NeverSleep/Noromaid-7b-v0.2
Noromaid-7b-v0.2 is a 7 billion parameter full finetune model developed collaboratively by IkariDev and Undi, based on the Mistral 0.1 7B architecture with a 4096 token context length. It is specifically optimized for roleplay (RP), erotic roleplay (ERP), and general conversational tasks. The model incorporates unique private datasets, including 'no_robots' for human-like behavior and 'Aesir Private RP' for fresh, non-repetitive roleplay data, distinguishing it from other models in its class.
Loading preview...
Noromaid-7b-v0.2: A Specialized 7B Finetune for Roleplay
Noromaid-7b-v0.2 is a collaborative 7 billion parameter full finetune model by IkariDev and Undi, built upon the Mistral 0.1 7B base. This model is explicitly designed as a test version, focusing on enhancing roleplay (RP), erotic roleplay (ERP), and general conversational capabilities.
Key Capabilities & Features
- Roleplay Optimization: Specifically finetuned for engaging in both general and erotic roleplay scenarios.
- Unique Training Data: Utilizes novel datasets, including the 'no_robots' dataset to foster more human-like behavior and the 'Aesir Private RP' dataset, which introduces fresh, never-before-used data to avoid common RP data repetition.
- Full Finetune: Underwent a full finetuning process for 2 epochs (4000 steps) on its base architecture.
- Flexible Prompting: Supports both a custom prompting format (with provided SillyTavern config files) and the standard Alpaca instruction format.
Good For
- Roleplay Applications: Ideal for developers and users seeking a model specialized in generating dynamic and nuanced roleplay interactions.
- Creative Writing: Its unique dataset and finetuning make it suitable for creative text generation, particularly in character-driven narratives.
- Exploratory Development: As a test version, it's valuable for those looking to experiment with specialized finetunes and provide feedback to the developers.