Based on the powerful computing capability of Alibaba Cloud, the deep learning solution provides you with an easy, open, and end-to-end deep learning service platform. This solution enables data scientists and algorithm engineers to quickly use Alibaba Cloud resources (including Elastic Compute Service (ECS) instances, GPU instances, Alibaba Cloud HPC, Object Storage Service (OSS), Elastic MapReduce, and Server Load Balancer) to perform data preparation, model development, model training, evaluation, prediction, and other tasks. This solution also easily transfers the deep learning capability to service APIs, accelerating the integration with business applications.

The deep learning solution has the following features:

  • Simple: Lowers the threshold for building and managing the deep learning platform.
  • Efficient: Improves the efficiency of heterogeneous computing resources, such as CPU and GPU, and provides unified user experience.
  • Open: Supports multiple mainstream deep learning frameworks, such as TensorFlow, Keras, and MXNet, and supports custom environments.
  • Full-cycle: Provides best practices for building end-to-end deep learning task process based on the powerful service system of Alibaba Cloud.
  • Service-oriented: Converts the deep learning capability to services, and easily integrates with applications on the cloud.

Start to use

  1. Prepare the environment.

    Create a container cluster. To use OSS data volumes to store data, Create an OSSFS data volume.

  2. Create a Jupyter environmentand Use Git to manage codes in the Jupyter environment.
  3. Run Standalone model training or Distributed Model Training to export the model.
  4. Use the exported model to Use TensorFlow Serving.