Multilingual Health LoRA for Qwen3.5-4B
A 4-bit QLoRA adapter fine-tuned from Qwen/Qwen3.5-4B on multilingual community-health question answering data.
Training
- Two Kaggle T4 GPUs with DDP
- LoRA rank 16, alpha 32
- Maximum sequence length 1024
- Training epochs 1
This is an educational/research artifact and is not a substitute for qualified medical advice.
