PrototypeCommon
Scottcjn/ram-coffers
LLM infrastructure cost reduction via NUMA-aware weight banking: 147 t/s (8.8x stock llama.cpp) on refurbished enterprise POWER8. Self-hosted inference, no cloud APIs. Part of the Proof of Physical AI stack.
- Owner
- Scottcjn (personal account)
- Stars
- Stars: 173
- Forks
- Forks: 59
- Language
- Python
- License
- AGPL-3.0
- Created
- 2026
- Topics
- llama-cppllmnumapower8ai-inferencedepinhebbianpowerpcproof-of-physical-aivec-perm
- Tier
- Prototype (100 to 249 ★) What this tier means
More repos by Scottcjn
All 12 repos by Scottcjn- Scottcjn/RustchainStars: 680PythonAlpha
- Scottcjn/bottubeStars: 365PythonAlpha
- Scottcjn/rustchain-bountiesStars: 289PythonAlpha
- Scottcjn/beacon-skillStars: 237PythonPrototype
- Scottcjn/grazer-skillStars: 197PythonPrototype
- Scottcjn/legend-of-elya-n64Stars: 142CPrototype
Nearby in Python
All Python repos- pretend1111/inko-skillStars: 173PythonPrototype
- SupraGSX/Forceware-382.69Stars: 173PythonPrototype
- Open-RUA/uni-robot-agentStars: 173PythonPrototype
- zohmg/zohmgStars: 173PythonPrototype
- tstone/django-uploadifyStars: 173PythonPrototype
- astraw/svg_stackStars: 173PythonPrototype