xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-15 18:35:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

enable --gpu-memory-utilization in benchmark_throughput.py (#3175)

Browse Source 
Co-authored-by: zixiao <shunli.dsl@alibaba-inc.com>
This commit is contained in:
Allen.Dou 2024-03-05 02:37:58 +08:00 committed by GitHub
parent 27a7b070db
commit 9cbc7e5f3b
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 14 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
benchmarks/benchmark_throughput.py
View File

@ -74,6 +74,7 @@ def run_vllm(
kv_cache_dtype: str, kv_cache_dtype: str,
device: str, device: str,
enable_prefix_caching: bool, enable_prefix_caching: bool,
gpu_memory_utilization: float = 0.9,
) -> float: ) -> float:
from vllm import LLM, SamplingParams from vllm import LLM, SamplingParams
llm = LLM(model=model, llm = LLM(model=model,
@ -84,6 +85,7 @@ def run_vllm(
trust_remote_code=trust_remote_code, trust_remote_code=trust_remote_code,
dtype=dtype, dtype=dtype,
max_model_len=max_model_len, max_model_len=max_model_len,
gpu_memory_utilization=gpu_memory_utilization,
enforce_eager=enforce_eager, enforce_eager=enforce_eager,
kv_cache_dtype=kv_cache_dtype, kv_cache_dtype=kv_cache_dtype,
device=device, device=device,
@ -206,13 +208,12 @@ def main(args: argparse.Namespace):
args.output_len) args.output_len)
if args.backend == "vllm": if args.backend == "vllm":
elapsed_time = run_vllm(requests, args.model, args.tokenizer, elapsed_time = run_vllm(
args.quantization, args.tensor_parallel_size, requests, args.model, args.tokenizer, args.quantization,
args.seed, args.n, args.use_beam_search, args.tensor_parallel_size, args.seed, args.n, args.use_beam_search,
args.trust_remote_code, args.dtype, args.trust_remote_code, args.dtype, args.max_model_len,
args.max_model_len, args.enforce_eager, args.enforce_eager, args.kv_cache_dtype, args.device,
args.kv_cache_dtype, args.device, args.enable_prefix_caching, args.gpu_memory_utilization)
args.enable_prefix_caching)
elif args.backend == "hf": elif args.backend == "hf":
assert args.tensor_parallel_size == 1 assert args.tensor_parallel_size == 1
elapsed_time = run_hf(requests, args.model, tokenizer, args.n, elapsed_time = run_hf(requests, args.model, tokenizer, args.n,
@ -287,6 +288,12 @@ if __name__ == "__main__":
'The "auto" option will use FP16 precision ' 'The "auto" option will use FP16 precision '
'for FP32 and FP16 models, and BF16 precision ' 'for FP32 and FP16 models, and BF16 precision '
'for BF16 models.') 'for BF16 models.')
parser.add_argument('--gpu-memory-utilization',
type=float,
default=0.9,
help='the fraction of GPU memory to be used for '
'the model executor, which can range from 0 to 1.'
'If unspecified, will use the default value of 0.9.')
parser.add_argument("--enforce-eager", parser.add_argument("--enforce-eager",
action="store_true", action="store_true",
help="enforce eager execution") help="enforce eager execution")