A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 13, 15:45 UTC
[CLI] add --max-tokens to `vllm complete` (#28109)
Signed-off-by: Iceber Gu <caiwei95@hotmail.com>
Iceber Gu committed
Nov 7, 2025 at 20:21 UTC
e0d6b4a867bd53edb33496756913c30480029ba8
1 file changed
+14
-6
vllm/entrypoints/cli/openai.py
+14
-6
@@ -195,10 +195,15 @@ class CompleteCommand(CLISubcommand):
195
def cmd(args: argparse.Namespace) -> None:
196
model_name, client = _interactive_cli(args)
197
198
+ kwargs = {
199
+ "model": model_name,
200
+ "stream": True,
201
+ }
202
+ if args.max_tokens:
203
+ kwargs["max_tokens"] = args.max_tokens
204
+
205
if args.quick:
199
- stream = client.completions.create(
200
- model=model_name, prompt=args.quick, stream=True
201
- )
206
+ stream = client.completions.create(prompt=args.quick, **kwargs)
207
_print_completion_stream(stream)
208
return
209
@@ -208,15 +213,18 @@ class CompleteCommand(CLISubcommand):
213
input_prompt = input("> ")
214
except EOFError:
215
break
211
- stream = client.completions.create(
212
- model=model_name, prompt=input_prompt, stream=True
213
- )
216
+ stream = client.completions.create(prompt=input_prompt, **kwargs)
217
_print_completion_stream(stream)
218
219
@staticmethod
220
def add_cli_args(parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
221
"""Add CLI arguments for the complete command."""
222
_add_query_options(parser)
223
+ parser.add_argument(
224
+ "--max-tokens",
225
+ type=int,
226
+ help="Maximum number of tokens to generate per output sequence.",
227
+ )
228
parser.add_argument(
229
"-q",
230
"--quick",