Differentiation between sequence and item-level datasets

The changes in this PR introduce a new category of Nucleus Dataset: the scene dataset. The purpose of the is_scene parameter to the NucleusDataset constructor is to differentiate between datasets composed of sequences and datasets composed of individual items. Scenes can only be uploaded to datasets created by setting the is_scene=True, and vice versa, items can only be added to datasets where is_scene=False

Release Model CI SDK

Adds an extension to the Nucleus Client for newly introduced Model CI functionality. With the Model CI, extension, users can use Nucleus slice’s to define test cases that describe critical edge case scenarios. A unit test can have one or more evaluation criteria, which define how to apply evaluation functions to a unit test. The newly introduced classes are:

UnitTest
EvaluationFunction
UnitTest Evaluation
UnitTestMetric
Documentation can be found at: Nucleus API Reference

Happy Coding!

Default category taxonomies

Category annotations and predictions can be uploaded without first adding a taxonomy and without specifying taxonomy_name.

Public image indexing, API Docstrings, and stability

• Docstrings now automatically formatted to https://docs.nucleus.scale.com/en/latest/api/nucleus/index.html
• Async is now required for scene upload
• Async upload stability fixes
• Continuous indexing and Image indexing are now available to all users
• Enabling continuous indexing will run a backfill image indexing job

Object Indexing and Prediction Categorizations

• Users with permissions can now generate object indexes on their datasets
• Adds support for Category Predictions

Remove need to use model run abstraction

Starting with this release it should be possible to entirely avoid the model run abstraction.

Improve retries and error messaging

• Add retries for all endpoints on certain statuses
• Add more descriptive error message for downtime
• Clean up async job for custom indexing

Patch Release (patch to previous release)

Merge pull request #122 from scaleapi/jihan/bugfix

Fix autotag export api endpoint path

Export autotag updates

• Export object autotag
• Export autotag training set positives
• More details in export for autotag

Remove dataset item ids from ingest for better uniformity

0.1.19

Bump version for release