I'm trying to get HuggingFace to count the downloads of a model, but it refuses to count it.
The model in questions is uploaded here: https://huggingface.co/ibm-granite/granite-geospatial-wxc-downscaling/blob/main/pytorch_model.bin
Users either download via a direct GIT LFS link: https://huggingface.co/ibm-granite/granite-geospatial-wxc-downscaling/resolve/main/pytorch_model.bin
Or via huggingface_hub python API but neither are counted.
I've created a dummy config.yaml which is counted but obviously has nothing to do with that model.
The easiest thing is to make users of your model to download a config.json or config.yaml which is needed to run your model. So a gist of the hugging face documentation on that topic is the following:
You either stick to their predefined file names, which as of today are: config.json, config.yaml, hyperparams.yaml, and meta.yaml
So only downloads to those are counted.
OR: You create a pull request to some repository editing some typescript file to change this behavior for your repository.