wehub-resource-sync
44 lines
1.2 KiB
Python
44 lines
1.2 KiB
Python
# SPDX-License-Identifier: Apache-2.0
|
|
"""
|
|
Engine abstraction for oMLX inference.
|
|
|
|
Provides multiple engine implementations:
|
|
- BatchedEngine: Continuous batching for multiple concurrent users
|
|
- VLMBatchedEngine: Vision-language model engine with image support
|
|
- EmbeddingEngine: Batch embedding generation using mlx-embeddings
|
|
- RerankerEngine: Document reranking using SequenceClassification models
|
|
|
|
Also re-exports core engine components for backwards compatibility.
|
|
"""
|
|
|
|
# Re-export from parent engine.py for backwards compatibility
|
|
from ..engine_core import AsyncEngineCore, EngineConfig, EngineCore
|
|
from .base import BaseEngine, BaseNonStreamingEngine, GenerationOutput
|
|
from .batched import BatchedEngine
|
|
from .embedding import EmbeddingEngine
|
|
|
|
from .dflash import DFlashEngine
|
|
from .reranker import RerankerEngine
|
|
from .stt import STTEngine
|
|
from .sts import STSEngine
|
|
from .tts import TTSEngine
|
|
from .vlm import VLMBatchedEngine
|
|
|
|
__all__ = [
|
|
"BaseEngine",
|
|
"BaseNonStreamingEngine",
|
|
"GenerationOutput",
|
|
"BatchedEngine",
|
|
"DFlashEngine",
|
|
"VLMBatchedEngine",
|
|
"EmbeddingEngine",
|
|
"RerankerEngine",
|
|
"STTEngine",
|
|
"STSEngine",
|
|
"TTSEngine",
|
|
# Core engine components
|
|
"EngineCore",
|
|
"AsyncEngineCore",
|
|
"EngineConfig",
|
|
]
|