tomhu/RL4TG-Qwen2.5-3B-GRPO
The tomhu/RL4TG-Qwen2.5-3B-GRPO-2-Epochs model is a 3.1 billion parameter language model based on Qwen2.5-3B-Instruct, fine-tuned using Group Relative Policy Optimization (GRPO). It is specifically optimized for Java unit-test generation, trained on 1,582 mutation-eligible Defects4J samples. This model excels at generating unit tests for Java code, making it suitable for research in automated software testing.
Loading preview...
Model Overview
This model, tomhu/RL4TG-Qwen2.5-3B-GRPO-2-Epochs, is a specialized checkpoint derived from the Qwen/Qwen2.5-3B-Instruct base model. It has been fine-tuned using the Group Relative Policy Optimization (GRPO) method over 2 epochs, specifically targeting Java unit-test generation.
Key Capabilities
- Java Unit-Test Generation: Optimized for creating unit tests for Java codebases.
- GRPO Fine-tuning: Utilizes a reinforcement learning approach for improved performance in its specialized task.
- Research Focus: Primarily intended for research purposes in automated software testing and code generation.
- Defects4J Training: Trained on a dataset of 1,582 mutation-eligible Defects4J training samples, enhancing its relevance for real-world Java projects.
Usage Considerations
This model is designed for generating Java unit tests. Users should be aware that generated tests require compilation and execution within the target project's own build environment to validate their correctness and efficacy. The model uses the Qwen2.5 chat template for prompt formatting. Various training checkpoints are available as Git revisions, allowing for exploration of the training trajectory.