Dynamic loading a model.md

Dynamic loading a model

Authors: Jubin Jose

Aquila Hub serves compressor models to generate latent vector for an input data. Deep Learning models are really good knowledge compressors - hence the name "compressor model".

As seen in Aquila DB Schema specification, "encoder" key in a shema definition specified which model to be loaded and used to compress data for a particular database.

• An Aquila Hub node should validate CID of the schema definition with corresponding database name.
• On successful validation, an Aquila Hub node should parse value corresponding to the "encoder" key in the schema and validate it.
• On successful validation, an Aquila Hub node should download "compressor model" from the URL and should keep in local storage.
• After ensuring safe storage of the model on disk, an Aquila Hub node should load the model to main memory as a service.

a URL can be location addressed or content addressed.