UCloud AI service SDK This SDK include basic APIs to run AI inference services on a http server. It currently only support tensorflow, mxnet, keras, caffe, we will continue to add more platform into it.

UCloud AI train SDK This SDK include tools for packing AI train codes into docker It currently only support tensorflow, we will continue to add more platform into it.

UaiService Tensorflow (0.11.0 tested) Tensorflow（1.1.0 tested） MXNet(0.9.5 tested) Keras(1.2.0 tested) Caffe(1.0.0 tested)

UaiTrain Tensorflow（1.1.0 tested）

1. Install your deep learning python package, such as Tensorflow, MXNet, Keras, Caffe (tested version preferred)

2. Install UCloud Ufile SDK (https://docs.ucloud.cn/storage_cdn/ufile/tools)

    $ wget http://sdk.ufile.ucloud.com.cn/python_sdk.tar.gz
    $ tar zxvf python_sdk.tar.gz
    $ cd ufile-python
    $ sudo python setup.py install
   

3. Install uai-sdk

    $ sudo sh setup.sh install
   

4. Install Flask

    $ sudo pip install Flask
   

5. Install all other lib your inference code required

1. Write your inference code according to uai API (Please refer to ./uai/README.md
2. Put your code into ./http_server/
3. Now run your http service, the default port is 8080

https://github.com/ucloud/uai-sdk-httpserver

1. Choose your deep learning python package, such as Tensorflow, MXNet, Keras, Caffe (tested version preferred)
2. Find corresponding tool under uai_tools folder
3. Use xx_deploy.py, i.e. for tensorflow $ python tf_deploy.py --help See help docs: https://docs.ucloud.cn/ai/uai-service/index
4. Notes: ai_arch, language, os, os_deps，pip files should be under the same path as xx_deploy.py

1. Find corresponding tool under uaitrain_tools folder
2. Use xx_deploy.py, i.e. for tensorflow $ python tf_deploy.py --help See help docs: https://docs.ucloud.cn/ai/uai-train
3. Notes: when use pack tool, please ensure all your code is under the same path where the pack tool located