Update datasets.py

          w = np.reshape(w, (n_samples, -1))

      else:

        # Set labels to be zero, with zero weights

        y = np.zeros((n_samples, 1*n_tasks))

        y = np.zeros((n_samples, n_tasks))

        w = np.zeros_like(y)

    if ids is None:

      ids = np.arange(n_samples)