Switch back to NVIDIA/cutlass, and upgrade to v4.2.1

[jasl]

The issue has been fixed https://github.com/NVIDIA/cutlass/blob/v4.2.1/tools/util/include/cutlass/util/mixed_dtype_utils.hpp#L180-L184

So I switch back to the official repo and upgrade to v4.2.1, which is important for ARM Blackwell

[netlify[bot]]

❌ Deploy Preview for pytorch-fbgemm-docs failed.

Name	Link
🔨 Latest commit	38720415c2e5ce856f4ea3cabb5f92dc383a5e14
🔍 Latest deploy log	https://app.netlify.com/projects/pytorch-fbgemm-docs/deploys/68eff23fd160830008bbfe1a

[meta-cla[bot]]

Hi @jasl!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at [email protected]. Thanks!

[meta-cla[bot]]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!

[facebook-github-bot]

@q10 has imported this pull request. If you are a Meta employee, you can view this in D83578508.

[q10]

The original issue that forced us to use the cutlass fork remains:

2025-09-30T17:33:52.6122654Z /__w/FBGEMM/FBGEMM/fbgemm_gpu/../external/cutlass/tools/util/include/cutlass/util/mixed_dtype_utils.hpp(180): error: namespace "cutlass" has no member "int8_t"
2025-09-30T17:33:52.6124845Z     static_assert(cute::is_same_v<T, cutlass::int8_t> ||

[jasl]

The original issue that forced us to use the cutlass fork remains:

2025-09-30T17:33:52.6122654Z /__w/FBGEMM/FBGEMM/fbgemm_gpu/../external/cutlass/tools/util/include/cutlass/util/mixed_dtype_utils.hpp(180): error: namespace "cutlass" has no member "int8_t"
2025-09-30T17:33:52.6124845Z     static_assert(cute::is_same_v<T, cutlass::int8_t> ||

Ah, sorry, I read the change backwards... Could I leave the PR for a while, and I'm asking Nvidia staff for help

[jasl]

@johnnynunez This is blocking us from continuing the work on Jetson Thor. Could you help?

[q10]

Working around this with https://github.com/pytorch/FBGEMM/pull/4964, which points CUTLASS to https://github.com/jwfromm/cutlass/tree/v4.2.1-FBGEMM

[johnnynunez]

Working around this with #4964, which points CUTLASS to https://github.com/jwfromm/cutlass/tree/v4.2.1-FBGEMM

Thanks

[IwakuraRein]

@q10 Hi. Could you try adding #include "cutlass/util/type_traits.h" to that file and see if the issue remains? Thanks.

It looks like the standard int types only exist in cute namespace, not cutlass. I will create a pr to remove all cutlass::int8_t and just use int8_t like cutlass/numeric_conversion.h does.

[IwakuraRein]

FYI pr2684 has been merged into cutlass main branch

[jasl]

FYI pr2684 has been merged into cutlass main branch

Can I check out the https://github.com/NVIDIA/cutlass/commit/e6e2cc29f5e7611dfc6af0ed6409209df0068cf2 ?

[jasl]

I switched to your commit, it should be safe, because it's the first commit prior to v4.2.1