def __getitem__(self, item):
"""Override __getitem__ so that we can directly look up plugins via dict-like syntax"""
try:
target_name = parse_target_uri(item)
plugin_like = self.registry[target_name]
except KeyError:
msg = (
'No plugin found for managing model deployments to "{target}". '
'In order to deploy models to "{target}", find and install an appropriate '
"plugin from "
"https://mlflow.org/docs/latest/plugins.html#community-plugins using "
"your package manager (pip, conda etc).".format(target=item)
)
raise MlflowException(msg, error_code=RESOURCE_DOES_NOT_EXIST)
if isinstance(plugin_like, entrypoints.EntryPoint):
try:
plugin_obj = plugin_like.load()
except (AttributeError, ImportError) as exc:
raise RuntimeError('Failed to load the plugin "{}": {}'.format(item, str(exc)))
self.registry[item] = plugin_obj
else:
plugin_obj = plugin_like
# Testing whether the plugin is valid or not
expected = {"target_help", "run_local"}
deployment_classes = []
for name, obj in inspect.getmembers(plugin_obj):
if name in expected:
expected.remove(name)
elif (
inspect.isclass(obj)
and issubclass(obj, BaseDeploymentClient)
and not obj == BaseDeploymentClient
):
deployment_classes.append(name)
if len(expected) > 0:
raise MlflowException(
"Plugin registered for the target {} does not has all "
"the required interfaces. Raise an issue with the "
"plugin developers.\n"
"Missing interfaces: {}".format(item, expected),
error_code=INTERNAL_ERROR,
)
if len(deployment_classes) > 1:
raise MlflowException(
"Plugin registered for the target {} has more than one "
"child class of BaseDeploymentClient. Raise an issue with"
" the plugin developers. "
"Classes found are {}".format(item, deployment_classes)
)
elif len(deployment_classes) == 0:
raise MlflowException(
"Plugin registered for the target {} has no child class"
" of BaseDeploymentClient. Raise an issue with the "
"plugin developers".format(item)
)
return plugin_obj
def get_deploy_client(target_uri):
"""
Returns a subclass of :py:class:`mlflow.deployments.BaseDeploymentClient` exposing standard
APIs for deploying models to the specified target. See available deployment APIs
by calling ``help()`` on the returned object or viewing docs for
:py:class:`mlflow.deployments.BaseDeploymentClient`. You can also run
``mlflow deployments help -t <target-uri>`` via the CLI for more details on target-specific
configuration options.
:param target_uri: URI of target to deploy to.
.. code-block:: python
:caption: Example
from mlflow.deployments import get_deploy_client
import pandas as pd
client = get_deploy_client('redisai')
# Deploy the model stored at artifact path 'myModel' under run with ID 'someRunId'. The
# model artifacts are fetched from the current tracking server and then used for deployment.
client.create_deployment("spamDetector", "runs:/someRunId/myModel")
# Load a CSV of emails and score it against our deployment
emails_df = pd.read_csv("...")
prediction_df = client.predict_deployment("spamDetector", emails_df)
# List all deployments, get details of our particular deployment
print(client.list_deployments())
print(client.get_deployment("spamDetector"))
# Update our deployment to serve a different model
client.update_deployment("spamDetector", "runs:/anotherRunId/myModel")
# Delete our deployment
client.delete_deployment("spamDetector")
"""
target = parse_target_uri(target_uri)
plugin = plugin_store[target]
for _, obj in inspect.getmembers(plugin):
if inspect.isclass(obj):
if issubclass(
obj,
BaseDeploymentClient) and not obj == BaseDeploymentClient:
return obj(target_uri)