HoLo-FuSe: 3-arm checkpoints (EMA incl., optimizer stripped) + model card

README.md

@@ -0,0 +1,72 @@
+---
+license: cc-by-nc-4.0
+tags: [diffusion, ddpm, class-conditional, hsl, zero-parameter, research, image-generation]
+pipeline_tag: unconditional-image-generation
+---
+
+# HoLo-FuSe — frozen 0-parameter HSL substrate as a diffusion conditioning door
+
+**Honest framing first:** this is a **minimum-scale baseline training run** whose only purpose is to prove
+that **HSL** (Holistic Signal Language — a frozen, deterministic 27-D feature frame with **0 learned
+parameters**, a 4.6 KB LUT) can serve as the **conditioning substrate** of a verified diffusion carrier.
+Not SOTA, not a product, not "HSL beats embeddings". The carrier is a standard class-conditional DDPM;
+HSL is the thing under test. **FuSe = Frozen Substrate, fused into a verified baseline.**
+
+Code & full record: [Woojiggun/HoLo-FuSe](https://github.com/Woojiggun/HoLo-FuSe) ·
+live demo: [ggunio/HoLo-FuSe-demo](https://huggingface.co/spaces/ggunio/HoLo-FuSe-demo) ·
+the zero door: [hsl-embedding-zero](https://github.com/Woojiggun/hsl-embedding-zero) (PyPI) ·
+siblings: [HoLo_ZeRo](https://huggingface.co/ggunio/HoLo_ZeRo) (byte-LM),
+[HoLo-ToLk-STT](https://huggingface.co/ggunio/HoLo-ToLk-STT) (audio).
+
+## What was verified (seed-matched, same budget, step 14000)
+
+| arm | conditioning | result |
+|---|---|---|
+| `none` | unconditional | readable cat+dog faces, mixed |
+| `hsl` | frozen HSL 27-D (0 learned params) → small readout | "Cat"→cats, "Dog"→dogs |
+| `learned` | same-budget `nn.Embedding` control | "Cat"→cats, "Dog"→dogs |
+
+- Flipping the label on the **same initial noise** morphs the sample between species → conditioning works.
+- **hsl ≈ learned**: the frozen substrate steers class as well as the learned control. Claim is
+  *comparable*, **not better** — single seed set, qualitative.
+- Known artifact: a background color tint in **all** arms (under-training of a ~35M model at 14k steps;
+  a sampling sweep showed CFG / dynamic-thresholding does not remove it). Doesn't affect the comparison.
+
+## Files
+
+| file | content |
+|---|---|
+| `holofuse_hsl_128.pt` | HSL-conditioned arm (**the demo one**) |
+| `holofuse_learned_128.pt` | learned-embedding control arm |
+| `holofuse_none_128.pt` | unconditional baseline arm |
+
+Each ≈274 MB: `{model, cond, ema, step, arch}` — EMA included (sample from EMA), optimizer stripped.
+Arch: U-Net base128, ch_mults 1,2,2,2, attn@16, ~35M params; DDPM cosine T=250; CFG cond-drop 0.15.
+
+## Use
+
+```python
+import torch, types
+from huggingface_hub import hf_hub_download
+# UNet / HSLLabelCond / DDPM classes: experiment.py in the GitHub repo (or the demo Space)
+from experiment import build_model, HSLLabelCond, DDPM
+
+st = torch.load(hf_hub_download("ggunio/HoLo-FuSe", "holofuse_hsl_128.pt"), map_location="cpu")
+margs = types.SimpleNamespace(arch="unet", base=128, ch_mults="1,2,2,2", attn_res="16", num_res=2)
+model = build_model(margs, 3, 128, 128).eval()
+cond_enc = HSLLabelCond(128)
+params = list(model.parameters()) + list(cond_enc.parameters())
+with torch.no_grad():                          # sample from EMA weights
+    for p, e in zip(params, st["ema"]): p.copy_(e)
+
+ddpm = DDPM(T=250, dyn_thresh=0.99)
+cond = cond_enc(["Cat"] * 4); uncond = torch.zeros_like(cond)
+imgs = ddpm.sample(model, cond, 3, 128, 4, guidance=1.6, uncond=uncond)   # [-1,1], [4,3,128,128]
+```
+
+## Data & license
+
+Trained on [AFHQ](https://github.com/clovaai/stargan-v2) (StarGAN v2, Choi et al. 2020) animal faces
+at 128px (Cat 5153 / Dog 4739, via `zzsi/afhq512_16k`). AFHQ is **CC BY-NC 4.0**, therefore these
+**weights and their outputs are CC BY-NC 4.0 — non-commercial, research/demo only**.
+Training: 16k steps/arm on a single free Colab T4, crash-resumable harness.

holofuse_hsl_128.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00f2f8d69a77d03ae7160a48421bd8676a7071da2ed05a2517c80f7fe870d2bc
+size 274976970

holofuse_learned_128.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cae7b44a0a4d5661114a72523847dd20e9863250b214a626a7942cabc25fa4a
+size 274892000

holofuse_none_128.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a14ef14f7dba0707d1db87d2ea3102a11a033203d3496496ee079d00ebee39ac
+size 274620794