Saikrishna2511/java2py-qwen
Saikrishna2511/java2py-qwen is a 0.5 billion parameter Qwen2.5-Coder-0.5B-Instruct model, fine-tuned by Saikrishna2511, specifically for translating Java code into Python code. This model leverages a LoRA fine-tuning approach on Java-to-Python pairs from the AVATAR-TC dataset. It is optimized for direct code translation tasks, providing a specialized solution for migrating Java snippets to Python.
Loading preview...
Model Overview
Saikrishna2511/java2py-qwen is a specialized 0.5 billion parameter language model, fine-tuned from the Qwen2.5-Coder-0.5B-Instruct base model. Its primary function is to translate Java code into Python code, making it a focused tool for code migration tasks.
Key Capabilities
- Java to Python Translation: The model is specifically trained to convert Java code snippets into their Python equivalents. It uses a prompt format that clearly delineates the Java input and expects Python output.
- LoRA Fine-tuning: This checkpoint was developed using LoRA (Low-Rank Adaptation) with specific parameters (r=16, alpha=32) on the AVATAR-TC dataset, focusing on Java→Python pairs. The weights have been merged for direct inference.
Training Details
The model was fine-tuned on the Qwen/Qwen2.5-Coder-0.5B-Instruct base model using Java→Python pairs from the AVATAR-TC dataset. This targeted training ensures its proficiency in the specific task of cross-language code translation.
Limitations
- As a relatively small 0.5B parameter model, its translation quality can vary significantly with the complexity of the input Java code.
- For more versatile code-related tasks, such as natural language to Python or code to documentation, the multi-task checkpoint
Saikrishna2511/qwen-multitaskis recommended. - This model is not intended for production use without thorough additional evaluation and testing.