ML Kit stores your AutoML training datasets differently, depending on your project's billing plan. When you're on the Blaze plan, ML Kit creates a new Cloud Storage bucket in your project to store AutoML Vision Edge data. When you're on the Spark or Flame plan, ML Kit stores your AutoML Vision Edge data internally instead of using your project's Cloud Storage.
If you create a dataset while on the Spark or Flame plan and later upgrade to Blaze, your dataset will be available, but will still be subject to the limitations of the Spark plan (these datasets are labeled Spark datasets in the Firebase console). If you want your dataset to take advantage of Blaze features, such as unlimited training examples (billed by storage use), you'll have to migrate the Spark dataset to a new dataset.
To migrate a dataset:
Open the AutoML section of the Firebase console. (Select your project when prompted.)
On the dataset you want to migrate, click View to open the details page, then click Export dataset. You will download a zip file containing the dataset's training images and labels.
Create a new dataset by uploading the zip file. (See Train your model.)