Help with implementing tabular_learner

Hi,

I currently am working on a multi-variate classification problem where i have a target column with 6 possible outcomes. I am trying to implement the following however I believe there might be an issue with the learner (this is an assumption). The following is what i have implemented thus far:

dep_var = ‘target’
cont_names = cont_names
cat_names = cat_names
procs = [Normalize] #other pre-processing done in sklearn, just using normalize on contiuous values here

test = TabularList.from_df(df.iloc[test_index].copy(), path=path, cat_names=cat_names, cont_names=cont_names)

data = (TabularList.from_df(df, path=path, cat_names=cat_names, cont_names=cont_names, procs=procs)
.split_by_idx(test_index)
.label_from_df(cols=dep_var)
.add_test(test)
.databunch())
#this shows the dataset
data.show_batch(rows=10)

learn = tabular_learner(data, layers=[200,100], metrics=accuracy)
learn.fit(1, 1e-2)

When I get here I see an issue with the model. The following is the output.
epoch: 0
train_loss: nan
valid_loss:nan
accuracy: 0.098435
time: 00:43

Am i missing anything when doing training on 6 possible categories? Any direction would be greatly appreciated!

1 Like