def test_ie_core_class():
input_shape = [1, 3, 4, 4]
param = ng.parameter(input_shape, np.float32, name="parameter")
relu = ng.relu(param, name="relu")
func = Function([relu], [param], "test")
func.get_ordered_ops()[2].friendly_name = "friendly"
cnn_network = ov.IENetwork(func)
ie_core = ov.Core()
ie_core.set_config({}, device_name="CPU")
executable_network = ie_core.load_network(cnn_network, "CPU", {})
td = TensorDesc("FP32", input_shape, "NCHW")
# from IPython import embed; embed()
request = executable_network.create_infer_request()
input_data = np.random.rand(*input_shape) - 0.5
expected_output = np.maximum(0.0, input_data)
input_blob = Blob(td, input_data)
request.set_input({"parameter": input_blob})
request.infer()
result = request.get_blob("relu").buffer
assert np.allclose(result, expected_output)
def create_relu(input_shape):
import ngraph as ng
input_shape = ng.impl.PartialShape(input_shape)
param = ng.parameter(input_shape, dtype=np.float32, name="data")
result = ng.relu(param, name="out")
function = ng.Function(result, [param], "TestFunction")
return function
def simple_if(condition_val):
condition = ng.constant(condition_val, dtype=np.bool)
# then_body
X_t = ng.parameter([2], np.float32, "X")
Y_t = ng.parameter([2], np.float32, "Y")
then_mul = ng.multiply(X_t, Y_t)
then_body_res_1 = ng.result(then_mul)
then_body = GraphBody([X_t, Y_t], [then_body_res_1])
then_body_inputs = [
TensorIteratorInvariantInputDesc(1, 0),
TensorIteratorInvariantInputDesc(2, 1)
]
then_body_outputs = [TensorIteratorBodyOutputDesc(0, 0)]
# else_body
X_e = ng.parameter([2], np.float32, "X")
Y_e = ng.parameter([2], np.float32, "Y")
add_e = ng.add(X_e, Y_e)
else_body_res_1 = ng.result(add_e)
else_body = GraphBody([X_e, Y_e], [else_body_res_1])
else_body_inputs = [
TensorIteratorInvariantInputDesc(1, 0),
TensorIteratorInvariantInputDesc(2, 1)
]
else_body_outputs = [TensorIteratorBodyOutputDesc(0, 0)]
X = ng.constant([3, 4], dtype=np.float32)
Y = ng.constant([2, 1], dtype=np.float32)
if_node = ng.if_op(condition, [X, Y], (then_body, else_body),
(then_body_inputs, else_body_inputs),
(then_body_outputs, else_body_outputs))
relu = ng.relu(if_node)
return relu
def get_test_cnnnetwork():
element_type = Type.f32
param = Parameter(element_type, Shape([1, 3, 22, 22]))
relu = ng.relu(param)
func = Function([relu], [param], 'test')
caps = Function.to_capsule(func)
cnnNetwork = IENetwork(caps)
assert cnnNetwork != None
return cnnNetwork
def test_create_IENetwork_from_nGraph():
element_type = Type.f32
param = Parameter(element_type, Shape([1, 3, 22, 22]))
relu = ng.relu(param)
func = Function([relu], [param], "test")
cnnNetwork = ov.IENetwork(func)
assert cnnNetwork is not None
func2 = cnnNetwork.get_function()
assert func2 is not None
assert len(func2.get_ops()) == 3
def test_CreateIENetworkFromNGraph():
element_type = Type.f32
param = Parameter(element_type, Shape([1, 3, 22, 22]))
relu = ng.relu(param)
func = Function([relu], [param], 'test')
caps = Function.to_capsule(func)
cnnNetwork = IENetwork(caps)
assert cnnNetwork != None
assert ng.function_from_cnn(cnnNetwork) != None
assert len(cnnNetwork.layers) == 2
def get_test_cnnnetwork():
param = ng.parameter(Shape([1, 3, 22, 22]), name="parameter")
relu = ng.relu(param)
res = ng.result(relu, name='result')
func = Function([res], [param], 'test')
caps = Function.to_capsule(func)
cnnNetwork = IENetwork(caps)
assert cnnNetwork != None
return cnnNetwork
def test_runtime_info():
test_shape = PartialShape([1, 1, 1, 1])
test_type = Type.f32
test_param = Parameter(test_type, test_shape)
relu_node = ng.relu(test_param)
runtime_info = relu_node.get_rt_info()
runtime_info["affinity"] = "test_affinity"
relu_node.set_friendly_name("testReLU")
runtime_info_after = relu_node.get_rt_info()
assert runtime_info_after["affinity"] == "test_affinity"
def test_get_IENetwork_from_nGraph():
element_type = Type.f32
param = Parameter(element_type, Shape([1, 3, 22, 22]))
relu = ng.relu(param)
func = Function([relu], [param], 'test')
caps = Function.to_capsule(func)
cnnNetwork = IENetwork(caps)
assert cnnNetwork != None
assert ng.function_from_cnn(cnnNetwork) != None
func2 = ng.function_from_cnn(cnnNetwork)
assert func2 != None
def test_offline_api():
element_type = Type.f32
param = Parameter(element_type, Shape([1, 3, 22, 22]))
relu = ng.relu(param)
func = Function([relu], [param], 'test')
caps = Function.to_capsule(func)
cnnNetwork = IENetwork(caps)
assert cnnNetwork != None
ApplyMOCTransformations(cnnNetwork, False)
func2 = ng.function_from_cnn(cnnNetwork)
assert func2 != None
assert len(func2.get_ops()) == 3
def test_mutiple_outputs():
input_shape = [4, 4]
input_data = np.arange(-8, 8).reshape(input_shape)
expected_output = np.split(input_data, 2, axis=1)[0]
expected_output[expected_output < 0] = 0
test_param = ng.parameter(input_shape, dtype=np.float32, name="A")
split = ng.split(test_param, axis=1, num_splits=2)
split_first_output = split.output(0)
relu = ng.relu(split_first_output)
runtime = get_runtime()
computation = runtime.computation(relu, test_param)
output = computation(input_data)
assert np.equal(output, expected_output).all()
def simple_if_without_parameters(condition_val):
condition = ng.constant(condition_val, dtype=np.bool)
# then_body
then_constant = ng.constant(0.7, dtype=np.float)
then_body_res_1 = ng.result(then_constant)
then_body = GraphBody([], [then_body_res_1])
then_body_inputs = []
then_body_outputs = [TensorIteratorBodyOutputDesc(0, 0)]
# else_body
else_const = ng.constant(9.0, dtype=np.float)
else_body_res_1 = ng.result(else_const)
else_body = GraphBody([], [else_body_res_1])
else_body_inputs = []
else_body_outputs = [TensorIteratorBodyOutputDesc(0, 0)]
if_node = ng.if_op(condition, [], (then_body, else_body),
(then_body_inputs, else_body_inputs),
(then_body_outputs, else_body_outputs))
relu = ng.relu(if_node)
return relu
def test_runtime_info():
test_shape = PartialShape([1, 3, 22, 22])
test_type = Type.f32
test_param = Parameter(test_type, test_shape)
relu_node = ng.relu(test_param)
runtime_info = relu_node.get_rt_info()
runtime_info["affinity"] = "test_affinity"
relu_node.set_friendly_name("testReLU")
runtime_info_after = relu_node.get_rt_info()
assert runtime_info == runtime_info_after
params = [test_param]
results = [relu_node]
ng_function = Function(results, params, "testFunc")
capsule = Function.to_capsule(ng_function)
cnn_network = IENetwork(capsule)
cnn_layer = cnn_network.layers["testReLU"]
assert cnn_layer is not None
assert cnn_layer.affinity == "test_affinity"
def create_ngraph_function(args) -> Function:
weights = np.fromfile(args.model, dtype=np.float32)
weights_offset = 0
padding_begin = [0, 0]
padding_end = [0, 0]
# input
input_shape = [64, 1, 28, 28]
param_node = ngraph.parameter(input_shape, np.float32, 'Parameter')
# convolution 1
conv_1_kernel_shape, conv_1_kernel_length = shape_and_length([20, 1, 5, 5])
conv_1_kernel = ngraph.constant(
weights[0:conv_1_kernel_length].reshape(conv_1_kernel_shape))
weights_offset += conv_1_kernel_length
conv_1_node = ngraph.convolution(param_node, conv_1_kernel, [1, 1],
padding_begin, padding_end, [1, 1])
# add 1
add_1_kernel_shape, add_1_kernel_length = shape_and_length([1, 20, 1, 1])
add_1_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
add_1_kernel_length].reshape(add_1_kernel_shape))
weights_offset += add_1_kernel_length
add_1_node = ngraph.add(conv_1_node, add_1_kernel)
# maxpool 1
maxpool_1_node = ngraph.max_pool(add_1_node, [2, 2], padding_begin,
padding_end, [2, 2], 'ceil', None)
# convolution 2
conv_2_kernel_shape, conv_2_kernel_length = shape_and_length(
[50, 20, 5, 5])
conv_2_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
conv_2_kernel_length].reshape(conv_2_kernel_shape))
weights_offset += conv_2_kernel_length
conv_2_node = ngraph.convolution(maxpool_1_node, conv_2_kernel, [1, 1],
padding_begin, padding_end, [1, 1])
# add 2
add_2_kernel_shape, add_2_kernel_length = shape_and_length([1, 50, 1, 1])
add_2_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
add_2_kernel_length].reshape(add_2_kernel_shape))
weights_offset += add_2_kernel_length
add_2_node = ngraph.add(conv_2_node, add_2_kernel)
# maxpool 2
maxpool_2_node = ngraph.max_pool(add_2_node, [2, 2], padding_begin,
padding_end, [2, 2], 'ceil', None)
# reshape 1
reshape_1_dims, reshape_1_length = shape_and_length([2])
# workaround to get int64 weights from float32 ndarray w/o unnecessary copying
dtype_weights = np.frombuffer(weights[weights_offset:weights_offset +
2 * reshape_1_length],
dtype=np.int64)
reshape_1_kernel = ngraph.constant(dtype_weights)
weights_offset += 2 * reshape_1_length
reshape_1_node = ngraph.reshape(maxpool_2_node, reshape_1_kernel, True)
# matmul 1
matmul_1_kernel_shape, matmul_1_kernel_length = shape_and_length(
[500, 800])
matmul_1_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
matmul_1_kernel_length].reshape(matmul_1_kernel_shape))
weights_offset += matmul_1_kernel_length
matmul_1_node = ngraph.matmul(reshape_1_node, matmul_1_kernel, False, True)
# add 3
add_3_kernel_shape, add_3_kernel_length = shape_and_length([1, 500])
add_3_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
add_3_kernel_length].reshape(add_3_kernel_shape))
weights_offset += add_3_kernel_length
add_3_node = ngraph.add(matmul_1_node, add_3_kernel)
# ReLU
relu_node = ngraph.relu(add_3_node)
# reshape 2
reshape_2_kernel = ngraph.constant(dtype_weights)
reshape_2_node = ngraph.reshape(relu_node, reshape_2_kernel, True)
# matmul 2
matmul_2_kernel_shape, matmul_2_kernel_length = shape_and_length([10, 500])
matmul_2_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
matmul_2_kernel_length].reshape(matmul_2_kernel_shape))
weights_offset += matmul_2_kernel_length
matmul_2_node = ngraph.matmul(reshape_2_node, matmul_2_kernel, False, True)
# add 4
add_4_kernel_shape, add_4_kernel_length = shape_and_length([1, 10])
add_4_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
add_4_kernel_length].reshape(add_4_kernel_shape))
weights_offset += add_4_kernel_length
add_4_node = ngraph.add(matmul_2_node, add_4_kernel)
# softmax
softmax_axis = 1
softmax_node = ngraph.softmax(add_4_node, softmax_axis)
# result
result_node = ngraph.result(softmax_node)
# nGraph function
function = Function(result_node, [param_node], 'lenet')
return function
def Relu(onnx_node,
ng_inputs): # type: (NodeWrapper, List[NgraphNode]) -> NgraphNode
"""Apply the Relu function, f(x) = max(0, x) to the input tensor elementwise."""
return ng.relu(ng_inputs[0])
def create_ngraph_function(args: argparse.Namespace) -> ngraph.impl.Function:
"""Create a network on the fly from the source code using ngraph"""
def shape_and_length(shape: list) -> typing.Tuple[list, int]:
length = reduce(lambda x, y: x * y, shape)
return shape, length
weights = np.fromfile(args.model, dtype=np.float32)
weights_offset = 0
padding_begin = padding_end = [0, 0]
# input
input_shape = [64, 1, 28, 28]
param_node = ngraph.parameter(input_shape, np.float32, 'Parameter')
# convolution 1
conv_1_kernel_shape, conv_1_kernel_length = shape_and_length([20, 1, 5, 5])
conv_1_kernel = ngraph.constant(
weights[0:conv_1_kernel_length].reshape(conv_1_kernel_shape))
weights_offset += conv_1_kernel_length
conv_1_node = ngraph.convolution(param_node, conv_1_kernel, [1, 1],
padding_begin, padding_end, [1, 1])
# add 1
add_1_kernel_shape, add_1_kernel_length = shape_and_length([1, 20, 1, 1])
add_1_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
add_1_kernel_length].reshape(add_1_kernel_shape), )
weights_offset += add_1_kernel_length
add_1_node = ngraph.add(conv_1_node, add_1_kernel)
# maxpool 1
maxpool_1_node = ngraph.max_pool(add_1_node, [2, 2], padding_begin,
padding_end, [2, 2], 'ceil', None)
# convolution 2
conv_2_kernel_shape, conv_2_kernel_length = shape_and_length(
[50, 20, 5, 5])
conv_2_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
conv_2_kernel_length].reshape(conv_2_kernel_shape), )
weights_offset += conv_2_kernel_length
conv_2_node = ngraph.convolution(maxpool_1_node, conv_2_kernel, [1, 1],
padding_begin, padding_end, [1, 1])
# add 2
add_2_kernel_shape, add_2_kernel_length = shape_and_length([1, 50, 1, 1])
add_2_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
add_2_kernel_length].reshape(add_2_kernel_shape), )
weights_offset += add_2_kernel_length
add_2_node = ngraph.add(conv_2_node, add_2_kernel)
# maxpool 2
maxpool_2_node = ngraph.max_pool(add_2_node, [2, 2], padding_begin,
padding_end, [2, 2], 'ceil', None)
# reshape 1
reshape_1_dims, reshape_1_length = shape_and_length([2])
# workaround to get int64 weights from float32 ndarray w/o unnecessary copying
dtype_weights = np.frombuffer(
weights[weights_offset:weights_offset + 2 * reshape_1_length],
dtype=np.int64,
)
reshape_1_kernel = ngraph.constant(dtype_weights)
weights_offset += 2 * reshape_1_length
reshape_1_node = ngraph.reshape(maxpool_2_node, reshape_1_kernel, True)
# matmul 1
matmul_1_kernel_shape, matmul_1_kernel_length = shape_and_length(
[500, 800])
matmul_1_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
matmul_1_kernel_length].reshape(matmul_1_kernel_shape), )
weights_offset += matmul_1_kernel_length
matmul_1_node = ngraph.matmul(reshape_1_node, matmul_1_kernel, False, True)
# add 3
add_3_kernel_shape, add_3_kernel_length = shape_and_length([1, 500])
add_3_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
add_3_kernel_length].reshape(add_3_kernel_shape), )
weights_offset += add_3_kernel_length
add_3_node = ngraph.add(matmul_1_node, add_3_kernel)
# ReLU
relu_node = ngraph.relu(add_3_node)
# reshape 2
reshape_2_kernel = ngraph.constant(dtype_weights)
reshape_2_node = ngraph.reshape(relu_node, reshape_2_kernel, True)
# matmul 2
matmul_2_kernel_shape, matmul_2_kernel_length = shape_and_length([10, 500])
matmul_2_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
matmul_2_kernel_length].reshape(matmul_2_kernel_shape), )
weights_offset += matmul_2_kernel_length
matmul_2_node = ngraph.matmul(reshape_2_node, matmul_2_kernel, False, True)
# add 4
add_4_kernel_shape, add_4_kernel_length = shape_and_length([1, 10])
add_4_kernel = ngraph.constant(
weights[weights_offset:weights_offset +
add_4_kernel_length].reshape(add_4_kernel_shape), )
weights_offset += add_4_kernel_length
add_4_node = ngraph.add(matmul_2_node, add_4_kernel)
# softmax
softmax_axis = 1
softmax_node = ngraph.softmax(add_4_node, softmax_axis)
# result
result_node = ngraph.result(softmax_node)
return ngraph.impl.Function(result_node, [param_node], 'lenet')
