Spaces:
Configuration error
Configuration error
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,10 +1,23 @@
|
|
| 1 |
-
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# AIOS — CPU-Native LLM Inference Architecture
|
| 2 |
+
|
| 3 |
+
A memory residency controller and Model Contract for deploying
|
| 4 |
+
large language models efficiently on CPU hardware.
|
| 5 |
+
|
| 6 |
+
## What We're Building
|
| 7 |
+
AIOS addresses the memory bandwidth bottleneck in CPU inference
|
| 8 |
+
through weight aliasing, sparsity maps, KV cache tiering, and
|
| 9 |
+
activation chunking — targeting 7B+ models on hardware
|
| 10 |
+
organizations already own.
|
| 11 |
+
|
| 12 |
+
## Current State
|
| 13 |
+
Framework and specification published. Runtime implementation
|
| 14 |
+
is the primary contribution opportunity.
|
| 15 |
+
|
| 16 |
+
## Links
|
| 17 |
+
- Paper: https://ssrn.com/abstract=6467298
|
| 18 |
+
- GitHub: https://github.com/acasavaraju/AIOS
|
| 19 |
+
- Primary validation target: GitHub Issue #2
|
| 20 |
+
|
| 21 |
+
## How to Contribute
|
| 22 |
+
Clone the repo, run validation/compliance.py on any GGUF model,
|
| 23 |
+
post results to the relevant GitHub issue.
|