Skip to content

Revert "[Feature] specify model in config.yaml (#14855)" #15293

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Mar 21, 2025
Merged

Revert "[Feature] specify model in config.yaml (#14855)" #15293

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 1 addition & 3 deletions docs/source/serving/openai_compatible_server.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -184,7 +184,6 @@ For example:
```yaml
# config.yaml

model: meta-llama/Llama-3.1-8B-Instruct
host: "127.0.0.1"
port: 6379
uvicorn-log-level: "info"
Expand All @@ -193,13 +192,12 @@ uvicorn-log-level: "info"
To use the above config file:

```bash
vllm serve --config config.yaml
vllm serve SOME_MODEL --config config.yaml
```

:::{note}
In case an argument is supplied simultaneously using command line and the config file, the value from the command line will take precedence.
The order of priorities is `command line > config file values > defaults`.
e.g. `vllm serve SOME_MODEL --config config.yaml`, SOME_MODEL takes precedence over `model` in config file.
:::

## API Reference
Expand Down
7 changes: 0 additions & 7 deletions tests/config/test_config_with_model.yaml
View file
Open in desktop

This file was deleted.

12 changes: 0 additions & 12 deletions tests/conftest.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1121,15 +1121,3 @@ def pytest_collection_modifyitems(config, items):
for item in items:
if "optional" in item.keywords:
item.add_marker(skip_optional)


@pytest.fixture(scope="session")
def cli_config_file():
"""Return the path to the CLI config file."""
return os.path.join(_TEST_DIR, "config", "test_config.yaml")


@pytest.fixture(scope="session")
def cli_config_file_with_model():
"""Return the path to the CLI config file with model."""
return os.path.join(_TEST_DIR, "config", "test_config_with_model.yaml")
0 tests/config/test_config.yaml → tests/data/test_config.yaml
View file
Open in desktop
File renamed without changes.
53 changes: 11 additions & 42 deletions tests/test_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@

import pytest
import torch
from vllm_test_utils.monitor import monitor
from vllm_test_utils import monitor

from vllm.config import ParallelConfig, VllmConfig, set_current_vllm_config
from vllm.utils import (FlexibleArgumentParser, MemorySnapshot,
Expand Down Expand Up @@ -140,8 +140,7 @@ def parser():
def parser_with_config():
parser = FlexibleArgumentParser()
parser.add_argument('serve')
parser.add_argument('model_tag', nargs='?')
parser.add_argument('--model', type=str)
parser.add_argument('model_tag')
parser.add_argument('--served-model-name', type=str)
parser.add_argument('--config', type=str)
parser.add_argument('--port', type=int)
Expand Down Expand Up @@ -197,29 +196,29 @@ def test_missing_required_argument(parser):
parser.parse_args([])


def test_cli_override_to_config(parser_with_config, cli_config_file):
def test_cli_override_to_config(parser_with_config):
args = parser_with_config.parse_args([
'serve', 'mymodel', '--config', cli_config_file,
'serve', 'mymodel', '--config', './data/test_config.yaml',
'--tensor-parallel-size', '3'
])
assert args.tensor_parallel_size == 3
args = parser_with_config.parse_args([
'serve', 'mymodel', '--tensor-parallel-size', '3', '--config',
cli_config_file
'./data/test_config.yaml'
])
assert args.tensor_parallel_size == 3
assert args.port == 12312
args = parser_with_config.parse_args([
'serve', 'mymodel', '--tensor-parallel-size', '3', '--config',
cli_config_file, '--port', '666'
'./data/test_config.yaml', '--port', '666'
])
assert args.tensor_parallel_size == 3
assert args.port == 666


def test_config_args(parser_with_config, cli_config_file):
def test_config_args(parser_with_config):
args = parser_with_config.parse_args(
['serve', 'mymodel', '--config', cli_config_file])
['serve', 'mymodel', '--config', './data/test_config.yaml'])
assert args.tensor_parallel_size == 2
assert args.trust_remote_code
assert not args.multi_step_stream_outputs
Expand All @@ -241,9 +240,10 @@ def test_config_file(parser_with_config):
])


def test_no_model_tag(parser_with_config, cli_config_file):
def test_no_model_tag(parser_with_config):
with pytest.raises(ValueError):
parser_with_config.parse_args(['serve', '--config', cli_config_file])
parser_with_config.parse_args(
['serve', '--config', './data/test_config.yaml'])


# yapf: enable
Expand Down Expand Up @@ -476,34 +476,3 @@ def test_swap_dict_values(obj, key1, key2):
assert obj[key1] == original_obj[key2]
else:
assert key1 not in obj


def test_model_specification(parser_with_config,
cli_config_file,
cli_config_file_with_model):
# Test model in CLI takes precedence over config
args = parser_with_config.parse_args([
'serve', 'cli-model', '--config', cli_config_file_with_model
])
assert args.model_tag == 'cli-model'
assert args.served_model_name == 'mymodel'

# Test model from config file works
args = parser_with_config.parse_args([
'serve', '--config', cli_config_file_with_model
])
assert args.model == 'config-model'
assert args.served_model_name == 'mymodel'

# Test no model specified anywhere raises error
with pytest.raises(ValueError, match="No model specified!"):
parser_with_config.parse_args(['serve', '--config', cli_config_file])

# Test other config values are preserved
args = parser_with_config.parse_args([
'serve', 'cli-model', '--config', cli_config_file_with_model
])
assert args.tensor_parallel_size == 2
assert args.trust_remote_code is True
assert args.multi_step_stream_outputs is False
assert args.port == 12312
22 changes: 9 additions & 13 deletions vllm/entrypoints/cli/serve.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,16 +21,14 @@ def __init__(self):

@staticmethod
def cmd(args: argparse.Namespace) -> None:
# If model is specified in CLI (as positional arg), it takes precedence
if hasattr(args, 'model_tag') and args.model_tag is not None:
args.model = args.model_tag
# Otherwise use model from config (already in args.model)

# Check if we have a model specified somewhere
if args.model == EngineArgs.model: # Still has default value
# The default value of `--model`
if args.model != EngineArgs.model:
raise ValueError(
"With `vllm serve`, you should provide the model either as a "
"positional argument or in config file.")
"With `vllm serve`, you should provide the model as a "
"positional argument instead of via the `--model` option.")

# EngineArgs expects the model name to be passed as --model.
args.model = args.model_tag

uvloop.run(run_server(args))

Expand All @@ -43,12 +41,10 @@ def subparser_init(
serve_parser = subparsers.add_parser(
"serve",
help="Start the vLLM OpenAI Compatible API server",
usage="vllm serve [model_tag] [options]")
usage="vllm serve <model_tag> [options]")
serve_parser.add_argument("model_tag",
type=str,
nargs='?',
help="The model tag to serve "
"(optional if specified in config)")
help="The model tag to serve")
serve_parser.add_argument(
"--config",
type=str,
Expand Down
34 changes: 9 additions & 25 deletions vllm/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1265,29 +1265,19 @@ def _pull_args_from_config(self, args: list[str]) -> list[str]:
config_args = self._load_config_file(file_path)

# 0th index is for {serve,chat,complete}
# optionally followed by model_tag (only for serve)
# followed by model_tag (only for serve)
# followed by config args
# followed by rest of cli args.
# maintaining this order will enforce the precedence
# of cli > config > defaults
if args[0] == "serve":
model_in_cli = len(args) > 1 and not args[1].startswith('-')
model_in_config = any(arg == '--model' for arg in config_args)

if not model_in_cli and not model_in_config:
if index == 1:
raise ValueError(
"No model specified! Please specify model either in "
"command-line arguments or in config file.")

if model_in_cli:
# Model specified as positional arg, keep CLI version
args = [args[0]] + [
args[1]
] + config_args + args[2:index] + args[index + 2:]
else:
# No model in CLI, use config if available
args = [args[0]
] + config_args + args[1:index] + args[index + 2:]
"No model_tag specified! Please check your command-line"
" arguments.")
args = [args[0]] + [
args[1]
] + config_args + args[2:index] + args[index + 2:]
else:
args = [args[0]] + config_args + args[1:index] + args[index + 2:]

Expand All @@ -1305,7 +1295,9 @@ def _load_config_file(self, file_path: str) -> list[str]:
'--port': '12323',
'--tensor-parallel-size': '4'
]

"""

extension: str = file_path.split('.')[-1]
if extension not in ('yaml', 'yml'):
raise ValueError(
Expand All @@ -1330,15 +1322,7 @@ def _load_config_file(self, file_path: str) -> list[str]:
if isinstance(action, StoreBoolean)
]

# Skip model from config if it's provided as positional argument
skip_model = (hasattr(self, '_parsed_args') and self._parsed_args
and len(self._parsed_args) > 1
and self._parsed_args[0] == 'serve'
and not self._parsed_args[1].startswith('-'))

for key, value in config.items():
if skip_model and key == 'model':
continue
if isinstance(value, bool) and key not in store_boolean_arguments:
if value:
processed_args.append('--' + key)
Expand Down