alirezaaminzadeh/minizinc-repair-coder
The alirezaaminzadeh/minizinc-repair-coder is a 1.5 billion parameter instruction-tuned causal language model, fine-tuned from Qwen/Qwen2.5-Coder-1.5B-Instruct. Developed by alirezaaminzadeh, this model is specialized for code-related tasks, leveraging its base architecture's coding capabilities. With a 32K context length, it is optimized for code generation and repair scenarios.
Loading preview...
Model Overview
The alirezaaminzadeh/minizinc-repair-coder is a specialized language model, fine-tuned from the Qwen/Qwen2.5-Coder-1.5B-Instruct base model. This 1.5 billion parameter model is designed for code-centric applications, building upon the robust coding capabilities of its parent architecture. It features a substantial context length of 32,768 tokens, making it suitable for processing and generating longer code snippets and complex programming logic.
Key Capabilities
- Code-focused Instruction Following: Inherits and enhances the instruction-following abilities of the Qwen2.5-Coder series, specifically tailored for programming tasks.
- Code Generation and Repair: Optimized for generating new code and potentially assisting in the repair or modification of existing codebases.
- Extended Context Window: The 32K context length allows for handling larger code files or more extensive conversational turns related to coding problems.
Training Details
This model was trained using Supervised Fine-Tuning (SFT) with the TRL library, indicating a focus on aligning its outputs with specific task instructions. The training leveraged TRL version 1.9.2, Transformers 5.14.1, Pytorch 2.13.0, Datasets 5.0.1, and Tokenizers 0.22.2.
Good For
- Developers working on code generation tasks.
- Applications requiring code repair or suggestion functionalities.
- Scenarios where a model with a strong understanding of programming constructs and a large context window is beneficial.