Regression with MxNet 1.5.0: Cannot apply softmax to a tensor that is 1D. Received input: Symbol: broadcast_add30

[x] If running on MXNet, check that you are up-to-date with the latest version. The installation instructions can be found here

Crossposting apache/incubator-mxnet#15833

Loading VGG16 is now currently broken with MxNet 1.5.0 and Keras-MxNet 2.2.4.1

from keras.applications.vgg16 import VGG16
m = VGG16(weights=None, classes=365)

Produces

ValueError                                Traceback (most recent call last)
<ipython-input-2-bb997f6d4e3d> in <module>
----> 1 m = VGG16(weights=None, classes=365)

~/.conda/envs/user/lib/python3.6/site-packages/keras/applications/__init__.py in wrapper(*args, **kwargs)
     26             kwargs['models'] = models
     27             kwargs['utils'] = utils
---> 28         return base_fun(*args, **kwargs)
     29 
     30     return wrapper

~/.conda/envs/user/lib/python3.6/site-packages/keras/applications/vgg16.py in VGG16(*args, **kwargs)
      9 @keras_modules_injection
     10 def VGG16(*args, **kwargs):
---> 11     return vgg16.VGG16(*args, **kwargs)
     12 
     13 

~/.conda/envs/user/lib/python3.6/site-packages/keras_applications/vgg16.py in VGG16(include_top, weights, input_tensor, input_shape, pooling, classes, **kwargs)
    178         x = layers.Dense(4096, activation='relu', name='fc1')(x)
    179         x = layers.Dense(4096, activation='relu', name='fc2')(x)
--> 180         x = layers.Dense(classes, activation='softmax', name='predictions')(x)
    181     else:
    182         if pooling == 'avg':

~/.conda/envs/user/lib/python3.6/site-packages/keras/engine/base_layer.py in __call__(self, inputs, **kwargs)
    468             # Actually call the layer,
    469             # collecting output(s), mask(s), and shape(s).
--> 470             output = self.call(inputs, **kwargs)
    471             output_mask = self.compute_mask(inputs, previous_mask)
    472 

~/.conda/envs/user/lib/python3.6/site-packages/keras/layers/core.py in call(self, inputs)
    893             output = K.bias_add(output, self.bias, data_format='channels_last')
    894         if self.activation is not None:
--> 895             output = self.activation(output)
    896         return output
    897 

~/.conda/envs/user/lib/python3.6/site-packages/keras/activations.py in softmax(x, axis)
     34     else:
     35         raise ValueError('Cannot apply softmax to a tensor that is 1D. '
---> 36                          'Received input: %s' % x)
     37 
     38 

ValueError: Cannot apply softmax to a tensor that is 1D. Received input: Symbol: broadcast_add30

awslabs / keras-apache-mxnet

Regression with MxNet 1.5.0: Cannot apply softmax to a tensor that is 1D. Received input: Symbol: broadcast_add30 #243