OBLITERATUS/Qwen2.5-Coder-7B-Instruct-OBLITERATED
OBLITERATUS/Qwen2.5-Coder-7B-Instruct-OBLITERATED is a 7.6 billion parameter instruction-tuned causal language model based on the Qwen2.5-Coder-7B-Instruct architecture, developed by Qwen and further processed by OBLITERATUS. This model has been modified using the 'advanced' method via the OBLITERATUS tool, which specializes in removing refusal behavior from language models through activation engineering. With a context length of 32768 tokens, it is primarily optimized for code generation and instruction-following tasks, now with reduced refusal tendencies.
Loading preview...
Overview
OBLITERATUS/Qwen2.5-Coder-7B-Instruct-OBLITERATED is a 7.6 billion parameter instruction-tuned model derived from Qwen/Qwen2.5-Coder-7B-Instruct. This version has undergone a specific modification process using the OBLITERATUS tool, which employs activation engineering to remove refusal behaviors from the base model. The modification method used is described as 'advanced'.
Key Characteristics
- Base Model: Qwen2.5-Coder-7B-Instruct, known for its coding capabilities.
- Parameter Count: 7.6 billion parameters.
- Context Length: Supports a substantial context window of 32768 tokens.
- Modification: Processed by OBLITERATUS to mitigate refusal behaviors, enhancing its utility for direct instruction following.
Use Cases
This model is particularly suited for applications requiring a code-centric instruction-following model that is less prone to refusing certain prompts. Its 'abliterated' nature makes it potentially more direct in responding to a wider range of instructions, especially in coding and technical domains where explicit and uninhibited responses are desired. Developers can integrate it using standard Hugging Face transformers library for tasks like code generation, debugging, and technical Q&A.