AX1 · pip4.2
- params
- 138.6M
- mixer
- phasor retention · rΔ·cos(θΔ)
- shape
- 12 layers · 16 heads · d896
- context
- 256 tokens (AX0Tok, 4k vocab)
- runtime
- CPU-only · Python + NumPy
- license
- GPL-3.0
Raised on 2,040 pre-1929 novels. Talks a little like an old book, and honestly says “I don’t know.”
get the weights on hugging face ↗