Fix include after ptx move

[mmarcinkiewicz]

Description

Please include a brief summary of the changes, relevant motivation and context.

Because:

• 4670342 (PR [Common] Move arch specific helpers in ptx.cuh to a new header #3392, "Move arch specific helpers in ptx.cuh to a new header") moved float_to_e8m0 out of util/ptx.cuh into the new util/ptx_arch_spec.cuh and swapped the include in all 9 pre-existing callers.
• 8d9325f (PR [Common/PyTorch] Fused grouped MXFP8 requantization #3359, "Fused grouped MXFP8 requantization") is its direct child — git rev-parse 8d9325f^ == 4670342 — branched before the move, adding a file that calls ptx::float_to_e8m0 while including only ptx.cuh.

Type of change

• Documentation change (change only to the documentation, either a fix or a new content)
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Infra/Build change
• Code refactoring

Changes

Please list the changes introduced in this PR:

• Fix include ptx_arch_spec.cuh in transformer_engine/common/cast/fused_group_requantize.cu

Checklist:

• I have read and followed the contributing guidelines
• The functionality is complete
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

[greptile-apps]

Greptile Summary

This PR fixes compilation of the fused grouped MXFP8 requantization source after an architecture-specific PTX helper was moved.

• Replaces the direct ptx.cuh include with ptx_arch_spec.cuh.
• Preserves access to general PTX declarations through the new header’s transitive include while restoring access to float_to_e8m0.

Confidence Score: 5/5

The PR appears safe to merge, with no actionable correctness, security, or build issues identified.

The new architecture-specific header provides the moved helper, transitively retains the previous PTX declarations, and is used by an architecture-specific CUDA compilation unit as intended.

Important Files Changed

Filename	Overview
transformer_engine/common/cast/fused_group_requantize.cu	Updates the PTX helper include to the intended architecture-specific header without dropping declarations from the original header.

_{Reviews (1): Last reviewed commit: "Fix include after ptx move" | Re-trigger Greptile}