amewebstudio/mnemosyne-multimodal-fused

🧠 Mnemosyne-SCLM v3.0

Multimodal Stateful Language Model with Persistent Memory

Architecture

• LLM: Llama 3.2 3B Instruct
• Audio: Whisper Large v3
• Memory: Multi-Scale (ST=512D, LT=256D)
• EARCP: v3 with Online Learning
• Context: 8K tokens

Author

Mike Amega (Logo) - Ame Web Studio