Solve pytorch-triton and triton package contention

[tdophung]

Description

pytorch-triton and triton packages install to the same location at site-packages/triton, and triton does not work for pytorch's torch.compile() call as there are a few things pytorch has added onto their version of triton (creating pytorch-triton to make it work and validated it with the release of torch). However pytorch-triton should in theory (and experimented) still be compatible with how jax uses it*.

Fixes # (issue)

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

• Add new env var to control when to use pytorch-triton in jax
• switch pytorch back to using/checking for pytorch-triton by default
• Add documentation (comments) on this contention of packages

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

[tdophung]

This will soon be the same as main. as this change is made here in: #1921, to be merged. it is just in this PR so I can test triton calls locally with the nitghtly jax container without running into errors because of jax 0.8.2+

[greptile-apps]

Greptile Summary

This PR resolves package conflicts between pytorch-triton and triton by:

• PyTorch extensions: Now use pytorch-triton by default (required for torch.compile())
• JAX extensions: Use standard triton by default, with NVTE_USE_PYTORCH_TRITON=1 env var to opt into pytorch-triton for mixed environments
• Runtime detection: Added _detect_triton_package() and _check_triton_compatibility() to detect which triton variant is installed and warn users appropriately
• JAX 0.8.2+ compatibility: Added version check for gpu_triton.TritonKernel constructor API changes
• New API: Added get_triton_info() function to query the installed triton package details
• Documentation: Comprehensive docstrings explaining the package options and installation requirements

Confidence Score: 4/5

• This PR is safe to merge with minor documentation cleanup needed
• The changes are well-structured with comprehensive documentation and proper version detection logic. The only issue found is a placeholder text in a docstring. The core functionality for detecting triton packages and handling JAX API changes appears correct.
• build_tools/pytorch.py has placeholder text in docstring that should be fixed

Important Files Changed

Filename	Overview
build_tools/jax.py	Added NVTE_USE_PYTORCH_TRITON env var support to conditionally select triton or pytorch-triton package in test_requirements(). Clear documentation added.
build_tools/pytorch.py	Changed from 'triton' to 'pytorch-triton' dependency for PyTorch extensions. Added documentation warning about PyTorch index requirement. Contains placeholder text in docstring.
transformer_engine/jax/triton_extensions/init.py	Added comprehensive documentation for triton package options, env vars, and usage examples. Added get_triton_info() to documented API.
transformer_engine/jax/triton_extensions/utils.py	Added triton package detection (_detect_triton_package), compatibility check (_check_triton_compatibility), get_triton_info() API, and JAX 0.8.2+ version handling for TritonKernel API.

Sequence Diagram

sequenceDiagram
    participant User
    participant setup.py
    participant build_tools/jax.py
    participant build_tools/pytorch.py
    participant triton_extensions/utils.py

    Note over User,triton_extensions/utils.py: Package Installation Flow
    
    User->>setup.py: pip install transformer-engine[pytorch]
    setup.py->>build_tools/pytorch.py: install_requirements()
    build_tools/pytorch.py-->>setup.py: ["pytorch-triton", ...]
    
    User->>setup.py: pip install transformer-engine[jax]
    setup.py->>build_tools/jax.py: test_requirements()
    build_tools/jax.py->>build_tools/jax.py: Check NVTE_USE_PYTORCH_TRITON
    alt NVTE_USE_PYTORCH_TRITON=1
        build_tools/jax.py-->>setup.py: ["pytorch-triton", ...]
    else Default
        build_tools/jax.py-->>setup.py: ["triton", ...]
    end
    
    Note over User,triton_extensions/utils.py: Runtime Detection Flow
    
    User->>triton_extensions/utils.py: import triton_extensions
    triton_extensions/utils.py->>triton_extensions/utils.py: _detect_triton_package()
    triton_extensions/utils.py->>triton_extensions/utils.py: _check_triton_compatibility()
    alt Placeholder package detected
        triton_extensions/utils.py-->>User: ImportError
    else pytorch-triton without env var
        triton_extensions/utils.py-->>User: UserWarning
    else Valid triton
        triton_extensions/utils.py-->>User: Success
    end

Loading

[greptile-apps]

Additional Comments (1)

1. transformer_engine/jax/triton_extensions/utils.py, line 322 (link)

   syntax: Typo: compile.name should be compiled.name. The variable compile is not defined in this scope - only compiled exists from line 300. This will cause a NameError at runtime for JAX versions < 0.8.2.

_{4 files reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

[greptile-apps]

Additional Comments (1)

1. build_tools/pytorch.py, line 21 (link)

   style: Placeholder text <version??> should be replaced with an actual version (e.g., cu121 or cu124) or made generic.

_{4 files reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}