Supported Storage Types

This library currently supports several places where you can save your models. You specify the storage type when you create a ModelStore instance:

Supported storage types

Storage

Requires

Example code

AWS s3

The name of an existing s3 bucket

AWS Example

Azure Container

The name of an existing container

Azure Example

Google Cloud Storage

The name of an existing bucket

Cloud Storage Example

File system

A path

File system Example

File system storage

The file system model storage assumes that (a) the root directory exists, and (b) the library user has permission to write to it.

If you want to create the root directory if it does not exist, pass along the create_directory=True argument.

model_store = ModelStore.from_file_system(
  root_directory="/path/to/directory",
  create_directory=True,
)

Read-only access

The three cloud storage types (AWS s3, Google GCS, Azure Containers) assume that (a) the bucket/container exists, and (b) the library user has both read and write permissions.

As of 0.0.74, modelstore also supports read-only access to public Google Cloud Storage buckets.