Hello, I recently uploaded a private model to huggingface’s model hub. When I visit the page for the model it lists that my model has about 11 downloads, however I have not used the model at all I’ve only viewed its page as I’ve gone through the process of uploading the model. How are downloads calculated? Is loading the Inference API on the model’s page count as a download? If so would it be possible to separate this count from general downloads since the model is private?
Hi @jodiak, you’re right – loading the model using from_pretrained (which is what the hosted Inference API uses in the background) currently increases the download counter.
I’ll see if we can exclude those from the logs.
1 Like
Hi @julien-c ,
I am also curious about how the ‘download count’ is calculated.
Does it indicate the number of using from_pretrained in total or just a period of time?
I have a model that was shown to have ~300k downloads, but now it only shows ~100k.
Thank you.
1 Like
Hi @hfl-rc , did you find the reason of this drop?
Hi @publichealthsurveill,
Sorry, I’ve no idea how it was calculated.
The download numbers seem to fluctuate a lot and unstable. 