mistral.rs icon indicating copy to clipboard operation
mistral.rs copied to clipboard
verified publisher icon EricLBuehler

Add blockwise fp8 gemm kernel

Open EricLBuehler opened this issue 9 months ago • 2 comments

Summary

  • add CUDA kernels for blockwise FP8 matmul
  • wire up new FFI and rust bindings
  • provide helper fp8_blockwise_gemm and test
  • compile new kernels only on CUDA >= 8.0

Testing

  • cargo test -p mistralrs-quant test_blockwise_fp8_gemm --features=cuda (fails: failed to get candle-core as a dependency due to network issues)

Summary by CodeRabbit

  • New Features
    • Added support for blockwise FP8 matrix multiplication (GEMM) on CUDA, enabling efficient computation with FP8 weights and multiple input/output precisions (FP16, BF16, FP32).
    • Introduced a new operation for blockwise FP8 GEMM, accessible via a public function.
  • Tests
    • Added tests to validate the new blockwise FP8 GEMM operation against reference outputs.

EricLBuehler avatar May 17 '25 17:05 EricLBuehler

Walkthrough

This change introduces blockwise FP8 GEMM (general matrix multiplication) support into the codebase. It adds CUDA kernel implementations and dummy stubs, updates the build system for conditional compilation based on compute capability, extends the FFI interface, and implements a Rust-side operation with validation and tests.

Changes

File(s) Change Summary
mistralrs-quant/build.rs, mistralrs-quant/src/blockwise_fp8/ffi.rs Build script now conditionally compiles real or dummy blockwise FP8 GEMM kernels and updates the HAVE_BLOCKWISE_GEMM_KERNELS constant in FFI based on CUDA capability. FFI declarations for three new kernel launcher functions were added, along with the constant.
mistralrs-quant/kernels/blockwise_fp8/blockwise_fp8_gemm.cu Added CUDA kernel and three launcher functions for blockwise FP8 GEMM supporting FP16, BF16, and FP32 input/output. Implements matrix multiplication with blockwise scaling and mixed precision, exposing C-callable entry points.
mistralrs-quant/kernels/blockwise_fp8/blockwise_fp8_gemm_dummy.cu Added dummy CUDA file with stubbed launcher functions for blockwise FP8 GEMM, each asserting on use. Used when the compute capability is insufficient.
mistralrs-quant/src/blockwise_fp8/ops.rs Introduced Fp8BlockwiseGemm struct and implemented the CustomOp3 trait. Added the fp8_blockwise_gemm function, handling validation, kernel invocation, and output allocation for the new operation. Extended the test suite with a CUDA test for blockwise FP8 GEMM.

Sequence Diagram(s)

sequenceDiagram
    participant RustOp as Fp8BlockwiseGemm (Rust)
    participant FFI as FFI Layer
    participant CUDA as CUDA Kernel

    RustOp->>RustOp: Validate input tensors
    RustOp->>FFI: Call launch_gemm_fp8_blockwise_kernel_* (based on dtype)
    FFI->>CUDA: Launch CUDA kernel with pointers and parameters
    CUDA-->>FFI: Compute output in device memory
    FFI-->>RustOp: Return output buffer
    RustOp->>RustOp: Return output tensor and shape

Poem

In the warren where numbers hop and leap,
Blockwise FP8 kernels now dig deep!
CUDA streams and tensors align,
Matrix bunnies multiply just fine.
If your GPU’s old, don’t fret or cry—
The dummy kernel just says “Goodbye!”
🐇✨

[!WARNING]

Review ran into problems

🔥 Problems

Git: Failed to clone repository. Please run the @coderabbitai full review command to re-trigger a full review. If the issue persists, set path_filters to include or exclude specific files.

✨ Finishing Touches
  • [ ] 📝 Generate Docstrings
🪧 Tips

Chat

There are 3 ways to chat with CodeRabbit:

  • Review comments: Directly reply to a review comment made by CodeRabbit. Example:
    • I pushed a fix in commit <commit_id>, please review it.
    • Explain this complex logic.
    • Open a follow-up GitHub issue for this discussion.
  • Files and specific lines of code (under the "Files changed" tab): Tag @coderabbitai in a new review comment at the desired location with your query. Examples:
    • @coderabbitai explain this code block.
    • @coderabbitai modularize this function.
  • PR comments: Tag @coderabbitai in a new PR comment to ask questions about the PR branch. For the best results, please provide a very specific query, as very limited context is provided in this mode. Examples:
    • @coderabbitai gather interesting stats about this repository and render them as a table. Additionally, render a pie chart showing the language distribution in the codebase.
    • @coderabbitai read src/utils.ts and explain its main purpose.
    • @coderabbitai read the files in the src/scheduler package and generate a class diagram using mermaid and a README in the markdown format.
    • @coderabbitai help me debug CodeRabbit configuration file.

Support

Need help? Create a ticket on our support page for assistance with any issues or questions.

Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments.

CodeRabbit Commands (Invoked using PR comments)

  • @coderabbitai pause to pause the reviews on a PR.
  • @coderabbitai resume to resume the paused reviews.
  • @coderabbitai review to trigger an incremental review. This is useful when automatic reviews are disabled for the repository.
  • @coderabbitai full review to do a full review from scratch and review all the files again.
  • @coderabbitai summary to regenerate the summary of the PR.
  • @coderabbitai generate docstrings to generate docstrings for this PR.
  • @coderabbitai generate sequence diagram to generate a sequence diagram of the changes in this PR.
  • @coderabbitai resolve resolve all the CodeRabbit review comments.
  • @coderabbitai configuration to show the current CodeRabbit configuration for the repository.
  • @coderabbitai help to get help.

Other keywords and placeholders

  • Add @coderabbitai ignore anywhere in the PR description to prevent this PR from being reviewed.
  • Add @coderabbitai summary to generate the high-level summary at a specific location in the PR description.
  • Add @coderabbitai anywhere in the PR title to generate the title automatically.

CodeRabbit Configuration File (.coderabbit.yaml)

  • You can programmatically configure CodeRabbit by adding a .coderabbit.yaml file to the root of your repository.
  • Please see the configuration documentation for more information.
  • If your editor has YAML language server enabled, you can add the path at the top of this file to enable auto-completion and validation: # yaml-language-server: $schema=https://coderabbit.ai/integrations/schema.v2.json

Documentation and Community

  • Visit our Documentation for detailed information on how to use CodeRabbit.
  • Join our Discord Community to get help, request features, and share feedback.
  • Follow us on X/Twitter for updates and announcements.

coderabbitai[bot] avatar May 17 '25 17:05 coderabbitai[bot]

Code Metrics Report 
===============================================================================
 Language            Files        Lines         Code     Comments       Blanks
===============================================================================
 C Header                3           62           53            0            9
 Dockerfile              1           41           22           10            9
 JSON                   12          107          106            0            1
 Makefile                1            6            5            0            1
 Python                 84         3713         3163          140          410
 Shell                   1           63           26           18           19
 Plain Text              3         3723            0         2413         1310
 TOML                   19          557          512            6           39
 YAML                    2           21           19            2            0
-------------------------------------------------------------------------------
 Jupyter Notebooks       3            0            0            0            0
 |- Markdown             2           77           32           31           14
 |- Python               2          205          178            1           26
 (Total)                            282          210           32           40
-------------------------------------------------------------------------------
 Markdown               55         5002            0         3812         1190
 |- BASH                 8          104          101            0            3
 |- JSON                 1           12           12            0            0
 |- Python               7          121          109            0           12
 |- Rust                22          757          634            1          122
 |- TOML                 2           75           63            0           12
 (Total)                           6071          919         3813         1339
-------------------------------------------------------------------------------
 Rust                  378       126689       113088         2587        11014
 |- Markdown           171         2145           29         1913          203
 (Total)                         128834       113117         4500        11217
===============================================================================
 Total                 562       139984       116994         8988        14002
===============================================================================

github-actions[bot] avatar May 17 '25 17:05 github-actions[bot]