We continue to improve the Graft platform and deal with bugs, the highlights of which are listed below. As always if you spot an issue please contact Graft Support
- Support for clustering of unstructured data.
- Support for new embedding strategy that can process longer text documents by chunking-and-averaging instead of truncating.
- 2D projections for embeddings are now computed on-demand, only when a user needs them for downstream tasks.
- Improve performance of 2D projection calculation.
- Improve instructions for connecting to user’s S3 data sources.
- Remove “Run all jobs” workflow since some jobs are now optional.
- Quick start guide and video for Image Classifier creation
- Do not ingest empty data labels as NaN.
- Update Slack default entity to have correct data modalities.
- Data Source Details page: improvements in readability of config.