Firebase ML stores your AutoML training datasets differently, depending onyour project's pricing plan. When your project is on the Blaze pricing plan,Firebase ML creates a newCloud Storage bucket in your project to storeAutoML Vision Edge data. When your project is on the Spark pricing plan,Firebase ML stores your AutoML Vision Edge data internally instead of usingyour project'sCloud Storage.

Firebase ML's AutoML Vision Edge features are deprecated. Consider usingVertex AI to automatically train ML models, which you can eitherexport as TensorFlow Lite models for on-device use ordeploy for cloud-based inference.

If you create a dataset while on the Spark pricing plan and later upgrade to theBlaze plan, your dataset will be available, but will still be subject to thelimitations of the Spark plan (these datasets are labeledSpark datasets intheFirebase console). If you want your dataset to take advantage of Blazefeatures, such as unlimited training examples (billed by storage use), you'llhave to migrate the Spark dataset to a new dataset.

To migrate a dataset:

1. Open theAutoML section of theFirebase console. (Select your project when prompted.)

2. On the dataset you want to migrate, clickView to open the details page,then clickExport dataset. You will download a zip file containing thedataset's training images and labels.

3. Create a new dataset by uploading the zip file.(SeeTrain your model.)