maylore-ai/Temper-1-1.5B
Temper-1-1.5B by maylore-ai is a 1.5 billion parameter language model fine-tuned from Qwen2.5-Coder-1.5B-Instruct specifically for generating Rust code. It achieves a 50.5% pass@1 score on the HumanEval-Rust benchmark, outperforming Qwen2.5-Coder-3B, a model twice its size. This model excels at Rust code generation, making it suitable for developers focused on Rust-specific programming tasks.
Loading preview...
Overview
Temper-1-1.5B is a 1.5 billion parameter model developed by maylore-ai, fine-tuned from Qwen2.5-Coder-1.5B-Instruct with a focus on generating Rust code. It demonstrates strong performance in Rust programming tasks, notably surpassing larger models in specific benchmarks. The model was trained on 10,000 Rust examples generated by DeepSeek-V4-Flash, ensuring high-quality, compiler-and-test-passing code.
Key Capabilities
- Specialized Rust Code Generation: Achieves a 50.5% pass@1 rate on the HumanEval-Rust benchmark, outperforming Qwen2.5-Coder-3B.
- Efficient Performance: Provides competitive Rust code generation at a smaller model size (1.5B parameters).
- High-Quality Training Data: Trained on verified Rust examples, ensuring functional and correct code outputs.
Good for
- Developers requiring a dedicated model for generating Rust code snippets and functions.
- Applications focused on Rust development where model size and efficiency are important.
- Tasks involving completing Rust code based on English prompts.
Limitations
- Primarily trained for Rust code generation; performance on other languages or non-Rust related queries may be suboptimal.
- Not designed for fill-in-the-middle code completion.
- Limited training data for async code and web frameworks in Rust.