Sixel images don't render correctly if the output box has text mixed with images

[tnwei]

This is on v0.2.0, using --experimental-images option:

── Code [8] ──────────────────────
im_rec = transforms.ToPILImage(mode='RGB')(img_final[0])

im_rec

Out[8]:

torch.Size([1, 6, 256, 256])

<PIL.Image.Image image mode=RGB size=256x256>