| #!/usr/bin/env python3 | |
| # -*- coding: utf-8 -*- | |
| """ | |
| HOS Model Optimizer - 小模型优化工具包 | |
| 提供模型量化、推理、训练、合并和部署的完整工具链， | |
| 针对 8GB VRAM 场景进行了优化。 | |
| 主要功能： | |
| - 量化：GGUF、AWQ、GPTQ 等多种量化方法 | |
| - 推理：支持 llama-cpp、vLLM、SGLang 三种后端 | |
| - 训练：QLoRA/LoRA 微调 | |
| - 部署：一键部署 API 服务 | |
| - 配置：智能配置管理和优化 | |
| """ | |
| __version__ = "1.0.0" | |
| __author__ = "HOS Team" | |
| # 导出主要类 | |
| from .config import ConfigManager | |
| from .inference import UnifiedInferenceEngine | |
| from .quantize import QuantizationError | |
| from .train import TrainingConfig | |
| from .deploy import HardwareDetector, ConfigSelector, ServiceLauncher | |
| __all__ = [ | |
| "__version__", | |
| "__author__", | |
| "ConfigManager", | |
| "UnifiedInferenceEngine", | |
| "QuantizationError", | |
| "TrainingConfig", | |
| "HardwareDetector", | |
| "ConfigSelector", | |
| "ServiceLauncher", | |
| ] | |