async def test_predict_field_headers_async():
client = PredictionServiceAsyncClient(
credentials=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.name = "name/value"
# Mock the actual call within the gRPC stub, and fake the request.
with mock.patch.object(type(client._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", "name=name/value",) in kw["metadata"]
async def test_predict_async(transport: str = "grpc_asyncio"):
client = PredictionServiceAsyncClient(
credentials=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 = prediction_service.PredictRequest()
# Mock the actual call within the gRPC stub, and fake the request.
with mock.patch.object(type(client._client._transport.predict), "__call__") as call:
# Designate an appropriate return value for the call.
call.return_value = grpc_helpers_async.FakeUnaryUnaryCall(
prediction_service.PredictResponse()
)
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] == request
# Establish that the response is the type that we expect.
assert isinstance(response, prediction_service.PredictResponse)
def test_predict_flattened_error():
client = PredictionServiceClient(credentials=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(),
name="name_value",
payload=data_items.ExamplePayload(
image=data_items.Image(image_bytes=b"image_bytes_blob")
),
params={"key_value": "value_value"},
)
def predict(
self,
request: prediction_service.PredictRequest = None,
*,
name: str = None,
payload: data_items.ExamplePayload = None,
params: Sequence[prediction_service.PredictRequest.ParamsEntry] = None,
retry: retries.Retry = gapic_v1.method.DEFAULT,
timeout: float = None,
metadata: Sequence[Tuple[str, str]] = (),
) -> prediction_service.PredictResponse:
r"""Perform an online prediction. The prediction result will be
directly returned in the response. Available for following ML
problems, and their expected request payloads:
- Image Classification - Image in .JPEG, .GIF or .PNG format,
image_bytes up to 30MB.
- Image Object Detection - Image in .JPEG, .GIF or .PNG format,
image_bytes up to 30MB.
- Text Classification - TextSnippet, content up to 60,000
characters, UTF-8 encoded.
- Text Extraction - TextSnippet, content up to 30,000
characters, UTF-8 NFC encoded.
- Translation - TextSnippet, content up to 25,000 characters,
UTF-8 encoded.
- Tables - Row, with column values matching the columns of the
model, up to 5MB. Not available for FORECASTING
[prediction_type][google.cloud.automl.v1beta1.TablesModelMetadata.prediction_type].
- Text Sentiment - TextSnippet, content up 500 characters,
UTF-8 encoded.
Args:
request (google.cloud.automl_v1beta1.types.PredictRequest):
The request object. Request message for
[PredictionService.Predict][google.cloud.automl.v1beta1.PredictionService.Predict].
name (str):
Required. Name of the model requested
to serve the prediction.
This corresponds to the ``name`` field
on the ``request`` instance; if ``request`` is provided, this
should not be set.
payload (google.cloud.automl_v1beta1.types.ExamplePayload):
Required. Payload to perform a
prediction on. The payload must match
the problem type that the model was
trained to solve.
This corresponds to the ``payload`` field
on the ``request`` instance; if ``request`` is provided, this
should not be set.
params (Sequence[google.cloud.automl_v1beta1.types.PredictRequest.ParamsEntry]):
Additional domain-specific parameters, any string must
be up to 25000 characters long.
- For Image Classification:
``score_threshold`` - (float) A value from 0.0 to
1.0. When the model makes predictions for an image,
it will only produce results that have at least this
confidence score. The default is 0.5.
- For Image Object Detection: ``score_threshold`` -
(float) When Model detects objects on the image, it
will only produce bounding boxes which have at least
this confidence score. Value in 0 to 1 range, default
is 0.5. ``max_bounding_box_count`` - (int64) No more
than this number of bounding boxes will be returned
in the response. Default is 100, the requested value
may be limited by server.
- For Tables: feature_importance - (boolean) Whether
feature importance should be populated in the
returned TablesAnnotation. The default is false.
This corresponds to the ``params`` 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.automl_v1beta1.types.PredictResponse:
Response message for
[PredictionService.Predict][google.cloud.automl.v1beta1.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([name, payload, params])
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 name is not None:
request.name = name
if payload is not None:
request.payload = payload
if params is not None:
request.params = params
# 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(
(("name", request.name), )), )
# Send the request.
response = rpc(
request,
retry=retry,
timeout=timeout,
metadata=metadata,
)
# Done; return the response.
return response
