為什么這會給我 InvalidArgumentError？

Question

我正在嘗試在 keras 中構建以下堆疊沙漏模型

def create_hourglass_network(num_classes, num_stacks, num_channels, inres, outres, bottleneck):
    input = Input(shape=(inres[0], inres[1], 3))

    front_features = create_front_module(input, num_channels, bottleneck)

    head_next_stage = front_features

    outputs = []
    for i in range(num_stacks):
        head_next_stage, head_to_loss = hourglass_module(head_next_stage, num_classes, num_channels, bottleneck, i)
        outputs.append(head_to_loss)

    model = Model(inputs=input, outputs=outputs)
    rms = RMSprop(lr=5e-4)
    model.compile(optimizer=rms, loss=mean_squared_error, metrics=["accuracy"])

    return model


def hourglass_module(bottom, num_classes, num_channels, bottleneck, hgid):
    # create left features , f1, f2, f4, and f8
    left_features = create_left_half_blocks(bottom, bottleneck, hgid, num_channels)

    # create right features, connect with left features
    rf1 = create_right_half_blocks(left_features, bottleneck, hgid, num_channels)

    # add 1x1 conv with two heads, head_next_stage is sent to next stage
    # head_parts is used for intermediate supervision
    head_next_stage, head_parts = create_heads(bottom, rf1, num_classes, hgid, num_channels)

    return head_next_stage, head_parts


def bottleneck_block(bottom, num_out_channels, block_name):
    # skip layer
    if K.int_shape(bottom)[-1] == num_out_channels:
        _skip = bottom
    else:
        _skip = Conv2D(num_out_channels, kernel_size=(1, 1), activation='relu', padding='same',
                       name=block_name + 'skip')(bottom)

    # residual: 3 conv blocks,  [num_out_channels/2  -> num_out_channels/2 -> num_out_channels]
    _x = Conv2D(num_out_channels / 2, kernel_size=(1, 1), activation='relu', padding='same',
                name=block_name + '_conv_1x1_x1')(bottom)
    _x = BatchNormalization()(_x)
    _x = Conv2D(num_out_channels / 2, kernel_size=(3, 3), activation='relu', padding='same',
                name=block_name + '_conv_3x3_x2')(_x)
    _x = BatchNormalization()(_x)
    _x = Conv2D(num_out_channels, kernel_size=(1, 1), activation='relu', padding='same',
                name=block_name + '_conv_1x1_x3')(_x)
    _x = BatchNormalization()(_x)
    _x = Add(name=block_name + '_residual')([_skip, _x])

    return _x


def bottleneck_mobile(bottom, num_out_channels, block_name):
    # skip layer
    if K.int_shape(bottom)[-1] == num_out_channels:
        _skip = bottom
    else:
        _skip = SeparableConv2D(num_out_channels, kernel_size=(1, 1), activation='relu', padding='same',
                                name=block_name + 'skip')(bottom)

    # residual: 3 conv blocks,  [num_out_channels/2  -> num_out_channels/2 -> num_out_channels]
    _x = SeparableConv2D(num_out_channels / 2, kernel_size=(1, 1), activation='relu', padding='same',
                         name=block_name + '_conv_1x1_x1')(bottom)
    _x = BatchNormalization()(_x)
    _x = SeparableConv2D(num_out_channels / 2, kernel_size=(3, 3), activation='relu', padding='same',
                         name=block_name + '_conv_3x3_x2')(_x)
    _x = BatchNormalization()(_x)
    _x = SeparableConv2D(num_out_channels, kernel_size=(1, 1), activation='relu', padding='same',
                         name=block_name + '_conv_1x1_x3')(_x)
    _x = BatchNormalization()(_x)
    _x = Add(name=block_name + '_residual')([_skip, _x])

    return _x


def create_front_module(input, num_channels, bottleneck):
    # front module, input to 1/4 resolution
    # 1 7x7 conv + maxpooling
    # 3 residual block

    _x = Conv2D(64, kernel_size=(7, 7), strides=(2, 2), padding='same', activation='relu', name='front_conv_1x1_x1')(
        input)
    _x = BatchNormalization()(_x)

    _x = bottleneck(_x, num_channels // 2, 'front_residual_x1')
    _x = MaxPooling2D(pool_size=(2, 2), strides=(2, 2))(_x)

    _x = bottleneck(_x, num_channels // 2, 'front_residual_x2')
    _x = bottleneck(_x, num_channels, 'front_residual_x3')

    return _x


def create_left_half_blocks(bottom, bottleneck, hglayer, num_channels):
    # create left half blocks for hourglass module
    # f1, f2, f4 , f8 : 1, 1/2, 1/4 1/8 resolution

    hgname = 'hg' + str(hglayer)

    f1 = bottleneck(bottom, num_channels, hgname + '_l1')
    _x = MaxPooling2D(pool_size=(2, 2), strides=(2, 2))(f1)

    f2 = bottleneck(_x, num_channels, hgname + '_l2')
    _x = MaxPooling2D(pool_size=(2, 2), strides=(2, 2))(f2)

    f4 = bottleneck(_x, num_channels, hgname + '_l4')
    _x = MaxPooling2D(pool_size=(2, 2), strides=(2, 2))(f4)

    f8 = bottleneck(_x, num_channels, hgname + '_l8')

    return (f1, f2, f4, f8)


def connect_left_to_right(left, right, bottleneck, name, num_channels):
    '''
    :param left: connect left feature to right feature
    :param name: layer name
    :return:
    '''
    # left -> 1 bottlenect
    # right -> upsampling
    # Add   -> left + right

    _xleft = bottleneck(left, num_channels, name + '_connect')
    _xright = UpSampling2D()(right)
    add = Add()([_xleft, _xright])
    out = bottleneck(add, num_channels, name + '_connect_conv')
    return out


def bottom_layer(lf8, bottleneck, hgid, num_channels):
    # blocks in lowest resolution
    # 3 bottlenect blocks + Add

    lf8_connect = bottleneck(lf8, num_channels, str(hgid) + "_lf8")

    _x = bottleneck(lf8, num_channels, str(hgid) + "_lf8_x1")
    _x = bottleneck(_x, num_channels, str(hgid) + "_lf8_x2")
    _x = bottleneck(_x, num_channels, str(hgid) + "_lf8_x3")

    rf8 = Add()([_x, lf8_connect])

    return rf8


def create_right_half_blocks(leftfeatures, bottleneck, hglayer, num_channels):
    lf1, lf2, lf4, lf8 = leftfeatures

    rf8 = bottom_layer(lf8, bottleneck, hglayer, num_channels)

    rf4 = connect_left_to_right(lf4, rf8, bottleneck, 'hg' + str(hglayer) + '_rf4', num_channels)

    rf2 = connect_left_to_right(lf2, rf4, bottleneck, 'hg' + str(hglayer) + '_rf2', num_channels)

    rf1 = connect_left_to_right(lf1, rf2, bottleneck, 'hg' + str(hglayer) + '_rf1', num_channels)

    return rf1


def create_heads(prelayerfeatures, rf1, num_classes, hgid, num_channels):
    # two head, one head to next stage, one head to intermediate features
    head = Conv2D(num_channels, kernel_size=(1, 1), activation='relu', padding='same', name=str(hgid) + '_conv_1x1_x1')(
        rf1)
    head = BatchNormalization()(head)

    # for head as intermediate supervision, use 'linear' as activation.
    head_parts = Conv2D(num_classes, kernel_size=(1, 1), activation='linear', padding='same',
                        name=str(hgid) + '_conv_1x1_parts')(head)

    # use linear activation
    head = Conv2D(num_channels, kernel_size=(1, 1), activation='linear', padding='same',
                  name=str(hgid) + '_conv_1x1_x2')(head)
    head_m = Conv2D(num_channels, kernel_size=(1, 1), activation='linear', padding='same',
                    name=str(hgid) + '_conv_1x1_x3')(head_parts)

    head_next_stage = Add()([head, head_m, prelayerfeatures])
    return head_next_stage, head_parts

我正在使用以下方法構建模型：

model=create_hourglass_network(19,2,256,(256,256),(64,64),bottleneck_block)

我收到以下錯誤：

InvalidArgumentError: Value for attr 'T' of float is not in the list of allowed values: int32, int64

; NodeDef: {{node RandomUniform}}; Op<name=RandomUniform; signature=shape:T -> output:dtype; attr=seed:int,default=0; attr=seed2:int,default=0; attr=dtype:type,allowed=[DT_HALF, DT_BFLOAT16, DT_FLOAT, DT_DOUBLE]; attr=T:type,allowed=[DT_INT32, DT_INT64]; is_stateful=true> [Op:RandomUniform]

這究竟是指什么？ 將bottleneck_block 作為爭論傳遞是否有問題？ 如果您有任何想法，他們將不勝感激！

這是調用 Bottle_neck 時的完整堆棧跟蹤：

File "<ipython-input-11-2bee0ad641ac>", line 1, in <module>
    out=bottleneck_block(np.zeros((4,256,256,3)),256,"blah")

  File "C:\Users\alecd\untitled2.py", line 65, in bottleneck_block

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\keras\engine\base_layer.py", line 887, in __call__
    self._maybe_build(inputs)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\keras\engine\base_layer.py", line 2141, in _maybe_build
    self.build(input_shapes)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\keras\layers\convolutional.py", line 165, in build
    dtype=self.dtype)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\keras\engine\base_layer.py", line 522, in add_weight
    aggregation=aggregation)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\training\tracking\base.py", line 744, in _add_variable_with_custom_getter
    **kwargs_for_getter)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\keras\engine\base_layer_utils.py", line 139, in make_variable
    shape=variable_shape if variable_shape else None)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\ops\variables.py", line 258, in __call__
    return cls._variable_v1_call(*args, **kwargs)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\ops\variables.py", line 219, in _variable_v1_call
    shape=shape)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\ops\variables.py", line 197, in <lambda>
    previous_getter = lambda **kwargs: default_variable_creator(None, **kwargs)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\ops\variable_scope.py", line 2507, in default_variable_creator
    shape=shape)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\ops\variables.py", line 262, in __call__
    return super(VariableMetaclass, cls).__call__(*args, **kwargs)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\ops\resource_variable_ops.py", line 1406, in __init__
    distribute_strategy=distribute_strategy)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\ops\resource_variable_ops.py", line 1537, in _init_from_args
    initial_value() if init_from_fn else initial_value,

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\keras\engine\base_layer_utils.py", line 119, in <lambda>
    init_val = lambda: initializer(shape, dtype=dtype)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\ops\init_ops_v2.py", line 437, in __call__
    return self._random_generator.random_uniform(shape, -limit, limit, dtype)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\ops\init_ops_v2.py", line 800, in random_uniform
    shape=shape, minval=minval, maxval=maxval, dtype=dtype, seed=self.seed)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\ops\random_ops.py", line 245, in random_uniform
    rnd = gen_random_ops.random_uniform(shape, dtype, seed=seed1, seed2=seed2)

  File "C:\Users\alecd\Anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\ops\gen_random_ops.py", line 812, in random_uniform
    _six.raise_from(_core._status_to_exception(e.code, message), None)

  File "<string>", line 3, in raise_from

InvalidArgumentError: Value for attr 'T' of float is not in the list of allowed values: int32, int64
    ; NodeDef: {{node RandomUniform}}; Op<name=RandomUniform; signature=shape:T -> output:dtype; attr=seed:int,default=0; attr=seed2:int,default=0; attr=dtype:type,allowed=[DT_HALF, DT_BFLOAT16, DT_FLOAT, DT_DOUBLE]; attr=T:type,allowed=[DT_INT32, DT_INT64]; is_stateful=true> [Op:RandomUniform]

Answer 1

我在 colab 中運行了你的代碼，但我沒有遇到像你一樣的問題。 你可以在這里查看我的筆記本。 由於版本原因，可能是您的python環境錯誤。 您可以使用自己的包檢查 colab 版本。 在 colab tensorflow 和 keras 版本中給出，

另外，我在您的 tf & keras 版本中實現了它並解決了您的問題。 你可以在這里檢查。 希望，您的問題已解決。 您的問題是您使用的是 num_out_channels / 2 ，您應該使用 num_out_channels // 2 （整數不是浮點數）。