Definitions: inference, endpoint, Infinity, open-source libraries and Premium Support
Inference Providers are hosted services that let users, companies or organizations run inference through an API on machine learning models hosted on the Hub. The Inference Endpoint Service lets a Customer create, edit, manage and delete Managed Endpoints. The Infinity Service is a containerized solution for deploying optimized inference pipelines for Transformers Models. The open-source libraries are Hugging Face's software projects on GitHub. Premium Support is qualified information or materials sent by email or messaging service to answer questions on using the libraries or models.
Supporting quotations
run inference via application programming interface on machine learning models publicly or privately hosted on
the Hugging Face Hub service through which the Customer can create, edit, manage and delete Managed Endpoints
the containerized solution to deploy end-to-end optimized inference pipelines for state of the art Transformers Models
the Hugging Face open-source software projects available at
qualified information or any other materials provided by Hugging Face via email or any other instant messaging or communication service
Qualifications & exceptions
These are service descriptions; fees and scope for them are set under Payment and any Supplemental Terms or signed documents.
Not stated here
'Managed Endpoints' and 'Customer' are not separately defined in the supplied text.
Use & Processing · Context
Inference Providers and AutoTrain run on user models and training data hosted on the Hub
With the Inference Providers Service, users or their organizations run inference through an API on machine learning models hosted publicly or privately on the Model Hub. With the AutoTrain premium Service, users create Models from their own training data, and everything is automatically hosted privately on the Model Hub; the user may then share the Models publicly or serve them through the Inference Providers Service.
Supporting quotations
where you or your organization(s) can run inference via application programming interface on machine learning models publicly or privately hosted on our Model Hub
create state of the art Models from your own training data, everything being automatically hosted privately on the Model Hub.
Qualifications & exceptions
AutoTrain output starts private by default; sharing publicly is the user's choice ('You can then share them').
Not stated here
How inputs, outputs or training data are processed or retained beyond hosting is not described in these Terms; it may fall under the unsupplied Privacy Policy or Supplemental Terms.
Use & Processing · Permission or license
Hugging Face may aggregate, anonymize or learn from data about your use of the Services
Subject to its Privacy Policy, Hugging Face may aggregate, anonymize or otherwise learn from data relating to the user's use of the Services, and use the result to improve those Services.
Supporting quotations
In addition and subject to our Privacy Policy, we may aggregate, anonymize, or otherwise learn from data relating to your use of the Services, and use the foregoing to improve those Services.
Qualifications & exceptions
Expressly 'subject to our Privacy Policy', which is not supplied here.
The data is described as data 'relating to your use of the Services'; the Content license is a separate grant.
Not stated here
What 'otherwise learn from' includes, and whether it reaches user Content or only usage data, is not defined in these Terms.
Whether this includes training or improving machine learning models is not stated in this source.