| from typing import Dict, Optional |
| from importlib import import_module |
| from lightmem.configs.topic_segmenter.base import TopicSegmenterConfig |
|
|
| class TopicSegmenterFactory: |
| _MODEL_MAPPING: Dict[str, str] = { |
| "identity": "lightmem.factory.topic_segmenter.identity.IdentitySegmenter", |
| "llmlingua-2": "lightmem.factory.topic_segmenter.llmlingua_2.LlmLingua2Segmenter", |
| } |
|
|
| @classmethod |
| def from_config(cls, config: TopicSegmenterConfig, shared: bool, compressor: None): |
| """ |
| Instantiate a topic segmenter by dynamically importing the class based on config. |
| |
| Args: |
| config: TopicSegmenterConfig containing model name and specific configs |
| |
| Returns: |
| An instance of the requested segmenter model |
| |
| Raises: |
| ValueError: If model name is not supported or instantiation fails |
| ImportError: If the module or class cannot be imported |
| """ |
| model_name = config.model_name |
| |
| if model_name not in cls._MODEL_MAPPING: |
| raise ValueError( |
| f"Unsupported segmenter model: {model_name}. " |
| f"Supported models are: {list(cls._MODEL_MAPPING.keys())}" |
| ) |
|
|
| class_path = cls._MODEL_MAPPING[model_name] |
| |
| try: |
| module_path, class_name = class_path.rsplit('.', 1) |
| module = import_module(module_path) |
| segmenter_class = getattr(module, class_name) |
| if config.configs is None: |
| return segmenter_class() |
| else: |
| return segmenter_class(config=config.configs, shared = shared, compressor = compressor) |
| |
| except ImportError as e: |
| raise ImportError( |
| f"Could not import segmenter class '{class_path}': {str(e)}" |
| ) from e |
| except AttributeError as e: |
| raise ImportError( |
| f"Maybe class '{class_name}' not found in module '{module_path}': {str(e)}" |
| ) from e |
| except Exception as e: |
| raise ValueError( |
| f"Failed to instantiate {model_name} segmenter: {str(e)}" |
| ) from e |
|
|
