themex1380/Gemma-2-9B-Chinese-Chat-Uncensored

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:9BQuant:FP8Ctx Length:16kPublished:Dec 12, 2024License:mitArchitecture:Transformer0.0K Open Weights Warm

The themex1380/Gemma-2-9B-Chinese-Chat-Uncensored model is a 9 billion parameter, uncensored, Chinese-chat fine-tune of the Gemma-2-9B-Chinese-Chat base model. It was fine-tuned using the unsloth framework on the Jenna-50K-Alpaca-Uncensored dataset, making it suitable for conversational AI applications requiring less restrictive content generation in Chinese. With a context length of 16384 tokens, it is designed for extended Chinese dialogue.

Loading preview...

Model Overview

The themex1380/Gemma-2-9B-Chinese-Chat-Uncensored is a 9 billion parameter language model specifically fine-tuned for uncensored Chinese chat applications. It builds upon the shenzhi-wang/Gemma-2-9B-Chinese-Chat base model and leverages the efficient unsloth framework for its fine-tuning process.

Key Capabilities

  • Uncensored Chinese Chat: The model's primary differentiator is its fine-tuning on the V3N0M/Jenna-50K-Alpaca-Uncensored dataset, enabling it to generate responses without typical content restrictions in Chinese conversations.
  • Gemma-2 Architecture: Based on the Gemma-2 family, it benefits from a robust foundation for language understanding and generation.
  • Extended Context Window: Supports a context length of 16384 tokens, allowing for more coherent and longer-form dialogues.

Training Details

The fine-tuning was performed on a single A100 SXM 80G GPU, utilizing a 16vCPU and 251 GB RAM setup on Runpod.io. The training script and logs are available for review, providing transparency into the fine-tuning methodology.

Good For

  • Developers requiring a Chinese-language conversational AI that can handle a broader range of topics without built-in censorship.
  • Applications where creative or unrestricted dialogue generation in Chinese is a priority.

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