async def test_predict_field_headers_async():
client = PredictionServiceAsyncClient(
credentials=ga_credentials.AnonymousCredentials(),
)
# Any value that is part of the HTTP/1.1 URI should be sent as
# a field header. Set these to a non-empty value.
request = prediction_service.PredictRequest()
request.endpoint = "endpoint/value"
# Mock the actual call within the gRPC stub, and fake the request.
with mock.patch.object(type(client.transport.predict), "__call__") as call:
call.return_value = grpc_helpers_async.FakeUnaryUnaryCall(
prediction_service.PredictResponse()
)
await client.predict(request)
# Establish that the underlying gRPC stub method was called.
assert len(call.mock_calls)
_, args, _ = call.mock_calls[0]
assert args[0] == request
# Establish that the field header was sent.
_, _, kw = call.mock_calls[0]
assert ("x-goog-request-params", "endpoint=endpoint/value",) in kw["metadata"]
async def test_predict_async(
transport: str = "grpc_asyncio", request_type=prediction_service.PredictRequest
):
client = PredictionServiceAsyncClient(
credentials=ga_credentials.AnonymousCredentials(), transport=transport,
)
# Everything is optional in proto3 as far as the runtime is concerned,
# and we are mocking out the actual API, so just send an empty request.
request = request_type()
# Mock the actual call within the gRPC stub, and fake the request.
with mock.patch.object(type(client.transport.predict), "__call__") as call:
# Designate an appropriate return value for the call.
call.return_value = grpc_helpers_async.FakeUnaryUnaryCall(
prediction_service.PredictResponse(
deployed_model_id="deployed_model_id_value",
)
)
response = await client.predict(request)
# Establish that the underlying gRPC stub method was called.
assert len(call.mock_calls)
_, args, _ = call.mock_calls[0]
assert args[0] == prediction_service.PredictRequest()
# Establish that the response is the type that we expect.
assert isinstance(response, prediction_service.PredictResponse)
assert response.deployed_model_id == "deployed_model_id_value"
def test_predict_field_headers():
client = PredictionServiceClient(
credentials=ga_credentials.AnonymousCredentials(),
)
# Any value that is part of the HTTP/1.1 URI should be sent as
# a field header. Set these to a non-empty value.
request = prediction_service.PredictRequest()
request.endpoint = 'endpoint/value'
# Mock the actual call within the gRPC stub, and fake the request.
with mock.patch.object(
type(client.transport.predict),
'__call__') as call:
call.return_value = prediction_service.PredictResponse()
client.predict(request)
# Establish that the underlying gRPC stub method was called.
assert len(call.mock_calls) == 1
_, args, _ = call.mock_calls[0]
assert args[0] == request
# Establish that the field header was sent.
_, _, kw = call.mock_calls[0]
assert (
'x-goog-request-params',
'endpoint=endpoint/value',
) in kw['metadata']
def test_predict_flattened_error():
client = PredictionServiceClient(credentials=ga_credentials.AnonymousCredentials(),)
# Attempting to call a method with both a request object and flattened
# fields is an error.
with pytest.raises(ValueError):
client.predict(
prediction_service.PredictRequest(),
endpoint="endpoint_value",
instances=[struct_pb2.Value(null_value=struct_pb2.NullValue.NULL_VALUE)],
parameters=struct_pb2.Value(null_value=struct_pb2.NullValue.NULL_VALUE),
)
def test_predict_empty_call():
# This test is a coverage failsafe to make sure that totally empty calls,
# i.e. request == None and no flattened fields passed, work.
client = PredictionServiceClient(
credentials=ga_credentials.AnonymousCredentials(), transport="grpc",
)
# Mock the actual call within the gRPC stub, and fake the request.
with mock.patch.object(type(client.transport.predict), "__call__") as call:
client.predict()
call.assert_called()
_, args, _ = call.mock_calls[0]
assert args[0] == prediction_service.PredictRequest()
async def predict(
self,
request: prediction_service.PredictRequest = None,
*,
endpoint: str = None,
instances: Sequence[struct.Value] = None,
parameters: struct.Value = None,
retry: retries.Retry = gapic_v1.method.DEFAULT,
timeout: float = None,
metadata: Sequence[Tuple[str, str]] = (),
) -> prediction_service.PredictResponse:
r"""Perform an online prediction.
Args:
request (:class:`google.cloud.aiplatform_v1.types.PredictRequest`):
The request object. Request message for
``PredictionService.Predict``.
endpoint (:class:`str`):
Required. The name of the Endpoint requested to serve
the prediction. Format:
``projects/{project}/locations/{location}/endpoints/{endpoint}``
This corresponds to the ``endpoint`` field
on the ``request`` instance; if ``request`` is provided, this
should not be set.
instances (:class:`Sequence[google.protobuf.struct_pb2.Value]`):
Required. The instances that are the input to the
prediction call. A DeployedModel may have an upper limit
on the number of instances it supports per request, and
when it is exceeded the prediction call errors in case
of AutoML Models, or, in case of customer created
Models, the behaviour is as documented by that Model.
The schema of any single instance may be specified via
Endpoint's DeployedModels'
[Model's][google.cloud.aiplatform.v1.DeployedModel.model]
[PredictSchemata's][google.cloud.aiplatform.v1.Model.predict_schemata]
``instance_schema_uri``.
This corresponds to the ``instances`` field
on the ``request`` instance; if ``request`` is provided, this
should not be set.
parameters (:class:`google.protobuf.struct_pb2.Value`):
The parameters that govern the prediction. The schema of
the parameters may be specified via Endpoint's
DeployedModels' [Model's
][google.cloud.aiplatform.v1.DeployedModel.model]
[PredictSchemata's][google.cloud.aiplatform.v1.Model.predict_schemata]
``parameters_schema_uri``.
This corresponds to the ``parameters`` field
on the ``request`` instance; if ``request`` is provided, this
should not be set.
retry (google.api_core.retry.Retry): Designation of what errors, if any,
should be retried.
timeout (float): The timeout for this request.
metadata (Sequence[Tuple[str, str]]): Strings which should be
sent along with the request as metadata.
Returns:
google.cloud.aiplatform_v1.types.PredictResponse:
Response message for
``PredictionService.Predict``.
"""
# Create or coerce a protobuf request object.
# Sanity check: If we got a request object, we should *not* have
# gotten any keyword arguments that map to the request.
has_flattened_params = any([endpoint, instances, parameters])
if request is not None and has_flattened_params:
raise ValueError("If the `request` argument is set, then none of "
"the individual field arguments should be set.")
request = prediction_service.PredictRequest(request)
# If we have keyword arguments corresponding to fields on the
# request, apply these.
if endpoint is not None:
request.endpoint = endpoint
if parameters is not None:
request.parameters = parameters
if instances:
request.instances.extend(instances)
# Wrap the RPC method; this adds retry and timeout information,
# and friendly error handling.
rpc = gapic_v1.method_async.wrap_method(
self._client._transport.predict,
default_timeout=None,
client_info=DEFAULT_CLIENT_INFO,
)
# Certain fields should be provided within the metadata header;
# add these here.
metadata = tuple(metadata) + (gapic_v1.routing_header.to_grpc_metadata(
(("endpoint", request.endpoint), )), )
# Send the request.
response = await rpc(
request,
retry=retry,
timeout=timeout,
metadata=metadata,
)
# Done; return the response.
return response
def predict(
self,
request: Union[prediction_service.PredictRequest, dict] = None,
*,
endpoint: str = None,
instances: Sequence[struct_pb2.Value] = None,
parameters: struct_pb2.Value = None,
retry: OptionalRetry = gapic_v1.method.DEFAULT,
timeout: float = None,
metadata: Sequence[Tuple[str, str]] = (),
) -> prediction_service.PredictResponse:
r"""Perform an online prediction.
.. code-block::
from google.cloud import aiplatform_v1
def sample_predict():
# Create a client
client = aiplatform_v1.PredictionServiceClient()
# Initialize request argument(s)
instances = aiplatform_v1.Value()
instances.null_value = "NULL_VALUE"
request = aiplatform_v1.PredictRequest(
endpoint="endpoint_value",
instances=instances,
)
# Make the request
response = client.predict(request=request)
# Handle the response
print(response)
Args:
request (Union[google.cloud.aiplatform_v1.types.PredictRequest, dict]):
The request object. Request message for
[PredictionService.Predict][google.cloud.aiplatform.v1.PredictionService.Predict].
endpoint (str):
Required. The name of the Endpoint requested to serve
the prediction. Format:
``projects/{project}/locations/{location}/endpoints/{endpoint}``
This corresponds to the ``endpoint`` field
on the ``request`` instance; if ``request`` is provided, this
should not be set.
instances (Sequence[google.protobuf.struct_pb2.Value]):
Required. The instances that are the input to the
prediction call. A DeployedModel may have an upper limit
on the number of instances it supports per request, and
when it is exceeded the prediction call errors in case
of AutoML Models, or, in case of customer created
Models, the behaviour is as documented by that Model.
The schema of any single instance may be specified via
Endpoint's DeployedModels'
[Model's][google.cloud.aiplatform.v1.DeployedModel.model]
[PredictSchemata's][google.cloud.aiplatform.v1.Model.predict_schemata]
[instance_schema_uri][google.cloud.aiplatform.v1.PredictSchemata.instance_schema_uri].
This corresponds to the ``instances`` field
on the ``request`` instance; if ``request`` is provided, this
should not be set.
parameters (google.protobuf.struct_pb2.Value):
The parameters that govern the prediction. The schema of
the parameters may be specified via Endpoint's
DeployedModels' [Model's
][google.cloud.aiplatform.v1.DeployedModel.model]
[PredictSchemata's][google.cloud.aiplatform.v1.Model.predict_schemata]
[parameters_schema_uri][google.cloud.aiplatform.v1.PredictSchemata.parameters_schema_uri].
This corresponds to the ``parameters`` field
on the ``request`` instance; if ``request`` is provided, this
should not be set.
retry (google.api_core.retry.Retry): Designation of what errors, if any,
should be retried.
timeout (float): The timeout for this request.
metadata (Sequence[Tuple[str, str]]): Strings which should be
sent along with the request as metadata.
Returns:
google.cloud.aiplatform_v1.types.PredictResponse:
Response message for
[PredictionService.Predict][google.cloud.aiplatform.v1.PredictionService.Predict].
"""
# Create or coerce a protobuf request object.
# Quick check: If we got a request object, we should *not* have
# gotten any keyword arguments that map to the request.
has_flattened_params = any([endpoint, instances, parameters])
if request is not None and has_flattened_params:
raise ValueError(
"If the `request` argument is set, then none of "
"the individual field arguments should be set."
)
# Minor optimization to avoid making a copy if the user passes
# in a prediction_service.PredictRequest.
# There's no risk of modifying the input as we've already verified
# there are no flattened fields.
if not isinstance(request, prediction_service.PredictRequest):
request = prediction_service.PredictRequest(request)
# If we have keyword arguments corresponding to fields on the
# request, apply these.
if endpoint is not None:
request.endpoint = endpoint
if instances is not None:
request.instances.extend(instances)
if parameters is not None:
request.parameters = parameters
# Wrap the RPC method; this adds retry and timeout information,
# and friendly error handling.
rpc = self._transport._wrapped_methods[self._transport.predict]
# Certain fields should be provided within the metadata header;
# add these here.
metadata = tuple(metadata) + (
gapic_v1.routing_header.to_grpc_metadata((("endpoint", request.endpoint),)),
)
# Send the request.
response = rpc(request, retry=retry, timeout=timeout, metadata=metadata,)
# Done; return the response.
return response
