waylay-ml-adapter-sdk 0.1.0

ML_adapter plugin for waylay-sdk-py.

waylay-ml-adapter-sdk

ML Model Adapter plugin for waylay-sdk.

Installation

# installs the ml tool, and the required sdk modules:
pip install waylay-ml-adapter-sdk

# install the adapter module you need:
pip install waylay-ml-adapter-sklearn[dill]

Classes exported by ml_adapter.sdk

The module ml_adapter.sdk exports the following classes

MLTool

ml_adapter.sdk.tool.MLTool

MLAdapter utility service for the waylay client.

Helps creating waylay webscripts and plugs that wrap a machine learning model.

Loaded as tool with name ml_tool in the python sdk.

Example

# create and test a simple model
import numpy as np
test_data = [1,2,3]
expected_result = [2,4,6]
doubler = lambda x: x*2
assert np.array_equal(
    np.array(expected_result),
    doubler(np.array(test_data))
)

# wrap in an adapter, test remoting
from ml_adapter.numpy import V1NumpyModelAdapter
adapter=V1NumpyModelAdapter(model=doubler)
test_resp == await adapter.call({"instances": test_data.tolist()})
assert test_resp['predictions'] = expected_result.tolist()

# use the ml_tool to deploy and test a webscript
# configure logging to see what is happening
import logging
logging.basicConfig(level='INFO')
from waylay.sdk import WaylayClient
client = WaylayClient.from_profile('demo')

ref = await client.ml_tool.create_webscript(
    adapter, name='MyMLWebscript', draft=True
)
ref = await client.ml_tool.wait_until_ready(ref)
result = await client.ml_tool.test_webscript(ref, test_data)
if expected_result.to_list() == result:
    await client.ml_tool.publish(ref)
else:
    await client.ml_tool.remove(ref)

See also

• waylay-sdk the Waylay python SDK.
• waylay-ml-adapter-sklearn ML adapter for scikit-learn models.
• waylay-ml-adapter-torch ML adapter for pytorch models.
• waylay-ml-adapter-base provides the basic ML adapter infrastructure.
• waylay-ml-adapter-api defines the remote data interfaces.