Spaces:
Sleeping
Sleeping
Update README.md
Browse files
README.md
CHANGED
|
@@ -9,6 +9,100 @@ app_file: app.py
|
|
| 9 |
pinned: false
|
| 10 |
license: other
|
| 11 |
short_description: Live coherence + throughput benchmark (no precomputed result
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
---
|
| 13 |
|
| 14 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
pinned: false
|
| 10 |
license: other
|
| 11 |
short_description: Live coherence + throughput benchmark (no precomputed result
|
| 12 |
+
thumbnail: >-
|
| 13 |
+
https://cdn-uploads.huggingface.co/production/uploads/685edcb04796127b024b4805/Sr9bUuYsRuDtVZnb-7kke.png
|
| 14 |
+
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
# Coherent Compute Engine
|
| 18 |
+
|
| 19 |
+
This Space demonstrates a **live, coherent compute engine** running entirely on CPU.
|
| 20 |
+
|
| 21 |
+
No precomputed benchmarks.
|
| 22 |
+
No synthetic scaling.
|
| 23 |
+
No hidden accelerators.
|
| 24 |
+
|
| 25 |
+
Every result shown is produced **at runtime** on the machine hosting this Space.
|
| 26 |
+
|
| 27 |
+
---
|
| 28 |
+
|
| 29 |
+
## What Is Being Measured
|
| 30 |
+
|
| 31 |
+
Each “item” is defined as:
|
| 32 |
+
|
| 33 |
+
> One coherent update of a state vector `[Ψ, E, L]` for a single oscillator.
|
| 34 |
+
|
| 35 |
+
The engine evolves millions of oscillators using:
|
| 36 |
+
- Numerically stable
|
| 37 |
+
- Branchless
|
| 38 |
+
- Energy-retentive updates
|
| 39 |
+
|
| 40 |
+
Metrics reported:
|
| 41 |
+
- **Throughput** — coherent updates per second
|
| 42 |
+
- **Coherence** — stability between initial and final state
|
| 43 |
+
- **Mean Energy** — energy retention across the system
|
| 44 |
+
- **Elapsed Time** — real wall-clock time
|
| 45 |
+
|
| 46 |
+
---
|
| 47 |
+
|
| 48 |
+
## Why This Matters
|
| 49 |
+
|
| 50 |
+
Most benchmarks optimize for raw FLOPs.
|
| 51 |
+
This engine optimizes for **coherent progression under constraint**.
|
| 52 |
+
|
| 53 |
+
It shows that:
|
| 54 |
+
- High throughput does not require GPUs
|
| 55 |
+
- Stability can be preserved at scale
|
| 56 |
+
- Energy retention can remain high even under load
|
| 57 |
+
|
| 58 |
+
This is **architecture**, not a paper claim.
|
| 59 |
+
|
| 60 |
+
---
|
| 61 |
+
|
| 62 |
+
## What This Is Not
|
| 63 |
+
|
| 64 |
+
- Not a neural network
|
| 65 |
+
- Not a pre-trained model
|
| 66 |
+
- Not a simulation with fixed outputs
|
| 67 |
+
- Not a GPU demo
|
| 68 |
+
|
| 69 |
+
It is a **compute engine**, measured live.
|
| 70 |
+
|
| 71 |
---
|
| 72 |
|
| 73 |
+
## Reproducibility
|
| 74 |
+
|
| 75 |
+
You can:
|
| 76 |
+
- Change oscillator count
|
| 77 |
+
- Change step count
|
| 78 |
+
- Rerun at any time
|
| 79 |
+
|
| 80 |
+
Results will change depending on:
|
| 81 |
+
- Host CPU
|
| 82 |
+
- Available cores
|
| 83 |
+
- Runtime conditions
|
| 84 |
+
|
| 85 |
+
That is the point.
|
| 86 |
+
|
| 87 |
+
---
|
| 88 |
+
|
| 89 |
+
## Ethics & Transparency
|
| 90 |
+
|
| 91 |
+
No data is uploaded.
|
| 92 |
+
No background processes.
|
| 93 |
+
No hidden scaling.
|
| 94 |
+
|
| 95 |
+
What you see is what runs.
|
| 96 |
+
|
| 97 |
+
---
|
| 98 |
+
|
| 99 |
+
## Author
|
| 100 |
+
|
| 101 |
+
Built as part of the **Rendered Frame Theory (RFT)** engineering work.
|
| 102 |
+
|
| 103 |
+
Focus:
|
| 104 |
+
- Truthful measurement
|
| 105 |
+
- Energy-aware compute
|
| 106 |
+
- Coherent system design
|
| 107 |
+
|
| 108 |
+
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|