don't understand this line about multi-label-text-classification-using-bert HOT 3 OPEN

You can fix the latter issue by changing for (i, prediction) in enumerate(predictions): in "create_output" to for prediction in predictions:. The index "i" isn't used.

And about the test splitting: I have no clue where @javaidnabi31 got those numbers, but I changed the splitting of the data like this:

split_1 = int(0.8 * len(train))
split_2 = int(0.9 * len(train))

x_train = train[:split_1]
x_eval = train[split_1:split_2]
x_test = train[split_2:]

This will split the dataset of "train" into 80% train, 10% eval and 10% test data.

from multi-label-text-classification-using-bert.

Actually 10k i used just to tun a quicker run. Otherwise you can use x_train= train[:SIZE_TRAIN]

from multi-label-text-classification-using-bert.

I ran it on Kaggle https://www.kaggle.com/javaidnabi/toxic-comment-classification-using-bert

from multi-label-text-classification-using-bert.