Changelog

Version 0.0.1

Datastores:

• Added
  • Support for encoding and indexing in a very convenient way (using pip packages) Multi-gpu support for encoding with SBERT models.
  • Tutorials for an easy understanding of datastores.

Models:

• Added
  • ONNX models: BERT, Roberta, T5 and BART
  • Automated deployment of ONNX models
  • Improved the model management server and added API to update the parameters of deployed models. This makes it easier to use GPU inference.
• Fixed
  • Issues in the generation pipeline of adapter-transformers

Explainability:

• Add checklist tests for Commonsense QA

UI:

• Added
  • Landing page
  • Feedback page
  • Commonsense explainability data
• Fixed
  • When providing context on an open domain skill the context was visualized instead of the prediction documents Mobile landscape view used default safe area inserts

Skills:

• Added
  • documentation of skills, skill-manager, and skill-API
• Updated
  • Changes to .env file setup for better secrets management

Documentation:

• Added:
  • Local setup of SQuARE platform
• Updated
  • project README with the new additions
  • documentation website to align with the new changes