Scikit-learn version bump and new TWC recipes
Myst Platform Release (2022-03-24)
Hello Myst Platform users,
Please update your myst-alpha package at your earliest convenience (instructions here). If you have seen errors resulting from use of The Weather Company recipes, updating your myst-alpha package should fix this.
Because of a scikit-learn version bump we have introduced, graphs that rely on scikit-learn-based connectors (Linear Regression, Extra Trees, Elastic Net, Random Forest) will not generate predictions between today's release and the graph’s next fit. This translates to just before 5pm PT to 5pm PT, if your graphs are configured to train at 5pm PT (midnight UTC).
⚡️ Enhancements:
- As mentioned above, we bumped up our scikit-learn version – in preparation for introducing additional model connectors (coming soon 👀). Graphs that rely on Linear Regression, Extra Trees, Elastic Net or Random Forest will fail between the release and their next fit. We strive to limit changes like this to once or twice per year. While we are in alpha, we will tolerate failing predictions between release and training. Rest assured that when we release in beta, we plan to handle these changes more gracefully. Thank you for your patience here!
- You can now use The Weather Company recipes (featured in this tutorial) to build time series more easily for cloud coverage, dew point temperature, and wind chill temperature. Previously, these recipes only supported temperature, relative humidity, wind direction, and wind speed.
- Before this release, XGBoost in the web UI – but not in the client library – allowed you to specify whether you wanted to predict and fit on null values. Now, you can specify these two parameters (
fit_on_null_valuesandpredict_on_null_values) for your XGBoost models through the client library as well.
As always, please do not hesitate to reach out via the chat embedded in the Myst Platform or email us at [email protected] with questions or feedback.
Thank you!
Ellery and the Myst team