Modelpublic

dhanesh-hf/Jarvis-Titan-V14-MoE-Merged

sourceHugging Faceotherupdated 12d agoView on Hugging Face

0likes334downloads

Model Card

🌐 J.A.R.V.I.S. Titan V14 MoE (Merged 14.8B Model)

Principal Investigator: Dhanesh Architecture: 14.8B DeepSeekMoE (8 Routed Experts + 1 Shared Expert) Format: Standalone Unified Model Safetensors

⚡ Overview

Jarvis-Titan-V14-MoE-Merged is the fully fused, production checkpoint integrating the base attention backbone with the V14 adapted MoE replacement layers (full_moe_replacement.safetensors).

—Total Parameters: 14.8B
—Active Parameters: ~3.2B
—Layers: 28
—Attention: Grouped-Query Attention (GQA, 28:4)

🛡️ License

Governed by the J.A.R.V.I.S. Titan Proprietary Research License (JTRL-v1.0). Academic non-commercial evaluation only. See `LICENSE`.