Muilti output regression about saint HOT 1 CLOSED

I believe I solved this issue. For anyone who is experiencing the same problem.
The main problem is with the shapes so you need to do several small changes to make it work.

First go to the script train.py and change the y_dim variable from 1 to the number of classes you have (notice that if the task is set to regression the code will force y_dim to be 1, you need to change it).

Second go to the data_openml.py and change the line:

'data': y[indices].reshape(-1, 1)

in the function data_split to

'data': y[indices].reshape(-1, y_dim)

where y_dim is the number of classes you have.

Third in the data_openml.py script in the __init__ method of the DataSetCatCon class you need to change the lines:

self.cls = np.zeros_like(self.y,dtype=int)
self.cls_mask = np.ones_like(self.y,dtype=int)

to

self.cls = np.zeros(shape=(len(self.y), 1), dtype=int)
self.cls_mask = np.ones(shape=(len(self.y), 1), dtype=int)

This was because we change the shape of y and we are adding a cls token column to the features, so those two lines need to create a cls token column in the length of the number of data points, but if we leave them as they are we will not get a column but a matrix with the size of the number of data points to the number of classes.

Notice that if you just change the numbers it might create a problem later if you wish to train the model on other tasks or different class numbers so you might need to change it back if you want to train the model on different tasks.

from saint.