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

Revert "[bench] Support common prefix len config (for decode-only bench)" (#31240)

Browse Source 
Signed-off-by: Ming Yang <minos.future@gmail.com>
This commit is contained in:
Ming Yang
2025-12-23 21:17:23 -08:00
committed by GitHub
parent 3ce791ac77
commit 3bb9561928
2 changed files with 0 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
vllm/benchmarks/datasets.py
View File

@@ -1847,7 +1847,6 @@ def get_samples(args, tokenizer: TokenizerLike) -> list[SampleRequest]:
random_seed=args.seed, random_seed=args.seed,
dataset_path=args.dataset_path, dataset_path=args.dataset_path,
disable_shuffle=args.disable_shuffle, disable_shuffle=args.disable_shuffle,
prefix_len=args.common_prefix_len,
).sample( ).sample(
tokenizer=tokenizer, tokenizer=tokenizer,
num_requests=args.num_prompts, num_requests=args.num_prompts,

6
vllm/benchmarks/serve.py
View File

@@ -1281,12 +1281,6 @@ def add_cli_args(parser: argparse.ArgumentParser):
help="Repetition penalty sampling parameter. Only has effect on " help="Repetition penalty sampling parameter. Only has effect on "
"openai-compatible backends.", "openai-compatible backends.",
) )
sampling_group.add_argument(
"--common-prefix-len",
type=int,
default=None,
help="Common prefix length shared by all prompts (used by random dataset)",
)
parser.add_argument( parser.add_argument(
"--served-model-name", "--served-model-name",