Chest-VLM Run 17 — Weighted loss only (balanced)
LoRA adapter fine-tuned on MedGemma 1.5 4B IT for dual-view chest X-ray structured ontology output (<Ontology v=2>).
Training objective: structure/content token weighting only (no study auxiliary loss).
Recommended when: best overall balanced discrimination (default demo config).
| Metric | Value |
|---|---|
| Study accuracy | 0.757 |
| Study F1 | 0.735 |
| Precision (Normal) | 0.671 |
| Finding macro F1 | 0.223 |
Usage
from peft import PeftModel
# Load base google/medgemma-1.5-4b-it, then:
model = PeftModel.from_pretrained(model, "maximehpe/chest-vlm-run17")
Or download via the project:
python script/download_model.py --run_name 17_weighted_loss_no_study
Not for clinical use. Research prototype only.
- Downloads last month
- 14
Model tree for maximehpe/chest-vlm-run17
Base model
google/medgemma-1.5-4b-it