VCoder is a Python-focused coding assistant fine-tuned from Qwen2.5-Coder-3B-Instruct using LoRA and Unsloth.
The model was trained on 15,000 Python instruction-response examples from the Python Code Instructions 15K dataset and optimized for Python code generation, problem solving, debugging, and algorithm implementation.
The model was trained using parameter-efficient fine-tuning (LoRA), allowing adaptation of the base model while keeping computational requirements low.
The model was evaluated against the original Qwen2.5-Coder-3B-Instruct on HumanEval coding tasks.
This demonstrates that the fine-tuned model performs better on Python coding tasks than the original base model.
1prompt = """
2### Instruction:
3Write a Python function to reverse a string.
4
5### Input:
6
7### Response:
8"""
1def reverse_string(text):
2 return text[::-1]
The dataset contains instruction-response pairs focused on Python programming tasks including:
1@misc{vcoder2026,
2 title={VCoder: Python Code Generation Model},
3 author={Varunesh V, Prawin R K, Sarguru N},
4 year={2026},
5 base_model={Qwen2.5-Coder-3B-Instruct}
6}
