https://a2ml.org/dev/quickstart.html
Features & Enhancements:
- A2ml support for Oversampling method, Trails per worker and Class weight [154](https://github.com/augerai/a2ml/issues/154)
- make Azure experiment and cluster names to comply with Azure naming policies [65](https://github.com/augerai/a2ml/issues/65)
- A2ML.predict call should allow a prediction ID to be passed to it [350](https://github.com/augerai/a2ml/issues/350)
- A2ML predict API should return `prediction_group_id` [346](https://github.com/augerai/a2ml/issues/346)
- allow deleting predictions and actuals from review [343](https://github.com/augerai/a2ml/issues/343)
- Support retry with exponential backoff if Azure web services return a 401 [334](https://github.com/augerai/a2ml/issues/334)
- Support pd.DataFrame as input/output for API methods [326](https://github.com/augerai/a2ml/issues/326)
- Upgrade pandas to v1.0+ [324](https://github.com/augerai/a2ml/issues/324)
- add evaluate_stop_task to hub_tasks [307](https://github.com/augerai/a2ml/issues/307)
- Check which primary metrics Azure support and update doc [303](https://github.com/augerai/a2ml/issues/303)
- Support model undeploy [282](https://github.com/augerai/a2ml/issues/282)
- treat T and F as true and false boolean values on importing data [269](https://github.com/augerai/a2ml/issues/269)
- provider arg should be a choice instead of a string [267](https://github.com/augerai/a2ml/issues/267)
- Remove actuals method and provide example how to use actual for one value [258](https://github.com/augerai/a2ml/issues/258)
- Support parquet data files [254](https://github.com/augerai/a2ml/issues/254)
- Leaderboard for Auger should contain error description if status is error [247](https://github.com/augerai/a2ml/issues/247)
- Should not have ‘auger’ or ‘azure’ in the results if method was run for one provider [235](https://github.com/augerai/a2ml/issues/235)
- Actual and actuals should not calculate score and return error if one or more ids are not exist [234](https://github.com/augerai/a2ml/issues/234)
- add A2ML.actual with just one value [232](https://github.com/augerai/a2ml/issues/232)
- host actuals update for credit app script on Heroku to run every day [231](https://github.com/augerai/a2ml/issues/231)
- actuals should take an array of hashes [228](https://github.com/augerai/a2ml/issues/228)
- Azure: add review model celery tasks to run from hub [216](https://github.com/augerai/a2ml/issues/216)
- add a2ml --version [204](https://github.com/augerai/a2ml/issues/204)
- Support disabling progress spinner [198](https://github.com/augerai/a2ml/issues/198)
- Setup CI/CD w/ kubernetes [194](https://github.com/augerai/a2ml/issues/194)
- Migrate get statistics & related code from auger-ml [193](https://github.com/augerai/a2ml/issues/193)
- Azure: support local review model [190](https://github.com/augerai/a2ml/issues/190)
- RemoteRunner should support multiple providers [179](https://github.com/augerai/a2ml/issues/179)
- Remove redundant content from README [156](https://github.com/augerai/a2ml/issues/156)
- Provider configuration for a2ml server [139](https://github.com/augerai/a2ml/issues/139)
- Add optional provider argument to deploy [137](https://github.com/augerai/a2ml/issues/137)
- Move review model tasks from auger-ml to a2ml [97](https://github.com/augerai/a2ml/issues/97)
- Make predict output file name unique [64](https://github.com/augerai/a2ml/issues/64)
Bug Fixes:
- Warning message about hyperdrive loading with azureml_run_type_providers [173](https://github.com/augerai/a2ml/issues/173)
- DatasetValidationError on import_data_task in Azure [373](https://github.com/augerai/a2ml/issues/373)
- New s3fs release (0.5.0) breaks our usage of boto [362](https://github.com/augerai/a2ml/issues/362)
- r2 metric for actual return > 1 [347](https://github.com/augerai/a2ml/issues/347)
- Add retry to s3 operations [344](https://github.com/augerai/a2ml/issues/344)
- High CPU usage with count_actuals_by_prediction_id_task, distribution_chart_stats_task and score_model_performance_daily_task [336](https://github.com/augerai/a2ml/issues/336)
- threshold should be ignored for regression and timeseries [331](https://github.com/augerai/a2ml/issues/331)
- wait_for_status does not report error [320](https://github.com/augerai/a2ml/issues/320)
- Store only actual and prediction_id on score_actuals task [314](https://github.com/augerai/a2ml/issues/314)
- Auger. cluster update [304](https://github.com/augerai/a2ml/issues/304)
- monitor_evaluate_task should update error information in session [302](https://github.com/augerai/a2ml/issues/302)
- Undeploy model should remove folder from s3 [295](https://github.com/augerai/a2ml/issues/295)
- Azure predict proba issue [293](https://github.com/augerai/a2ml/issues/293)
- AZURE_CREDENTIALS_PATH is not used [288](https://github.com/augerai/a2ml/issues/288)
- Model deploy via cli fails with missing positional argument 'review' [280](https://github.com/augerai/a2ml/issues/280)
- Auger and Azure return back probabilities in different ways [270](https://github.com/augerai/a2ml/issues/270)
- Documentation doesn't render function signatures correctly with sphinx 3.1.x [245](https://github.com/augerai/a2ml/issues/245)
- Code block formatting broken for predict and select methods [241](https://github.com/augerai/a2ml/issues/241)
- build_review_data doesnt need to be exposed to user [230](https://github.com/augerai/a2ml/issues/230)
- actuals doc has a bad example [229](https://github.com/augerai/a2ml/issues/229)
- Python API calls should return the 404 rest api call error message from provider [222](https://github.com/augerai/a2ml/issues/222)
- a2ml.predict() comes back with oh so many possible things from azure [221](https://github.com/augerai/a2ml/issues/221)
- Predict documentation [217](https://github.com/augerai/a2ml/issues/217)
- Azure: Predict with probabilities for boolean values [213](https://github.com/augerai/a2ml/issues/213)
- fix pip show a2ml [205](https://github.com/augerai/a2ml/issues/205)
- predict doesn't return file name where predictions are stored [202](https://github.com/augerai/a2ml/issues/202)
- ERROR: Could not find a version that satisfies the requirement onnxruntime==1.0.0 [201](https://github.com/augerai/a2ml/issues/201)
- ModuleNotFoundError: No module named 'arff' [200](https://github.com/augerai/a2ml/issues/200)
- Azure: predict with categorical target return classes codes for target [186](https://github.com/augerai/a2ml/issues/186)
- Doc: fix a2ml.evaluate [181](https://github.com/augerai/a2ml/issues/181)
- Status returned from evaluate has different cases for providers [145](https://github.com/augerai/a2ml/issues/145)
- a2ml evaluate response [117](https://github.com/augerai/a2ml/issues/117)