Pay-per-token Inference #30
Description
Currently, models can be requested directly via API for free on a time basis. A per-token monitoring system should be implemented, either tied to user/worker stats stored in the DHT or through some other dedicated monitor.
Validators currently maintain a list of popular Hugging Face models that have been used on the network, and these models are automatically loaded and available for free use. These free api jobs could be the target of the token monitoring and payment system.