xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-09 16:35:43 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Benchmarks] Throw usage error when using dataset-name random and dataset-path together (#24819)

Browse Source 
Signed-off-by: Ye (Charlotte) Qi <yeq@meta.com>
This commit is contained in:
Ye (Charlotte) Qi 2025-09-14 10:50:01 -07:00 committed by GitHub
parent 1177dd53e9
commit ff68035932
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 22 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
vllm/benchmarks/datasets.py
View File

@ -11,6 +11,7 @@ generation. Supported dataset types include:
- HuggingFace
- VisionArena
"""
import argparse
import ast
import base64
import io
@ -1019,6 +1020,25 @@ class ShareGPTDataset(BenchmarkDataset):
return samples
class _ValidateDatasetArgs(argparse.Action):
"""Argparse action to validate dataset name and path compatibility."""
def __call__(self, parser, namespace, values, option_string=None):
setattr(namespace, self.dest, values)
# Get current values of both dataset_name and dataset_path
dataset_name = getattr(namespace, 'dataset_name', 'random')
dataset_path = getattr(namespace, 'dataset_path', None)
# Validate the combination
if dataset_name == "random" and dataset_path is not None:
parser.error(
"Cannot use 'random' dataset with --dataset-path. "
"Please specify the appropriate --dataset-name (e.g., "
"'sharegpt', 'custom', 'sonnet') for your dataset file: "
f"{dataset_path}"
)
def add_dataset_parser(parser: FlexibleArgumentParser):
parser.add_argument("--seed", type=int, default=0)
parser.add_argument(
@ -1031,6 +1051,7 @@ def add_dataset_parser(parser: FlexibleArgumentParser):
"--dataset-name",
type=str,
default="random",
action=_ValidateDatasetArgs,
choices=[
"sharegpt", "burstgpt", "sonnet", "random", "random-mm", "hf",
"custom", "prefix_repetition", "spec_bench"
@ -1046,6 +1067,7 @@ def add_dataset_parser(parser: FlexibleArgumentParser):
"--dataset-path",
type=str,
default=None,
action=_ValidateDatasetArgs,
help="Path to the sharegpt/sonnet dataset. "
"Or the huggingface dataset ID if using HF dataset.",
)