Overview of Moonlit-Umbra-12B

Moonlit-Umbra-12B is a 12 billion parameter language model developed by Vortex5. It was created using a custom merge method, specifically the 'saef' merge technique, combining eleven different 12B models. Notable base models include Muse-12B, magnum-v4-12b, NeonMaid-12B-v2, and Violet_Twilight-v0.2, among others. The merge configuration specifies parameters such as paradox: 0.45, strength: 1.0, boost: 0.5, and modes: 2, with bfloat16 dtype. The tokenizer used is sourced from Vortex5/Scarlet-Seraph-12B.

Key Capabilities

• Model Merging: Utilizes a custom 'saef' merge method to combine multiple 12B models, aiming to synthesize their strengths.
• Specialized Training: Built upon a diverse set of base models, suggesting a broad range of underlying knowledge and generative styles.

Intended Use Cases

Moonlit-Umbra-12B is specifically designed and optimized for:

• Creative Writing: Generating imaginative and coherent text for various creative projects.
• Storytelling: Crafting narratives, developing plots, and creating engaging story content.
• Roleplay: Facilitating interactive and dynamic roleplaying scenarios with nuanced character interactions.