6 Downloads Updated yesterday
ollama run ram1234598766/Cesium2
Updated yesterday
yesterday
e3afad084c3f · 3.1GB ·
Modular Orchestrated Reasoning with Pattern-adaptive Hot-swappable skills
Fine-tuned Qwen2.5-1.5B-Instruct with 14 novel subsystems — dynamic MoE expert expansion, Mixture-of-Depths layer skipping, multi-head chain-of-thought reasoning, and a plugin architecture. Trained free on a Kaggle P100 in ~30 minutes.
ollama run ram1234598766/Cesium2
| Parameter | Value |
|---|---|
| Temperature | 0.7 |
| Top P | 0.9 |
| Top K | 50 |
| Context | 8192 |
Apache-2.0