Tabular_learner(dls) error


I am doing my own stuff to learn how the tabular part works.

In my DataFrame I do not have any zero values (I have stock trading data)… however I got the below error …

Where should I look or what I missed … ?

You can do some preprocessing before you model or train. Check fill missing values and normalize.

1 Like

Normilize did help but run into another error :slight_smile:

just a guess but seems like you have negative values and try to get log of it!

1 Like

Ok, I do have negative values as I was using 0/1 but got error on zero devision (of course) and changed that to -1 (-1/1) … Maybe I need to pic a positive number like 2 …
I would not maybe use boolean (True/False)?? This data column is used in the y_names…

I might need to refresh my math skills :smiley:

I suggest you take a look at docs with examples.

there says Note : Sometimes with tabular data, your y 's may be encoded (such as 0 and 1). In such a case you should explicitly pass y_block = CategoryBlock in your constructor so fastai won’t presume you are doing regression.

1 Like

Thanks, I will do so…
I also need to work on the data before throwing it into a model…

Anyway I got the thing working and getting some perditions out… however it is clear that the data needs work before I will continue with the model…

My thinking is that I will repopulate the Tabular Pandas with preliminary data, but the add some columns with calculations … I wonder if there are some good samples on how to add columns into Tabular Pandas which are calculated form the other columns in Tabular Pandas.

Anyway thank you much for the help :slight_smile: