3 ms·
The only difference is the label, really. The underlying transformer architecture and the approach of using a codebook is identical to a large language model. T
by ShamelessC 3y ago
The only difference is the label, really. The underlying transformer architecture and the approach of using a codebook is identical to a large language model. The same approach was also used originally for image generation in DALL-E 1.