The geometric structure does contribute to the output, about 6-7%. It's essentially a small nudge improvement system.
As it stands, this system is TOPICAL at best, and it still helped. Just not enough.
This can be gradually or greatly improved with the correct steps, including expanded control, more encoding curation, and a full battery of analysis.
Automodel Now Available
image
# ── Test AutoModel loading + generation ──
from transformers import AutoModel
from torchvision.utils import save_image, make_grid
model = AutoModel.from_pretrained(
"AbstractPhil/geolip-diffusion-proto", trust_remote_code=True
).cuda()
print(f"Params: {sum(p.numel() for p in model.parameters()):,}")
print(f"Relay diagnostics: {model.get_relay_diagnostics()}")
# Generate per-class samples
class_names = ['plane','auto','bird','cat','deer','dog','frog','horse','ship','truck']
all_imgs = []
for c in range(10):
imgs = model.sample(n_samples=4, class_label=c)
all_imgs.append(imgs)
grid = make_grid(torch.cat(all_imgs), nrow=4)
save_image(grid, "automodel_test.png")
print("✓ Saved automodel_test.png — 4 per class, 10 classes")
GeoLIP Diffusion Prototype
Flow matching diffusion with constellation relay as geometric regulator.
This is an experimental prototype exploring whether fixed geometric reference frames
(constellation anchors on the unit hypersphere) can regulate the internal geometry
of a diffusion model's denoising network during generation.
Gated residual (gate init ≈ 0.047) scales the feature map
Key property: the relay preserves 99.4% geometric fidelity through 16
stacked layers where vanilla attention preserves only 7.4%. It acts as a
geometric checkpoint that prevents representation drift at the normalized
manifold boundaries between network blocks.
What This Tests
The hypothesis: diffusion models discover that noise is a deterministic
routing system (DDIM proved this — same seed always produces same image).
The constellation operates on the same principle — fixed geometric anchors
as a reference frame that noise/data routes through. By inserting the relay
at the bottleneck, we test whether explicit geometric regulation improves
or changes the flow matching dynamics.
Empirical Findings (from this research session)
Finding
Result
CV ≈ 0.20 is the natural pentachoron volume regularity of S^15
Confirmed across all precisions, 1-bit to fp64
Effective geometric dimension of trained models ≈ 16
Confirmed across 17+ architectures
Relay preserves 99.4% cos_to_orig through 16 layers
vs 7.4% for attention alone
fp8 triangulation preserves geometry perfectly
CV identical to fp32 at d=16
Noise transforms are classifiable as deterministic routing
