uknowae58/bambara-finetune

# Bambara Fine-tuning Fine-tuning Qwen3.5 on Bambara language using Unsloth. ## Requirements - GPU with 5GB+ VRAM - Python 3.10+ - PyTorch 2.0+ ## Setup ```bash pip install unsloth datasets trl ``` ## Dataset - oza75/bambara-lm-qa - 824k Q&A pairs in Bambara ## Usage ```bash python train.py ``` ## Model - Base: Qwen/Qwen3-0.5B - Method: LoRA (r=16) - Max seq length: 2048 ## Hardware Tested on: - RTX 5090 32GB - RTX 4090 24GB ## License MIT