尝试为GatedConv2D和GatedDeConv2D实现自定义图层时，出现此错误。 生成器：
inp = Input(shape=(self.vars.INP_SHAPE[0], self.vars.INP_SHAPE[1], 9))
cnum = 64
x1, mask1 = self.GatedConv2D(inp, cnum, (7, 7), (2,2), use_lrn=False)
x2, mask2 = self.GatedConv2D(x1, 2*cnum, (5, 5), (2, 2))
x3, mask3 = self.GatedConv2D(x2, 4*cnum, (5, 5), (2, 2))
x4, mask4 = self.GatedConv2D(x3, 8*cnum, (3, 3), (2, 2))
x5, mask5 = self.GatedConv2D(x4, 8*cnum, (3, 3), (2, 2))
x6, mask6 = self.GatedConv2D(x5, 8*cnum, (3, 3), (2, 2))
x7, mask7 = self.GatedConv2D(x6, 8*cnum, (3, 3), (2, 2))
x7, _ = self.GatedConv2D(x7, 8*cnum, (3, 3), (1, 1), dilation=2)
x7, _ = self.GatedConv2D(x7, 8*cnum, (3, 3), (1, 1), dilation=4)
x7, _ = self.GatedConv2D(x7, 8*cnum, (3, 3), (1, 1), dilation=8)
x7, _ = self.GatedConv2D(x7, 8*cnum, (3, 3), (1, 1), dilation=16)
x8, _ = self.GatedDeConv2D(x7, [self.vars.SCFEGAN_BATCH_SIZE, int(self.vars.INP_SHAPE[0]/64), int(self.vars.INP_SHAPE[1]/64), 8*cnum])
x8 = Concatenate(axis=0)([x6, x8])
x8, mask8 = self.GatedConv2D(x8, 8*cnum, (3, 3), (1, 1))
x9, _ = self.GatedDeConv2D(x8, [self.vars.SCFEGAN_BATCH_SIZE, int(self.vars.INP_SHAPE[0]/32), int(self.vars.INP_SHAPE[1]/32), 8*cnum])
x9 = Concatenate(axis=0)([x5, x9])
x9, mask9 = self.GatedConv2D(x9, 8*cnum, (3, 3), (1, 1))
x10, _ = self.GatedDeConv2D(x9, [self.vars.SCFEGAN_BATCH_SIZE, int(self.vars.INP_SHAPE[0]/16), int(self.vars.INP_SHAPE[1]/16), 8*cnum])
x10 = Concatenate(axis=0)([x4, x10])
x10, mask10 = self.GatedConv2D(x10, 8*cnum, (3, 3), (1, 1))
x11, _ = self.GatedDeConv2D(x10, [self.vars.SCFEGAN_BATCH_SIZE, int(self.vars.INP_SHAPE[0]/8), int(self.vars.INP_SHAPE[1]/8), 4*cnum])
x11 = Concatenate(axis=0)([x3, x11])
x11, mask11 = self.GatedConv2D(x11, 4*cnum, (3, 3), (1, 1))
x12, _ = self.GatedDeConv2D(x11, [self.vars.SCFEGAN_BATCH_SIZE, int(self.vars.INP_SHAPE[0]/4), int(self.vars.INP_SHAPE[1]/4), 2*cnum])
x12 = Concatenate(axis=0)([x2, x12])
x12, mask12 = self.GatedConv2D(x12, 2*cnum, (3, 3), (1, 1))
x13, _ = self.GatedDeConv2D(x12, [self.vars.SCFEGAN_BATCH_SIZE, int(self.vars.INP_SHAPE[0]/2), int(self.vars.INP_SHAPE[1]/2), cnum])
x13 = Concatenate(axis=0)([x1, x13])
x13, mask13 = self.GatedConv2D(x13, cnum, (3, 3), (1, 1))
x14, _ = self.GatedDeConv2D(x13, [self.vars.SCFEGAN_BATCH_SIZE, int(self.vars.INP_SHAPE[0]), int(self.vars.INP_SHAPE[1]), 9])
x14 = Concatenate(axis=0)([inp, x14])
x14, mask14 = self.GatedConv2D(x14, 3, (3, 3), (1, 1))
x14 = Activation('tanh')(x14)
model = Model(inputs=inp, outputs=[x14, mask14])
GatedConv2D：
def GatedConv2D(self, x, filters, kernel_size, strides, dilation=1, activation='leaky_relu', use_lrn=True):
inp = x
x = Conv2D(filters=filters, kernel_size=kernel_size, strides=strides, dilation_rate=dilation, padding='same')(x)
if use_lrn:
x = LRNLayer()(x)
if activation == 'leaky_relu':
x = LeakyReLU()(x)
else:
x = Activation(activation)('x')
g = Conv2D(filters=filters, kernel_size=kernel_size, strides=strides, padding='same', dilation_rate=dilation)(inp)
g = Activation('sigmoid')(g)
x = multiply([x, g])
return x, g
GatedDeConv2D：
class GatedDeConv(Layer):
def __init__(self, out_shape, kernel_size, strides, std_dev):
super(GatedDeConv, self).__init__()
self.out_shape = out_shape
self.kernel_size = kernel_size
self.strides = strides
self.std_dev = std_dev
def call(self, x):
inp = x
kernel = K.random_uniform_variable(shape=(self.kernel_size[0], self.kernel_size[1], self.out_shape[-1], int(x.get_shape()[-1])), low=0, high=1)
deconv = K.conv2d_transpose(x, kernel=kernel, strides=self.strides, output_shape=self.out_shape, padding='same')
biases = K.zeros(shape=(self.out_shape[-1]))
deconv = K.reshape(K.bias_add(deconv, biases), deconv.get_shape())
deconv = LeakyReLU()(deconv)
g = K.conv2d_transpose(inp, kernel, output_shape=self.out_shape, strides=self.strides, padding='same')
biases2 = K.zeros(shape=(self.out_shape[-1]))
g = K.reshape(K.bias_add(g, biases2), deconv.get_shape())
g = K.sigmoid(g)
deconv = tf.multiply(deconv, g)
outputs = [deconv, g]
output_shapes = self.compute_output_shape(x.shape)
for output, shape in zip(outputs, output_shapes):
output._keras_shape = shape
return [deconv, g]
def compute_output_shape(self, input_shape):
return [self.out_shape, self.out_shape]
def compute_mask(self, input, input_mask=None):
return 2 * [None]
LRNLayer：
class LRNLayer(Layer):
def __init__(self, alpha=1e-4, beta=0.75, k=2, n=5):
super(LRNLayer, self).__init__()
self.alpha = alpha
self.beta = beta
self.k = k
self.n = n
def call(self, x):
op = []
nc = np.shape(x)[-1]
for i in range(nc):
sq = K.sum((x[:,:,:,max(0, int(i-self.n/2)):min(nc-1, i+int(self.n/2))+1]) ** 2)
op.append(x[:,:,:,i]/((self.k + self.alpha * sq) ** self.beta))
op = tf.convert_to_tensor(op)
op = tf.transpose(op, perm=[1,2,3,0])
op_shape = self.compute_output_shape(np.shape(x))
op._keras_shape = op_shape
return op
def compute_output_shape(self, input_shape):
return input_shape
def compute_mask(self, input, input_mask):
return 1*[None]
鉴别器：
inp = Input(shape=self.vars.INP_SHAPE)
x = Conv2D(filters=64, kernel_size=(3, 3), strides=(2, 2), activation='relu')(inp)
x = ZeroPadding2D(padding=(1, 1))(x)
x = Conv2D(filters=128, kernel_size=(3, 3), strides=(2, 2), activation='relu')(x)
x = ZeroPadding2D(padding=(1, 1))(x)
x = Conv2D(filters=256, kernel_size=(3, 3), strides=(2, 2), activation='relu')(x)
x = ZeroPadding2D(padding=(1, 1))(x)
x = Conv2D(filters=256, kernel_size=(3, 3), strides=(2, 2), activation='relu')(x)
x = ZeroPadding2D(padding=(1, 1))(x)
x = Conv2D(filters=256, kernel_size=(3, 3), strides=(2, 2), activation='relu')(x)
x = ZeroPadding2D(padding=(1, 1))(x)
x = Conv2D(filters=256, kernel_size=(3, 3), strides=(2, 2), activation='relu')(x)
model = Model(inputs=inp, outputs=x)
错误：
---------------------------------------------------------------------------
InvalidArgumentError Traceback (most recent call last)
<ipython-input-14-f1979e2d5f34> in <module>()
1 listdir('/var')
----> 2 model.train()
3 # except:
4 # print('Failed')
5 get_ipython().system('cat /var/log/colab-jupyter.log')
<ipython-input-9-790e38d234ac> in train(self)
332 fakes = np.zeros((self.vars.SCFEGAN_BATCH_SIZE, *self.vars.SCFEGAN_DISC_OP_SIZE))
333
--> 334 gen_imgs = self.generator.predict(inp)
335
336 cmp_images = self.complete_imgs(images, masks, gen_imgs)
/usr/local/lib/python3.6/dist-packages/keras/engine/training.py in predict(self, x, batch_size, verbose, steps)
1167 batch_size=batch_size,
1168 verbose=verbose,
-> 1169 steps=steps)
1170
1171 def train_on_batch(self, x, y,
/usr/local/lib/python3.6/dist-packages/keras/engine/training_arrays.py in predict_loop(model, f, ins, batch_size, verbose, steps)
292 ins_batch[i] = ins_batch[i].toarray()
293
--> 294 batch_outs = f(ins_batch)
295 batch_outs = to_list(batch_outs)
296 if batch_index == 0:
/usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py in __call__(self, inputs)
2713 return self._legacy_call(inputs)
2714
-> 2715 return self._call(inputs)
2716 else:
2717 if py_any(is_tensor(x) for x in inputs):
/usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py in _call(self, inputs)
2673 fetched = self._callable_fn(*array_vals, run_metadata=self.run_metadata)
2674 else:
-> 2675 fetched = self._callable_fn(*array_vals)
2676 return fetched[:len(self.outputs)]
2677
/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py in __call__(self, *args, **kwargs)
1437 ret = tf_session.TF_SessionRunCallable(
1438 self._session._session, self._handle, args, status,
-> 1439 run_metadata_ptr)
1440 if run_metadata:
1441 proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/errors_impl.py in __exit__(self, type_arg, value_arg, traceback_arg)
526 None, None,
527 compat.as_text(c_api.TF_Message(self.status.status)),
--> 528 c_api.TF_GetCode(self.status.status))
529 # Delete the underlying status object from memory otherwise it stays alive
530 # as there is a reference to status from this from the traceback due to
InvalidArgumentError: Conv2DCustomBackpropInput: input and out_backprop must have the same batch sizeinput batch: 1outbackprop batch: 2 batch_dim: 0
[[{{node gated_de_conv_8/conv2d_transpose_1}}]]
我没有找到针对我的情况的任何解决方案。 提前致谢。 我正在尝试实现SCFEGAN模型并面临此错误。 Tensorflow版本：1.13.1 Keras版本：2.2.4 Python版本：3.6