| | --- |
| | license: apache-2.0 |
| | --- |
| | |
| | She's not ready yet, this will need to be halted. However, smaller experiments related to AE-Interpolation will begin shortly |
| |
|
| | # Update; Lyra is essentially the geofractal router |
| |
|
| | I took pretty much every concept I could remember and built it into a fair approximation of their potentials together. |
| |
|
| | There are many missing concepts, but the ongoing experiments are showing MM-VAE lyra was not a one-off experiment. Wide manifolds are showing progress and require ramping up. |
| |
|
| | # Fault 1; Invalid latent constraints |
| |
|
| | Geofractal structure with Beatrix assumes certain rules under expectations of geometric constraints. |
| |
|
| | The Flux AE does not provide these necessary constraints, and the outcome of both the SD15 and Flux1 AE show that the constraints bloat the concerns rather than condense. |
| |
|
| | Geometric architectures must condense restraints to enable the full compression capacity afforded by the Flux1 AE architecture, which means I need to run multiple experiments and train a proper interface for this before it will function correctly. |
| |
|
| | The prototypes begin today with branched behavior and MM-VAE Lyra miniature composites. The notebook will be in the repo with the rest. |
| |
|
| | # Experimental start |
| | Say hello to multi-spectrum oscillation velocity prediction flow-matching fusion. |
| |
|
| | The oscillation system replaces the denoising process. |
| |
|
| |  |
| |
|
| | This is the functional prototype because I'm too tired to stay up. |
