Upload hos_optimizer/__init__.py with huggingface_hub
Browse files- hos_optimizer/__init__.py +37 -0
hos_optimizer/__init__.py
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
# -*- coding: utf-8 -*-
|
| 3 |
+
"""
|
| 4 |
+
HOS Model Optimizer - 小模型优化工具包
|
| 5 |
+
|
| 6 |
+
提供模型量化、推理、训练、合并和部署的完整工具链,
|
| 7 |
+
针对 8GB VRAM 场景进行了优化。
|
| 8 |
+
|
| 9 |
+
主要功能:
|
| 10 |
+
- 量化:GGUF、AWQ、GPTQ 等多种量化方法
|
| 11 |
+
- 推理:支持 llama-cpp、vLLM、SGLang 三种后端
|
| 12 |
+
- 训练:QLoRA/LoRA 微调
|
| 13 |
+
- 部署:一键部署 API 服务
|
| 14 |
+
- 配置:智能配置管理和优化
|
| 15 |
+
"""
|
| 16 |
+
|
| 17 |
+
__version__ = "1.0.0"
|
| 18 |
+
__author__ = "HOS Team"
|
| 19 |
+
|
| 20 |
+
# 导出主要类
|
| 21 |
+
from .config import ConfigManager
|
| 22 |
+
from .inference import UnifiedInferenceEngine
|
| 23 |
+
from .quantize import QuantizationError
|
| 24 |
+
from .train import TrainingConfig
|
| 25 |
+
from .deploy import HardwareDetector, ConfigSelector, ServiceLauncher
|
| 26 |
+
|
| 27 |
+
__all__ = [
|
| 28 |
+
"__version__",
|
| 29 |
+
"__author__",
|
| 30 |
+
"ConfigManager",
|
| 31 |
+
"UnifiedInferenceEngine",
|
| 32 |
+
"QuantizationError",
|
| 33 |
+
"TrainingConfig",
|
| 34 |
+
"HardwareDetector",
|
| 35 |
+
"ConfigSelector",
|
| 36 |
+
"ServiceLauncher",
|
| 37 |
+
]
|