Skip to main content

Highlights

  • Training Jobs now support multiple Datasets: You can now attach multiple datasets to a Training Job.
  • flexai source command deprecated: The flexai source command has been deprecated. You can now pass the --repository-url and --repository-revision flags directly to the flexai training run command.

Added

  • Support for multiple Datasets in Training Jobs: You can now attach multiple datasets to a Training Job, allowing for more complex training scenarios and improved data management.
  • flexai code-registry command: The new flexai code-registry command allows you to grant access to your code repositories to the FlexAI. It replaces the functionality of the deprecated flexai source command. The flexai code-registry command provides a more streamlined and efficient way to manage your code repositories within the FlexAI ecosystem.
  • flexai training run new flags: You can now pass the --repository-url and --repository-revision flags to the flexai training run command, allowing you to pass any public GitHub repository URL or the url of a private GitHub repository that has been registered with FlexAI using the flexai code-registry connect command.

Changed

  • flexai training run command: The flexai training run command has been updated to accept multiple instances of the -D / --dataset flag, which can be used to specify a Dataset name and also supports a value mapping to specify the name of the Dataset inside the /input/ directory. For example:
    • Key-value pair:
      Will be mounted as:
    • Dataset name:
      Will be mounted as:

Removed

  • flexai source command deprecated: The flexai source command has been deprecated in favor of the flexai code-registry command.