[Misc]add coding benchmark for speculative decoding (#15303)

Signed-off-by: CXIAAAAA <cxia0209@gmail.com>
This commit is contained in:
Chen Xia
2025-03-28 10:47:05 +08:00
committed by GitHub
parent 4ae17bf1e2
commit e7f720ea56
3 changed files with 101 additions and 21 deletions
+11 -5
View File
@@ -53,8 +53,9 @@ except ImportError:
from argparse import ArgumentParser as FlexibleArgumentParser
from benchmark_dataset import (BurstGPTDataset, HuggingFaceDataset,
RandomDataset, SampleRequest, ShareGPTDataset,
SonnetDataset, VisionArenaDataset)
InstructCoderDataset, RandomDataset,
SampleRequest, ShareGPTDataset, SonnetDataset,
VisionArenaDataset)
from benchmark_utils import convert_to_pytorch_benchmark_format, write_to_json
MILLISECONDS_TO_SECONDS_CONVERSION = 1000
@@ -588,9 +589,14 @@ def main(args: argparse.Namespace):
elif args.dataset_name == "hf":
# Choose between VisionArenaDataset
# and HuggingFaceDataset based on provided parameters.
dataset_class = (VisionArenaDataset if args.dataset_path
== VisionArenaDataset.VISION_ARENA_DATASET_PATH
and args.hf_subset is None else HuggingFaceDataset)
dataset_class = HuggingFaceDataset
if args.dataset_path == VisionArenaDataset.VISION_ARENA_DATASET_PATH:
assert args.hf_subset is None, "VisionArenaDataset needs hf_subset to be None." #noqa: E501
dataset_class = VisionArenaDataset
elif args.dataset_path == "likaixin/InstructCoder":
dataset_class = InstructCoderDataset
args.hf_split = "train"
input_requests = dataset_class(
dataset_path=args.dataset_path,
dataset_subset=args.hf_subset,