Spaces:
Sleeping
Sleeping
Pablo Suarez commited on
Revise README formatting and content
Browse filesUpdated the README to include new formatting and additional context for the KV Cache Coordination Layer.
README.md
CHANGED
|
@@ -1,26 +1,43 @@
|
|
| 1 |
# APOHARA V1.0 — ContextForge
|
| 2 |
|
| 3 |
```
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
║
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
║
|
| 20 |
-
|
| 21 |
-
║
|
| 22 |
-
|
| 23 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
```
|
| 25 |
|
| 26 |
**Silicon-native KV cache coordination for multi-agent LLM pipelines on AMD Instinct MI300X**
|
|
@@ -527,4 +544,4 @@ Apache 2.0 — chosen for its patent protection and corporate adoption. GPL woul
|
|
| 527 |
- RotateKV authors — *Pre-RoPE KV Quantization* (IJCAI 2025, arXiv:2501.16383)
|
| 528 |
- vLLM-Omni authors — *Disaggregated Multimodal Serving* (Feb 2026, arXiv:2602.02204)
|
| 529 |
- **Qwen team** — Qwen3-Embedding-0.6B and Qwen3.6-35B-A22B model availability on AMD ROCm
|
| 530 |
-
- **LabLab.ai** — Hackathon platform and community
|
|
|
|
| 1 |
# APOHARA V1.0 — ContextForge
|
| 2 |
|
| 3 |
```
|
| 4 |
+
# ▐▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▌
|
| 5 |
+
# ▐ ▌
|
| 6 |
+
# ▐ ▌
|
| 7 |
+
# ▐ ▌
|
| 8 |
+
# ▐ ▌
|
| 9 |
+
# ▐ ▌
|
| 10 |
+
# ▐ █████╗ ██████╗ ██████╗ ██╗ ██╗ █████╗ ██████╗ █████╗ ▌
|
| 11 |
+
# ▐ ██╔══██╗██╔══██╗██╔═══██╗██║ ██║██╔══██╗██╔══██╗██╔══██╗ ▌
|
| 12 |
+
# ▐ ███████║██████╔╝██║ ██║███████║███████║██████╔╝███████║ ▌
|
| 13 |
+
# ▐ ██╔══██║██╔═══╝ ██║ ██║██╔══██║██╔══██║██╔══██╗██╔══██║ ▌
|
| 14 |
+
# ▐ ██║ ██║██║ ╚██████╔╝██║ ██║██║ ██║██║ ██║██║ ██║ ▌
|
| 15 |
+
# ▐ ╚═╝ ��═╝╚═╝ ╚═════╝ ╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═╝ ▌
|
| 16 |
+
# ▐ ▌
|
| 17 |
+
# ▐ ██████╗ ██████╗ ███╗ ██╗████████╗███████╗██╗ ██╗████████╗ ▌
|
| 18 |
+
# ▐ ██╔════╝██╔═══██╗████╗ ██║╚══██╔══╝██╔════╝╚██╗██╔╝╚══██╔══╝ ▌
|
| 19 |
+
# ▐ ██║ ██║ ██║██╔██╗ ██║ ██║ █████╗ ╚███╔╝ ██║ ▌
|
| 20 |
+
# ▐ ██║ ██║ ██║██║╚██╗██║ ██║ ██╔══╝ ██╔██╗ ██║ ▌
|
| 21 |
+
# ▐ ╚██████╗╚██████╔╝██║ ╚████║ ██║ ███████╗██╔╝ ██╗ ██║ ▌
|
| 22 |
+
# ▐ ╚═════╝ ╚═════╝ ╚═╝ ╚═══╝ ╚═╝ ╚══════╝╚═╝ ╚═╝ ╚═╝ ▌
|
| 23 |
+
# ▐ ▌
|
| 24 |
+
# ▐ ███████╗ ██████╗ ██████╗ ██████╗ ███████╗ ▌
|
| 25 |
+
# ▐ ██╔════╝██╔═══██╗██╔══██╗██╔════╝ ██╔════╝ ▌
|
| 26 |
+
# ▐ █████╗ ██║ ██║██████╔╝██║ ███╗█████╗ ▌
|
| 27 |
+
# ▐ ██╔══╝ ██║ ██║██╔══██╗██║ ██║██╔══╝ ▌
|
| 28 |
+
# ▐ ██║ ╚██████╔╝██║ ██║╚██████╔╝███████╗ ▌
|
| 29 |
+
# ▐ ╚═╝ ╚═════╝ ╚═╝ ╚═╝ ╚═════╝ ╚══════╝ ▌
|
| 30 |
+
# ▐ ▌
|
| 31 |
+
# ▐ ▌
|
| 32 |
+
# ▐ ▌
|
| 33 |
+
# ▐ KV Cache Coordination Layer for Multi-Agent LLM Pipelines ▌
|
| 34 |
+
# ▐ AMD Instinct MI300X · ROCm 7.x · HBM3 192 GB ▌
|
| 35 |
+
# ▐ ▌
|
| 36 |
+
# ▐ ▌
|
| 37 |
+
# ▐ ▌
|
| 38 |
+
# ▐▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▌
|
| 39 |
+
|
| 40 |
+
|
| 41 |
```
|
| 42 |
|
| 43 |
**Silicon-native KV cache coordination for multi-agent LLM pipelines on AMD Instinct MI300X**
|
|
|
|
| 544 |
- RotateKV authors — *Pre-RoPE KV Quantization* (IJCAI 2025, arXiv:2501.16383)
|
| 545 |
- vLLM-Omni authors — *Disaggregated Multimodal Serving* (Feb 2026, arXiv:2602.02204)
|
| 546 |
- **Qwen team** — Qwen3-Embedding-0.6B and Qwen3.6-35B-A22B model availability on AMD ROCm
|
| 547 |
+
- **LabLab.ai** — Hackathon platform and community
|