def setUp(self):
self.op_type = "conv2d_fusion"
self.exhaustive_search = False
self.data_format = "AnyLayout"
self.dtype = np.float32
self.activation = 'relu'
self.add_bias = True
self.add_residual_data = True
self.init_group()
self.init_dilation()
self.init_test_case()
self.init_bias_residual()
self.init_activation()
self.set_search_method()
conv2d_param = {
'stride': self.stride,
'pad': self.pad,
'dilation': self.dilations
}
input = np.random.random(self.input_size).astype(self.dtype)
filter = np.random.random(self.filter_size).astype(self.dtype)
output = conv2d_forward_naive(input, filter, self.groups,
conv2d_param).astype(self.dtype)
self.inputs = {
'Input': OpTest.np_dtype_to_fluid_dtype(input),
'Filter': OpTest.np_dtype_to_fluid_dtype(filter)
}
if self.add_residual_data:
residual_data = np.random.random(output.shape).astype(self.dtype)
self.inputs['ResidualData'] = OpTest.np_dtype_to_fluid_dtype(
residual_data)
output += residual_data
if self.add_bias:
bias = np.random.random(self.filter_size[0]).astype(self.dtype)
self.inputs['Bias'] = OpTest.np_dtype_to_fluid_dtype(bias)
output = output + bias.reshape((1, bias.size, 1, 1))
assert self.activation in ['relu', 'identity']
if self.activation == 'relu':
output = np.maximum(output, 0)
self.attrs = {
'strides': self.stride,
'paddings': self.pad,
'groups': self.groups,
'dilations': self.dilations,
'data_format': self.data_format,
'exhaustive_search': self.exhaustive_search,
'activation': self.activation
}
self.outputs = {'Output': output}
def setUp(self):
self.op_type = "depthwise_conv2d"
self.dtype = np.float32
self.set_npu()
self.init_data_format()
self.init_data_type()
self.init_paddings()
self.init_test_case()
self.init_test_case_2()
conv2d_param = {
'stride': self.stride,
'pad': self.pad,
'dilation': self.dilations
}
input = np.random.random(self.input_size).astype(self.dtype)
filter = np.random.uniform(-1, 1, self.filter_size).astype(self.dtype)
output, _, _, _, _ = conv2d_forward_naive(input, filter, self.groups,
conv2d_param,
self.padding_algorithm,
self.data_format)
output = output.astype(self.dtype)
self.inputs = {
'Input': OpTest.np_dtype_to_fluid_dtype(input),
'Filter': OpTest.np_dtype_to_fluid_dtype(filter)
}
self.attrs = {
'strides': self.stride,
'paddings': self.pad,
'padding_algorithm': self.padding_algorithm,
'groups': self.groups,
'dilations': self.dilations,
'data_format': self.data_format
}
self.outputs = {'Output': output}
def setUp(self):
self.op_type = "conv2d_fusion"
self.exhaustive_search = False
self.data_format = "NCHW"
self.dtype = np.float32
self.activation = 'relu'
self.add_residual_data = True
self.split_channels = None
self.outputs = None
self.padding_algorithm = "EXIPLICIT"
self.init_group()
self.init_dilation()
self.init_test_case()
self.init_residual()
self.init_activation()
self.init_paddings()
self.set_search_method()
conv2d_param = {
'stride': self.stride,
'pad': self.pad,
'dilation': self.dilations
}
input = np.random.random(self.input_size).astype(self.dtype)
filter = np.random.random(self.filter_size).astype(self.dtype)
bias = np.random.random(self.filter_size[0]).astype(self.dtype)
self.output, _, _, _, _ = conv2d_forward_naive(input, filter,
self.groups,
conv2d_param,
self.padding_algorithm,
self.data_format)
self.output = self.output.astype(self.dtype)
self.inputs = {
'Input': OpTest.np_dtype_to_fluid_dtype(input),
'Filter': OpTest.np_dtype_to_fluid_dtype(filter),
'Bias': OpTest.np_dtype_to_fluid_dtype(bias)
}
if self.add_residual_data:
residual_data = np.random.random(self.output.shape).astype(
self.dtype)
self.inputs['ResidualData'] = OpTest.np_dtype_to_fluid_dtype(
residual_data)
self.output += residual_data
# Add bias
self.output = self.output + bias.reshape((1, bias.size, 1, 1))
assert self.activation in ['relu', 'identity']
if self.activation == 'relu':
self.output = np.maximum(self.output, 0)
self.attrs = {
'strides': self.stride,
'paddings': self.pad,
'groups': self.groups,
'dilations': self.dilations,
'data_format': self.data_format,
'exhaustive_search': self.exhaustive_search,
'activation': self.activation,
'padding_algorithm': self.padding_algorithm
}
if self.split_channels is not None:
self.attrs['split_channels'] = self.split_channels
self.outputs = {'Output': self.output}
self.set_outputs()