Skip to content
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

[torch.compile] Add torch inductor pass for fusing silu_and_mul with subsequent scaled_fp8_quant operations #10867

Open
wants to merge 35 commits into
base: main
Choose a base branch
from
Open

[torch.compile] Add torch inductor pass for fusing silu_and_mul with subsequent scaled_fp8_quant operations #10867

wants to merge 35 commits into from
+420 −16

Conversation

SageMoore
Copy link
Contributor

@SageMoore SageMoore commented Dec 3, 2024 •
edited by github-actions bot
Loading

Credit to @LucasWilkinson for the kernel.

This pass currently only supports static per-tensor quantization. Other quantization schemes will be included in a subsequent PRs.

I've attached some QPS sweeps that were run using neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8 on an H100. Generally speaking, this pass improves the TPOT of FP8 Llama by 2-3%. There are similar improvements with TTFT with the exception of 20QPS which is much (~2x) faster.

fused_results
torch_compile_results

@github-actions GitHub Actions
Copy link

github-actions bot commented Dec 3, 2024

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

  • Add ready label to the PR
  • Enable auto-merge.

🚀

@mergify mergify bot added the ci/build label Dec 3, 2024
@SageMoore
init
2e0031a 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
remove backend format changes
8a957c7 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
format
2913716 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
move activation_quant_kernels to the quantization dir
11c6fae 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
added replacement unit test
2dfecb5 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
added kernel unit test
702fa46 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
misc cleanup
583ff4c 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
move activation quant fusion to its own pass
e5680f7 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
update test
4b775c4 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
format
d5ff865 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
format
c970dec 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
format
596c445 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
format
7ab3e18 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
format
d347431 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
format
553d99c 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
format
774559d 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
format
e2fda7f 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore SageMoore force-pushed the sage/silu-mul-quant branch from 27be0bd to e2fda7f Compare December 6, 2024 20:34
@SageMoore SageMoore marked this pull request as ready for review December 6, 2024 20:36
@SageMoore
minor comment fix
6915fa2 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
minor updates
6d4b8d0 
Signed-off-by: Sage Moore <[email protected]>
tlrmchlsmth
tlrmchlsmth reviewed Dec 12, 2024
View reviewed changes
Copy link
Collaborator

@tlrmchlsmth tlrmchlsmth left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Focused on csrc/quantization/activation_kernels.cu. spotted a couple of potential int32_t overflows

csrc/core/math.hpp Outdated Show resolved Hide resolved
csrc/quantization/activation_kernels.cu Outdated Show resolved Hide resolved
csrc/quantization/activation_kernels.cu Show resolved Hide resolved
csrc/quantization/activation_kernels.cu Show resolved Hide resolved
@tlrmchlsmth tlrmchlsmth mentioned this pull request Dec 12, 2024
Merged
@mergify Mergify
Copy link

mergify bot commented Dec 18, 2024

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @SageMoore.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

@mergify mergify bot added the needs-rebase label Dec 18, 2024
tlrmchlsmth
tlrmchlsmth approved these changes Dec 18, 2024
View reviewed changes
Copy link
Collaborator

@tlrmchlsmth tlrmchlsmth left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

A couple more comments - LGTM if we can support non power of two hidden sizes

tests/compile/test_functionalization.py Outdated Show resolved Hide resolved
tests/compile/test_functionalization.py Outdated Show resolved Hide resolved
tests/kernels/test_fused_quant_activation.py Outdated Show resolved Hide resolved
csrc/quantization/activation_kernels.cu Show resolved Hide resolved
csrc/quantization/activation_kernels.cu Show resolved Hide resolved
@mergify mergify bot removed the needs-rebase label Dec 19, 2024
@SageMoore SageMoore force-pushed the sage/silu-mul-quant branch 2 times, most recently from 21c0f3d to 70bb71b Compare December 19, 2024 16:56
@mergify mergify bot added the frontend label Dec 19, 2024
@mergify Mergify
Copy link

mergify bot commented Dec 19, 2024

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @SageMoore.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

@mergify mergify bot added the needs-rebase label Dec 19, 2024
@mergify mergify bot removed the needs-rebase label Dec 19, 2024
@SageMoore
Copy link
Contributor Author

Apologies for the noise. I accidentally added my signature to a bunch of irrelevant commits which pulled them into the PR temporarily. Things should be sorted now.

@SageMoore
fix amd build
ec1290a 
Signed-off-by: Sage Moore <[email protected]>
ProExpertProg
ProExpertProg reviewed Dec 20, 2024
View reviewed changes
csrc/quantization/activation_kernels.cu Outdated Show resolved Hide resolved
csrc/quantization/activation_kernels.cu Show resolved Hide resolved
csrc/quantization/activation_kernels.cu
} // namespace vllm

// Launch activation, gating, and quantize kernel.
#define LAUNCH_ACTIVATION_GATE_KERNEL(KERNEL) \
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is there a reason this needs a macro?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I just copied what the existing act_and_mul kernel does. This allows us to just drop in kernels for the other activation functions. I'm in favor of keeping it.

tests/compile/test_functionalization.py Outdated Show resolved Hide resolved
youkaichao
youkaichao approved these changes Dec 20, 2024
View reviewed changes
Copy link
Member

@youkaichao youkaichao left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

glad to see more fusion passes, I will hand it over to @tlrmchlsmth and @ProExpertProg for detailed review.

@SageMoore
review comments and format
008b725 
Signed-off-by: Sage Moore <[email protected]>
@SageMoore
minor test fix
4a0ac7e 
Signed-off-by: Sage Moore <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@bnellnm bnellnm bnellnm left review comments

@ProExpertProg ProExpertProg ProExpertProg left review comments

@youkaichao youkaichao youkaichao approved these changes

@tlrmchlsmth tlrmchlsmth tlrmchlsmth approved these changes

@WoosukKwon WoosukKwon Awaiting requested review from WoosukKwon WoosukKwon is a code owner

@mgoin mgoin Awaiting requested review from mgoin

@robertgshaw2-neuralmagic robertgshaw2-neuralmagic Awaiting requested review from robertgshaw2-neuralmagic

@DarkLight1337 DarkLight1337 Awaiting requested review from DarkLight1337

@ywang96 ywang96 Awaiting requested review from ywang96

@simon-mo simon-mo Awaiting requested review from simon-mo

@njhill njhill Awaiting requested review from njhill

@comaniac comaniac Awaiting requested review from comaniac

@alexm-neuralmagic alexm-neuralmagic Awaiting requested review from alexm-neuralmagic

Assignees
No one assigned
Labels
ci/build frontend ready ONLY add when PR is ready to merge/full CI is needed
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
5 participants
@SageMoore @tlrmchlsmth @ProExpertProg @youkaichao @bnellnm