Model operations

This section shows you how to deploy code to the model zoo, to be available for other users in your organization.

Job code can be deployed to the platform model zoo, for others in the organization to use. In general terms the code is packaged to a settings schema and a Docker container. The schema is used to generate UI for settings selection on the platform. When the job is started the Docker container is run on a virtual server with drivers and hardware to support training of deep learning models.

Building a docker container

Containers running on the platform have few requirements.

• The entry point must start the training / test report

• The container application must support the following arguments

  • model_id / test_id; GUID for the job

  • api_key; Job access key

  • info_file; Settings for job

  • job_dir; location of job persistent storage directory

Start from the Telescope example application

Telescope application bitbucket repository