vllm-cpu 0.11.0.post2__cp312-cp312-manylinux_2_17_x86_64.whl

vllm/entrypoints/cli/__init__.py

@@ -0,0 +1,12 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+from vllm.entrypoints.cli.benchmark.latency import BenchmarkLatencySubcommand
+from vllm.entrypoints.cli.benchmark.serve import BenchmarkServingSubcommand
+from vllm.entrypoints.cli.benchmark.throughput import (
+    BenchmarkThroughputSubcommand)
+
+__all__: list[str] = [
+    "BenchmarkLatencySubcommand",
+    "BenchmarkServingSubcommand",
+    "BenchmarkThroughputSubcommand",
+]

vllm/entrypoints/cli/benchmark/base.py

@@ -0,0 +1,25 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import argparse
+
+from vllm.entrypoints.cli.types import CLISubcommand
+
+
+class BenchmarkSubcommandBase(CLISubcommand):
+    """ The base class of subcommands for vllm bench. """
+
+    help: str
+
+    @classmethod
+    def add_cli_args(cls, parser: argparse.ArgumentParser) -> None:
+        """Add the CLI arguments to the parser."""
+        raise NotImplementedError
+
+    @staticmethod
+    def cmd(args: argparse.Namespace) -> None:
+        """Run the benchmark.
+
+        Args:
+            args: The arguments to the command.
+        """
+        raise NotImplementedError

vllm/entrypoints/cli/benchmark/latency.py

@@ -0,0 +1,21 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import argparse
+
+from vllm.benchmarks.latency import add_cli_args, main
+from vllm.entrypoints.cli.benchmark.base import BenchmarkSubcommandBase
+
+
+class BenchmarkLatencySubcommand(BenchmarkSubcommandBase):
+    """ The `latency` subcommand for vllm bench. """
+
+    name = "latency"
+    help = "Benchmark the latency of a single batch of requests."
+
+    @classmethod
+    def add_cli_args(cls, parser: argparse.ArgumentParser) -> None:
+        add_cli_args(parser)
+
+    @staticmethod
+    def cmd(args: argparse.Namespace) -> None:
+        main(args)

vllm/entrypoints/cli/benchmark/main.py

@@ -0,0 +1,55 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+from __future__ import annotations
+
+import argparse
+import typing
+
+from vllm.entrypoints.cli.benchmark.base import BenchmarkSubcommandBase
+from vllm.entrypoints.cli.types import CLISubcommand
+from vllm.entrypoints.utils import VLLM_SUBCMD_PARSER_EPILOG
+
+if typing.TYPE_CHECKING:
+    from vllm.utils import FlexibleArgumentParser
+
+
+class BenchmarkSubcommand(CLISubcommand):
+    """ The `bench` subcommand for the vLLM CLI. """
+
+    name = "bench"
+    help = "vLLM bench subcommand."
+
+    @staticmethod
+    def cmd(args: argparse.Namespace) -> None:
+        args.dispatch_function(args)
+
+    def validate(self, args: argparse.Namespace) -> None:
+        pass
+
+    def subparser_init(
+            self,
+            subparsers: argparse._SubParsersAction) -> FlexibleArgumentParser:
+        bench_parser = subparsers.add_parser(
+            self.name,
+            description=self.help,
+            usage=f"vllm {self.name} <bench_type> [options]")
+        bench_subparsers = bench_parser.add_subparsers(required=True,
+                                                       dest="bench_type")
+
+        for cmd_cls in BenchmarkSubcommandBase.__subclasses__():
+            cmd_subparser = bench_subparsers.add_parser(
+                cmd_cls.name,
+                help=cmd_cls.help,
+                description=cmd_cls.help,
+                usage=f"vllm {self.name} {cmd_cls.name} [options]",
+            )
+            cmd_subparser.set_defaults(dispatch_function=cmd_cls.cmd)
+            cmd_cls.add_cli_args(cmd_subparser)
+            cmd_subparser.epilog = VLLM_SUBCMD_PARSER_EPILOG.format(
+                subcmd=f"{self.name} {cmd_cls.name}")
+        return bench_parser
+
+
+def cmd_init() -> list[CLISubcommand]:
+    return [BenchmarkSubcommand()]

vllm/entrypoints/cli/benchmark/serve.py

@@ -0,0 +1,21 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import argparse
+
+from vllm.benchmarks.serve import add_cli_args, main
+from vllm.entrypoints.cli.benchmark.base import BenchmarkSubcommandBase
+
+
+class BenchmarkServingSubcommand(BenchmarkSubcommandBase):
+    """ The `serve` subcommand for vllm bench. """
+
+    name = "serve"
+    help = "Benchmark the online serving throughput."
+
+    @classmethod
+    def add_cli_args(cls, parser: argparse.ArgumentParser) -> None:
+        add_cli_args(parser)
+
+    @staticmethod
+    def cmd(args: argparse.Namespace) -> None:
+        main(args)

vllm/entrypoints/cli/benchmark/throughput.py

@@ -0,0 +1,21 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import argparse
+
+from vllm.benchmarks.throughput import add_cli_args, main
+from vllm.entrypoints.cli.benchmark.base import BenchmarkSubcommandBase
+
+
+class BenchmarkThroughputSubcommand(BenchmarkSubcommandBase):
+    """ The `throughput` subcommand for vllm bench. """
+
+    name = "throughput"
+    help = "Benchmark offline inference throughput."
+
+    @classmethod
+    def add_cli_args(cls, parser: argparse.ArgumentParser) -> None:
+        add_cli_args(parser)
+
+    @staticmethod
+    def cmd(args: argparse.Namespace) -> None:
+        main(args)

vllm/entrypoints/cli/collect_env.py

@@ -0,0 +1,36 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+from __future__ import annotations
+
+import argparse
+import typing
+
+from vllm.collect_env import main as collect_env_main
+from vllm.entrypoints.cli.types import CLISubcommand
+
+if typing.TYPE_CHECKING:
+    from vllm.utils import FlexibleArgumentParser
+
+
+class CollectEnvSubcommand(CLISubcommand):
+    """The `collect-env` subcommand for the vLLM CLI. """
+    name = "collect-env"
+
+    @staticmethod
+    def cmd(args: argparse.Namespace) -> None:
+        """Collect information about the environment."""
+        collect_env_main()
+
+    def subparser_init(
+            self,
+            subparsers: argparse._SubParsersAction) -> FlexibleArgumentParser:
+        return subparsers.add_parser(
+            "collect-env",
+            help="Start collecting environment information.",
+            description="Start collecting environment information.",
+            usage="vllm collect-env")
+
+
+def cmd_init() -> list[CLISubcommand]:
+    return [CollectEnvSubcommand()]

vllm/entrypoints/cli/main.py

@@ -0,0 +1,60 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+'''The CLI entrypoints of vLLM
+
+Note that all future modules must be lazily loaded within main
+to avoid certain eager import breakage.'''
+from __future__ import annotations
+
+import importlib.metadata
+
+
+def main():
+    import vllm.entrypoints.cli.benchmark.main
+    import vllm.entrypoints.cli.collect_env
+    import vllm.entrypoints.cli.openai
+    import vllm.entrypoints.cli.run_batch
+    import vllm.entrypoints.cli.serve
+    from vllm.entrypoints.utils import VLLM_SUBCMD_PARSER_EPILOG, cli_env_setup
+    from vllm.utils import FlexibleArgumentParser
+
+    CMD_MODULES = [
+        vllm.entrypoints.cli.openai,
+        vllm.entrypoints.cli.serve,
+        vllm.entrypoints.cli.benchmark.main,
+        vllm.entrypoints.cli.collect_env,
+        vllm.entrypoints.cli.run_batch,
+    ]
+
+    cli_env_setup()
+
+    parser = FlexibleArgumentParser(
+        description="vLLM CLI",
+        epilog=VLLM_SUBCMD_PARSER_EPILOG.format(subcmd="[subcommand]"),
+    )
+    parser.add_argument(
+        '-v',
+        '--version',
+        action='version',
+        version=importlib.metadata.version('vllm'),
+    )
+    subparsers = parser.add_subparsers(required=False, dest="subparser")
+    cmds = {}
+    for cmd_module in CMD_MODULES:
+        new_cmds = cmd_module.cmd_init()
+        for cmd in new_cmds:
+            cmd.subparser_init(subparsers).set_defaults(
+                dispatch_function=cmd.cmd)
+            cmds[cmd.name] = cmd
+    args = parser.parse_args()
+    if args.subparser in cmds:
+        cmds[args.subparser].validate(args)
+
+    if hasattr(args, "dispatch_function"):
+        args.dispatch_function(args)
+    else:
+        parser.print_help()
+
+
+if __name__ == "__main__":
+    main()

vllm/entrypoints/cli/openai.py

@@ -0,0 +1,233 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+from __future__ import annotations
+
+import argparse
+import os
+import signal
+import sys
+from typing import TYPE_CHECKING
+
+from openai import OpenAI
+from openai.types.chat import ChatCompletionMessageParam
+
+from vllm.entrypoints.cli.types import CLISubcommand
+
+if TYPE_CHECKING:
+    from vllm.utils import FlexibleArgumentParser
+
+
+def _register_signal_handlers():
+
+    def signal_handler(sig, frame):
+        sys.exit(0)
+
+    signal.signal(signal.SIGINT, signal_handler)
+    signal.signal(signal.SIGTSTP, signal_handler)
+
+
+def _interactive_cli(args: argparse.Namespace) -> tuple[str, OpenAI]:
+    _register_signal_handlers()
+
+    base_url = args.url
+    api_key = args.api_key or os.environ.get("OPENAI_API_KEY", "EMPTY")
+    openai_client = OpenAI(api_key=api_key, base_url=base_url)
+
+    if args.model_name:
+        model_name = args.model_name
+    else:
+        available_models = openai_client.models.list()
+        model_name = available_models.data[0].id
+
+    print(f"Using model: {model_name}")
+
+    return model_name, openai_client
+
+
+def _print_chat_stream(stream) -> str:
+    output = ""
+    for chunk in stream:
+        delta = chunk.choices[0].delta
+        if delta.content:
+            output += delta.content
+            print(delta.content, end="", flush=True)
+    print()
+    return output
+
+
+def _print_completion_stream(stream) -> str:
+    output = ""
+    for chunk in stream:
+        text = chunk.choices[0].text
+        if text is not None:
+            output += text
+            print(text, end="", flush=True)
+    print()
+    return output
+
+
+def chat(system_prompt: str | None, model_name: str, client: OpenAI) -> None:
+    conversation: list[ChatCompletionMessageParam] = []
+    if system_prompt is not None:
+        conversation.append({"role": "system", "content": system_prompt})
+
+    print("Please enter a message for the chat model:")
+    while True:
+        try:
+            input_message = input("> ")
+        except EOFError:
+            break
+        conversation.append({"role": "user", "content": input_message})
+
+        stream = client.chat.completions.create(model=model_name,
+                                                messages=conversation,
+                                                stream=True)
+        output = _print_chat_stream(stream)
+        conversation.append({"role": "assistant", "content": output})
+
+
+def _add_query_options(
+        parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
+    parser.add_argument(
+        "--url",
+        type=str,
+        default="http://localhost:8000/v1",
+        help="url of the running OpenAI-Compatible RESTful API server")
+    parser.add_argument(
+        "--model-name",
+        type=str,
+        default=None,
+        help=("The model name used in prompt completion, default to "
+              "the first model in list models API call."))
+    parser.add_argument(
+        "--api-key",
+        type=str,
+        default=None,
+        help=(
+            "API key for OpenAI services. If provided, this api key "
+            "will overwrite the api key obtained through environment variables."
+        ))
+    return parser
+
+
+class ChatCommand(CLISubcommand):
+    """The `chat` subcommand for the vLLM CLI. """
+    name = "chat"
+
+    @staticmethod
+    def cmd(args: argparse.Namespace) -> None:
+        model_name, client = _interactive_cli(args)
+        system_prompt = args.system_prompt
+        conversation: list[ChatCompletionMessageParam] = []
+
+        if system_prompt is not None:
+            conversation.append({"role": "system", "content": system_prompt})
+
+        if args.quick:
+            conversation.append({"role": "user", "content": args.quick})
+
+            stream = client.chat.completions.create(model=model_name,
+                                                    messages=conversation,
+                                                    stream=True)
+            output = _print_chat_stream(stream)
+            conversation.append({"role": "assistant", "content": output})
+            return
+
+        print("Please enter a message for the chat model:")
+        while True:
+            try:
+                input_message = input("> ")
+            except EOFError:
+                break
+            conversation.append({"role": "user", "content": input_message})
+
+            stream = client.chat.completions.create(model=model_name,
+                                                    messages=conversation,
+                                                    stream=True)
+            output = _print_chat_stream(stream)
+            conversation.append({"role": "assistant", "content": output})
+
+    @staticmethod
+    def add_cli_args(parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
+        """Add CLI arguments for the chat command."""
+        _add_query_options(parser)
+        parser.add_argument(
+            "--system-prompt",
+            type=str,
+            default=None,
+            help=("The system prompt to be added to the chat template, "
+                  "used for models that support system prompts."))
+        parser.add_argument("-q",
+                            "--quick",
+                            type=str,
+                            metavar="MESSAGE",
+                            help=("Send a single prompt as MESSAGE "
+                                  "and print the response, then exit."))
+        return parser
+
+    def subparser_init(
+            self,
+            subparsers: argparse._SubParsersAction) -> FlexibleArgumentParser:
+        parser = subparsers.add_parser(
+            "chat",
+            help="Generate chat completions via the running API server.",
+            description="Generate chat completions via the running API server.",
+            usage="vllm chat [options]")
+        return ChatCommand.add_cli_args(parser)
+
+
+class CompleteCommand(CLISubcommand):
+    """The `complete` subcommand for the vLLM CLI. """
+    name = 'complete'
+
+    @staticmethod
+    def cmd(args: argparse.Namespace) -> None:
+        model_name, client = _interactive_cli(args)
+
+        if args.quick:
+            stream = client.completions.create(model=model_name,
+                                               prompt=args.quick,
+                                               stream=True)
+            _print_completion_stream(stream)
+            return
+
+        print("Please enter prompt to complete:")
+        while True:
+            try:
+                input_prompt = input("> ")
+            except EOFError:
+                break
+            stream = client.completions.create(model=model_name,
+                                               prompt=input_prompt,
+                                               stream=True)
+            _print_completion_stream(stream)
+
+    @staticmethod
+    def add_cli_args(parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
+        """Add CLI arguments for the complete command."""
+        _add_query_options(parser)
+        parser.add_argument(
+            "-q",
+            "--quick",
+            type=str,
+            metavar="PROMPT",
+            help=
+            "Send a single prompt and print the completion output, then exit.")
+        return parser
+
+    def subparser_init(
+            self,
+            subparsers: argparse._SubParsersAction) -> FlexibleArgumentParser:
+        parser = subparsers.add_parser(
+            "complete",
+            help=("Generate text completions based on the given prompt "
+                  "via the running API server."),
+            description=("Generate text completions based on the given prompt "
+                         "via the running API server."),
+            usage="vllm complete [options]")
+        return CompleteCommand.add_cli_args(parser)
+
+
+def cmd_init() -> list[CLISubcommand]:
+    return [ChatCommand(), CompleteCommand()]

vllm/entrypoints/cli/run_batch.py

@@ -0,0 +1,67 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+from __future__ import annotations
+
+import argparse
+import asyncio
+import importlib.metadata
+import typing
+
+from vllm.entrypoints.cli.types import CLISubcommand
+from vllm.entrypoints.utils import VLLM_SUBCMD_PARSER_EPILOG
+from vllm.logger import init_logger
+
+if typing.TYPE_CHECKING:
+    from vllm.utils import FlexibleArgumentParser
+
+logger = init_logger(__name__)
+
+
+class RunBatchSubcommand(CLISubcommand):
+    """The `run-batch` subcommand for vLLM CLI."""
+    name = "run-batch"
+
+    @staticmethod
+    def cmd(args: argparse.Namespace) -> None:
+        from vllm.entrypoints.openai.run_batch import main as run_batch_main
+
+        logger.info("vLLM batch processing API version %s",
+                    importlib.metadata.version("vllm"))
+        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:
+            from prometheus_client import start_http_server
+
+            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:
+        from vllm.entrypoints.openai.run_batch import make_arg_parser
+
+        run_batch_parser = subparsers.add_parser(
+            self.name,
+            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)
+        run_batch_parser.epilog = VLLM_SUBCMD_PARSER_EPILOG.format(
+            subcmd=self.name)
+        return run_batch_parser
+
+
+def cmd_init() -> list[CLISubcommand]:
+    return [RunBatchSubcommand()]