def test_modelinferrequest_to_types(model_infer_request):
inference_request = ModelInferRequestConverter.to_types(model_infer_request)
expected = types.InferenceRequest(
id="",
inputs=[
types.RequestInput(
name="input-0",
datatype="INT32",
shape=[3],
data=types.TensorData.parse_obj([1, 2, 3]),
),
types.RequestInput(
name="input-1",
datatype="INT32",
shape=[1],
data=types.TensorData.parse_obj([4]),
),
types.RequestInput(
name="input-2",
datatype="INT32",
shape=[2],
data=types.TensorData.parse_obj([5, 6]),
),
],
)
assert type(inference_request) is types.InferenceRequest
assert dict(inference_request) == dict(expected)
def generate_test_requests() -> List[types.InferenceRequest]:
contents_lens = np.power(2, np.arange(10, 16)).astype(int)
max_value = 9999
requests = []
for contents_len in contents_lens:
inputs = max_value * np.random.rand(contents_len)
requests.append(
types.InferenceRequest(inputs=[
types.RequestInput(
name="input-0",
shape=[contents_len],
datatype="FP32",
data=types.TensorData.parse_obj(inputs.tolist()),
)
]))
return requests
def test_modelinferrequest_to_types(model_infer_request):
inference_request = ModelInferRequestConverter.to_types(
model_infer_request)
expected = types.InferenceRequest(
id="",
inputs=[
types.RequestInput(
name="input-0",
datatype="INT32",
shape=[1, 3],
data=types.TensorData.parse_obj([1, 2, 3]),
parameters=types.Parameters(content_type="np"),
)
],
)
assert type(inference_request) is types.InferenceRequest
assert dict(inference_request) == dict(expected)
def extract_request(self) -> np.array:
inputs = self.request["inputs"][0]
default_codec = NumpyCodec()
return default_codec.decode(types.RequestInput(**inputs))