3 ms·
Textual inversion also takes an entire concept (including context, style, and 3D attributes) and fits it into 768 float32s / 4KB.
by minimaxir 4y ago
Textual inversion also takes an entire concept (including context, style, and 3D attributes) and fits it into 768 float32s / 4KB.
- lifthrasiir 4y agoThat's a different topic, as you need the entire model to "decompress" that 4 KB data. I'm talking about the model itself.