nbeerbower/Mistral-Nemo-Prism-12B-v7
nbeerbower/Mistral-Nemo-Prism-12B-v7 is a 12 billion parameter experimental language model developed by nbeerbower, based on the Mahou-1.5-mistral-nemo-12B-lorablated architecture. It was fine-tuned using ORPO on the Arkhaios-DPO and Purpura-DPO datasets with a 32768 token context length. The primary goal of this model is to reduce archaic language and purple prose while remaining uncensored, though the developer notes that LoRA fine-tuning did not fully resolve these language issues.
Loading preview...
Mistral-Nemo-Prism-12B-v7 Overview
This model, developed by nbeerbower, is an experimental 12 billion parameter language model built upon the Mahou-1.5-mistral-nemo-12B-lorablated base. It was fine-tuned using the ORPO method over 10 epochs on the Arkhaios-DPO and Purpura-DPO datasets, with an increased beta value of 2 for this specific version. The model supports a context length of 32768 tokens.
Key Capabilities
- Uncensored Output: Designed to be completely uncensored.
- Language Style Experimentation: Specifically aimed at reducing archaic language and "purple prose" in its generations.
Good For
- Experimental Use: Ideal for researchers and developers interested in exploring the effects of ORPO fine-tuning on language style and censorship.
- Understanding Fine-tuning Limitations: Useful for observing the challenges in completely altering deeply embedded language characteristics through LoRA fine-tuning.
Note: This is an experimental iteration and not intended for production-ready applications, as the developer indicates that some language issues persist despite the fine-tuning efforts.
Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.