20231088/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[Bugfix] Correctly call cudaProfilerStop in benchmarks script (#14183)

Browse Source 
Signed-off-by: Brayden Zhong <b8zhong@uwaterloo.ca>
This commit is contained in:
Brayden Zhong 2025-03-06 19:42:49 -05:00 committed by GitHub
parent ad60bbb2b2
commit c34eeec58d
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
6 changed files with 5 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
benchmarks/kernels/benchmark_layernorm.py
View File

@ -40,7 +40,7 @@ def main(num_tokens: int,
end_time = time.perf_counter() end_time = time.perf_counter()
if profile: if profile:
torch.cuda.cudart().cudaProfilerStart() torch.cuda.cudart().cudaProfilerStop()
return (end_time - start_time) / num_iters return (end_time - start_time) / num_iters
# Warmup. # Warmup.

1
benchmarks/kernels/benchmark_lora.py
View File

@ -153,7 +153,6 @@ def ref_group_gemm(ref_out: torch.Tensor, input: torch.Tensor,
result = torch.nn.functional.linear(x, w) result = torch.nn.functional.linear(x, w)
result *= scaling result *= scaling
out_list.append(result) out_list.append(result)
torch.cat(out_list, dim=0)
cat_result = torch.cat(out_list, dim=0) cat_result = torch.cat(out_list, dim=0)

3
benchmarks/kernels/benchmark_machete.py
View File

@ -45,7 +45,6 @@ def terse_type_name(dt):
torch.float16: "fp16", torch.float16: "fp16",
torch.int8: "int8", torch.int8: "int8",
torch.float8_e4m3fn: "fp8", torch.float8_e4m3fn: "fp8",
torch.bfloat16: "bf16",
torch.float: "float", torch.float: "float",
torch.int: "int", torch.int: "int",
}[dt] }[dt]
@ -259,7 +258,7 @@ def machete_create_bench_fn(bt: BenchmarkTensors,
return lambda: ops.machete_mm( return lambda: ops.machete_mm(
a=bt.a, a=bt.a,
b_q=bt.w_q, b_q=w_q,
b_type=bt.wtype, b_type=bt.wtype,
b_group_scales=bt.w_g_s, b_group_scales=bt.w_g_s,
b_group_zeros=w_g_zp, b_group_zeros=w_g_zp,

1
benchmarks/kernels/benchmark_moe.py
View File

@ -1,6 +1,7 @@
# SPDX-License-Identifier: Apache-2.0 # SPDX-License-Identifier: Apache-2.0
import argparse import argparse
import json
import time import time
from contextlib import nullcontext from contextlib import nullcontext
from datetime import datetime from datetime import datetime

2
benchmarks/kernels/benchmark_paged_attention.py
View File

@ -176,7 +176,7 @@ def main(
end_time = time.perf_counter() end_time = time.perf_counter()
if profile: if profile:
torch.cuda.cudart().cudaProfilerStart() torch.cuda.cudart().cudaProfilerStop()
return (end_time - start_time) / num_iters return (end_time - start_time) / num_iters
# Warmup. # Warmup.

2
benchmarks/kernels/benchmark_quant.py
View File

@ -40,7 +40,7 @@ def main(num_tokens: int,
end_time = time.perf_counter() end_time = time.perf_counter()
if profile: if profile:
torch.cuda.cudart().cudaProfilerStart() torch.cuda.cudart().cudaProfilerStop()
return (end_time - start_time) / num_iters return (end_time - start_time) / num_iters
# Warmup. # Warmup.