E X A O N E 3.5 2.4 B Instruct abliterated

Introduction

The EXAONE-3.5-2.4B-INSTRUCT-ABLITERATED model is an uncensored variant of the LGAI-EXAONE/EXAONE-3.5-2.4B-Instruct. It has been modified using a process called "abliteration" to remove refusals from the language model without using TransformerLens. This model supports English and Korean languages.

Architecture

The model is based on the Transformers library, employing a 2.4 billion parameter architecture. It is designed for text generation tasks and has been modified to bypass traditional refusal mechanisms, making it suitable for applications requiring more open-ended conversational capabilities.

Training

This version is a proof-of-concept implementation, derived from the original EXAONE-3.5-2.4B-Instruct model. The modification aims to enhance conversational flexibility by removing refusal responses typically present in large language models (LLMs).

Guide: Running Locally

To run the model locally, you can use the Ollama platform. Here are the basic steps:

1. Ensure you have the necessary environment set up to use Ollama.
2. Execute the following command to run the model:

   ollama run huihui_ai/exaone3.5-abliterated:2.4b
   

3. For optimal performance, it is recommended to use cloud GPUs, such as those provided by AWS, Google Cloud, or Azure.

License

The model is released under the exaone license. For more detailed licensing information, please refer to the provided LICENSE link.