# PMC Helper Model Checkpoints

These checkpoints contain helper models used in creating the PMC-2M dataset.

## Download Weights

Make sure to run this script before using the pipeline.

```python
from huggingface_hub import snapshot_download

weights_dir = snapshot_download(
    repo_id="vector-institute/pmc-helper-models",
    local_dir="openpmcvl/granular/checkpoints",
    allow_patterns=["*.pt", "*.pth"]
)
```

The weights will be downloaded to the specified `local_dir`. Your existing code can then load them from this location.
Change the `local_dir` to the path of the directory leading to `openpmcvl/granular/checkpoints` in your project.

## Citation
TBD

The original weights come from PMC-CLIP paper. Please cite them if using the weights.
```
@misc{lin2023pmcclipcontrastivelanguageimagepretraining,
      title={PMC-CLIP: Contrastive Language-Image Pre-training using Biomedical Documents}, 
      author={Weixiong Lin and Ziheng Zhao and Xiaoman Zhang and Chaoyi Wu and Ya Zhang and Yanfeng Wang and Weidi Xie},
      year={2023},
      eprint={2303.07240},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2303.07240}, 
}
```