You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For Behaviors users, I'll add a Table Question Answering feature. Similar to Prompt (and also Premium), it will let users ask questions about their behaviors in natural language. "Which behavior impacts my sleep the most" - though it might have to be rephrased like "which column impacts the sleep column most" (I'll add helper text to that effect). Here are the models, I'll first explore tapas-base-finetuned-wtq and (tapex-base)[https://huggingface.co/microsoft/tapex-base].
I'll probably get working on this in short order. I know a lot of people use Behaviors (aka Fields), probably more than even journal entries. A lot of users are coming from Habitica, just to get some correlation insights on their Tasks there. I'm hoping this will be a fun feature; even if it's less powerful than manually architected statistical models, like we have with the correlations feature via XGBoost currently; and the upcoming #23 (DoWhy). But it will be fun to explore nonetheless.
The text was updated successfully, but these errors were encountered:
For Behaviors users, I'll add a Table Question Answering feature. Similar to Prompt (and also Premium), it will let users ask questions about their behaviors in natural language. "Which behavior impacts my sleep the most" - though it might have to be rephrased like "which column impacts the sleep column most" (I'll add helper text to that effect). Here are the models, I'll first explore tapas-base-finetuned-wtq and (tapex-base)[https://huggingface.co/microsoft/tapex-base].
I'll probably get working on this in short order. I know a lot of people use Behaviors (aka Fields), probably more than even journal entries. A lot of users are coming from Habitica, just to get some correlation insights on their Tasks there. I'm hoping this will be a fun feature; even if it's less powerful than manually architected statistical models, like we have with the correlations feature via XGBoost currently; and the upcoming #23 (DoWhy). But it will be fun to explore nonetheless.
The text was updated successfully, but these errors were encountered: