机器学习平台的【模型管理】模块支持上传多种格式的模型，但是不同模型的目录结构规范有所不同，在创建模型或者模型版本时可参考如下规范。

CaffeModel
{MODEL_PATH}/
├── {NAME}.caffemodel
└── {NAME}.prototxt

MXNetParams
{MODEL_PATH}/
├── {NAME}.params
└── {NAME}-symbol.json

SavedModel
{MODEL_PATH}/
├── saved_model.pb
└── variables
    ├── variables.data-00000-of-00001
    └── variables.index

ONNX
{MODEL_PATH}/
└──{NAME}.onnx

TorchScript
{MODEL_PATH}/
└── {NAME}.pt

TensorRT
{MODEL_PATH}/
   └──{NAME}.[plan/engine]

scikit-learn
{MODEL_PATH}/
   └──{NAME}.[pkl/joblib]

XGBoost
{MODEL_PATH}/
   └──{NAME}.[model/json]
   
LightGBM
{MODEL_PATH}/
   └──{NAME}.txt