Skip to content

Commit 34c0c46

Browse files
committed
fix launch latency
1 parent e7930ca commit 34c0c46

File tree

3 files changed

+9
-0
lines changed

3 files changed

+9
-0
lines changed

benchmarks/nightly/autogen.yaml

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -140,3 +140,6 @@ rope_bwd:
140140
swiglu_bwd:
141141
op: swiglu
142142
args: --op swiglu --baseline torch_swiglu --metrics speedup --bwd --only liger_swiglu,torch_swiglu
143+
launch_latency:
144+
op: launch_latency
145+
args: --op launch_latency --metrics walltime

benchmarks/nightly/gen.py

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -76,6 +76,8 @@ def process_manual_options(
7676
run_configs[benchmark]["disabled"] = True
7777
for benchmark in extra_args:
7878
run_configs[benchmark]["args"] = extra_args[benchmark]["args"]
79+
for benchmark, benchmark_config in options.get("enabled", {}).items():
80+
run_configs[benchmark] = benchmark_config.copy()
7981
return run_configs
8082

8183

benchmarks/nightly/manual.yaml

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -7,6 +7,10 @@ disabled:
77
- fp8_gemm_fwd
88
- fp8_gemm_rowwise_fwd
99
- fp8_gemm_rowwise_grouped_fwd
10+
enabled:
11+
launch_latency:
12+
op: launch_latency
13+
args: --op launch_latency --metrics walltime
1014
extra_args:
1115
# triton_tutorial_flash_v2_opt does not work on Triton main branch
1216
bf16_flash_attention_fwd:

0 commit comments

Comments
 (0)