SkGufranAhmed/Huihui-gemma-3-270m-it-abliterated

Model Overview

This model, Huihui-gemma-3-270m-it-abliterated, is a GGUF quantized version of an uncensored, instruction-tuned variant of Google's gemma-3-270m-it model. The base model was fine-tuned by huihui.ai using the TRL framework to intentionally remove safety filtering, resulting in a model capable of generating sensitive or controversial content.

Key Characteristics

• Base Model: google/gemma-3-270m-it (270 million parameters).
• Abliterated/Uncensored: Fine-tuned to significantly reduce safety filtering, allowing for broader content generation.
• Quantization: Provided in GGUF format with F16 (16-bit float) and Q8 (8-bit) quantizations for efficient inference.
• Training: Fine-tuned using Supervised Fine-Tuning (SFT) with the TRL library on an English-only dataset.

Intended Use Cases

This model is primarily recommended for:

• Research and Experimentation: Exploring the capabilities and limitations of uncensored language models.
• Controlled Environments: Use in settings where outputs can be rigorously monitored and reviewed.

Important Usage Warnings

Users should be aware that this model:

• May generate sensitive, controversial, or inappropriate content due to reduced safety filtering.
• Is not suitable for all audiences or public-facing applications.
• Requires users to ensure compliance with local laws and ethical standards.
• Lacks default safety guarantees; huihui.ai and SkGufranAhmed bear no responsibility for consequences arising from its use.