FLAGS = parser.parse_args()
# We use a simple model that takes 2 input tensors of 16 integers
# each and returns 2 output tensors of 16 integers each. One
# output tensor is the element-wise sum of the inputs and one
# output is the element-wise difference.
model_name = "simple_int8"
model_version = ""
batch_size = 1
# Create gRPC stub for communicating with the server
channel = grpc.insecure_channel(FLAGS.url)
grpc_stub = service_pb2_grpc.GRPCInferenceServiceStub(channel)
# Generate the request
request = service_pb2.ModelInferRequest()
request.model_name = model_name
request.model_version = model_version
# Input data
input0_data = [i for i in range(16)]
input1_data = [1 for i in range(16)]
# Populate the inputs in inference request
input0 = service_pb2.ModelInferRequest().InferInputTensor()
input0.name = "INPUT0"
input0.datatype = "INT8"
input0.shape.extend([1, 16])
input0.contents.int_contents[:] = input0_data
input1 = service_pb2.ModelInferRequest().InferInputTensor()
def requestGenerator(input_name, output_name, c, h, w, format, dtype, FLAGS,
result_filenames, supports_batching):
request = service_pb2.ModelInferRequest()
request.model_name = FLAGS.model_name
request.model_version = FLAGS.model_version
filenames = []
if os.path.isdir(FLAGS.image_filename):
filenames = [
os.path.join(FLAGS.image_filename, f)
for f in os.listdir(FLAGS.image_filename)
if os.path.isfile(os.path.join(FLAGS.image_filename, f))
]
else:
filenames = [
FLAGS.image_filename,
]
filenames.sort()
output = service_pb2.ModelInferRequest().InferRequestedOutputTensor()
output.name = output_name
output.parameters['classification'].int64_param = FLAGS.classes
request.outputs.extend([output])
input = service_pb2.ModelInferRequest().InferInputTensor()
input.name = input_name
input.datatype = dtype
if format == mc.ModelInput.FORMAT_NHWC:
input.shape.extend(
[FLAGS.batch_size, h, w, c] if supports_batching else [h, w, c])
else:
input.shape.extend(
[FLAGS.batch_size, c, h, w] if supports_batching else [c, h, w])
# Preprocess image into input data according to model requirements
# Preprocess the images into input data according to model
# requirements
image_data = []
for filename in filenames:
img = Image.open(filename)
image_data.append(preprocess(img, format, dtype, c, h, w,
FLAGS.scaling))
# Send requests of FLAGS.batch_size images. If the number of
# images isn't an exact multiple of FLAGS.batch_size then just
# start over with the first images until the batch is filled.
image_idx = 0
last_request = False
while not last_request:
input_bytes = None
input_filenames = []
request.ClearField("inputs")
request.ClearField("raw_input_contents")
for idx in range(FLAGS.batch_size):
input_filenames.append(filenames[image_idx])
if input_bytes is None:
input_bytes = image_data[image_idx].tobytes()
else:
input_bytes += image_data[image_idx].tobytes()
image_idx = (image_idx + 1) % len(image_data)
if image_idx == 0:
last_request = True
request.inputs.extend([input])
result_filenames.append(input_filenames)
request.raw_input_contents.extend([input_bytes])
yield request
