CodeLlama-7B QLoRA — Python Code Completion
Fine-tuned on CodeSearchNet Python with LoRA (rank=8) · Evaluated on HumanEval
| pass@1 |
pass@5 |
pass@10 |
| 26.83% |
35.91% |
38.41% |
Pre-computed outputs from fine-tuned CodeLlama-7B + QLoRA model (inference requires GPU)
Model: sedaklc/codellama-7b-qlora-humaneval