🐛 Fixes

• Updating base_request_to_hub method to allow handling of responses without json
  • PR: #155

📚 Documentation

• Adding more informative error messaging due to potential JWT and Clerk data discrepancy
  • PR: #144

🧹 Chores

• Adding more informative error messaging due to potential JWT and Clerk data discrepancy
  • PR: #144

🚀 Features

• optimizing time complexity of split field validator
  • PR: #148

🚀 Features

• Allow an empty train partition to support test-set only (e.g. zero-shot) benchmarks
  • PR: #135

🐛 Fixes

• Fix: PolarisFileSystem uses the dataset name instead of the dataset slug, causing a 404
  • PR: #146

🚀 Features

• Adding a loading indicator for upload/download operations
  • PR: #140

🐛 Fixes

• Only adding zarrRootPath field to datasets that use Zarr
  • PR: #137
• Fix undefined split during checksum calc with multiple test sets
  • PR: #141

🐛 Fixes

• fix: Issue where a None token breaks the login flow
  • PR: #136

🚀 Features

• Include a checksum for Zarr archives in the Dataset checksum
  • PR: #102
• feat: JWT-based authentication for Hub API requests
  • PR: #122

🐛 Fixes

• fix: Errors in login and owner parsing
  • PR: #129
• Fix auth token being None during login
  • PR: #130

🐛 Fixes

• No longer manually upgrade mkdocs-jupyter through PyPi in the CICD
  • PR: #123

🧹 Chores

• Remove the polaris.curation module. This code now lives in a separate repository called auroris
  • PR #124

📚 Documentation

• Use existing benchmarks and datasets in docs
  • PR: #117
• Remove polyfill
  • PR: #118

🚀 Features

• Optimize the data loader for datasets that fit in memory or on a local disk
  • PR: #96
• feat: Update settings for production environment
  • PR: #106

🐛 Fixes

• Fix UserWarning during serialization and DeprecationWarning
  • PR: #107
• fix: Client ignoring default .env file
  • PR: #108

• Fixed bug in release process