bugfix
This commit is contained in:
parent
089bc1af1f
commit
4a685b6a78
@ -6,6 +6,7 @@ import argparse
|
||||
def get_args():
|
||||
parser = argparse.ArgumentParser(description="Example script using argparse")
|
||||
parser.add_argument('--gpus', '-g', type=str, required=False, default='0', help='Identify GPU id, default is 0, comma split')
|
||||
parser.add_argument("--stream", action="store_true", help="是否流式输出")
|
||||
parser.add_argument('modelpath', type=str, help='Path to model folder')
|
||||
args = parser.parse_args()
|
||||
return args
|
||||
@ -16,7 +17,8 @@ def main():
|
||||
os.environ['CUDA_LAUNCH_BLOCKING'] = '1'
|
||||
gpus = args.gpus.split(',')
|
||||
cnt=len(gpus)
|
||||
cmdline = f'/share/vllm-0.8.5/bin/python -m llmengine.chatllm --model {args.modelpath} --gpus {cnt}'
|
||||
stream=' --stream' if args.stream else ' '
|
||||
cmdline = f'/share/vllm-0.8.5/bin/python -m llmengine.chatllm --model {args.modelpath} --gpus {cnt} {stream}'
|
||||
print(args, cmdline)
|
||||
os.system(cmdline)
|
||||
|
||||
|
Loading…
Reference in New Issue
Block a user