crestf411/L3.1-8B-Slush-v1.1

L3.1-8B-Slush-v1.1 by crestf411 is an 8 billion parameter Llama 3.1-based model with a 32768 token context length, specifically engineered to enhance creativity, writing, and roleplaying capabilities. It utilizes a two-stage training process involving LoRA dropout and a custom merge, aiming to improve imagination while retaining the base model's intelligence. This model is optimized for creative text generation and interactive roleplay scenarios.