nbeerbower/Mistral-Nemo-Gutenberg-Doppel-12B-v2

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:12BQuant:FP8Ctx Length:32kPublished:Oct 4, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

nbeerbower/Mistral-Nemo-Gutenberg-Doppel-12B-v2 is a 12 billion parameter language model based on the Mistral-Nemo architecture. It was fine-tuned by nbeerbower using the ORPO method on the Gutenberg-DPO and gutenberg2-dpo datasets. This model is specialized for tasks benefiting from its unique training on literary and DPO-aligned text, making it suitable for nuanced text generation and understanding.

Loading preview...

Model Overview

nbeerbower/Mistral-Nemo-Gutenberg-Doppel-12B-v2 is a 12 billion parameter language model derived from the axolotl-ai-co/romulus-mistral-nemo-12b-simpo base model. This iteration has been specifically fine-tuned by nbeerbower.

Training Details

The model underwent ORPO (Odds Ratio Preference Optimization) tuning for 3 epochs, utilizing two A100 GPUs. The training leveraged two distinct datasets:

  • jondurbin/gutenberg-dpo-v0.1
  • nbeerbower/gutenberg2-dpo

This specialized training approach, combining the Mistral-Nemo architecture with DPO-aligned literary datasets, suggests a focus on generating high-quality, preference-aligned text, potentially with a literary or narrative style.

Potential Use Cases

  • Creative Writing: Generating stories, poems, or other literary content.
  • Text Refinement: Improving the style and coherence of existing text.
  • Preference-Aligned Generation: Tasks where output quality is judged by human preference, benefiting from DPO training.

Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
top_p
top_k
frequency_penalty
presence_penalty
repetition_penalty
min_p