Training with array type label column

Hi All,
I have structured training data as a pandas dataframe object. The data looks like below.

The label column is my target and I want to train a convolutional layer with this data. The input values are normalized between 0 to 1. The problem is with the label column. It takes an array of classes.
I am unable to feed this data to a convolution layer. Can anyone please help me with how I can use this data for training a model?


I’m not sure I understand the issue correctly, as you wouldn’t pass the targets to the model (or conv layer as mentioned in your post).
Assuming you are working on a multi-label classification, you could create a multi-hot encoded target and use nn.BCEWithLogitsLoss as the criterion. Each active class would then be represented by a 1 in the target, while inactive classes would get a 0.