Fredithefish/Guanaco-13B-Uncensored

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:13BQuant:FP8Ctx Length:4kPublished:Sep 7, 2023License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

Guanaco-13B-Uncensored is a 13 billion parameter causal language model developed by Fredithefish, fine-tuned on the Llama-2-13B base architecture. It was trained for 4 epochs on the Unfiltered Guanaco Dataset, specifically designed to provide responses without content filtering or censorship. This model excels at generating uncensored English text and does not perform well with other languages.

Loading preview...

Guanaco-13B-Uncensored Overview

Guanaco-13B-Uncensored is a 13 billion parameter language model developed by Fredithefish, built upon the Llama-2-13B base model. It has been fine-tuned for 4 epochs using the V2 of the Unfiltered Guanaco Dataset. The primary characteristic of this model is its design to generate responses without any content filtering or censorship, providing direct answers without denials.

Key Capabilities

  • Uncensored Content Generation: Specifically trained to produce responses without content restrictions.
  • English Language Focus: Optimized for English text generation; performance in other languages is limited.
  • Llama-2 Base: Leverages the robust architecture of Llama-2-13B.

Good For

  • Use cases requiring unfiltered or uncensored text output.
  • Applications where direct, unrestricted responses are preferred.
  • English-centric text generation tasks where content filtering is undesirable.

Limitations

  • Does not perform well with languages other than English.

Special thanks are extended to AutoMeta for providing the necessary computing power for training, and to TheBloke for creating the GGUF and GPTQ quantizations for this model.

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