Kelvin000010191/reasoner-coder-7b-merge
Kelvin000010191/reasoner-coder-7b-merge is a 7.6 billion parameter language model created by Kelvin000010191, merging Qwen/Qwen2.5-Coder-7B-Instruct with open-thoughts/OpenThinker3-7B using the DARE TIES method. This model is designed to combine the coding capabilities of the Qwen Coder base with the reasoning abilities of OpenThinker3. It is optimized for tasks requiring both code generation and logical problem-solving, leveraging a 32768 token context length.
Loading preview...
Model Overview
Kelvin000010191/reasoner-coder-7b-merge is a 7.6 billion parameter language model developed by Kelvin000010191. It was created using the mergekit tool, specifically employing the DARE TIES merge method.
Key Capabilities
This model is a strategic merge designed to combine the strengths of its constituent models:
- Coding Proficiency: Inherits strong code generation and understanding capabilities from its base model, Qwen/Qwen2.5-Coder-7B-Instruct.
- Reasoning Abilities: Integrates reasoning enhancements from open-thoughts/OpenThinker3-7B.
- Extended Context: Maintains a substantial context window of 32768 tokens, beneficial for complex coding or reasoning tasks.
Merge Details
The merge process utilized Qwen/Qwen2.5-Coder-7B-Instruct as the base model. The open-thoughts/OpenThinker3-7B model was incorporated with specific parameter weighting, where embedding and lm_head layers were excluded from the merge, and other layers received a weight of 0.5. The tokenizer from the Qwen Coder base was retained, ensuring compatibility with its ChatML template and tool-call special tokens.
Good For
- Code Generation: Generating and understanding programming code.
- Logical Problem Solving: Tasks requiring analytical and reasoning skills.
- Combined Code & Reasoning: Use cases where both coding and complex logical thought are necessary, such as debugging, algorithm design, or explaining code logic.