I’ve been going over Hugging Face for research purposes, we’ve been looking over several datasets and models. We started doing this sometime ago last year during September and October. However, recently I checked this again, and it seems like these commit histories have changed.
For instance, we looked over gem back in October of last year, but it shows that its commit history started on Jan 25.
I am using Hugging Face as a use case for research purposes, I want to publish this research eventually, however people will ask questions about this, so I was wondering if someone could offer an explanation for this.
1 Like
Many times, people do changes in the repositories metadata to help with discoverability or consistency across models/datasets. Other times, the model card or dataset sheet are extended/improved. For example, the last change of GEM - Update files from the datasets library (from 1.17.0) · gem at d5a0674 - is just fixing a typo in the dataset metadata
I see, but what does that imply for the commit histories? Do these updates mean that the commit histories are replaced?
For the record, I just want to know in order to log this information as a justification for the procedures we’re taking. We’ve been observing the commit histories and want to be able to explain why they may change over time.