---
tags:
- fine-tuning
- lora
- humaneval
- qwen3
license: apache-2.0
---

# Training Scripts

Fine-tuning script for reproducing the HuggingFace fine-tuning challenge.

## Script

**train_humaneval_clean.py** - Fine-tune Qwen3-0.6B to beat base on HumanEval.

## Usage

```bash
hf jobs uv run   --flavor a10g-small   --timeout 14400   --secrets HF_TOKEN   "https://huggingface.co/passagereptile455/training-scripts/resolve/main/train_humaneval_clean.py"
```

## Results

See [qwen3-humaneval-sft](https://huggingface.co/passagereptile455/qwen3-humaneval-sft) for the trained model and full results.

## Notes

Script uses PEP 723 inline dependencies - no local setup required.