Model Overview

cgato/Nemo-12b-Humanize-SFT-v0.2.5-KTO is a 12 billion parameter language model that has undergone a two-stage optimization process. Initially, it was fine-tuned (SFT) on a refined dataset, followed by further training using Kahneman-Tversky Optimization (KTO) on a separate dataset. This iterative approach aims to enhance the model's conversational quality and coherence.

Key Features

• 12 Billion Parameters: A moderately sized model offering a balance between performance and computational efficiency.
• 32768 Token Context Length: Supports extensive conversations and detailed prompts, crucial for complex interactions.
• KTO Optimization: Utilizes Kahneman-Tversky Optimization for improved response generation, building upon its supervised fine-tuning.
• ChatML Format: Employs the ChatML format, including system prompts, for structured and effective dialogue management.

Recommended Use Cases

• Roleplay Scenarios: Specifically optimized for roleplay, leveraging the ChatML format and system prompts for character definition and interaction.
• Conversational AI: Suitable for general conversational agents requiring coherent and context-aware responses.
• Extended Dialogue Generation: Benefits from its large context window, enabling long-form text generation and maintaining context over many turns.