What’s New: MindsDB February 2021

Read on to see all of the new and improved features added in the latest version of MindsDB in February.

Improvements

  • Move MindsDB configuration storage to the sqlite database instead of configuration files
  • Allow dot notation in MongoDB find function
  • Added batch predictions for MongoDB API
  • Improve the MongoDB find parsing when user provides list as argument
  • Improved error tracking using Sentry
  • Improvements to killing the process on API error
  • New option to filter outliers in the target column when training a predictor.
  • Improved error messaging around API calls
  • Added GPU support for LightGBM mixer
  • Time series normalization based on group-by combinations
  • Added new encoder to improve time series forecasting performance.
  • improved gpu performance using round-robin scheduling

Bug fixes

  • Fix for checking the connection to Mongo Atlas
  • Fixed parsing for nested JSON as datasource
  • Fixed parsing of nested documents from MongoDB
  • Fixed issue with using TLS or SRV in mongo connection string
  • Fixed check connection option on existing connections
  • Minor bug fixes around new LightGBM mixer
  • Fix time series RNN batching to avoid out of memory issue
  • Fixed IndexError when using text encoder
  • Fix for ‘force_disable_cache’ in time series task error
2 Likes