From af5991210d4f3d5067cfe378dd35c49c6e9f8ded Mon Sep 17 00:00:00 2001 From: xiaoguoguo626807 <100397923+xiaoguoguo626807@users.noreply.github.com> Date: Mon, 19 Dec 2022 12:35:56 +0800 Subject: [PATCH] =?UTF-8?q?=E3=80=90fluid=20API=20clear=E3=80=91move=20con?= =?UTF-8?q?v2d=20(#49025)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * move conv2d * code-style and build bug * restore c++ * code-style * rm fluid init conv2d * remove nn.conv2d * modify doc --- python/paddle/fluid/contrib/layers/nn.py | 4 +- .../slim/tests/test_quantization_pass.py | 4 +- .../tests/test_image_classification_fp16.py | 2 +- .../tests/test_multi_precision_fp16_train.py | 2 +- .../contrib/tests/test_quantize_transpiler.py | 2 +- python/paddle/fluid/layers/nn.py | 334 ------------------ python/paddle/fluid/nets.py | 4 +- python/paddle/fluid/profiler.py | 4 +- .../tests/book/test_image_classification.py | 2 +- .../tests/unittests/asp/asp_pruning_base.py | 2 +- .../asp/test_asp_customized_pruning.py | 2 +- .../unittests/asp/test_asp_optimize_static.py | 2 +- .../unittests/asp/test_asp_pruning_static.py | 2 +- .../tests/unittests/asp/test_asp_save_load.py | 2 +- .../fluid/tests/unittests/dist_se_resnext.py | 2 +- .../dygraph_to_static/test_ptb_lm_v2.py | 2 +- .../tests/unittests/ipu/test_conv_op_ipu.py | 2 +- .../tests/unittests/ipu/test_print_op_ipu.py | 2 +- .../unittests/ipu/test_reduce_x_op_ipu.py | 2 +- .../test_mkldnn_conv_activation_fuse_pass.py | 3 +- .../test_mkldnn_conv_bias_fuse_pass.py | 10 +- .../ir/inference/test_trt_conv_pass.py | 4 +- .../test_trt_conv_quant_dequant_pass.py | 4 +- .../ir/inference/test_trt_dynamic_shape.py | 3 +- .../inference/test_trt_tuned_dynamic_shape.py | 2 +- .../unittests/mlu/sync_batch_norm_op_mlu.py | 2 +- .../unittests/npu/sync_batch_norm_op_npu.py | 2 +- .../fluid/tests/unittests/seresnext_net.py | 2 +- .../fluid/tests/unittests/test_adam_op.py | 2 +- .../fluid/tests/unittests/test_adamax_api.py | 2 +- .../fluid/tests/unittests/test_adamw_op.py | 2 +- .../fluid/tests/unittests/test_backward.py | 2 +- .../fluid/tests/unittests/test_conv2d_api.py | 36 +- .../tests/unittests/test_conv2d_layer.py | 2 +- .../fluid/tests/unittests/test_conv2d_op.py | 4 +- .../tests/unittests/test_conv_nn_grad.py | 18 +- .../tests/unittests/test_functional_conv2d.py | 4 +- .../test_functional_conv2d_transpose.py | 2 +- .../tests/unittests/test_fuse_bn_act_pass.py | 2 +- .../unittests/test_fuse_bn_add_act_pass.py | 8 +- .../test_fuse_relu_depthwise_conv_pass.py | 4 +- .../test_imperative_load_static_param.py | 4 +- .../tests/unittests/test_inference_api.py | 2 +- .../fluid/tests/unittests/test_initializer.py | 4 +- .../fluid/tests/unittests/test_layers.py | 2 +- .../test_parallel_executor_fetch_feed.py | 4 +- .../tests/unittests/test_set_value_op.py | 2 +- .../unittests/test_sync_batch_norm_op.py | 2 +- .../tests/unittests/xpu/test_adamw_op_xpu.py | 2 +- python/paddle/static/__init__.py | 4 + python/paddle/static/nn/__init__.py | 2 +- python/paddle/static/nn/common.py | 318 +++++++++++++++++ 52 files changed, 417 insertions(+), 427 deletions(-) diff --git a/python/paddle/fluid/contrib/layers/nn.py b/python/paddle/fluid/contrib/layers/nn.py index b6c4688dac..abb6e1ed88 100644 --- a/python/paddle/fluid/contrib/layers/nn.py +++ b/python/paddle/fluid/contrib/layers/nn.py @@ -1974,7 +1974,7 @@ def fused_bn_add_act( with fluid.program_guard(main_program, startup_program): x = fluid.layers.data(name='x', shape=[1, 28, 28], dtype='float32') y = fluid.layers.data(name="y", shape=[1], dtype='int64') - conv1_1 = fluid.layers.conv2d( + conv1_1 = paddle.static.nn.conv2d( input=x, filter_size=3, num_filters=32, @@ -1983,7 +1983,7 @@ def fused_bn_add_act( act=None, bias_attr=False, data_format='NHWC') - conv1_2 = fluid.layers.conv2d( + conv1_2 = paddle.static.nn.conv2d( input=x, filter_size=3, num_filters=32, diff --git a/python/paddle/fluid/contrib/slim/tests/test_quantization_pass.py b/python/paddle/fluid/contrib/slim/tests/test_quantization_pass.py index 78310d1ade..db89a4ee1e 100644 --- a/python/paddle/fluid/contrib/slim/tests/test_quantization_pass.py +++ b/python/paddle/fluid/contrib/slim/tests/test_quantization_pass.py @@ -50,7 +50,7 @@ def residual_block(num, quant_skip_pattern=None): def conv_bn_layer( input, ch_out, filter_size, stride, padding, act='relu', bias_attr=False ): - tmp = fluid.layers.conv2d( + tmp = paddle.static.nn.conv2d( input=input, filter_size=filter_size, num_filters=ch_out, @@ -709,7 +709,7 @@ def quant_dequant_residual_block(num, quant_skip_pattern=None): def conv_bn_layer( input, ch_out, filter_size, stride, padding, act='relu', bias_attr=False ): - tmp = fluid.layers.conv2d( + tmp = paddle.static.nn.conv2d( input=input, filter_size=filter_size, num_filters=ch_out, diff --git a/python/paddle/fluid/contrib/tests/test_image_classification_fp16.py b/python/paddle/fluid/contrib/tests/test_image_classification_fp16.py index b6e243fbf2..117877073a 100644 --- a/python/paddle/fluid/contrib/tests/test_image_classification_fp16.py +++ b/python/paddle/fluid/contrib/tests/test_image_classification_fp16.py @@ -32,7 +32,7 @@ def resnet_cifar10(input, depth=32): def conv_bn_layer( input, ch_out, filter_size, stride, padding, act='relu', bias_attr=False ): - tmp = fluid.layers.conv2d( + tmp = paddle.static.nn.conv2d( input=input, filter_size=filter_size, num_filters=ch_out, diff --git a/python/paddle/fluid/contrib/tests/test_multi_precision_fp16_train.py b/python/paddle/fluid/contrib/tests/test_multi_precision_fp16_train.py index 6da6d86dc5..4264a94e9e 100644 --- a/python/paddle/fluid/contrib/tests/test_multi_precision_fp16_train.py +++ b/python/paddle/fluid/contrib/tests/test_multi_precision_fp16_train.py @@ -50,7 +50,7 @@ def resnet_cifar10(input, depth=32): def conv_bn_layer( input, ch_out, filter_size, stride, padding, act='relu', bias_attr=False ): - tmp = fluid.layers.conv2d( + tmp = paddle.static.nn.conv2d( input=input, filter_size=filter_size, num_filters=ch_out, diff --git a/python/paddle/fluid/contrib/tests/test_quantize_transpiler.py b/python/paddle/fluid/contrib/tests/test_quantize_transpiler.py index b890d995a3..18b17ea98e 100644 --- a/python/paddle/fluid/contrib/tests/test_quantize_transpiler.py +++ b/python/paddle/fluid/contrib/tests/test_quantize_transpiler.py @@ -41,7 +41,7 @@ def residual_block(num): def conv_bn_layer( input, ch_out, filter_size, stride, padding, act='relu', bias_attr=False ): - tmp = fluid.layers.conv2d( + tmp = paddle.static.nn.conv2d( input=input, filter_size=filter_size, num_filters=ch_out, diff --git a/python/paddle/fluid/layers/nn.py b/python/paddle/fluid/layers/nn.py index 65f02001b2..543ae8744f 100644 --- a/python/paddle/fluid/layers/nn.py +++ b/python/paddle/fluid/layers/nn.py @@ -65,7 +65,6 @@ from collections.abc import Iterable __all__ = [ 'fc', 'embedding', - 'conv2d', 'row_conv', 'layer_norm', 'spectral_norm', @@ -745,339 +744,6 @@ def _pull_box_sparse( return outs -def conv2d( - input, - num_filters, - filter_size, - stride=1, - padding=0, - dilation=1, - groups=None, - param_attr=None, - bias_attr=None, - use_cudnn=True, - act=None, - name=None, - data_format="NCHW", -): - r""" - :api_attr: Static Graph - - The convolution2D layer calculates the output based on the input, filter - and strides, paddings, dilations, groups parameters. Input and - Output are in NCHW or NHWC format, where N is batch size, C is the number of - channels, H is the height of the feature, and W is the width of the feature. - Filter is in MCHW format, where M is the number of output image channels, - C is the number of input image channels, H is the height of the filter, - and W is the width of the filter. If the groups is greater than 1, - C will equal the number of input image channels divided by the groups. - Please refer to UFLDL's `convolution - `_ - for more details. - If bias attribution and activation type are provided, bias is added to the - output of the convolution, and the corresponding activation function is - applied to the final result. - - For each input :math:`X`, the equation is: - - .. math:: - - Out = \sigma (W \\ast X + b) - - Where: - - * :math:`X`: Input value, a tensor with NCHW or NHWC format. - * :math:`W`: Filter value, a tensor with MCHW format. - * :math:`\\ast`: Convolution operation. - * :math:`b`: Bias value, a 2-D tensor with shape [M, 1]. - * :math:`\\sigma`: Activation function. - * :math:`Out`: Output value, the shape of :math:`Out` and :math:`X` may be different. - - Example: - - - Input: - - Input shape: :math:`(N, C_{in}, H_{in}, W_{in})` - - Filter shape: :math:`(C_{out}, C_{in}, H_f, W_f)` - - - Output: - - Output shape: :math:`(N, C_{out}, H_{out}, W_{out})` - - Where - - .. math:: - - H_{out}&= \\frac{(H_{in} + 2 * paddings[0] - (dilations[0] * (H_f - 1) + 1))}{strides[0]} + 1 \\\\ - W_{out}&= \\frac{(W_{in} + 2 * paddings[1] - (dilations[1] * (W_f - 1) + 1))}{strides[1]} + 1 - - Args: - input (Tensor): The input is 4-D Tensor with shape [N, C, H, W], the data type - of input is float16 or float32 or float64. - num_filters(int): The number of filter. It is as same as the output - image channel. - filter_size (int|tuple): The filter size. If filter_size - is a tuple, it must contain two integers, (filter_size_height, - filter_size_width). Otherwise, filter_size_height = filter_size_width =\ - filter_size. - stride (int|tuple): The stride size. It means the stride in convolution. - If stride is a tuple, it must contain two integers, (stride_height, stride_width). - Otherwise, stride_height = stride_width = stride. Default: stride = 1. - padding (string|int|list|tuple): The padding size. It means the number of zero-paddings - on both sides for each dimension.If `padding` is a string, either 'VALID' or - 'SAME' which is the padding algorithm. If padding size is a tuple or list, - it could be in three forms: `[pad_height, pad_width]` or - `[pad_height_top, pad_height_bottom, pad_width_left, pad_width_right]`, and when - `data_format` is `"NCHW"`, `padding` can be in the form `[[0,0], [0,0], - [pad_height_top, pad_height_bottom], [pad_width_left, pad_width_right]]`. - when `data_format` is `"NHWC"`, `pool_padding` can be in the form - `[[0,0], [pad_height_top, pad_height_bottom], [pad_width_left, pad_width_right], [0,0]]`. - Default: padding = 0. - dilation (int|tuple): The dilation size. It means the spacing between the kernel - points. If dilation is a tuple, it must contain two integers, (dilation_height, - dilation_width). Otherwise, dilation_height = dilation_width = dilation. - Default: dilation = 1. - groups (int): The groups number of the Conv2d Layer. According to grouped - convolution in Alex Krizhevsky's Deep CNN paper: when group=2, - the first half of the filters is only connected to the first half - of the input channels, while the second half of the filters is only - connected to the second half of the input channels. Default: groups=1. - param_attr (ParamAttr|None): The parameter attribute for learnable parameters/weights - of conv2d. If it is set to None or one attribute of ParamAttr, conv2d - will create ParamAttr as param_attr. If the Initializer of the param_attr - is not set, the parameter is initialized with :math:`Normal(0.0, std)`, - and the :math:`std` is :math:`(\\frac{2.0 }{filter\_elem\_num})^{0.5}`. Default: None. - bias_attr (ParamAttr|bool|None): The parameter attribute for the bias of conv2d. - If it is set to False, no bias will be added to the output units. - If it is set to None or one attribute of ParamAttr, conv2d - will create ParamAttr as bias_attr. If the Initializer of the bias_attr - is not set, the bias is initialized zero. Default: None. - use_cudnn (bool): Use cudnn kernel or not, it is valid only when the cudnn - library is installed. Default: True - act (str): Activation type, if it is set to None, activation is not appended. - Default: None - name(str|None): For detailed information, please refer - to :ref:`api_guide_Name`. Usually name is no need to set and - None by default. - data_format (str, optional): Specify the data format of the input, and the data format of the output - will be consistent with that of the input. An optional string from: `"NCHW"`, `"NHWC"`. - The default is `"NCHW"`. When it is `"NCHW"`, the data is stored in the order of: - `[batch_size, input_channels, input_height, input_width]`. - - Returns: - A Tensor representing the conv2d, whose data type is the - same with input. If act is None, the tensor storing the convolution - result, and if act is not None, the tensor storing convolution - and non-linearity activation result. - - Raises: - ValueError: If the type of `use_cudnn` is not bool. - ValueError: If `data_format` is not "NCHW" or "NHWC". - ValueError: If the channel dimmention of the input is less than or equal to zero. - ValueError: If `padding` is a string, but not "SAME" or "VALID". - ValueError: If `padding` is a tuple, but the element corresponding to the input's batch size is not 0 - or the element corresponding to the input's channel is not 0. - ShapeError: If the input is not 4-D Tensor. - ShapeError: If the input's dimension size and filter's dimension size not equal. - ShapeError: If the dimension size of input minus the size of `stride` is not 2. - ShapeError: If the number of input channels is not equal to filter's channels * groups. - ShapeError: If the number of output channels is not be divided by groups. - - Examples: - .. code-block:: python - - import paddle - paddle.enable_static() - - data = paddle.static.data(name='data', shape=[None, 3, 32, 32], dtype='float32') - conv2d = paddle.static.nn.conv2d(input=data, num_filters=2, filter_size=3, act="relu") - print(conv2d.shape) # [-1, 2, 30, 30] - """ - - check_variable_and_dtype( - input, 'input', ['float16', 'float32', 'float64'], 'conv2d' - ) - if len(input.shape) != 4: - raise ValueError( - "Input size should be 4, " - "but received {}".format(len(input.shape)) - ) - num_channels = input.shape[1] - if not isinstance(use_cudnn, bool): - raise ValueError( - "Attr(use_cudnn) should be True or False. Received " - "Attr(use_cudnn): %s. " % str(use_cudnn) - ) - - if data_format not in ["NCHW", "NHWC"]: - raise ValueError( - "Attr(data_format) should be 'NCHW' or 'NHWC'. Received " - "Attr(data_format): %s." % str(data_format) - ) - - channel_last = data_format == "NHWC" - num_channels = input.shape[3] if channel_last else input.shape[1] - if num_channels < 0: - raise ValueError( - "The channel dimmention of the input(%s) should be defined. " - "Received: %s." % (str(input.shape), str(num_channels)) - ) - assert param_attr is not False, "param_attr should not be False here." - - if groups is None: - num_filter_channels = num_channels - elif groups <= 0: - raise ValueError( - "the groups of input must be greater than 0, " - "but received the groups of input is {}".format(groups) - ) - else: - if num_channels % groups != 0: - raise ValueError( - "the channel of input must be divisible by groups," - "received: the channel of input is {}, the shape of input is {}" - ", the groups is {}".format(num_channels, input.shape, groups) - ) - num_filter_channels = num_channels // groups - - l_type = 'conv2d' - if ( - num_channels == groups - and num_filters % num_channels == 0 - and not use_cudnn - ): - l_type = 'depthwise_conv2d' - - if ( - num_channels == groups - and num_filters % num_channels == 0 - and core.is_compiled_with_rocm() - ): - l_type = 'depthwise_conv2d' - - # NPU only supports depthwise_conv2d when "input_channel = output_channel = groups" - if core.is_compiled_with_npu(): - if num_channels == groups and num_channels == num_filters: - l_type = 'depthwise_conv2d' - else: - l_type = 'conv2d' - - helper = LayerHelper(l_type, **locals()) - dtype = helper.input_dtype() - - filter_size = utils.convert_to_list(filter_size, 2, 'filter_size') - stride = utils.convert_to_list(stride, 2, 'stride') - dilation = utils.convert_to_list(dilation, 2, 'dilation') - - # padding - def _update_padding(padding, data_format): - def is_list_or_tuple(ele): - if isinstance(ele, list) or isinstance(ele, tuple): - return True - return False - - if is_list_or_tuple(padding) and len(padding) == 4: - if is_list_or_tuple(padding[0]) and (data_format == "NCHW"): - if not (padding[0] == [0, 0] and padding[1] == [0, 0]): - raise ValueError( - "Non-zero padding(%s) in the batch or channel dimensions " - "is not supported." % str(padding) - ) - padding = padding[2:4] - padding = [ele for a_list in padding for ele in a_list] - elif is_list_or_tuple(padding[0]) and (data_format == "NHWC"): - if not (padding[0] == [0, 0] and padding[3] == [0, 0]): - raise ValueError( - "Non-zero padding(%s) in the batch or channel dimensions " - "is not supported." % str(padding) - ) - padding = padding[1:3] - padding = [ele for a_list in padding for ele in a_list] - padding = utils.convert_to_list(padding, 4, 'padding') - if utils._is_symmetric_padding(padding, 2): - padding = [padding[0], padding[2]] - - else: - padding = utils.convert_to_list(padding, 2, 'padding') - - return padding - - padding_algorithm = "EXPLICIT" - if isinstance(padding, str): - padding = padding.upper() - if padding not in ["SAME", "VALID"]: - raise ValueError( - "Unknown padding: '%s'. It can only be 'SAME' or 'VALID'." - % str(padding) - ) - if padding == "VALID": - padding_algorithm = "VALID" - padding = [0, 0] - elif padding == "SAME": - padding_algorithm = "SAME" - padding = [0, 0] - - padding = _update_padding(padding, data_format) - - filter_shape = [num_filters, int(num_filter_channels)] + filter_size - - def _get_default_param_initializer(): - filter_elem_num = filter_size[0] * filter_size[1] * num_channels - if filter_elem_num <= 0: - raise ValueError( - "Invalid filter number, excepted number is larger than 0, but" - " received {}, please check the input shape and " - "filter size.".format(filter_elem_num) - ) - std = (2.0 / filter_elem_num) ** 0.5 - return Normal(0.0, std, 0) - - filter_param = helper.create_parameter( - attr=helper.param_attr, - shape=filter_shape, - dtype=dtype, - default_initializer=_get_default_param_initializer(), - ) - - pre_bias = helper.create_variable_for_type_inference(dtype) - - if ( - core.is_compiled_with_cuda() - and paddle.fluid.get_flags("FLAGS_conv2d_disable_cudnn")[ - "FLAGS_conv2d_disable_cudnn" - ] - ): - use_cudnn = False - - helper.append_op( - type=l_type, - inputs={ - 'Input': input, - 'Filter': filter_param, - }, - outputs={"Output": pre_bias}, - attrs={ - 'strides': stride, - 'paddings': padding, - 'dilations': dilation, - 'groups': groups, - 'use_cudnn': use_cudnn, - 'use_mkldnn': False, - 'fuse_relu_before_depthwise_conv': False, - "padding_algorithm": padding_algorithm, - "data_format": data_format, - }, - ) - - if data_format == 'NCHW': - pre_act = helper.append_bias_op(pre_bias, dim_start=1, dim_end=2) - else: - pre_act = helper.append_bias_op(pre_bias, dim_start=3, dim_end=4) - - return helper.append_activation(pre_act) - - @templatedoc() def layer_norm( input, diff --git a/python/paddle/fluid/nets.py b/python/paddle/fluid/nets.py index 0c42f07b39..92efd25430 100644 --- a/python/paddle/fluid/nets.py +++ b/python/paddle/fluid/nets.py @@ -119,7 +119,7 @@ def simple_img_conv_pool( pool_stride=2, act="relu") """ - conv_out = layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=input, num_filters=num_filters, filter_size=filter_size, @@ -246,7 +246,7 @@ def img_conv_group( if conv_with_batchnorm[i]: local_conv_act = None - tmp = layers.conv2d( + tmp = paddle.static.nn.conv2d( input=tmp, num_filters=conv_num_filter[i], filter_size=conv_filter_size[i], diff --git a/python/paddle/fluid/profiler.py b/python/paddle/fluid/profiler.py index 6ebb2f9a08..9b29b01fd6 100644 --- a/python/paddle/fluid/profiler.py +++ b/python/paddle/fluid/profiler.py @@ -88,7 +88,7 @@ def npu_profiler(output_file, config=None): epoc = 8 dshape = [4, 3, 28, 28] data = fluid.data(name='data', shape=[None, 3, 28, 28], dtype='float32') - conv = fluid.layers.conv2d(data, 20, 3, stride=[1, 1], padding=[1, 1]) + conv = paddle.static.nn.conv2d(data, 20, 3, stride=[1, 1], padding=[1, 1]) place = fluid.NPUPlace(0) exe = fluid.Executor(place) @@ -338,7 +338,7 @@ def profiler( epoc = 8 dshape = [4, 3, 28, 28] data = fluid.data(name='data', shape=[None, 3, 28, 28], dtype='float32') - conv = fluid.layers.conv2d(data, 20, 3, stride=[1, 1], padding=[1, 1]) + conv = paddle.static.nn.conv2d(data, 20, 3, stride=[1, 1], padding=[1, 1]) place = fluid.CPUPlace() exe = fluid.Executor(place) diff --git a/python/paddle/fluid/tests/book/test_image_classification.py b/python/paddle/fluid/tests/book/test_image_classification.py index b6d32a68da..f20beb731c 100644 --- a/python/paddle/fluid/tests/book/test_image_classification.py +++ b/python/paddle/fluid/tests/book/test_image_classification.py @@ -31,7 +31,7 @@ def resnet_cifar10(input, depth=32): def conv_bn_layer( input, ch_out, filter_size, stride, padding, act='relu', bias_attr=False ): - tmp = fluid.layers.conv2d( + tmp = paddle.static.nn.conv2d( input=input, filter_size=filter_size, num_filters=ch_out, diff --git a/python/paddle/fluid/tests/unittests/asp/asp_pruning_base.py b/python/paddle/fluid/tests/unittests/asp/asp_pruning_base.py index b092f3e16a..08c530241c 100644 --- a/python/paddle/fluid/tests/unittests/asp/asp_pruning_base.py +++ b/python/paddle/fluid/tests/unittests/asp/asp_pruning_base.py @@ -35,7 +35,7 @@ class TestASPHelperPruningBase(unittest.TestCase): name='img', shape=[None, 3, 32, 32], dtype='float32' ) label = fluid.data(name='label', shape=[None, 1], dtype='int64') - hidden = fluid.layers.conv2d( + hidden = paddle.static.nn.conv2d( input=img, num_filters=4, filter_size=3, padding=2, act="relu" ) hidden = fluid.layers.fc(input=hidden, size=32, act='relu') diff --git a/python/paddle/fluid/tests/unittests/asp/test_asp_customized_pruning.py b/python/paddle/fluid/tests/unittests/asp/test_asp_customized_pruning.py index 2e337ffef2..dc8527377b 100644 --- a/python/paddle/fluid/tests/unittests/asp/test_asp_customized_pruning.py +++ b/python/paddle/fluid/tests/unittests/asp/test_asp_customized_pruning.py @@ -202,7 +202,7 @@ class TestASPStaticCustomerizedPruneFunc(unittest.TestCase): name='img', shape=[None, 3, 32, 32], dtype='float32' ) label = fluid.data(name='label', shape=[None, 1], dtype='int64') - hidden = fluid.layers.conv2d( + hidden = paddle.static.nn.conv2d( input=img, num_filters=4, filter_size=3, padding=2, act="relu" ) hidden = fluid.layers.fc( diff --git a/python/paddle/fluid/tests/unittests/asp/test_asp_optimize_static.py b/python/paddle/fluid/tests/unittests/asp/test_asp_optimize_static.py index 99e63caa7f..d34129eced 100644 --- a/python/paddle/fluid/tests/unittests/asp/test_asp_optimize_static.py +++ b/python/paddle/fluid/tests/unittests/asp/test_asp_optimize_static.py @@ -35,7 +35,7 @@ class TestASPStaticOptimize(unittest.TestCase): name='img', shape=[None, 3, 24, 24], dtype='float32' ) label = fluid.data(name='label', shape=[None, 1], dtype='int64') - hidden = fluid.layers.conv2d( + hidden = paddle.static.nn.conv2d( input=img, num_filters=4, filter_size=3, padding=2, act="relu" ) hidden = fluid.layers.fc(input=hidden, size=32, act='relu') diff --git a/python/paddle/fluid/tests/unittests/asp/test_asp_pruning_static.py b/python/paddle/fluid/tests/unittests/asp/test_asp_pruning_static.py index 2951c1a7ff..82c2afc299 100644 --- a/python/paddle/fluid/tests/unittests/asp/test_asp_pruning_static.py +++ b/python/paddle/fluid/tests/unittests/asp/test_asp_pruning_static.py @@ -35,7 +35,7 @@ class TestASPStaticPruningBase(unittest.TestCase): name='img', shape=[None, 3, 24, 24], dtype='float32' ) label = fluid.data(name='label', shape=[None, 1], dtype='int64') - hidden = fluid.layers.conv2d( + hidden = paddle.static.nn.conv2d( input=img, num_filters=2, filter_size=3, padding=2, act="relu" ) hidden = fluid.layers.fc(input=hidden, size=32, act='softmax') diff --git a/python/paddle/fluid/tests/unittests/asp/test_asp_save_load.py b/python/paddle/fluid/tests/unittests/asp/test_asp_save_load.py index 57c77c15ba..30d3155788 100644 --- a/python/paddle/fluid/tests/unittests/asp/test_asp_save_load.py +++ b/python/paddle/fluid/tests/unittests/asp/test_asp_save_load.py @@ -136,7 +136,7 @@ class TestASPStaticOptimize(unittest.TestCase): name='img', shape=[None, 3, 32, 32], dtype='float32' ) label = fluid.data(name='label', shape=[None, 1], dtype='int64') - hidden = fluid.layers.conv2d( + hidden = paddle.static.nn.conv2d( input=img, num_filters=4, filter_size=3, padding=2, act="relu" ) hidden = fluid.layers.fc(input=hidden, size=32, act='relu') diff --git a/python/paddle/fluid/tests/unittests/dist_se_resnext.py b/python/paddle/fluid/tests/unittests/dist_se_resnext.py index 2b0867bc39..526b580f40 100644 --- a/python/paddle/fluid/tests/unittests/dist_se_resnext.py +++ b/python/paddle/fluid/tests/unittests/dist_se_resnext.py @@ -164,7 +164,7 @@ class SE_ResNeXt: def conv_bn_layer( self, input, num_filters, filter_size, stride=1, groups=1, act=None ): - conv = fluid.layers.conv2d( + conv = paddle.static.nn.conv2d( input=input, num_filters=num_filters, filter_size=filter_size, diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_ptb_lm_v2.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_ptb_lm_v2.py index 60712aeda7..b28aa1a1c2 100644 --- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_ptb_lm_v2.py +++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_ptb_lm_v2.py @@ -215,7 +215,7 @@ class PtbModel(paddle.nn.Layer): ) loss = paddle.reshape(loss, shape=[-1, self.num_steps]) loss = paddle.mean(loss, axis=[0]) - loss = paddle.paddle.sum(loss) + loss = paddle.sum(loss) return loss, last_hidden, last_cell diff --git a/python/paddle/fluid/tests/unittests/ipu/test_conv_op_ipu.py b/python/paddle/fluid/tests/unittests/ipu/test_conv_op_ipu.py index 48a5d0712d..a61059bec5 100644 --- a/python/paddle/fluid/tests/unittests/ipu/test_conv_op_ipu.py +++ b/python/paddle/fluid/tests/unittests/ipu/test_conv_op_ipu.py @@ -56,7 +56,7 @@ class TestBase(IPUOpTest): x = paddle.static.data( name=self.feed_list[0], shape=self.feed_shape[0], dtype='float32' ) - x = paddle.fluid.layers.conv2d(x, **self.attrs) + x = paddle.static.nn.conv2d(x, **self.attrs) self.fetch_list = [x.name] def run_model(self, exec_mode): diff --git a/python/paddle/fluid/tests/unittests/ipu/test_print_op_ipu.py b/python/paddle/fluid/tests/unittests/ipu/test_print_op_ipu.py index 27b012eb1b..529cb0e4c8 100644 --- a/python/paddle/fluid/tests/unittests/ipu/test_print_op_ipu.py +++ b/python/paddle/fluid/tests/unittests/ipu/test_print_op_ipu.py @@ -54,7 +54,7 @@ class TestBase(IPUOpTest): shape=self.feed_shape[0], dtype=self.feed_dtype[0], ) - out = paddle.fluid.layers.conv2d(x, num_filters=3, filter_size=3) + out = paddle.static.nn.conv2d(x, num_filters=3, filter_size=3) out = paddle.fluid.layers.Print(out, **self.attrs) if self.is_training: diff --git a/python/paddle/fluid/tests/unittests/ipu/test_reduce_x_op_ipu.py b/python/paddle/fluid/tests/unittests/ipu/test_reduce_x_op_ipu.py index 5aa8da4ada..7f9a4de041 100644 --- a/python/paddle/fluid/tests/unittests/ipu/test_reduce_x_op_ipu.py +++ b/python/paddle/fluid/tests/unittests/ipu/test_reduce_x_op_ipu.py @@ -134,7 +134,7 @@ class TestMin(TestMean): class TestSum(TestMean): def set_test_op(self): - self.op = paddle.paddle.sum + self.op = paddle.sum class TestLogsumexp(TestMean): diff --git a/python/paddle/fluid/tests/unittests/ir/inference/test_mkldnn_conv_activation_fuse_pass.py b/python/paddle/fluid/tests/unittests/ir/inference/test_mkldnn_conv_activation_fuse_pass.py index e4ff00d9cf..f8d6f69132 100644 --- a/python/paddle/fluid/tests/unittests/ir/inference/test_mkldnn_conv_activation_fuse_pass.py +++ b/python/paddle/fluid/tests/unittests/ir/inference/test_mkldnn_conv_activation_fuse_pass.py @@ -17,6 +17,7 @@ import unittest import numpy as np from inference_pass_test import InferencePassTest +import paddle import paddle.fluid as fluid from paddle.fluid.core import PassVersionChecker @@ -28,7 +29,7 @@ class ConvActivationMkldnnFusePassTest(InferencePassTest): data = fluid.data( name="data", shape=[-1, 3, 100, 100], dtype="float32" ) - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( data, num_filters=self.conv_num_filters, filter_size=self.conv_filter_size, diff --git a/python/paddle/fluid/tests/unittests/ir/inference/test_mkldnn_conv_bias_fuse_pass.py b/python/paddle/fluid/tests/unittests/ir/inference/test_mkldnn_conv_bias_fuse_pass.py index 922b84c6d2..1f32de177e 100644 --- a/python/paddle/fluid/tests/unittests/ir/inference/test_mkldnn_conv_bias_fuse_pass.py +++ b/python/paddle/fluid/tests/unittests/ir/inference/test_mkldnn_conv_bias_fuse_pass.py @@ -33,7 +33,7 @@ class ConvBiasMkldnnFusePassSamePadTest(InferencePassTest): initializer=fluid.initializer.Xavier(uniform=False), learning_rate=0.001, ) - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=data, num_filters=3, filter_size=3, @@ -66,7 +66,7 @@ class ConvBiasMkldnnFusePassValidPadTest(ConvBiasMkldnnFusePassSamePadTest): initializer=fluid.initializer.Xavier(uniform=False), learning_rate=0.001, ) - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=data, num_filters=3, filter_size=3, @@ -92,7 +92,7 @@ class ConvBiasMkldnnFusePassExplictPadTest(ConvBiasMkldnnFusePassSamePadTest): initializer=fluid.initializer.Xavier(uniform=False), learning_rate=0.001, ) - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=data, num_filters=3, filter_size=3, @@ -117,7 +117,7 @@ class ConvBiasMkldnnFusePassGroupTest(ConvBiasMkldnnFusePassSamePadTest): initializer=fluid.initializer.Xavier(uniform=False), learning_rate=0.001, ) - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=data, num_filters=3, filter_size=3, @@ -148,7 +148,7 @@ class ConvBiasMkldnnFusePassDialtionsGroupsTest( initializer=fluid.initializer.Xavier(uniform=False), learning_rate=0.001, ) - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=data, num_filters=3, filter_size=3, diff --git a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_conv_pass.py b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_conv_pass.py index 9a78d0929a..9b6ab8287f 100644 --- a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_conv_pass.py +++ b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_conv_pass.py @@ -33,7 +33,7 @@ class TensorRTSubgraphPassConvTest(InferencePassTest): data = fluid.data( name="data", shape=[-1, 6, 64, 64], dtype="float32" ) - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=data, num_filters=self.conv_num_filters, filter_size=self.conv_filter_size, @@ -210,7 +210,7 @@ class DynamicShapeTensorRTSubgraphPassConvTest(InferencePassTest): data = fluid.data( name="data", shape=[-1, 6, -1, -1], dtype="float32" ) - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=data, num_filters=self.conv_num_filters, filter_size=self.conv_filter_size, diff --git a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_conv_quant_dequant_pass.py b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_conv_quant_dequant_pass.py index 20b6068f23..b34e128ddf 100644 --- a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_conv_quant_dequant_pass.py +++ b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_conv_quant_dequant_pass.py @@ -35,7 +35,7 @@ class QuantDequantTensorRTSubgraphPassConvTest(QuantDequantTest): data_reshape = paddle.reshape(self.data, shape=[1, 4, 14, 14]) self.label = fluid.data(name='label', shape=[1, 1], dtype='int64') label_shape = paddle.reshape(self.label, shape=[1, 1, 1]) - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=data_reshape, num_filters=self.conv_num_filters, filter_size=self.conv_filter_size, @@ -150,7 +150,7 @@ class DynamicShapeQuantDequantTensorRTSubgraphPassConvTest(QuantDequantTest): data_reshape = paddle.reshape(self.data, shape=[1, 4, 14, 14]) self.label = fluid.data(name='label', shape=[1, 1], dtype='int64') label_shape = paddle.reshape(self.label, shape=[1, 1, 1]) - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=data_reshape, num_filters=self.conv_num_filters, filter_size=self.conv_filter_size, diff --git a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_dynamic_shape.py b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_dynamic_shape.py index 2e5a80c8fc..98ca955ee9 100644 --- a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_dynamic_shape.py +++ b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_dynamic_shape.py @@ -17,6 +17,7 @@ import unittest import numpy as np from inference_pass_test import InferencePassTest +import paddle import paddle.fluid as fluid import paddle.fluid.core as core from paddle.fluid.core import AnalysisConfig @@ -28,7 +29,7 @@ class TRTDynamicShapeTest(InferencePassTest): data = fluid.data( name="data", shape=[-1, 3, 16, 16], dtype="float32" ) - out = fluid.layers.conv2d( + out = paddle.static.nn.conv2d( input=data, num_filters=3, filter_size=3, diff --git a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_tuned_dynamic_shape.py b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_tuned_dynamic_shape.py index 1173f6e4ba..cb587a8a80 100644 --- a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_tuned_dynamic_shape.py +++ b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_tuned_dynamic_shape.py @@ -34,7 +34,7 @@ class TRTTunedDynamicShapeTest(unittest.TestCase): data = fluid.data( name="data", shape=[-1, 6, 64, 64], dtype="float32" ) - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=data, num_filters=3, filter_size=3, diff --git a/python/paddle/fluid/tests/unittests/mlu/sync_batch_norm_op_mlu.py b/python/paddle/fluid/tests/unittests/mlu/sync_batch_norm_op_mlu.py index 6412c4b5f5..20b3a00cc9 100644 --- a/python/paddle/fluid/tests/unittests/mlu/sync_batch_norm_op_mlu.py +++ b/python/paddle/fluid/tests/unittests/mlu/sync_batch_norm_op_mlu.py @@ -76,7 +76,7 @@ class TestSyncBatchNormOpTraining(TestSyncBatchNormRunnerBase): dtype=self.dtype, append_batch_size=False, ) - conv = fluid.layers.conv2d( + conv = paddle.static.nn.conv2d( input=data, num_filters=32, filter_size=1, diff --git a/python/paddle/fluid/tests/unittests/npu/sync_batch_norm_op_npu.py b/python/paddle/fluid/tests/unittests/npu/sync_batch_norm_op_npu.py index 49d4f92bdf..72c822ad3a 100644 --- a/python/paddle/fluid/tests/unittests/npu/sync_batch_norm_op_npu.py +++ b/python/paddle/fluid/tests/unittests/npu/sync_batch_norm_op_npu.py @@ -80,7 +80,7 @@ class TestSyncBatchNormOpTraining(TestSyncBatchNormRunnerBase): dtype=self.dtype, append_batch_size=False, ) - conv = fluid.layers.conv2d( + conv = paddle.static.nn.conv2d( input=data, num_filters=32, filter_size=1, diff --git a/python/paddle/fluid/tests/unittests/seresnext_net.py b/python/paddle/fluid/tests/unittests/seresnext_net.py index 4417df03aa..0c04d61121 100644 --- a/python/paddle/fluid/tests/unittests/seresnext_net.py +++ b/python/paddle/fluid/tests/unittests/seresnext_net.py @@ -68,7 +68,7 @@ def squeeze_excitation(input, num_channels, reduction_ratio): def conv_bn_layer( input, num_filters, filter_size, stride=1, groups=1, act=None ): - conv = fluid.layers.conv2d( + conv = paddle.static.nn.conv2d( input=input, num_filters=num_filters, filter_size=filter_size, diff --git a/python/paddle/fluid/tests/unittests/test_adam_op.py b/python/paddle/fluid/tests/unittests/test_adam_op.py index 9b71fc59f1..a8d41a3d85 100644 --- a/python/paddle/fluid/tests/unittests/test_adam_op.py +++ b/python/paddle/fluid/tests/unittests/test_adam_op.py @@ -613,7 +613,7 @@ class TestAdamOpV2(unittest.TestCase): with fluid.program_guard(train_prog, startup): with fluid.unique_name.guard(): data = fluid.data(name="data", shape=shape) - conv = fluid.layers.conv2d(data, 8, 3) + conv = paddle.static.nn.conv2d(data, 8, 3) loss = paddle.mean(conv) beta1 = paddle.static.create_global_var( diff --git a/python/paddle/fluid/tests/unittests/test_adamax_api.py b/python/paddle/fluid/tests/unittests/test_adamax_api.py index 1b1e658def..47c2791e46 100644 --- a/python/paddle/fluid/tests/unittests/test_adamax_api.py +++ b/python/paddle/fluid/tests/unittests/test_adamax_api.py @@ -52,7 +52,7 @@ class TestAdamaxAPI(unittest.TestCase): with fluid.program_guard(train_prog, startup): with fluid.unique_name.guard(): data = fluid.data(name="data", shape=shape) - conv = fluid.layers.conv2d(data, 8, 3) + conv = paddle.static.nn.conv2d(data, 8, 3) loss = paddle.mean(conv) beta1 = 0.85 beta2 = 0.95 diff --git a/python/paddle/fluid/tests/unittests/test_adamw_op.py b/python/paddle/fluid/tests/unittests/test_adamw_op.py index 59943cbeb3..acee63d1ae 100644 --- a/python/paddle/fluid/tests/unittests/test_adamw_op.py +++ b/python/paddle/fluid/tests/unittests/test_adamw_op.py @@ -209,7 +209,7 @@ class TestAdamWOp(unittest.TestCase): with fluid.program_guard(train_prog, startup): with fluid.unique_name.guard(): data = fluid.data(name="data", shape=shape) - conv = fluid.layers.conv2d(data, 8, 3) + conv = paddle.static.nn.conv2d(data, 8, 3) loss = paddle.mean(conv) beta1 = paddle.static.create_global_var( diff --git a/python/paddle/fluid/tests/unittests/test_backward.py b/python/paddle/fluid/tests/unittests/test_backward.py index 2409e7a1ad..e126a67bf5 100644 --- a/python/paddle/fluid/tests/unittests/test_backward.py +++ b/python/paddle/fluid/tests/unittests/test_backward.py @@ -285,7 +285,7 @@ class TestGradientsError(unittest.TestCase): def test_error(self): x = fluid.data(name='x', shape=[None, 2, 8, 8], dtype='float32') x.stop_gradient = False - conv = fluid.layers.conv2d(x, 4, 1, bias_attr=False) + conv = paddle.static.nn.conv2d(x, 4, 1, bias_attr=False) y = F.relu(conv) with self.assertRaises(TypeError): diff --git a/python/paddle/fluid/tests/unittests/test_conv2d_api.py b/python/paddle/fluid/tests/unittests/test_conv2d_api.py index ef1035d09c..16bf938d1c 100644 --- a/python/paddle/fluid/tests/unittests/test_conv2d_api.py +++ b/python/paddle/fluid/tests/unittests/test_conv2d_api.py @@ -40,7 +40,7 @@ class TestConv2DAPI(unittest.TestCase): dtype="float32", ) - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input_NHWC, num_filters=3, filter_size=[3, 3], @@ -51,7 +51,7 @@ class TestConv2DAPI(unittest.TestCase): data_format="NCHW", ) - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input_NCHW, num_filters=3, filter_size=[3, 3], @@ -62,7 +62,7 @@ class TestConv2DAPI(unittest.TestCase): data_format="NCHW", ) - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input_NCHW, num_filters=3, filter_size=[3, 3], @@ -73,7 +73,7 @@ class TestConv2DAPI(unittest.TestCase): data_format="NCHW", ) - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input_NHWC, num_filters=3, filter_size=[3, 3], @@ -84,7 +84,7 @@ class TestConv2DAPI(unittest.TestCase): data_format="NHWC", ) - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input_NCHW, num_filters=3, filter_size=[3, 3], @@ -95,7 +95,7 @@ class TestConv2DAPI(unittest.TestCase): data_format="NCHW", ) - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input_NCHW, num_filters=3, filter_size=[3, 3], @@ -129,7 +129,7 @@ class TestConv2DAPI_Error(unittest.TestCase): # ValueError: cudnn def run_1(): - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input, num_filters=3, filter_size=[3, 3], @@ -145,7 +145,7 @@ class TestConv2DAPI_Error(unittest.TestCase): # ValueError: data_format def run_2(): - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input, num_filters=3, filter_size=[3, 3], @@ -161,7 +161,7 @@ class TestConv2DAPI_Error(unittest.TestCase): # ValueError: padding def run_3(): - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input, num_filters=3, filter_size=[3, 3], @@ -176,7 +176,7 @@ class TestConv2DAPI_Error(unittest.TestCase): self.assertRaises(ValueError, run_3) def run_4(): - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input, num_filters=3, filter_size=[3, 3], @@ -191,7 +191,7 @@ class TestConv2DAPI_Error(unittest.TestCase): self.assertRaises(ValueError, run_4) def run_5(): - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input, num_filters=3, filter_size=[3, 3], @@ -214,7 +214,7 @@ class TestConv2DAPI_Error(unittest.TestCase): ) def run_6(): - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=x, num_filters=3, filter_size=[3, 3], @@ -230,7 +230,7 @@ class TestConv2DAPI_Error(unittest.TestCase): # ValueError: groups def run_7(): - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input, num_filters=3, filter_size=[3, 3], @@ -246,7 +246,7 @@ class TestConv2DAPI_Error(unittest.TestCase): # ValueError: filter num def run_8(): - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input, num_filters=0, filter_size=0, @@ -262,7 +262,7 @@ class TestConv2DAPI_Error(unittest.TestCase): # ValueError: groups def run_9(): - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input, num_filters=0, filter_size=0, @@ -278,7 +278,7 @@ class TestConv2DAPI_Error(unittest.TestCase): # ValueError: stride def run_10(): - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input, num_filters=1, filter_size=1, @@ -302,7 +302,7 @@ class TestConv2DAPI_Error(unittest.TestCase): # ValueError: cudnn def run_1(): - fluid.layers.conv2d( + paddle.static.nn.conv2d( input=input, num_filters=0, filter_size=0, @@ -331,7 +331,7 @@ class TestConv2DEnviron(unittest.TestCase): name="inputs", dtype="float32", ) - result = fluid.layers.conv2d( + result = paddle.static.nn.conv2d( input=inputs, num_filters=4, filter_size=[3, 3], diff --git a/python/paddle/fluid/tests/unittests/test_conv2d_layer.py b/python/paddle/fluid/tests/unittests/test_conv2d_layer.py index bdaaac3b8c..fea765e6f4 100644 --- a/python/paddle/fluid/tests/unittests/test_conv2d_layer.py +++ b/python/paddle/fluid/tests/unittests/test_conv2d_layer.py @@ -127,7 +127,7 @@ class Conv2DTestCase(unittest.TestCase): else: padding = self.padding - y_var = fluid.layers.conv2d( + y_var = paddle.static.nn.conv2d( x_var, self.num_filters, self.filter_size, diff --git a/python/paddle/fluid/tests/unittests/test_conv2d_op.py b/python/paddle/fluid/tests/unittests/test_conv2d_op.py index ac2947a64a..ca7017a4ad 100644 --- a/python/paddle/fluid/tests/unittests/test_conv2d_op.py +++ b/python/paddle/fluid/tests/unittests/test_conv2d_op.py @@ -706,7 +706,7 @@ class TestConv2DOpError(unittest.TestCase): x1 = fluid.create_lod_tensor( np.array([-1, 3, 5, 5]), [[1, 1, 1, 1]], fluid.CPUPlace() ) - fluid.layers.conv2d(x1, 1, 1) + paddle.static.nn.conv2d(x1, 1, 1) self.assertRaises(TypeError, test_Variable) @@ -716,7 +716,7 @@ class TestConv2DOpError(unittest.TestCase): x2 = fluid.layers.data( name='x2', shape=[3, 4, 5, 6], dtype="int32" ) - fluid.layers.conv2d(x2, 1, 1) + paddle.static.nn.conv2d(x2, 1, 1) self.assertRaises(TypeError, test_dtype) diff --git a/python/paddle/fluid/tests/unittests/test_conv_nn_grad.py b/python/paddle/fluid/tests/unittests/test_conv_nn_grad.py index 890fe3b10c..2bb99a00be 100644 --- a/python/paddle/fluid/tests/unittests/test_conv_nn_grad.py +++ b/python/paddle/fluid/tests/unittests/test_conv_nn_grad.py @@ -31,7 +31,7 @@ class TestConvDoubleGradCheck(unittest.TestCase): eps = 0.005 dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64 x = layers.data('x', shape, False, dtype) - y = layers.conv2d(x, 2, 1, groups=1, bias_attr=False) + y = paddle.static.nn.conv2d(x, 2, 1, groups=1, bias_attr=False) x_arr = np.random.uniform(-1, 1, shape).astype(dtype) w = fluid.default_main_program().global_block().all_parameters() @@ -58,7 +58,7 @@ class TestConvDoubleGradCheckTest0(unittest.TestCase): eps = 0.005 dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64 x = layers.data('x', shape, False, dtype) - y = layers.conv2d(x, 2, 1, bias_attr=False) + y = paddle.static.nn.conv2d(x, 2, 1, bias_attr=False) x_arr = np.random.uniform(-1, 1, shape).astype(dtype) w = fluid.default_main_program().global_block().all_parameters() @@ -84,7 +84,7 @@ class TestConvDoubleGradCheckTest1(unittest.TestCase): eps = 0.005 dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64 x = layers.data('x', shape, False, dtype) - y = layers.conv2d(x, 2, 1, padding=1, bias_attr=False) + y = paddle.static.nn.conv2d(x, 2, 1, padding=1, bias_attr=False) x_arr = np.random.uniform(-1, 1, shape).astype(dtype) w = fluid.default_main_program().global_block().all_parameters() @@ -163,7 +163,7 @@ class TestConv2DoubleGradCheck_AsyPadding(unittest.TestCase): eps = 0.005 dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64 x = layers.data('x', shape, False, dtype) - y = layers.conv2d( + y = paddle.static.nn.conv2d( input=x, num_filters=2, filter_size=1, @@ -196,7 +196,7 @@ class TestConv2DoubleGradCheck_PaddingSAME(unittest.TestCase): eps = 0.005 dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64 x = layers.data('x', shape, False, dtype) - y = layers.conv2d( + y = paddle.static.nn.conv2d( input=x, num_filters=2, filter_size=1, @@ -229,7 +229,7 @@ class TestConv2DoubleGradCheck_PaddingVALID(unittest.TestCase): eps = 0.005 dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64 x = layers.data('x', shape, False, dtype) - y = layers.conv2d( + y = paddle.static.nn.conv2d( input=x, num_filters=2, filter_size=1, @@ -262,7 +262,7 @@ class TestConv2DoubleGradCheck_ChannelLast(unittest.TestCase): eps = 0.005 dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64 x = layers.data('x', shape, False, dtype) - y = layers.conv2d( + y = paddle.static.nn.conv2d( input=x, num_filters=2, filter_size=1, @@ -297,7 +297,7 @@ class TestConv2DoubleGradCheck_ChannelLast_AsyPadding(unittest.TestCase): eps = 0.005 dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64 x = layers.data('x', shape, False, dtype) - y = layers.conv2d( + y = paddle.static.nn.conv2d( input=x, num_filters=2, filter_size=1, @@ -507,7 +507,7 @@ class TestDepthWiseConvDoubleGradCheck(unittest.TestCase): # use_cudnn == False # groups == filters # num_filters % num_channels == 0 - y = layers.conv2d( + y = paddle.static.nn.conv2d( x, shape[1], 1, groups=shape[1], bias_attr=False, use_cudnn=False ) x_arr = np.random.uniform(-1, 1, shape).astype(dtype) diff --git a/python/paddle/fluid/tests/unittests/test_functional_conv2d.py b/python/paddle/fluid/tests/unittests/test_functional_conv2d.py index 2e2dcd5baf..ce83ba62ac 100644 --- a/python/paddle/fluid/tests/unittests/test_functional_conv2d.py +++ b/python/paddle/fluid/tests/unittests/test_functional_conv2d.py @@ -89,7 +89,7 @@ class TestFunctionalConv2D(TestCase): (-1, self.in_channels, -1, -1), dtype=self.dtype, ) - y = fluid.layers.conv2d( + y = paddle.static.nn.conv2d( x, self.out_channels, self.filter_shape, @@ -507,7 +507,7 @@ class TestFunctionalConv2DErrorCase12(TestCase): with fluid.unique_name.guard(): with fluid.program_guard(main, start): x = fluid.data("input", self.input.shape, dtype=paddle.float32) - y = fluid.layers.conv2d( + y = paddle.static.nn.conv2d( x, self.num_filters, self.filter_size, diff --git a/python/paddle/fluid/tests/unittests/test_functional_conv2d_transpose.py b/python/paddle/fluid/tests/unittests/test_functional_conv2d_transpose.py index a66713eaee..c87a5c77d7 100644 --- a/python/paddle/fluid/tests/unittests/test_functional_conv2d_transpose.py +++ b/python/paddle/fluid/tests/unittests/test_functional_conv2d_transpose.py @@ -527,7 +527,7 @@ class TestFunctionalConv2DErrorCase10(TestCase): with fluid.unique_name.guard(): with fluid.program_guard(main, start): x = fluid.data("input", self.input.shape, dtype=paddle.float32) - y = fluid.layers.conv2d( + y = paddle.static.nn.conv2d( x, self.num_filters, self.filter_size, diff --git a/python/paddle/fluid/tests/unittests/test_fuse_bn_act_pass.py b/python/paddle/fluid/tests/unittests/test_fuse_bn_act_pass.py index 066443ebcf..7826880ac6 100644 --- a/python/paddle/fluid/tests/unittests/test_fuse_bn_act_pass.py +++ b/python/paddle/fluid/tests/unittests/test_fuse_bn_act_pass.py @@ -23,7 +23,7 @@ class TestFuseBatchNormActPass(unittest.TestCase): with fluid.program_guard(main_program, startup_program): x = fluid.layers.data(name='x', shape=[1, 28, 28], dtype='float32') y = fluid.layers.data(name="y", shape=[1], dtype='int64') - hidden1 = fluid.layers.conv2d( + hidden1 = paddle.static.nn.conv2d( input=x, filter_size=3, num_filters=16, diff --git a/python/paddle/fluid/tests/unittests/test_fuse_bn_add_act_pass.py b/python/paddle/fluid/tests/unittests/test_fuse_bn_add_act_pass.py index 503433df13..a193a81d6c 100644 --- a/python/paddle/fluid/tests/unittests/test_fuse_bn_add_act_pass.py +++ b/python/paddle/fluid/tests/unittests/test_fuse_bn_add_act_pass.py @@ -66,7 +66,7 @@ class TestFusedBnAddActAPI(unittest.TestCase): with fluid.program_guard(main_program, startup_program): x = fluid.layers.data(name='x', shape=[1, 28, 28], dtype='float32') y = fluid.layers.data(name="y", shape=[1], dtype='int64') - conv1_1 = fluid.layers.conv2d( + conv1_1 = paddle.static.nn.conv2d( input=x, filter_size=3, num_filters=32, @@ -77,7 +77,7 @@ class TestFusedBnAddActAPI(unittest.TestCase): bias_attr=False, data_format='NHWC', ) - conv1_2 = fluid.layers.conv2d( + conv1_2 = paddle.static.nn.conv2d( input=x, filter_size=3, num_filters=32, @@ -125,7 +125,7 @@ class TestFusedBnAddActAPI(unittest.TestCase): with fluid.program_guard(main_program, startup_program): x = fluid.layers.data(name='x', shape=[1, 28, 28], dtype='float32') y = fluid.layers.data(name="y", shape=[1], dtype='int64') - conv1_1 = fluid.layers.conv2d( + conv1_1 = paddle.static.nn.conv2d( input=x, filter_size=3, num_filters=32, @@ -143,7 +143,7 @@ class TestFusedBnAddActAPI(unittest.TestCase): act=None, data_layout='NHWC', ) - conv1_2 = fluid.layers.conv2d( + conv1_2 = paddle.static.nn.conv2d( input=conv1_1, filter_size=1, num_filters=32, diff --git a/python/paddle/fluid/tests/unittests/test_fuse_relu_depthwise_conv_pass.py b/python/paddle/fluid/tests/unittests/test_fuse_relu_depthwise_conv_pass.py index 649b61af7b..c074a1fe0d 100644 --- a/python/paddle/fluid/tests/unittests/test_fuse_relu_depthwise_conv_pass.py +++ b/python/paddle/fluid/tests/unittests/test_fuse_relu_depthwise_conv_pass.py @@ -29,7 +29,7 @@ def norm(*args, **kargs): def sep_conv(input, channel, stride, filter, dilation=1, act=None): # with scope('depthwise'): - input = fluid.layers.conv2d( + input = paddle.static.nn.conv2d( input, input.shape[1], filter, @@ -44,7 +44,7 @@ def sep_conv(input, channel, stride, filter, dilation=1, act=None): if act: input = act(input) # with scope('pointwise'): - input = fluid.layers.conv2d( + input = paddle.static.nn.conv2d( input, channel, 1, 1, groups=1, padding=0, bias_attr=False ) input = norm(input) diff --git a/python/paddle/fluid/tests/unittests/test_imperative_load_static_param.py b/python/paddle/fluid/tests/unittests/test_imperative_load_static_param.py index adc1b37770..dab14dbc22 100644 --- a/python/paddle/fluid/tests/unittests/test_imperative_load_static_param.py +++ b/python/paddle/fluid/tests/unittests/test_imperative_load_static_param.py @@ -34,10 +34,10 @@ class TestDygraphLoadStatic(unittest.TestCase): fc_out1 = fluid.layers.fc(a, 10) fc_out2 = fluid.layers.fc(a, 20) - conv_out_1 = fluid.layers.conv2d( + conv_out_1 = paddle.static.nn.conv2d( conv_in, num_filters=10, filter_size=5, act="relu" ) - conv_out_2 = fluid.layers.conv2d( + conv_out_2 = paddle.static.nn.conv2d( conv_in, num_filters=10, filter_size=5, act="relu" ) diff --git a/python/paddle/fluid/tests/unittests/test_inference_api.py b/python/paddle/fluid/tests/unittests/test_inference_api.py index dff2c83623..289146035b 100644 --- a/python/paddle/fluid/tests/unittests/test_inference_api.py +++ b/python/paddle/fluid/tests/unittests/test_inference_api.py @@ -80,7 +80,7 @@ def get_sample_model(): startup_program = fluid.Program() with fluid.program_guard(main_program, startup_program): data = fluid.data(name="data", shape=[-1, 6, 64, 64], dtype="float32") - conv_out = fluid.layers.conv2d( + conv_out = paddle.static.nn.conv2d( input=data, num_filters=3, filter_size=3, diff --git a/python/paddle/fluid/tests/unittests/test_initializer.py b/python/paddle/fluid/tests/unittests/test_initializer.py index 42d670a54d..bae17622e2 100644 --- a/python/paddle/fluid/tests/unittests/test_initializer.py +++ b/python/paddle/fluid/tests/unittests/test_initializer.py @@ -661,7 +661,7 @@ class TestSetGlobalInitializer(unittest.TestCase): with fluid.program_guard(main_prog, startup_prog): x = fluid.data(name="x", shape=[1, 3, 32, 32]) # default initilizer of param in layers.conv2d is NormalInitializer - conv = fluid.layers.conv2d(x, 5, 3) + conv = paddle.static.nn.conv2d(x, 5, 3) block = startup_prog.global_block() self.assertEqual(len(block.ops), 2) @@ -689,7 +689,7 @@ class TestSetGlobalInitializer(unittest.TestCase): with fluid.program_guard(main_prog, startup_prog): x = fluid.data(name="x", shape=[1, 3, 32, 32]) # default initilizer of bias in layers.conv2d is ConstantInitializer - conv = fluid.layers.conv2d(x, 5, 3) + conv = paddle.static.nn.conv2d(x, 5, 3) block = startup_prog.global_block() self.assertEqual(len(block.ops), 2) diff --git a/python/paddle/fluid/tests/unittests/test_layers.py b/python/paddle/fluid/tests/unittests/test_layers.py index a6af7fecff..f99f4fd71d 100644 --- a/python/paddle/fluid/tests/unittests/test_layers.py +++ b/python/paddle/fluid/tests/unittests/test_layers.py @@ -2805,7 +2805,7 @@ class TestBook(LayerTest): images = layers.data( name='pixel', shape=[3, 48, 48], dtype='float32' ) - return layers.conv2d( + return paddle.static.nn.conv2d( input=images, num_filters=3, filter_size=[4, 4] ) diff --git a/python/paddle/fluid/tests/unittests/test_parallel_executor_fetch_feed.py b/python/paddle/fluid/tests/unittests/test_parallel_executor_fetch_feed.py index 668297eaea..992ecbda46 100644 --- a/python/paddle/fluid/tests/unittests/test_parallel_executor_fetch_feed.py +++ b/python/paddle/fluid/tests/unittests/test_parallel_executor_fetch_feed.py @@ -25,10 +25,10 @@ from paddle.fluid import compiler def Lenet(data, class_dim): - conv1 = fluid.layers.conv2d(data, 4, 5, 1, act=None) + conv1 = paddle.static.nn.conv2d(data, 4, 5, 1, act=None) bn1 = paddle.static.nn.batch_norm(conv1, act='relu') pool1 = paddle.nn.functional.max_pool2d(bn1, 2, 2) - conv2 = fluid.layers.conv2d(pool1, 16, 5, 1, act=None) + conv2 = paddle.static.nn.conv2d(pool1, 16, 5, 1, act=None) bn2 = paddle.static.nn.batch_norm(conv2, act='relu') pool2 = paddle.nn.functional.max_pool2d(bn2, 2, 2) diff --git a/python/paddle/fluid/tests/unittests/test_set_value_op.py b/python/paddle/fluid/tests/unittests/test_set_value_op.py index 664a7c11d5..385269164c 100644 --- a/python/paddle/fluid/tests/unittests/test_set_value_op.py +++ b/python/paddle/fluid/tests/unittests/test_set_value_op.py @@ -1437,7 +1437,7 @@ class TestGradientTruncated(unittest.TestCase): # set_value_grad_op will not be run during backward. y, value = op(x) y2 = y + 1 - loss = paddle.paddle.sum(y2) + loss = paddle.sum(y2) sgd = paddle.optimizer.Adam() sgd.minimize(loss) place = ( diff --git a/python/paddle/fluid/tests/unittests/test_sync_batch_norm_op.py b/python/paddle/fluid/tests/unittests/test_sync_batch_norm_op.py index 10755fb729..1b8f912bb5 100644 --- a/python/paddle/fluid/tests/unittests/test_sync_batch_norm_op.py +++ b/python/paddle/fluid/tests/unittests/test_sync_batch_norm_op.py @@ -73,7 +73,7 @@ class TestSyncBatchNormOpTraining(unittest.TestCase): dtype=self.dtype, append_batch_size=False, ) - conv = fluid.layers.conv2d( + conv = paddle.static.nn.conv2d( input=data, num_filters=32, filter_size=1, diff --git a/python/paddle/fluid/tests/unittests/xpu/test_adamw_op_xpu.py b/python/paddle/fluid/tests/unittests/xpu/test_adamw_op_xpu.py index 63b838d1c6..5b628498f0 100644 --- a/python/paddle/fluid/tests/unittests/xpu/test_adamw_op_xpu.py +++ b/python/paddle/fluid/tests/unittests/xpu/test_adamw_op_xpu.py @@ -196,7 +196,7 @@ class XPUTestAdamwOp2(XPUOpTestWrapper): with fluid.program_guard(train_prog, startup): with fluid.unique_name.guard(): data = fluid.data(name="data", shape=shape) - conv = fluid.layers.conv2d(data, 8, 3) + conv = paddle.static.nn.conv2d(data, 8, 3) loss = paddle.mean(conv) beta1 = paddle.static.create_global_var( diff --git a/python/paddle/static/__init__.py b/python/paddle/static/__init__.py index 292a1da119..0692e2bc69 100644 --- a/python/paddle/static/__init__.py +++ b/python/paddle/static/__init__.py @@ -74,11 +74,15 @@ from ..fluid.io import batch # noqa: F401 from ..fluid.contrib.layers import ctr_metric_bundle # noqa: F401 from ..fluid.layers import exponential_decay # noqa: F401 +from .nn.common import batch_norm # noqa: F401 +from .nn.common import conv2d # noqa: F401 from .nn.metric import auc # noqa: F401 from .nn.metric import accuracy # noqa: F401 __all__ = [ # noqa 'append_backward', + 'batch_norm', + 'conv2d', 'gradients', 'Executor', 'global_scope', diff --git a/python/paddle/static/nn/__init__.py b/python/paddle/static/nn/__init__.py index d4b0d921be..4fc00892dd 100755 --- a/python/paddle/static/nn/__init__.py +++ b/python/paddle/static/nn/__init__.py @@ -19,6 +19,7 @@ from .common import data_norm # noqa: F401 from .common import continuous_value_model # noqa: F401 from .common import group_norm # noqa: F401 from .common import deform_conv2d # noqa: F401 +from .common import conv2d # noqa: F401 from .common import conv3d # noqa: F401 from .common import conv2d_transpose # noqa: F401 from .common import conv3d_transpose # noqa: F401 @@ -30,7 +31,6 @@ from .control_flow import ( from .common import bilinear_tensor_product # noqa: F401 from .common import py_func # noqa: F401 from ...tensor.creation import create_parameter # noqa: F401 -from ...fluid.layers import conv2d # noqa: F401 from ...fluid.layers import layer_norm # noqa: F401 from .loss import nce # noqa: F401 from .common import prelu # noqa: F401 diff --git a/python/paddle/static/nn/common.py b/python/paddle/static/nn/common.py index 687e4545e3..13a0595d4f 100644 --- a/python/paddle/static/nn/common.py +++ b/python/paddle/static/nn/common.py @@ -685,6 +685,324 @@ def group_norm( return helper.append_activation(group_norm_out) +def conv2d( + input, + num_filters, + filter_size, + stride=1, + padding=0, + dilation=1, + groups=None, + param_attr=None, + bias_attr=None, + use_cudnn=True, + act=None, + name=None, + data_format="NCHW", +): + r""" + The convolution2D layer calculates the output based on the input, filter + and strides, paddings, dilations, groups parameters. Input and + Output are in NCHW or NHWC format, where N is batch size, C is the number of + channels, H is the height of the feature, and W is the width of the feature. + Filter is in MCHW format, where M is the number of output image channels, + C is the number of input image channels, H is the height of the filter, + and W is the width of the filter. If the groups is greater than 1, + C will equal the number of input image channels divided by the groups. + Please refer to UFLDL's `convolution + `_ + for more details. + If bias attribution and activation type are provided, bias is added to the + output of the convolution, and the corresponding activation function is + applied to the final result. + + For each input :math:`X`, the equation is: + + .. math:: + + Out = \sigma (W \\ast X + b) + + Where: + + * :math:`X`: Input value, a tensor with NCHW or NHWC format. + * :math:`W`: Filter value, a tensor with MCHW format. + * :math:`\\ast`: Convolution operation. + * :math:`b`: Bias value, a 2-D tensor with shape [M, 1]. + * :math:`\\sigma`: Activation function. + * :math:`Out`: Output value, the shape of :math:`Out` and :math:`X` may be different. + + Example: + + - Input: + + Input shape: :math:`(N, C_{in}, H_{in}, W_{in})` + + Filter shape: :math:`(C_{out}, C_{in}, H_f, W_f)` + + - Output: + + Output shape: :math:`(N, C_{out}, H_{out}, W_{out})` + + Where + + .. math:: + + H_{out}&= \\frac{(H_{in} + 2 * paddings[0] - (dilations[0] * (H_f - 1) + 1))}{strides[0]} + 1 \\\\ + W_{out}&= \\frac{(W_{in} + 2 * paddings[1] - (dilations[1] * (W_f - 1) + 1))}{strides[1]} + 1 + + Args: + input (Tensor): The input is 4-D Tensor with shape [N, C, H, W], the data type + of input is float16 or float32 or float64. + num_filters(int): The number of filter. It is as same as the output + image channel. + filter_size (int|tuple): The filter size. If filter_size + is a tuple, it must contain two integers, (filter_size_height, + filter_size_width). Otherwise, filter_size_height = filter_size_width =\ + filter_size. + stride (int|tuple, optional): The stride size. It means the stride in convolution. + If stride is a tuple, it must contain two integers, (stride_height, stride_width). + Otherwise, stride_height = stride_width = stride. Default: stride = 1. + padding (string|int|list|tuple, optional): The padding size. It means the number of zero-paddings + on both sides for each dimension.If `padding` is a string, either 'VALID' or + 'SAME' which is the padding algorithm. If padding size is a tuple or list, + it could be in three forms: `[pad_height, pad_width]` or + `[pad_height_top, pad_height_bottom, pad_width_left, pad_width_right]`, and when + `data_format` is `"NCHW"`, `padding` can be in the form `[[0,0], [0,0], + [pad_height_top, pad_height_bottom], [pad_width_left, pad_width_right]]`. + when `data_format` is `"NHWC"`, `pool_padding` can be in the form + `[[0,0], [pad_height_top, pad_height_bottom], [pad_width_left, pad_width_right], [0,0]]`. + Default: padding = 0. + dilation (int|tuple, optional): The dilation size. It means the spacing between the kernel + points. If dilation is a tuple, it must contain two integers, (dilation_height, + dilation_width). Otherwise, dilation_height = dilation_width = dilation. + Default: dilation = 1. + groups (int, optional): The groups number of the Conv2d Layer. According to grouped + convolution in Alex Krizhevsky's Deep CNN paper: when group=2, + the first half of the filters is only connected to the first half + of the input channels, while the second half of the filters is only + connected to the second half of the input channels. Default: groups=1. + param_attr (ParamAttr|None, optional): The parameter attribute for learnable parameters/weights + of conv2d. If it is set to None or one attribute of ParamAttr, conv2d + will create ParamAttr as param_attr. If the Initializer of the param_attr + is not set, the parameter is initialized with :math:`Normal(0.0, std)`, + and the :math:`std` is :math:`(\\frac{2.0 }{filter\_elem\_num})^{0.5}`. Default: None. + bias_attr (ParamAttr|bool|None, optional): The parameter attribute for the bias of conv2d. + If it is set to False, no bias will be added to the output units. + If it is set to None or one attribute of ParamAttr, conv2d + will create ParamAttr as bias_attr. If the Initializer of the bias_attr + is not set, the bias is initialized zero. Default: None. + use_cudnn (bool, optional): Use cudnn kernel or not, it is valid only when the cudnn + library is installed. Default: True + act (str, optional): Activation type, if it is set to None, activation is not appended. + Default: None + name(str|None, optional): For detailed information, please refer + to :ref:`api_guide_Name`. Usually name is no need to set and + None by default. + data_format (str, optional): Specify the data format of the input, and the data format of the output + will be consistent with that of the input. An optional string from: `"NCHW"`, `"NHWC"`. + The default is `"NCHW"`. When it is `"NCHW"`, the data is stored in the order of: + `[batch_size, input_channels, input_height, input_width]`. + + Returns: + A Tensor representing the conv2d, whose data type is the + same with input. If act is None, the tensor storing the convolution + result, and if act is not None, the tensor storing convolution + and non-linearity activation result. + + Examples: + .. code-block:: python + + import paddle + paddle.enable_static() + + data = paddle.static.data(name='data', shape=[None, 3, 32, 32], dtype='float32') + conv2d = paddle.static.nn.conv2d(input=data, num_filters=2, filter_size=3, act="relu") + print(conv2d.shape) # [-1, 2, 30, 30] + """ + + check_variable_and_dtype( + input, 'input', ['float16', 'float32', 'float64'], 'conv2d' + ) + if len(input.shape) != 4: + raise ValueError( + "Input size should be 4, " + "but received {}".format(len(input.shape)) + ) + num_channels = input.shape[1] + if not isinstance(use_cudnn, bool): + raise ValueError( + "Attr(use_cudnn) should be True or False. Received " + "Attr(use_cudnn): %s. " % str(use_cudnn) + ) + + if data_format not in ["NCHW", "NHWC"]: + raise ValueError( + "Attr(data_format) should be 'NCHW' or 'NHWC'. Received " + "Attr(data_format): %s." % str(data_format) + ) + + channel_last = data_format == "NHWC" + num_channels = input.shape[3] if channel_last else input.shape[1] + if num_channels < 0: + raise ValueError( + "The channel dimmention of the input(%s) should be defined. " + "Received: %s." % (str(input.shape), str(num_channels)) + ) + assert param_attr is not False, "param_attr should not be False here." + + if groups is None: + num_filter_channels = num_channels + elif groups <= 0: + raise ValueError( + "the groups of input must be greater than 0, " + "but received the groups of input is {}".format(groups) + ) + else: + if num_channels % groups != 0: + raise ValueError( + "the channel of input must be divisible by groups," + "received: the channel of input is {}, the shape of input is {}" + ", the groups is {}".format(num_channels, input.shape, groups) + ) + num_filter_channels = num_channels // groups + + l_type = 'conv2d' + if ( + num_channels == groups + and num_filters % num_channels == 0 + and not use_cudnn + ): + l_type = 'depthwise_conv2d' + + if ( + num_channels == groups + and num_filters % num_channels == 0 + and core.is_compiled_with_rocm() + ): + l_type = 'depthwise_conv2d' + + # NPU only supports depthwise_conv2d when "input_channel = output_channel = groups" + if core.is_compiled_with_npu(): + if num_channels == groups and num_channels == num_filters: + l_type = 'depthwise_conv2d' + else: + l_type = 'conv2d' + + helper = LayerHelper(l_type, **locals()) + dtype = helper.input_dtype() + + filter_size = utils.convert_to_list(filter_size, 2, 'filter_size') + stride = utils.convert_to_list(stride, 2, 'stride') + dilation = utils.convert_to_list(dilation, 2, 'dilation') + + # padding + def _update_padding(padding, data_format): + def is_list_or_tuple(ele): + if isinstance(ele, list) or isinstance(ele, tuple): + return True + return False + + if is_list_or_tuple(padding) and len(padding) == 4: + if is_list_or_tuple(padding[0]) and (data_format == "NCHW"): + if not (padding[0] == [0, 0] and padding[1] == [0, 0]): + raise ValueError( + "Non-zero padding(%s) in the batch or channel dimensions " + "is not supported." % str(padding) + ) + padding = padding[2:4] + padding = [ele for a_list in padding for ele in a_list] + elif is_list_or_tuple(padding[0]) and (data_format == "NHWC"): + if not (padding[0] == [0, 0] and padding[3] == [0, 0]): + raise ValueError( + "Non-zero padding(%s) in the batch or channel dimensions " + "is not supported." % str(padding) + ) + padding = padding[1:3] + padding = [ele for a_list in padding for ele in a_list] + padding = utils.convert_to_list(padding, 4, 'padding') + if utils._is_symmetric_padding(padding, 2): + padding = [padding[0], padding[2]] + + else: + padding = utils.convert_to_list(padding, 2, 'padding') + + return padding + + padding_algorithm = "EXPLICIT" + if isinstance(padding, str): + padding = padding.upper() + if padding not in ["SAME", "VALID"]: + raise ValueError( + "Unknown padding: '%s'. It can only be 'SAME' or 'VALID'." + % str(padding) + ) + if padding == "VALID": + padding_algorithm = "VALID" + padding = [0, 0] + elif padding == "SAME": + padding_algorithm = "SAME" + padding = [0, 0] + + padding = _update_padding(padding, data_format) + + filter_shape = [num_filters, int(num_filter_channels)] + filter_size + + def _get_default_param_initializer(): + filter_elem_num = filter_size[0] * filter_size[1] * num_channels + if filter_elem_num <= 0: + raise ValueError( + "Invalid filter number, excepted number is larger than 0, but" + " received {}, please check the input shape and " + "filter size.".format(filter_elem_num) + ) + std = (2.0 / filter_elem_num) ** 0.5 + return Normal(0.0, std, 0) + + filter_param = helper.create_parameter( + attr=helper.param_attr, + shape=filter_shape, + dtype=dtype, + default_initializer=_get_default_param_initializer(), + ) + + pre_bias = helper.create_variable_for_type_inference(dtype) + + if ( + core.is_compiled_with_cuda() + and paddle.fluid.get_flags("FLAGS_conv2d_disable_cudnn")[ + "FLAGS_conv2d_disable_cudnn" + ] + ): + use_cudnn = False + + helper.append_op( + type=l_type, + inputs={ + 'Input': input, + 'Filter': filter_param, + }, + outputs={"Output": pre_bias}, + attrs={ + 'strides': stride, + 'paddings': padding, + 'dilations': dilation, + 'groups': groups, + 'use_cudnn': use_cudnn, + 'use_mkldnn': False, + 'fuse_relu_before_depthwise_conv': False, + "padding_algorithm": padding_algorithm, + "data_format": data_format, + }, + ) + + if data_format == 'NCHW': + pre_act = helper.append_bias_op(pre_bias, dim_start=1, dim_end=2) + else: + pre_act = helper.append_bias_op(pre_bias, dim_start=3, dim_end=4) + + return helper.append_activation(pre_act) + + def conv3d( input, num_filters, -- GitLab