Keras: `Unknown entry in loss dictionary` when load_model with Sequential model

[serser]

I am with Tensorflow 1.7 and Keras 2.1.6.

My model was trained on Model with a Sequential built-in, I wrote a min case to reproduce the error:

import keras
from keras.models import Sequential, Model, load_model
from keras.layers import Input,Conv2D, BatchNormalization
from keras.losses import binary_crossentropy
from keras.utils.vis_utils import plot_model

seq = Sequential([
            Conv2D(64, (3, 3), activation='relu', input_shape=(240,320,1), padding='same',name='conv'),
BatchNormalization()], name='conv_seq')

_input = Input((240,320,1), name='input')

model = Model(inputs=_input, outputs=seq(_input))
model.compile(optimizer="adam",loss={'conv_seq': binary_crossentropy})

plot_model(model,to_file='seq_plot.png',show_shapes=True, show_layer_names=True)

model.save('seq.hd5')

After training, while I tried to load_model:

_model = load_model('seq.hd5')

it throws error:

/usr/local/lib/python2.7/site-packages/keras/models.pyc in load_model(filepath, custom_objects, compile)
    299                       metrics=metrics,
    300                       loss_weights=loss_weights,
--> 301                       sample_weight_mode=sample_weight_mode)
    302 
    303         # Set optimizer weights.

/usr/local/lib/python2.7/site-packages/keras/engine/training.pyc in compile(self, optimizer, loss, metrics, loss_weights, sample_weight_mode, weighted_metrics, target_tensors, **kwargs)
    614                                      'dictionary: "' + name + '". '
    615                                      'Only expected the following keys: ' +
--> 616                                      str(self.output_names))
    617             loss_functions = []
    618             for name in self.output_names:

Unknown entry in loss dictionary: "conv_seq". Only expected the following keys: ['sequential_1']

It seems my so named conv_seq sequential layer is not recognized when reloading, it is named as sequential_1 instead. Is this a bug?

[serser]

Hello folks, I've checked the Keras source and discovered when performing load_model or model_from_config, the name however is not passed to Sequential class which caused such error. Thus, I added some code to mitigate this.

# keras/models.py, L1404 
    @classmethod
    def from_config(cls, config, custom_objects=None,name=None):
        if 'class_name' not in config[0] or config[0]['class_name'] == 'Merge':
            return cls.legacy_from_config(config)

        model = cls(name=name)

    ...

and in

# keras/utils/generic_utils.py, Line 143
                if 'name' in config:
                    return cls.from_config(config['config'],
                                           custom_objects=dict(list(_GLOBAL_CUSTOM_OBJECTS.items()) +
                                                               list(custom_objects.items())),
                                           name=config['name'])

[MaharshiYeluri01]

try naming all the layers, it should work fine