【发布时间】：2020-02-28 05:44:13
【问题描述】：
我已经使用 keras-resnet 0.2.0 (python3) 创建了一个 ResNet1D 模型，并且在许多时期都没有任何问题地拟合了我的数据，但是在保存之后，然后只需尝试重新读取模型（通过 load_model） ，我得到一个张量形状不匹配错误：
Traceback (most recent call last):
File "/usr/local/lib/python3.6/site-packages/tensorflow_core/python/framework/ops.py", line 1607, in _create_c_op
c_op = c_api.TF_FinishOperation(op_desc)
tensorflow.python.framework.errors_impl.InvalidArgumentError: Shape must be rank 3 but is rank 4 for 'padding_conv1_2/Pad' (op: 'Pad') with input shapes: [1,?,30,2], [3,2].
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "<stdin>", line 2, in <module>
File "/usr/local/lib/python3.6/site-packages/keras/engine/saving.py", line 661, in model_from_json
return deserialize(config, custom_objects=custom_objects)
File "/usr/local/lib/python3.6/site-packages/keras/layers/__init__.py", line 168, in deserialize
printable_module_name='layer')
File "/usr/local/lib/python3.6/site-packages/keras/utils/generic_utils.py", line 147, in deserialize_keras_object
list(custom_objects.items())))
File "/usr/local/lib/python3.6/site-packages/keras/engine/network.py", line 1107, in from_config
return cls(inputs=input_tensors, outputs=output_tensors, name=name)
File "/usr/local/lib/python3.6/site-packages/keras_resnet/models/_1d.py", line 184, in __init__
**kwargs
File "/usr/local/lib/python3.6/site-packages/keras_resnet/models/_1d.py", line 82, in __init__
x = keras.layers.ZeroPadding1D(padding=3, name="padding_conv1")(inputs)
File "/usr/local/lib/python3.6/site-packages/keras/engine/base_layer.py", line 489, in __call__
output = self.call(inputs, **kwargs)
File "/usr/local/lib/python3.6/site-packages/keras/layers/convolutional.py", line 2151, in call
return K.temporal_padding(inputs, padding=self.padding[0])
File "/usr/local/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py", line 2759, in temporal_padding
return tf.pad(x, pattern)
File "/usr/local/lib/python3.6/site-packages/tensorflow_core/python/ops/array_ops.py", line 2840, in pad
result = gen_array_ops.pad(tensor, paddings, name=name)
File "/usr/local/lib/python3.6/site-packages/tensorflow_core/python/ops/gen_array_ops.py", line 6399, in pad
"Pad", input=input, paddings=paddings, name=name)
File "/usr/local/lib/python3.6/site-packages/tensorflow_core/python/framework/op_def_library.py", line 794, in _apply_op_helper
op_def=op_def)
File "/usr/local/lib/python3.6/site-packages/tensorflow_core/python/util/deprecation.py", line 507, in new_func
return func(*args, **kwargs)
File "/usr/local/lib/python3.6/site-packages/tensorflow_core/python/framework/ops.py", line 3357, in create_op
attrs, op_def, compute_device)
File "/usr/local/lib/python3.6/site-packages/tensorflow_core/python/framework/ops.py", line 3426, in _create_op_internal
op_def=op_def)
File "/usr/local/lib/python3.6/site-packages/tensorflow_core/python/framework/ops.py", line 1770, in __init__
control_input_ops)
File "/usr/local/lib/python3.6/site-packages/tensorflow_core/python/framework/ops.py", line 1610, in _create_c_op
raise ValueError(str(e))
ValueError: Shape must be rank 3 but is rank 4 for 'padding_conv1_2/Pad' (op: 'Pad') with input shapes: [1,?,30,2], [3,2].
我已尽可能地缩减模型并将模型单独保存为 JSON（21kb，无权重），在此 github repository 中。
我可以使用以下 sn-p 复制错误。 Keras-resnet 需要安装在 python3 中，并且您需要来自存储库的 ck.json 文件。下面，为 model_from_json() 提供了一个自定义对象字典，因为该模型包含一个自定义层。
from keras.models import model_from_json
import keras_resnet
from keras_resnet.models import ResNet1D18
with open('ck.json', 'r') as f:
model = model_from_json(f.read(), {'ResNet1D18': keras_resnet.models.ResNet1D18})
我对此很陌生，所以我希望我只是做了一些愚蠢的事情，但问题似乎不是模型本身的形状不匹配，因为我可以创建模型、拟合数据并保存模型没有问题。 读取已保存的模型会引发错误。从下面的模型摘要中，第一个 ZeroPadding1D 层的输入形状是 (?, 30, 2) 但保存的模型如何将其转换为 (1, ?, 30, 2) ，如上面的错误？
提前感谢您的帮助！
模型总结为：
Model: "resnet1d18_1"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_1 (InputLayer) (None, 30, 2) 0
__________________________________________________________________________________________________
padding_conv1 (ZeroPadding1D) (None, 36, 2) 0 input_1[0][0]
__________________________________________________________________________________________________
conv1 (Conv1D) (None, 15, 64) 896 padding_conv1[0][0]
__________________________________________________________________________________________________
bn_conv1 (BatchNormalization) (None, 15, 64) 256 conv1[0][0]
__________________________________________________________________________________________________
conv1_relu (Activation) (None, 15, 64) 0 bn_conv1[0][0]
__________________________________________________________________________________________________
pool1 (MaxPooling1D) (None, 8, 64) 0 conv1_relu[0][0]
__________________________________________________________________________________________________
padding2a_branch2a (ZeroPadding (None, 10, 64) 0 pool1[0][0]
__________________________________________________________________________________________________
res2a_branch2a (Conv1D) (None, 8, 64) 12288 padding2a_branch2a[0][0]
__________________________________________________________________________________________________
bn2a_branch2a (BatchNormalizati (None, 8, 64) 256 res2a_branch2a[0][0]
__________________________________________________________________________________________________
res2a_branch2a_relu (Activation (None, 8, 64) 0 bn2a_branch2a[0][0]
__________________________________________________________________________________________________
padding2a_branch2b (ZeroPadding (None, 10, 64) 0 res2a_branch2a_relu[0][0]
__________________________________________________________________________________________________
res2a_branch2b (Conv1D) (None, 8, 64) 12288 padding2a_branch2b[0][0]
__________________________________________________________________________________________________
res2a_branch1 (Conv1D) (None, 8, 64) 4096 pool1[0][0]
__________________________________________________________________________________________________
bn2a_branch2b (BatchNormalizati (None, 8, 64) 256 res2a_branch2b[0][0]
__________________________________________________________________________________________________
bn2a_branch1 (BatchNormalizatio (None, 8, 64) 256 res2a_branch1[0][0]
__________________________________________________________________________________________________
res2a (Add) (None, 8, 64) 0 bn2a_branch2b[0][0]
bn2a_branch1[0][0]
__________________________________________________________________________________________________
res2a_relu (Activation) (None, 8, 64) 0 res2a[0][0]
__________________________________________________________________________________________________
pool5 (GlobalAveragePooling1D) (None, 64) 0 res2a_relu[0][0]
__________________________________________________________________________________________________
fc1000 (Dense) (None, 1) 65 pool5[0][0]
==================================================================================================
Total params: 30,657
Trainable params: 30,145
Non-trainable params: 512
【问题讨论】：
标签： python-3.x keras save load resnet