3 ms·

This also means local fine-tuning is possible. Expect to see an explosion of new things like we did with Stable Diffusion, limited to some extent by the ~0.7 o

This also means local fine-tuning is possible. Expect to see an explosion of new things like we did with Stable Diffusion, limited to some extent by the ~0.7 order of magnitude more VRAM required.

bioemerl 4y ago
Does it? I would have expected compression losses to make training really hard.
- Miraste 4y ago
  The compression is optional.