Skip to content

vllm.entrypoints.cli.run_batch

RunBatchSubcommand

Bases: CLISubcommand

The run-batch subcommand for vLLM CLI.

Source code in vllm/entrypoints/cli/run_batch.py 
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
class RunBatchSubcommand(CLISubcommand):
    """The `run-batch` subcommand for vLLM CLI."""

    def __init__(self):
        self.name = "run-batch"
        super().__init__()

    @staticmethod
    def cmd(args: argparse.Namespace) -> None:
        logger.info("vLLM batch processing API version %s", VLLM_VERSION)
        logger.info("args: %s", args)

        # Start the Prometheus metrics server.
        # LLMEngine uses the Prometheus client
        # to publish metrics at the /metrics endpoint.
        if args.enable_metrics:
            logger.info("Prometheus metrics enabled")
            start_http_server(port=args.port, addr=args.url)
        else:
            logger.info("Prometheus metrics disabled")

        asyncio.run(run_batch_main(args))

    def subparser_init(
            self,
            subparsers: argparse._SubParsersAction) -> FlexibleArgumentParser:
        run_batch_parser = subparsers.add_parser(
            "run-batch",
            help="Run batch prompts and write results to file.",
            description=(
                "Run batch prompts using vLLM's OpenAI-compatible API.\n"
                "Supports local or HTTP input/output files."),
            usage=
            "vllm run-batch -i INPUT.jsonl -o OUTPUT.jsonl --model <model>",
        )
        run_batch_parser = make_arg_parser(run_batch_parser)
        show_filtered_argument_or_group_from_help(run_batch_parser,
                                                  "run-batch")
        run_batch_parser.epilog = VLLM_SUBCMD_PARSER_EPILOG
        return run_batch_parser

name instance-attribute 

name = 'run-batch'

__init__ 

__init__()
Source code in vllm/entrypoints/cli/run_batch.py 
22
23
24
def __init__(self):
    self.name = "run-batch"
    super().__init__()

cmd staticmethod 

cmd(args: Namespace) -> None
Source code in vllm/entrypoints/cli/run_batch.py 
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
@staticmethod
def cmd(args: argparse.Namespace) -> None:
    logger.info("vLLM batch processing API version %s", VLLM_VERSION)
    logger.info("args: %s", args)

    # Start the Prometheus metrics server.
    # LLMEngine uses the Prometheus client
    # to publish metrics at the /metrics endpoint.
    if args.enable_metrics:
        logger.info("Prometheus metrics enabled")
        start_http_server(port=args.port, addr=args.url)
    else:
        logger.info("Prometheus metrics disabled")

    asyncio.run(run_batch_main(args))

subparser_init 

subparser_init(
    subparsers: _SubParsersAction,
) -> FlexibleArgumentParser
Source code in vllm/entrypoints/cli/run_batch.py 
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
def subparser_init(
        self,
        subparsers: argparse._SubParsersAction) -> FlexibleArgumentParser:
    run_batch_parser = subparsers.add_parser(
        "run-batch",
        help="Run batch prompts and write results to file.",
        description=(
            "Run batch prompts using vLLM's OpenAI-compatible API.\n"
            "Supports local or HTTP input/output files."),
        usage=
        "vllm run-batch -i INPUT.jsonl -o OUTPUT.jsonl --model <model>",
    )
    run_batch_parser = make_arg_parser(run_batch_parser)
    show_filtered_argument_or_group_from_help(run_batch_parser,
                                              "run-batch")
    run_batch_parser.epilog = VLLM_SUBCMD_PARSER_EPILOG
    return run_batch_parser

cmd_init 

cmd_init() -> list[CLISubcommand]
Source code in vllm/entrypoints/cli/run_batch.py 
61
62
def cmd_init() -> list[CLISubcommand]:
    return [RunBatchSubcommand()]