retrieval/README.md

# Finetuning CLIP reward model

```bash
python train_pl.py --cfg clip_negative_text --id clip_negative_text
```