Accelerated Data Science v2.5.9 is released

  • Added the following framework-specific classes for fast and easy model deployment:
    • AutoMLModel
    • SKlearnModel
    • XGBoostModel
    • LightGBMModel
    • PyTorchModel
    • TensorFlowModel
  • Added the GenericModel class for frameworks not included in the preceding list.
  • You can now deploy your models using the methods in these new classes:
    • .prepare(): Creates, runtime.yaml, model info, and schema files for running your model after deployment, and adds the model artifacts to an artifact directory.
    • .verify(): Helps test your model locally, before deploying it from the model catalog to an endpoint.
    • .save(): Saves the model and model artifacts to the model catalog.
    • .deploy(): Deploys the model from model catalog to a REST endpoint.
    • .predict(): Calls the endpoint and creates inferences from the deployed model.
  • Added support to create jobs with managed egress.
  • Fixed bug in jobs, where log entries were being dropped when there were a large number logs in a short period of time. Now you can list all logs with jobwatch().

See the Data Science notebook examples for deploying TensorFlow, XGBoost, LightGBM, and PyTorch models using these new classes. And for more information, see Data ScienceADS SDKocifs SDK, and our Data Science blog