vllm-project--vllm-omni
64 行
2.2 KiB
Python
64 行
2.2 KiB
Python
"""
|
|
vLLM-Omni: Multi-modality models inference and serving with
|
|
non-autoregressive structures.
|
|
|
|
This package extends vLLM beyond traditional text-based, autoregressive
|
|
generation to support multi-modality models with non-autoregressive
|
|
structures and non-textual outputs.
|
|
|
|
Architecture:
|
|
- 🟡 Modified: vLLM components modified for multimodal support
|
|
- 🔴 Added: New components for multimodal and non-autoregressive
|
|
processing
|
|
"""
|
|
|
|
# We import version early, because it will warn if vLLM / vLLM Omni
|
|
# are not using the same major + minor version (if vLLM is installed).
|
|
# We should do this before applying patch, because vLLM imports might
|
|
# throw in patch if the versions differ.
|
|
from .version import __version__, __version_tuple__ # isort:skip # noqa: F401
|
|
|
|
try:
|
|
from . import patch # noqa: F401
|
|
except ModuleNotFoundError as exc: # pragma: no cover - optional dependency
|
|
if exc.name != "vllm":
|
|
raise
|
|
# Allow importing vllm_omni without vllm (e.g., documentation builds)
|
|
patch = None # type: ignore
|
|
|
|
# Register custom configs (AutoConfig, AutoTokenizer) as early as possible.
|
|
from vllm_omni.transformers_utils import configs as _configs # noqa: F401, E402
|
|
from vllm_omni.transformers_utils import parsers as _parsers # noqa: F401, E402
|
|
|
|
from .config import OmniModelConfig
|
|
|
|
|
|
def __getattr__(name: str):
|
|
# Lazy import for AsyncOmni and Omni to avoid pulling in heavy
|
|
# dependencies (vllm model_loader → fused_moe → pynvml) at package
|
|
# import time. This prevents crashes in lightweight subprocesses
|
|
# (e.g. model-architecture inspection) that lack a CUDA context.
|
|
# See: https://github.com/vllm-project/vllm-omni/issues/1793
|
|
if name == "AsyncOmni":
|
|
from .entrypoints.async_omni import AsyncOmni
|
|
|
|
return AsyncOmni
|
|
if name == "Omni":
|
|
from .entrypoints.omni import Omni
|
|
|
|
return Omni
|
|
raise AttributeError(f"module {__name__!r} has no attribute {name!r}")
|
|
|
|
|
|
__all__ = [
|
|
"__version__",
|
|
"__version_tuple__",
|
|
# Main components
|
|
"Omni",
|
|
"AsyncOmni",
|
|
# Configuration
|
|
"OmniModelConfig",
|
|
# All other components are available through their respective modules
|
|
# processors.*, schedulers.*, executors.*, etc.
|
|
]
|