and your PR looks excellent! 🚀

[conda-forge-admin]

Hi! This is the friendly automated conda-forge-linting service.

I wanted to let you know that I linted all conda-recipes in your PR (recipes/mamba-ssm/recipe.yaml) and found some lint.

Here's what I've got...

For recipes/mamba-ssm/recipe.yaml:

• ❌ Non noarch packages should have python requirement without any version constraints.
• ❌ Non noarch packages should have python requirement without any version constraints.

_{This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/14747268959. Examine the logs at this URL for more detail.}

[conda-forge-admin]

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 (recipes/mamba-ssm/recipe.yaml) and found it was in an excellent condition.

[weiji14]

There's an RuntimeError on the import mamba_ssm test, e.g. at https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=1236098&view=logs&j=df9408f7-e795-500f-143d-aa7735a0ed8b&t=47addc80-ccbf-566b-ee3a-66cdb8436937&l=1297:

 Installing test environment
 ✔ Successfully updated the test environment
 + python $SRC_DIR/conda_build_script.py
 Traceback (most recent call last):
   File "$SRC_DIR/conda_build_script.py", line 1, in <module>
     import mamba_ssm
   File "$PREFIX/lib/python3.12/site-packages/mamba_ssm/__init__.py", line 3, in <module>
     from mamba_ssm.ops.selective_scan_interface import selective_scan_fn, mamba_inner_fn
   File "$PREFIX/lib/python3.12/site-packages/mamba_ssm/ops/selective_scan_interface.py", line 16, in <module>
     from mamba_ssm.ops.triton.layer_norm import _layer_norm_fwd
   File "$PREFIX/lib/python3.12/site-packages/mamba_ssm/ops/triton/layer_norm.py", line 169, in <module>
     @triton.autotune(
      ^^^^^^^^^^^^^^^^
   File "$PREFIX/lib/python3.12/site-packages/triton/runtime/autotuner.py", line 352, in decorator
     return Autotuner(fn, fn.arg_names, configs, key, reset_to_zero, restore_value, pre_hook=pre_hook,
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   File "$PREFIX/lib/python3.12/site-packages/triton/runtime/autotuner.py", line 127, in __init__
     self.do_bench = driver.active.get_benchmarker()
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   File "$PREFIX/lib/python3.12/site-packages/triton/runtime/driver.py", line 23, in __getattr__
     self._initialize_obj()
   File "$PREFIX/lib/python3.12/site-packages/triton/runtime/driver.py", line 20, in _initialize_obj
     self._obj = self._init_fn()
                 ^^^^^^^^^^^^^^^
   File "$PREFIX/lib/python3.12/site-packages/triton/runtime/driver.py", line 8, in _create_driver
     raise RuntimeError(f"{len(actives)} active drivers ({actives}). There should only be one.")
 RuntimeError: 0 active drivers ([]). There should only be one.
 × error Script failed with status 1
 × error Work directory: '/tmp/.tmpKdXIy2'
 × error To debug the build, run it manually in the work directory (execute the `./conda_build.sh` or `conda_build.bat` script)

Should have been fixed in triton-lang/triton#6570 (see also triton-lang/triton#6150). Could either skip the import check (import should technically work on a CUDA-enabled device), or wait for a new release of triton>3.3.0 with that patch? Edit: found another workaround by pinning to triton<3.2.0, at the expense of only supporting pytorch=2.5.

[conda-forge-admin]

Hi! This is the friendly automated conda-forge-linting service.

I wanted to let you know that I linted all conda-recipes in your PR (recipes/mamba-ssm/recipe.yaml) and found some lint.

Here's what I've got...

For recipes/mamba-ssm/recipe.yaml:

• ❌ The recipe must have some tests.

_{This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/14764654406. Examine the logs at this URL for more detail.}

[conda-forge-admin]

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 (recipes/mamba-ssm/recipe.yaml) and found it was in an excellent condition.

[weiji14]

@conda-forge/help-python, ready for review!

[weiji14]

Gentle ping again @conda-forge/help-python to see if this can be reviewed and merged 🙏

[ocefpaf]

I'm not a pytorch cuda user, but can't this be noarch?

[weiji14]

I think this CUDA-dependent package is Linux only. And actually, the reason I'm packaging this is to get builds working on arm64 which aren't available on the release page at https://github.com/state-spaces/mamba/releases/tag/v2.2.4, so it's probably not a noarch package.

[ocefpaf]

Shouldn't you need compilers then? @jakirkham or @isuruf you have more experience with cuda recipes. Can you chime in and help us?

[danielnachun]

I made some suggestions for what I think are the basic CUDA requirements but there will probably be more needed.

[danielnachun]

Can we add ninja as a test-only dependency here?

[weiji14]

Don't think there's a way to add test-only dependencies with v1 recipes for python tests - https://rattler.build/latest/reference/recipe_file/#python-tests, only able to set the python-version.

[weiji14]

Gentle nudge again @danielnachun, are you ok with me keeping pip check disabled, or would you prefer another solution from conda-forge/ninja-feedstock#26?

[danielnachun]

At this point you can probably just patch out ninja from the pyproject.toml - if pip check doesn't work for this package, it also won't work for any Python package that ever depends on this, which is more of a concern. A simple sed -i 's/"ninja",//g' pyproject.toml should hopefully let pip check pass.

[weiji14]

Ok, I've added the patch, and Python 3.10 build's pip check seems to have passed. Just need to wait for the other matrix CI builds to finish.

[weiji14]

Linter error seems to be due to #30936, and windows build error can be ignored since this recipe is for CUDA-enabled Linux builds only.

[danielnachun]

Suggested change

	skip:
	# mamba-ssm requires CUDA (or ROCm) GPU
	- cuda_compiler_version == "None"
	skip:
	# mamba-ssm requires CUDA (or ROCm) GPU
	- cuda_compiler_version == "None"
	- not linux

I think we should also just be skipping macOS and Windows altogether here since this requires a CUDA compiler.

[danielnachun]

Linter error seems to be due to #30936, and windows build error can be ignored since this recipe is for CUDA-enabled Linux builds only.

That does appear to be unrelated, but let's give the upstream team a chance to respond since that issue so new just to make sure there isn't some weird cryptic issue with the recipe itself here.

[danielnachun]

My apologies but it appears a separate PR for this was already merged: #30861. In that case we'll have to close this, but please feel free to add yourself as a maintainer to the feedstock at https://github.com/conda-forge/mamba-ssm-feedstock.

[weiji14]

Oh great, I'll try and bring the changes from here into https://github.com/conda-forge/mamba-ssm-feedstock then. Thanks again for your time.