Model training in the SDK
This guide details the process for training models outside the Snorkel Flow UI using the Python Software Development Kit (SDK). This method is particularly useful for users seeking to implement specialized models not included in the default model offerings, or to apply advanced modeling techniques not supported by the UI. Training with the SDK provides the flexibility to experiment with various model types using programmatically labeled data from Snorkel Flow.
The Snorkel Flow SDK allows for seamless integration with any modeling pipeline, enabling the training of custom models with datasets curated in Snorkel Flow. After training, models and their predictions can be registered back to the platform, allowing users to conduct detailed error analyses and make iterative refinements to their models and datasets, enhancing overall model performance.
For more information, see the following resources: