Raven 8B v1: Uncensored Llama-3.1-Nemotron-8B Finetune

Raven 8B v1 is an 8 billion parameter language model developed by DarkArtsForge, based on the Llama-3.1-Nemotron-8B architecture. This model is notable for being a fully uncensored finetune, specifically trained on a small dataset derived from the Edgar Allan Poe corpus (DarkArtsForge/Poe_v1).

Key Characteristics

• Base Model: Llama-3.1-Nemotron-8B
• Parameter Count: 8 billion
• Training: Finetuned for 5 epochs using PMPF on a specialized Poe dataset.
• Uncensored Nature: Designed to produce narratives and roleplay content without typical content filters.

Intended Use Cases

• Creative Writing: Excels in generating dark, gothic, or thematically intense narratives.
• Roleplay: Suitable for roleplaying scenarios that may involve violent or graphic erotic content.

Important Considerations

• Content Warning: Users are advised that this model can generate explicit and graphic content. System prompts should be adjusted carefully to manage output.
• Chat Template: It is recommended to use the Llama 3 chat template for optimal performance.