[Bugfix] Fix kernel benchmark (#33752)

Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
This commit is contained in:
Jee Jee Li
2026-02-22 21:18:08 -08:00
committed by GitHub
parent 987506bca6
commit 7291d1b288
14 changed files with 26 additions and 0 deletions
@@ -13,6 +13,7 @@ from torch.utils.benchmark import Measurement as TMeasurement
from tqdm import tqdm
import vllm._custom_ops as ops
from vllm.benchmarks.lib.utils import default_vllm_config
from vllm.model_executor.layers.layernorm import RMSNorm
from vllm.model_executor.layers.quantization.utils.fp8_utils import (
per_token_group_quant_fp8,
@@ -291,6 +292,7 @@ def print_timers(timers: Iterable[TMeasurement]):
compare.print()
@default_vllm_config()
def main():
torch.set_default_device("cuda")
bench_params = get_bench_params()