FloatMSECriterion_updateOutput received an invalid combination of arguments

Code was working yesterday, but getting this error now when using ColumnarModelData learner.

Here is the relevant code:

md = ColumnarModelData.from_data_frame(PATH, val_idxs, train_df, y, cat_flds=cat_vars, bs=128)
learner = md.get_learner(emb_szs, len(train_df.columns)-len(cat_vars), 
                  0.04, 1, [1000, 500], [0.001, 0.01], y_range=y_range)


You have to add .astype(np.float32) after y, since we no longer automatically cast that dependent variable.


Ah ok … thanks! I wasn’t aware of the change.

That could be because I didn’t tell anyone… :wink:

(I was hoping to find a way to avoid it breaking anything but ran out of time)