davidfoss commited on
Commit
d7c82f0
·
verified ·
1 Parent(s): 0b3fcbc

Update unified Origami Preview model card, wirkung.cls figures, and Family Paper PDF

Browse files
.gitattributes CHANGED
@@ -345,3 +345,7 @@ assets/banner.png filter=lfs diff=lfs merge=lfs -text
345
  assets/keyframes.png filter=lfs diff=lfs merge=lfs -text
346
  assets/mrail_network.png filter=lfs diff=lfs merge=lfs -text
347
  assets/weights_vs_origami.png filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
345
  assets/keyframes.png filter=lfs diff=lfs merge=lfs -text
346
  assets/mrail_network.png filter=lfs diff=lfs merge=lfs -text
347
  assets/weights_vs_origami.png filter=lfs diff=lfs merge=lfs -text
348
+ paper/Foss_2026_Origami_HCGM.pdf filter=lfs diff=lfs merge=lfs -text
349
+ assets/architecture.png filter=lfs diff=lfs merge=lfs -text
350
+ assets/mechanism.png filter=lfs diff=lfs merge=lfs -text
351
+ assets/mrail.png filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -4,140 +4,138 @@ license_name: wirkung-model-and-runtime-1.0
4
  license_link: LICENSE
5
  library_name: wirkung
6
  pipeline_tag: text-to-video
7
- base_model:
8
- - MiniMaxAI/MiniMax-H3
9
- base_model_relation: quantized
10
  language:
11
  - en
12
- - zh
13
  tags:
14
  - wirkung
15
  - origami
16
  - preview
 
 
 
 
 
 
17
  - hcgm
18
  - m-rail
19
- - sgpu
20
- - hmoe
21
- - constant-state
22
- - text-to-video
23
  - audio-video
24
- - flow-matching
25
- - zero-gpu
26
  ---
27
 
28
- ![Standalone joint audio-video flow-matching Holographic Causal Graph Machine for continuous uncut generation in constant memory.](assets/banner.png)
29
 
30
- [Paper](https://doi.org/10.13140/RG.2.2.28206.68167) · [Runtime (PyPI)](https://pypi.org/project/wirkung/) · [Website](https://wirku.ng) · [Base weights](https://huggingface.co/MiniMaxAI/MiniMax-H3) · [License](https://huggingface.co/wirkung/Origami-H3/blob/main/LICENSE)
31
 
32
- **Origami-H3 (Preview)** is a standalone joint audio-video flow-matching Holographic Causal Graph Machine (HCGM) compiled in closed form from [`MiniMaxAI/MiniMax-H3`](https://huggingface.co/MiniMaxAI/MiniMax-H3) with zero gradient updates and zero fine-tuning steps. Folding the coupled audio-video Diffusion Transformer (`DiT`), distilled few-step `Turbo-LoRA` trajectory adapters, hypernetwork `VSA` modulation gates, and streaming spatiotemporal video-and-audio `VAE` decoders into self-describing `.o3f` operators, Origami-H3 generates synchronized video and audio directly from NVMe under [`wirkung`](https://pypi.org/project/wirkung/) with no GPU requirement, no `ComfyUI` or `diffusers` stack, and flat `O(1)` memory across continuous uncut takes.
 
 
33
 
34
  ```bash
35
  pip install -U wirkung && wkng pull wirkung/Origami-H3 -o ./Origami-H3
36
- wkng run ./Origami-H3 -p "Evening rain on neon cobblestones." -o take.mp4
37
  ```
38
 
39
- ## Architectural Overview: Upstream vs. Origami
40
 
41
- | Architectural Dimension | Conventional AV Diffusion Stack | `Origami-H3` (Wirkung HCGM Preview) | Structural Invariant |
42
- |---|---|---|---|
43
- | Compilation & Adaptation | Pretrained joint AV flow stack | **Closed-form Parseval spectral fold** | Zero gradient or training steps |
44
- | Execution Orchestration | Multi-GPU HBM + PyTorch / ComfyUI | **Standalone native `mmap` `.o3f` engine** | Zero GPU VRAM, zero ComfyUI / diffusers |
45
- | Nonlinear Block Fidelity | Fragile under naive low-bit compression | **Parseval row-norm conservation + crystals** | Preserves nonlinear SwiGLU flow dynamics |
46
- | Trajectory & Gate Organs | Separate dense LoRA & hypernetwork files | **In-graph Walsh-folded `Turbo-LoRA` & `VSA`** | Unified demand-paged operator graph |
47
- | Spatiotemporal Decoding | Full-clip latent buffers in GPU VRAM | **Folded streaming video & audio `VAE`** | Bounded memory during latent decoding |
48
- | Temporal Generation Horizon | Short fixed clips with boundary resets | **Continuous uncut `M-Rail` continuation** | Flat `O(1)` carrier state across long takes |
49
- | Fleet Reuse | Full denoising recomputation per job | **Shared `M-Rail` (`X = X_known + R`)** | Associative trajectory & conditioning reuse |
 
50
 
51
- ![Disk-mapped execution and constant session state](assets/weights_vs_origami.png)
52
- <sub><b>Figure 1.</b> Execution substrate and live trajectory memory. Conventional audio-video diffusion pipelines load multi-organ checkpoints into GPU HBM and scale memory with clip duration, whereas Origami-H3 streams `.o3f` operators via OS `mmap` and holds live trajectory state strictly flat (`O(1)`) across continuous uncut horizons.</sub>
53
 
54
- ## Signature HCGM Architecture & Distinguishing Capabilities
55
 
56
- - **The File Is the Machine (Zero `ComfyUI`, Zero `PyTorch`)**: Every organ of the joint audio-video synthesis pipeline, including conditioning refiners, coupled flow-matching `DiT` blocks, few-step `Turbo-LoRA` adapters, `VSA` modulation gates, and both video and audio `VAE` decoders, executes directly from read-only OS `mmap` pages under the standalone `wirkung` runtime.
57
- - **Parseval Spectral Stabilization on Nonlinear Flow Blocks**: Naive low-bit compression collapses coupled audio-video flow trajectories inside nonlinear SwiGLU blocks. Origami-H3 enforces Parseval row-norm conservation paired with projective spectral crystals so flow velocities remain coherent from pure Gaussian noise to decoded media.
58
- - **Continuous Uncut Horizon in Flat `O(1)` State**: Rather than resetting latent state at short clip boundaries, `M-Rail` sideline observers and smooth boundary continuation maintain temporal motion and acoustic phase continuity inside a constant memory envelope.
59
 
60
- ![HCGM organ and operator architecture](assets/architecture_organs.png)
61
- <sub><b>Figure 2.</b> Three-stage full-width HCGM topology of Origami-H3: Parseval-stabilized conditioning and folded hypernetwork gates, coupled audio-video flow `DiT` with in-graph `Turbo-LoRA`, and streaming spatiotemporal `VAE` readout for continuous uncut takes.</sub>
62
 
63
- ## Shared M-Rail (`X = X_known + R`), sGPU Fleet Economics & HMoE
 
 
 
64
 
65
- ![Shared M-Rail, sGPU fleet, and HMoE topology](assets/mrail_network.png)
66
- <sub><b>Figure 3.</b> Shared `M-Rail` fleet topology for Origami-H3. Conditioning spans and crystallized trajectory anchors replay directly (`R = 0`), while novel creative residuals (`R != 0`) are solved once and shared across CPU and GPU nodes.</sub>
67
 
68
- Origami-H3 participates natively in the **Shared `M-Rail`** trajectory fabric and bundled **`sGPU`** (`sgpu`, `cpuda`, `rail`) deterministic execution stack:
69
 
70
- 1. **Warm Up Once, Render Indefinitely on Disk-Backed CPU Fleets**: Expressing synthesis trajectories as `X = X_known + R`, studios and enterprises can optionally warm up recurring visual styles, conditioning anchors, and scene templates on burst GPUs once, shut GPUs down, and render ongoing workloads on disk-backed CPU nodes.
71
- 2. **Collective Associative Residual Closure**: Newly computed residual branches (`R != 0`) commit verified `ANCHOR` / `DELTA` stations and `OP_COPY` morphisms to Shared `M-Rail` (`wkng maps merge`), allowing peer machines to resume directly from verified stations.
72
- 3. **Heavy Mixture of Experts (`HMoE`) Multimodal World Specialist**: Mounted on the synchronous `M-Rail` `cstar` bus alongside `Origami-9B`, `Origami-27B`, `Origami-31B`, and `Origami-78B`, Origami-H3 serves as the dedicated joint audio-video synthesis and world-rendering macro-expert orchestrated directly by reasoning HCGMs.
73
 
74
- ## Runtime CLI & Python SDK
75
 
76
- Install the standalone native runtime from PyPI (`Python >= 3.10`, `numpy` as the sole Python dependency; zero PyTorch, zero CUDA toolkit, zero `ComfyUI`, zero `diffusers`):
 
77
 
78
- ```bash
79
- # Check host execution capabilities and verify compiled AV operator shards
80
- wkng doctor
81
- sgpu doctor
82
- wkng verify ./Origami-H3
83
 
84
- # Merge shared M-Rail trajectory maps across studio or cloud nodes
85
- wkng maps merge ./peer_node_maps --into ~/.wkng/maps
 
 
 
 
 
86
  ```
87
 
88
  ```python
89
  from wirkung import Engine
90
 
91
- engine = Engine("./Origami-H3")
92
- result = engine.generate(
93
- "Mechanical watch escapement ticking in warm studio light.",
94
- output_path="./watch_take.mp4",
95
- )
96
- print(result)
97
  ```
98
 
99
- ## Repository Contents
100
-
101
- - `bundle.json`, `config.json`, `generation_config.json`: HCGM joint audio-video flow-matching bundle schema and architecture metadata.
102
- - `layers/`: Parseval-stabilized `.o3f` operators across all coupled `DiT` blocks, token refiners, conditioning projections, Walsh-folded `Turbo-LoRA` trajectory adapters, and folded `VSA` modulation gates.
103
- - `vae/`: compiled `.o3f` operators and normalization buffers for the streaming spatiotemporal video `VAE` and synchronized `Audio-VAE`.
104
- - `exact_core.safetensors`: exact modulation norms, timestep embeddings, and trajectory carrier anchors.
105
- - `LICENSE`, `NOTICE`: Wirkung Model and Runtime License Agreement 1.0 and upstream MiniMax-H3 attribution notice.
106
 
107
- ## Enterprise, Custom Models & Research Inquiries
 
 
 
 
 
 
108
 
109
- Wirkung partners with studios, enterprises, AI labs, and infrastructure providers on:
110
 
111
- - **Custom Model Compilations & Preview Access**: Closed-form HCGM folding of proprietary multimodal, video, audio, or language checkpoints into zero-GPU, constant-state operator artifacts.
112
- - **Exact Formula & Algorithm Synthesis (`Wirkung`)**: Symbolic and spectral operator discovery replacing dense compute bottlenecks with certified closed-form kernels.
113
- - **Fleet & Inference-Provider Partnerships**: Production deployments of Shared `M-Rail`, disk-backed `sGPU` CPU cloud fabrics, and multi-specialist `HMoE` routing.
114
- - **Research & Academic Collaboration**: Joint investigations into constant-state generative flow models, continuous-horizon synthesis, and reusable computation.
115
 
116
- Reach us directly:
117
- - **Email:** [`contact@wirku.ng`](mailto:contact@wirku.ng)
118
- - **X:** [`https://x.com/FossDT`](https://x.com/FossDT)
119
- - **Website:** [`https://wirku.ng`](https://wirku.ng)
120
 
121
  ## Papers
122
 
123
- - **[The Holographic Causal Graph Machine: Compiling a Pretrained Transformer into a Constant-State Model without Training](https://doi.org/10.13140/RG.2.2.28206.68167)**. David Tom Foss, 2026.
124
- - **[Compute Once: Constant-State Language Models Make Answers Reusable and Agents Cheap](https://doi.org/10.13140/RG.2.2.18140.35202)**. David Tom Foss, 2026.
125
- - **[Crystallized Attention: Fixed-State Carriers and Exact Index Correction for the Full-Attention Heads of a 27B Hybrid Language Model](https://doi.org/10.13140/RG.2.2.21299.18725)**. David Tom Foss, 2026.
126
- - **[Scarcity as Method. Attention, Replaced: Exact Forms, Constant State, and Reusable Inference](https://doi.org/10.13140/RG.2.2.34347.40489)**. David Tom Foss, 2026.
127
 
128
  ## Citation
129
 
130
  ```bibtex
131
- @misc{foss2026origamih3,
132
- author = {David Tom Foss},
133
- title = {{Origami-H3: Parseval-Stabilized HCGM Compilation and
134
- Continuous M-Rail Synthesis of Audio-Video Flow}},
135
- year = {2026},
136
- howpublished = {Wirkung Inc., \url{https://huggingface.co/wirkung/Origami-H3}},
137
- note = {DOI 10.13140/RG.2.2.28206.68167}
138
  }
139
  ```
140
 
141
  ## License
142
 
143
- Licensed under the **Wirkung Model and Runtime License Agreement 1.0** ([`LICENSE`](LICENSE)). Free for non-commercial research, academic evaluation, and personal use in compiled Object form; commercial or production deployment of Origami models, `wkng`, or `sgpu` requires a commercial license from Wirkung Inc. ([`contact@wirku.ng`](mailto:contact@wirku.ng)). Compiled in closed form without training from [`MiniMaxAI/MiniMax-H3`](https://huggingface.co/MiniMaxAI/MiniMax-H3), subject to upstream MiniMax-H3 license terms (see [`NOTICE`](NOTICE)).
 
4
  license_link: LICENSE
5
  library_name: wirkung
6
  pipeline_tag: text-to-video
7
+ num_parameters: 21600000000
 
 
8
  language:
9
  - en
10
+ - de
11
  tags:
12
  - wirkung
13
  - origami
14
  - preview
15
+ - training-free
16
+ - lossless-folding
17
+ - constant-memory
18
+ - zero-vram
19
+ - cpu-inference
20
+ - npu
21
  - hcgm
22
  - m-rail
23
+ - 22b
24
+ - video-generation
 
 
25
  - audio-video
26
+ - diffusion-transformer
 
27
  ---
28
 
29
+ ![MiniMax-H3 21.6B AV diffusion folded without retraining from 44.89 GB to 10.38 GB (4.33x), with a constant state across frames.](assets/banner.png)
30
 
31
+ [Paper](paper/Foss_2026_Origami_HCGM.pdf) · [Runtime (PyPI)](https://pypi.org/project/wirkung/) · [Wirkung](https://wirku.ng) · [Base weights](https://github.com/MiniMax-AI) · [Contact](mailto:contact@wirku.ng) · [License](LICENSE)
32
 
33
+ **`Origami-H3` is a 21.60B-parameter Holographic Causal Graph Machine (HCGM) Preview compiled in closed form from `MiniMax-H3` with zero retraining, zero fine-tuning, and zero gradient steps. It folds the 44.89 GB source checkpoint into 10.38 GB (4.33x compression), executes directly from NVMe/SSD storage via OS `mmap` with `0 GB` GPU VRAM and `0 GB` dedicated weight-RAM, and keeps live state strictly constant ($O(1)$) across arbitrary horizons on CPU, NPU, `sGPU`, Metal, and standalone PTX.**
34
+
35
+ `Origami-H3` preserves the full `MiniMax-H3` architecture (50 joint audio-video Diffusion Transformer (DiT) blocks, 2 token refiners, a fused rank-128 Walsh-6 Turbo-LoRA, a 24-block 3D `SoL-H3-VAE`, and an `Audio-VAE`) while turning a passive weight checkpoint into a self-contained binary file executed under one cross-silicon contract. It succeeds our first-generation constant-state releases published under [`huggingface.co/tfwnotops`](https://huggingface.co/tfwnotops) by removing `vLLM`, `MLX`, and `Transformers` runtime dependencies altogether. This repository is a **Preview** release: structural byte footprints, fold ratios, and constant-memory invariants below are fixed properties of the compiled artifact, while hardware-dependent throughput figures are omitted from the card and documented in the [Origami Family Paper](paper/Foss_2026_Origami_HCGM.pdf).
36
 
37
  ```bash
38
  pip install -U wirkung && wkng pull wirkung/Origami-H3 -o ./Origami-H3
39
+ wkng verify ./Origami-H3 --full && wkng info ./Origami-H3
40
  ```
41
 
42
+ ## Source checkpoint against Origami-H3
43
 
44
+ | Property | Source (`MiniMax-H3`) | `Origami-H3` (HCGM Preview) |
45
+ |:--|--:|--:|
46
+ | Parameters | 21.60B | **21.60B** (identical graph topology) |
47
+ | Retraining / fine-tuning steps | Pretrained baseline | **0** (closed-form compilation) |
48
+ | Checkpoint footprint on disk | 44.89 GB BF16 | **10.38 GB** (**4.33x** smaller; 307 `.o3f` DiT + 3D/Audio VAE operator files + `exact_core.safetensors`) |
49
+ | Required GPU VRAM | Multi-tens of GB HBM | **0 GB** (runs on plain CPU / NPU / `sGPU` or optional GPU) |
50
+ | Dedicated weight-RAM | Pinned in host / device RAM | **0 GB** (read-only OS `mmap` directly from SSD/NVMe) |
51
+ | Live session / horizon state | Unbounded multi-window 3D spatiotemporal latent buffers (exceeds 16 GB unified memory) | **Constant $O(1)$ horizon carrier across continuous uncut audio-video windows** |
52
+ | Operator & readout path | Dense per-step matrix scan | 4-step shifted Euler flow matching + `M-Rail` physical/geometric sideline verification |
53
+ | Runtime dependencies | PyTorch / `vLLM` / `MLX` / CUDA | **`wirkung` + `numpy`** (standalone C11, NPU, `sGPU`, Metal, PTX) |
54
 
55
+ ![Physical byte anatomy and sequence-horizon state scaling of Origami-H3](assets/architecture.png)
 
56
 
57
+ Figure 1. (a) Physical byte scale contrasting the 44.89 GB BF16 `MiniMax-H3` pipeline against the 10.38 GB `Origami-H3` `.o3f` bundle mapped directly from SSD (`0 GB` dedicated GPU VRAM, `0 GB` weight-RAM). (b) Live spatiotemporal state across consecutive latent audio-video windows: unbounded window caching against the constant $O(1)$ `Origami-H3` horizon state.
58
 
59
+ ## Zero-retraining compilation and constant-state architecture
 
 
60
 
61
+ Every Origami model is produced by the same four-part compilation pipeline without gradient descent:
 
62
 
63
+ 1. **The file is the machine (disk-resident `mmap` execution).** An HCGM model is an immutable binary artifact on disk. The operating system maps operator pages read-only on demand from SSD or NVMe storage, requiring neither GPU VRAM nor weight-deserialization buffers in host RAM. Under our `v3-exakt` dyadic integer reduction contract, CPU (`AVX2` / `NEON`), general NPUs, embedded `sGPU` (`cpu_cuda`), Apple Metal, and driver-API PTX evaluate identical canonical reductions rather than drifting floating-point kernel approximations.
64
+ 2. **Closed-form Parseval + projective-crystal folding (`.o3f`, 4.33x without retraining).** Compiled directly from the frozen 44.89 GB BF16 `MiniMax-H3` pipeline with zero gradient steps into 10.38 GB across 307 `.o3f` operator files, `Origami-H3` pairs Parseval row-norm conservation with rank-48 boundary crystals (blocks 0..2 and 47..49), rank-32 interior crystals (blocks 3..46), gauge-balanced QR-SVD rank-128 Walsh-6 `rANS` Turbo-LoRA, and a folded 97-operator 3D `SoL-H3-VAE`, preserving nonlinear SwiGLU velocity fields on 16 GB consumer hardware.
65
+ 3. **Constant $O(1)$ uncut horizon state synthesized via `Wirkung`.** Using `Wirkung`, our internal exact structural-law and closed-form operator synthesis engine, `Origami-H3` replaces hard scene-cut window resets with a bounded $O(1)$ spatiotemporal carrier across consecutive 4-step shifted Euler windows while parallel constant-state `M-Rail` observers check 3D geometric, physical, and audio-visual invariants in the same weight stream.
66
+ 4. **Certified readout and online trajectory decomposition ($X = X_{\mathrm{known}} + R$).** Rather than re-evaluating every parameter blindly for isolated single tokens, HCGM decomposes live hidden trajectories into known low-rank structure $X_{\mathrm{known}}$ (cosine trajectory banks, multi-order suffix tables, and rank-1 Sherman-Morrison-Woodbury updates) plus novel residual $R$, verifying multiple candidate steps in one weight pass with zero greedy output drift.
67
 
68
+ ![Constant-memory read geometry and folded operator execution in Origami-H3](assets/mechanism.png)
 
69
 
70
+ Figure 2. (a) Continuous multi-window trajectory coherence under uncut `Origami-H3` `M-Rail` flow against open-loop window-boundary drift. (b) Single-pass `.o3f` DiT operator combining Parseval `FWHT_128` 4-class bitplanes and Walsh-6 LoRA with a rank-$r$ projective crystal ($U_r V_r^\top$) feeding the folded 3D and Audio VAEs.
71
 
72
+ ## Shared trajectory assimilation (M-Rail), GPU-less server fleets and multi-model routing (HMoE)
 
 
73
 
74
+ Beyond single-node execution, `Origami-H3` attaches natively to **Shared `M-Rail`** (`.mrail` v3), a content-addressed associative trajectory manifold shared across processes or across an enterprise network:
75
 
76
+ - **Warm up once on one GPU, run permanently on disk-only CPU/NPU servers.** Because HCGM models require only SSD storage and megabytes of live state, an organization can run a single GPU (or a short temporary cloud GPU lease) until its recurring domain workflows are crystallized into a shared `.mrail` file, then shut the GPU down and serve frontier models across commodity CPU/NPU VMs and CPU-only servers. Known trajectories execute via `Spur` copy morphisms or resume in $O(1)$ space from `Station` (`ANCHOR` + `DELTA`) snapshots; whenever any server encounters an unseen task ($R \neq 0$), computing and committing that residual once makes the solution immediately available to every other node on the rail.
77
+ - **Heavy Mixture of Experts (`HMoE`).** Because idle disk-mapped models consume zero weight-RAM, `M-Rail` routes across complete compiled Origami models (`Origami-9B`, `Origami-27B`, `Origami-31B`, `Origami-78B`, `Origami-H3`), zero-recompile static-graph NPU specialist pools, and sub-2 KB exact algebraic group organs on a synchronous station bus (`cstar`) under deterministic check gates.
78
 
79
+ ![Shared M-Rail GPU-less server fleet topology and multi-session memory scaling for Origami-H3](assets/mrail.png)
80
+
81
+ Figure 3. (a) Shared `M-Rail` (`.mrail` v3) fleet architecture: optional temporary GPU warmup feeds a permanent zero-GPU CPU/NPU/`sGPU` server fleet where any novel residual $+R$ solved once commits back to all nodes alongside `HMoE` routing on `cstar`. (b) Total resident RAM/VRAM across $N = 1..32$ concurrent 32k-token sessions: pinned weights plus $N \cdot \mathrm{KV}$ against shared read-only `mmap` plus $N \cdot O(1)$ constant state.
82
+
83
+ ## How to run
84
 
85
+ Requires Python 3.10+ and the `wirkung` runtime package (`pip install -U wirkung`), which installs the `wkng`, `wirkung`, `sgpu`, and `rail` CLI entrypoints.
86
+
87
+ ```bash
88
+ pip install -U wirkung
89
+ wkng pull wirkung/Origami-H3 -o ./Origami-H3
90
+ wkng info ./Origami-H3
91
+ wkng verify ./Origami-H3 --full
92
  ```
93
 
94
  ```python
95
  from wirkung import Engine
96
 
97
+ eng = Engine("./Origami-H3")
98
+ print(eng.info())
 
 
 
 
99
  ```
100
 
101
+ ## Files
 
 
 
 
 
 
102
 
103
+ - `layers/*.o3f`, `vae/*.o3f`: 307 folded HCGM DiT, Walsh-6 LoRA, 3D `SoL-H3-VAE`, and `Audio-VAE` operator files (10.38 GB total bundle)
104
+ - `exact_core.safetensors`: exact modulation, norm, and conditioning scalars
105
+ - `bundle.json`: shard inventory and SHA-256 digests verified by `wkng verify`
106
+ - `config.json`, `generation_config.json`: pipeline and 4-step Euler schedule configuration
107
+ - `paper/Foss_2026_Origami_HCGM.pdf`: the Origami Model Family technical report
108
+ - `assets/`: `banner.png`, `architecture.png`, `mechanism.png`, `mrail.png`
109
+ - `LICENSE`, `NOTICE`: Wirkung Model and Runtime License 1.0 and upstream attribution
110
 
111
+ ## Inquiries and collaboration
112
 
113
+ We welcome inquiries regarding custom zero-retraining HCGM model compilation, closed-form formula and exact algorithm synthesis via `Wirkung`, Shared `M-Rail` / zero-GPU enterprise fleet deployments, and research collaboration:
 
 
 
114
 
115
+ - **Email:** [contact@wirku.ng](mailto:contact@wirku.ng)
116
+ - **X (Twitter):** [https://x.com/FossDT](https://x.com/FossDT)
117
+ - **Web:** [https://wirku.ng](https://wirku.ng)
 
118
 
119
  ## Papers
120
 
121
+ - **[The Origami Model Family: Compiling Pretrained Foundation Models into Holographic Causal Graph Machines](paper/Foss_2026_Origami_HCGM.pdf)**. David Tom Foss, 2026. Included directly in this repository at `paper/Foss_2026_Origami_HCGM.pdf`.
122
+ - **[Holographic Causal Graph Machines: Source-Free Constant-State Execution of Pretrained Transformers](https://doi.org/10.5281/zenodo.19040842)**. David Tom Foss, 2026.
123
+ - **[Crystallized Attention: Constant-Memory Exact Store and Exponential-Mode Carriers](https://doi.org/10.5281/zenodo.19058017)**. David Tom Foss, 2026.
124
+ - **[Intelligence Is Not Made of Scarcity: Constant-Memory Causal Execution and Associative M-Rail](https://doi.org/10.13140/RG.2.2.28823.43689)**. David Tom Foss, 2026.
125
 
126
  ## Citation
127
 
128
  ```bibtex
129
+ @techreport{foss2026origami,
130
+ title = {The Origami Model Family: Compiling Pretrained Foundation Models into Holographic Causal Graph Machines},
131
+ author = {Foss, David Tom},
132
+ institution = {Wirkung Inc.},
133
+ year = {2026},
134
+ month = nov,
135
+ url = {https://wirku.ng}
136
  }
137
  ```
138
 
139
  ## License
140
 
141
+ Released under the [Wirkung Model and Runtime License 1.0](LICENSE). Compiled in closed form without training steps from Upstream `MiniMax-H3` terms and notices (`NOTICE`); full attribution and closed-form transformation notices are recorded in [`NOTICE`](NOTICE).
assets/{weights_vs_origami.png → architecture.png} RENAMED
File without changes
assets/architecture_organs.pdf DELETED
Binary file (33.5 kB)
 
assets/banner.png CHANGED

Git LFS Details

  • SHA256: 4bcb2c3cdfe4e100cc92b17a2a23a77bf57227fce998011c9ed4007dcf225282
  • Pointer size: 131 Bytes
  • Size of remote file: 210 kB

Git LFS Details

  • SHA256: 1ff1f05649f393064c816a3c4463b408d0b5804942f066b031d27f463bbd019e
  • Pointer size: 131 Bytes
  • Size of remote file: 212 kB
assets/{architecture_organs.png → mechanism.png} RENAMED
File without changes
assets/{mrail_network.png → mrail.png} RENAMED
File without changes
assets/mrail_network.pdf DELETED
Binary file (29.6 kB)
 
assets/weights_vs_origami.pdf DELETED
Binary file (46 kB)
 
paper/Foss_2026_Origami_HCGM.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:72360f42140569826578d8c872b993a5be6ec1b43015d2536b82ec7e33ff9868
3
+ size 1716287