I would like to cache huggingface models on my organisation’s nexus repository while our server doesn’t have internet directly to pull the model. Apprececiate if anyone can share the details
See the Installation — transformers 4.0.0 documentation
If you want to avoid downloading models from the hub you can put the files into the cache directory, with nexus you could obviously put the model files in a JAR file and have a script unpack the models into the right place.
I have read the document but not showing in details steps. Please advise is there any UR L that i can configure in nexus repo?
I don’t think you can put a URL to a file if it’s not already JAR file, I think you have to make a JAR file with the model files (like make a maven project that has the files in src/main/resources and then deploy the resulting JAR to your nexus server.
You’ll need some script that unpacks the files in the JAR to the cache directory which is not hard because a JAR file is just a ZIP file that is named “.jar” and has certain files in it.