- You can now label MP3 and WAV files in our new Audio editor. You can use this native audio editor to perform classification and short transcription tasks. Timer, Benchmarks, Consensus, and MAL are supported. Visit our Audio editor docs to learn more.
- Our [Model Diagnostics](doc:model-diagnostics] & Catalog features are now in GA. With these new parts of our platform, you can understand where your model is performing poorly and select data to label that will improve those problem areas.
- We restructured a couple of our core features to focus on the Data Row rather than the Label.
- The Label browser is now the Data browser (beta). When in review mode, the annotations will show in a read-only state. To make any edits to the annotations on the Data Row, the user must explicitly hit the “Edit” button.
- The new Data rows activity table (beta) allows you to easily view, filter, and manage the Data rows in your project.
- Our new Labeling functions feature allows you to find Data Rows that are more similar or less similar to a particular Data Row by adjusting a sliding scale. While functions are typically less accurate than models or human annotators, they can be used to quickly enrich data to better support data curation.
Our latest version of the Python SDK is v3.8.0. It includes the following updates. See our full changelog in Github for more details.
- ModelRun.upsert_data_rows(): Add data rows to a model run without also attaching labels.
- OperationNotAllowedException: Raised when users hit resource limits or are not allowed to use a particular operation.
ModelRun.upsert_labels(): Blocks until the upsert job is complete. Error messages have been improved
Organization.invite_limit()are no longer experimental
AnnotationGroupwas renamed to
ModelRun.delete_annotation_groups()was renamed to
ModelRun.annotation_groups()was renamed to
DataRowMetadataFieldno longer relies on pydantic for field validation and coercion. This prevents unintended type coercions from occurring.