Code Generation on Code Benchmarks (test)
34.1HumanEvalCodeLlama-Velocitune
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CodeLlama-VelocituneModel=CodeLlama-7B, Training=Dynamic domain weights adjustment2024.11 | 34.1 | 44.4 | 39.3 | |
| CodeLlamaModel=CodeLlama-7B2024.11 | 30.5 | 43.2 | 36.8 | |
| CodeLlama-BaselineModel=CodeLlama-7B, Training=Static domain mixture2024.11 | 26.2 | 44.8 | 35.5 |