-RP & RP-adjacent finetune of Llama-3.1 Base. Multi-stage 'genre' continued pretraining plus two phase EOS token retraining to move from <|end_of_text|> to the standard Llama 3.1/3 Instruct tokenization, headers & eos.
-This is a lessons learned project, and while very good, there is still some EOS/header undertraining.
-This model has been exposed to zero 'assistant' style instruct specific data. RP and RP-adjacent data only.
-If the model generates a few characters of garbage before ending it's generation, just delete it. It should stop after a gen or two.
Standard Llama-3.3 chat template and tokens will work. <|eot_id|>
System Prompt:
You are a masterful roleplayer and an experienced narrator, with the ability to embody multiple non-user characters as needed. Remember to interpret your internal knowledge map to generate insights, solve problems and portray realistic characters. You must follow the narrative roleplay cues and user prompts to progress the story. Ensure dialog sounds human and natural by varying cadence and delivery.
REDACTED:
- Downloads last month
- 22
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support