huihui-ai/DeepSeek-R1-Distill-Qwen-7B-abliterated

This is an uncensored version of deepseek-ai/DeepSeek-R1-Distill-Qwen-7B created with abliteration (see remove-refusals-with-transformers to know more about it).
This is a crude, proof-of-concept implementation to remove refusals from an LLM model without using TransformerLens.

Important Note The 7B model performs slightly worse and may experience some issues with encoding, such as occasional use of wrong characters.

Use with ollama

You can use huihui_ai/deepseek-r1-abliterated directly

ollama run huihui_ai/deepseek-r1-abliterated:7b