biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Performance] Move apply_w8a8_block_fp8_linear to an op class (#24666)

Browse Source 
Signed-off-by: ElizaWszola <ewszola@redhat.com>
Signed-off-by: ElizaWszola <elizaw.9289@gmail.com>
Signed-off-by: Luka Govedič <lgovedic@redhat.com>
Signed-off-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Luka Govedič <lgovedic@redhat.com>
This commit is contained in:
ElizaWszola
2025-09-23 21:03:10 +02:00
committed by GitHub
parent 8c1c81a3de
commit 63400259d0
14 changed files with 341 additions and 201 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
benchmarks/cutlass_benchmarks/w8a8_benchmarks.py
View File

@@ -17,7 +17,7 @@ from weight_shapes import WEIGHT_SHAPES
from vllm import _custom_ops as ops
from vllm.model_executor.layers.quantization.utils.fp8_utils import (
w8a8_block_fp8_matmul,
w8a8_triton_block_scaled_mm,
)
from vllm.utils import FlexibleArgumentParser, cdiv
@@ -158,7 +158,7 @@ def bench_fp8(
"cutlass_fp8_fp8_fp16_scaled_mm_bias": lambda: ops.cutlass_scaled_mm(
a, b, scale_a, scale_b, torch.float16, bias.to(dtype=torch.float16)
),
"triton_fp8_fp8_fp16_scaled_mm_blockwise": lambda: w8a8_block_fp8_matmul(
"triton_fp8_fp8_fp16_scaled_mm_blockwise": lambda: w8a8_triton_block_scaled_mm(
a_cont, b.t(), block_scale_a, block_scale_b.t(), (128, 128)
),
"cutlass_fp8_fp8_fp16_scaled_mm_blockwise": lambda: ops.cutlass_scaled_mm(