Skip to content

cuda with fused int4 #16154

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Draft
mergennachin wants to merge 2 commits into
base: main
Choose a base branch
from
Draft

cuda with fused int4 #16154

mergennachin wants to merge 2 commits into from

Conversation

@mergennachin
Copy link
Contributor

@mergennachin mergennachin commented Dec 9, 2025

  • GPU Device Caching for Encoder Output in CUDA Backend
  • Fused INT4 weight-only quantized matmul pass for CUDA backend

@mergennachin
GPU Device Caching for Encoder Output in CUDA Backend
0e13ae6 
Add CUDA GPU caching functionality for encoder outputs to improve
performance in ASR applications by avoiding redundant computation.

Key changes:
- Add GPU caching mechanism in cuda_backend.cpp with RAII management
- Add clear_stored_tensor option for cache control
- Add encoder output caching support in ASR runner
@pytorch-bot
Copy link

pytorch-bot bot commented Dec 9, 2025 •
edited
Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/16154

Note: Links to docs will display an error until the docs builds have been completed.

❌ 5 New Failures, 1 Unrelated Failure

As of commit 8f11434 with merge base 6cca6e6 (image):

NEW FAILURES - The following jobs have failed:

UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@meta-cla meta-cla bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Dec 9, 2025
@github-actions
Copy link

github-actions bot commented Dec 9, 2025

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

@mergennachin mergennachin force-pushed the cuda_with_fused_int4 branch 3 times, most recently from 59621ef to d4782c6 Compare December 9, 2025 18:49
@mergennachin
Fused INT4 weight-only quantized matmul pass for CUDA backend
8f11434 
Add fusion pass that combines multiple int4pack_mm operations sharing the
same input tensor into a single fused operation, reducing kernel launch
overhead for LLM attention (Q/K/V) and MLP (Gate/Up) projections.

Key changes:
- Add FuseInt4WeightOnlyQuantMatmulPass in backends/cuda/passes/
- Add CSEPass before fusion to merge duplicate preprocessing chains
- Fix AotiBackend.preprocess to properly handle PassResult from passes
  that return new graph_modules (using _update_exported_program_graph_module)
- Add comprehensive tests for the fusion pass
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

Copilot code review Copilot Awaiting requested review from Copilot Copilot will automatically review once the pull request is marked ready for review

At least 1 approving review is required to merge this pull request.

Assignees

No one assigned

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@mergennachin