* Return list of Vespa responses as output to `app.feed_batch` (https://github.com/vespa-engine/pyvespa/pull/361)
0.22.0
* `ListwiseRankingFramework` now accepts .csv file path as input to learn from data that are too large to load into memory (https://github.com/vespa-engine/pyvespa/pull/350) * `ListwiseRankingFramework` fit functions now accept an optional `hyperparameters` argument. If a dict of hyperparameter values is provided the method will skip the hyperparameter search and use the provided values to fit the model. (https://github.com/vespa-engine/pyvespa/pull/352) * Add parallel distribution strategy to the `ListwiseRankingFramework` hyperparameter search. (https://github.com/vespa-engine/pyvespa/pull/354)
0.21.0
* Remove duplicate documents when collecting Vespa features with `app.collect_training_data_point` (https://github.com/vespa-engine/pyvespa/pull/345) * Create `app.collect_vespa_features` and `app.store_vespa_features` that use `app.query_batch` to speedup Vespa feature collection. (https://github.com/vespa-engine/pyvespa/pull/347) * `app.collect_training_data_point` and `app.collect_training_data` will be deprecated in future releases and should be replaced by `app.collect_vespa_features` and `app.store_vespa_features`.
0.20.0
* Fix bug (https://github.com/vespa-engine/pyvespa/issues/341) that requires users to install ml packages when importing the application module. (https://github.com/vespa-engine/pyvespa/pull/342) * Improve feed_batch robustness by processing the batch in mini-batches and adding a retry layer (https://github.com/vespa-engine/pyvespa/pull/340)