[FIXED] Weird Nan loss for custom Keras loss

Issue

I'm trying to implement a custom loss in Keras but can't get it to work.

I have implemented it in numpy and with keras.backend:

def log_rmse_np(y_true, y_pred):
    d_i = np.log(y_pred) -  np.log(y_true)
    loss1 = (np.sum(np.square(d_i))/np.size(d_i))
    loss2 = ((np.square(np.sum(d_i)))/(2 * np.square(np.size(d_i))))
    loss = loss1 - loss2
    print('np_loss =  %s - %s = %s'%(loss1, loss2, loss))
    return loss

def log_rmse(y_true, y_pred):
    d_i = (K.log(y_pred) -  K.log(y_true))
    loss1 = K.mean(K.square(d_i))
    loss2 = K.square(K.sum(K.flatten(d_i),axis=-1))/(K.cast_to_floatx(2) * K.square(K.cast_to_floatx(K.int_shape(K.flatten(d_i))[0])))
    loss = loss1 - loss2
    return loss

When I test and compare the losses with the following function everything seems to work just fine.

def check_loss(_shape):
    if _shape == '2d':
        shape = (6, 7)
    elif _shape == '3d':
        shape = (5, 6, 7)
    elif _shape == '4d':
        shape = (8, 5, 6, 7)
    elif _shape == '5d':
        shape = (9, 8, 5, 6, 7)

    y_a = np.random.random(shape)
    y_b = np.random.random(shape)

    out1 = K.eval(log_rmse(K.variable(y_a), K.variable(y_b)))
    out2 = log_rmse_np(y_a, y_b)

    print('shapes:', str(out1.shape), str(out2.shape))
    print('types: ', type(out1), type(out2))
    print('log_rmse:    ', np.linalg.norm(out1))
    print('log_rmse_np: ', np.linalg.norm(out2))
    print('difference:  ', np.linalg.norm(out1-out2))
    assert out1.shape == out2.shape
    #assert out1.shape == shape[-1]

def test_loss():
    shape_list = ['2d', '3d', '4d', '5d']
    for _shape in shape_list:
        check_loss(_shape)
        print ('======================')

test_loss()

The above code prints:

np_loss =  1.34490449177 - 0.000229461787517 = 1.34467502998
shapes: () ()
types:  <class 'numpy.float32'> <class 'numpy.float64'>
log_rmse:     1.34468
log_rmse_np:  1.34467502998
difference:   3.41081509703e-08
======================
np_loss =  1.68258448859 - 7.67580654591e-05 = 1.68250773052
shapes: () ()
types:  <class 'numpy.float32'> <class 'numpy.float64'>
log_rmse:     1.68251
log_rmse_np:  1.68250773052
difference:   1.42057615005e-07
======================
np_loss =  1.99736933814 - 0.00386228512295 = 1.99350705302
shapes: () ()
types:  <class 'numpy.float32'> <class 'numpy.float64'>
log_rmse:     1.99351
log_rmse_np:  1.99350705302
difference:   2.53924863358e-08
======================
np_loss =  1.95178217182 - 1.60006871892e-05 = 1.95176617114
shapes: () ()
types:  <class 'numpy.float32'> <class 'numpy.float64'>
log_rmse:     1.95177
log_rmse_np:  1.95176617114
difference:   3.78277884572e-08
======================

I never get an exception when I compile and fit my model with this loss and when I run the model with the 'adam'-loss everything works fine. However with this loss keras keeps showing a nan-loss:

Epoch 1/10000
 17/256 [>.............................] - ETA: 124s - loss: nan

Kind of stuck here... Am I doing something wrong?

Using Tensorflow 1.4 on Ubuntu 16.04

Update:

After a suggestion by Marcin Możejko I updated the code but unfortunately the training loss is still Nan:

def get_log_rmse(normalization_constant):
    def log_rmse(y_true, y_pred):
        d_i = (K.log(y_pred) -  K.log(y_true))
        loss1 = K.mean(K.square(d_i))
        loss2 = K.square(K.sum(K.flatten(d_i),axis=-1))/K.cast_to_floatx(2 * normalization_constant ** 2)
        loss = loss1 - loss2
        return loss
    return log_rmse

Then the model is compiled via:

model.compile(optimizer='adam', loss=get_log_rmse(batch_size))

Update 2:

The model summary looks like this:

Layer (type)                 Output Shape              Param #   
=================================================================
input_2 (InputLayer)         (None, 160, 256, 3)       0         
_________________________________________________________________
block1_conv1 (Conv2D)        (None, 160, 256, 64)      1792      
_________________________________________________________________
block1_conv2 (Conv2D)        (None, 160, 256, 64)      36928     
_________________________________________________________________
block1_pool (MaxPooling2D)   (None, 80, 128, 64)       0         
_________________________________________________________________
block2_conv1 (Conv2D)        (None, 80, 128, 128)      73856     
_________________________________________________________________
block2_conv2 (Conv2D)        (None, 80, 128, 128)      147584    
_________________________________________________________________
block2_pool (MaxPooling2D)   (None, 40, 64, 128)       0         
_________________________________________________________________
block3_conv1 (Conv2D)        (None, 40, 64, 256)       295168    
_________________________________________________________________
block3_conv2 (Conv2D)        (None, 40, 64, 256)       590080    
_________________________________________________________________
block3_conv3 (Conv2D)        (None, 40, 64, 256)       590080    
_________________________________________________________________
block3_conv4 (Conv2D)        (None, 40, 64, 256)       590080    
_________________________________________________________________
block3_pool (MaxPooling2D)   (None, 20, 32, 256)       0         
_________________________________________________________________
block4_conv1 (Conv2D)        (None, 20, 32, 512)       1180160   
_________________________________________________________________
block4_conv2 (Conv2D)        (None, 20, 32, 512)       2359808   
_________________________________________________________________
block4_conv3 (Conv2D)        (None, 20, 32, 512)       2359808   
_________________________________________________________________
block4_conv4 (Conv2D)        (None, 20, 32, 512)       2359808   
_________________________________________________________________
block4_pool (MaxPooling2D)   (None, 10, 16, 512)       0         
_________________________________________________________________
conv2d_transpose_5 (Conv2DTr (None, 10, 16, 128)       1048704   
_________________________________________________________________
up_sampling2d_5 (UpSampling2 (None, 20, 32, 128)       0         
_________________________________________________________________
conv2d_transpose_6 (Conv2DTr (None, 20, 32, 64)        131136    
_________________________________________________________________
up_sampling2d_6 (UpSampling2 (None, 40, 64, 64)        0         
_________________________________________________________________
conv2d_transpose_7 (Conv2DTr (None, 40, 64, 32)        32800     
_________________________________________________________________
up_sampling2d_7 (UpSampling2 (None, 80, 128, 32)       0         
_________________________________________________________________
conv2d_transpose_8 (Conv2DTr (None, 80, 128, 16)       8208      
_________________________________________________________________
up_sampling2d_8 (UpSampling2 (None, 160, 256, 16)      0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 160, 256, 1)       401       
=================================================================
Total params: 11,806,401
Trainable params: 11,806,401
Non-trainable params: 0

Update 3:

Sample y_true:

Solution

The problem is with this part:

K.cast_to_floatx(K.int_shape(K.flatten(d_i))[0]

as a loss function is compiled before any shape is provided - this expression evaluates to None and this where your error comes from. I tried to set batch_input_shape instead of input_shape but this didn't work also (probably due to the way keras compiles the model). I advice to set this number to a constant in the following manner:

def get_log_rmse(normalization_constant):
    def log_rmse(y_true, y_pred):
        d_i = (K.log(y_pred) -  K.log(y_true))
        loss1 = K.mean(K.square(d_i))
        loss2 = K.square(
            K.sum(
                K.flatten(d_i),axis=-1))/(K.cast_to_floatx(
                    2 * normalization_constant ** 2) 
        loss = loss1 - loss2
        return loss
    return log_rmse

and then compile:

model.compile(..., loss=get_log_rmse(normalization_constant))

I guess that normalization_constant is equal to a batch_size but I'm not sure so I've made it generic.

Update:

After a suggestion by Marcin Możejko I updated the code but unfortunately the training loss is still Nan:

def get_log_rmse(normalization_constant):
    def log_rmse(y_true, y_pred):
        d_i = (K.log(y_pred) -  K.log(y_true))
        loss1 = K.mean(K.square(d_i))
        loss2 = K.square(K.sum(K.flatten(d_i),axis=-1))/K.cast_to_floatx(2 * normalization_constant ** 2)
        loss = loss1 - loss2
        return loss
    return log_rmse

Then the model is compiled via:

model.compile(optimizer='adam', loss=get_log_rmse(batch_size))

Update 2:

The model definition looks like this:

input_shape = (160, 256, 3)
print('Input_shape: %s'%str(input_shape))
base_model = keras.applications.vgg19.VGG19(include_top=False, weights='imagenet', 
                               input_tensor=None, input_shape=input_shape, 
                               pooling=None, # None, 'avg', 'max'
                               classes=1000)
for i in range(5):
    base_model.layers.pop()
base_model = Model(inputs=base_model.input, outputs=base_model.get_layer('block4_pool').output)
print('VGG19 output_shape: ' + str(base_model.output_shape))

x = Deconv(128, kernel_size=(4, 4), strides=1, padding='same', activation='relu')(base_model.output)
x = UpSampling2D((2, 2))(x)
x = Deconv(64, kernel_size=(4, 4), strides=1, padding='same', activation='relu')(x)
x = UpSampling2D((2, 2))(x)
x = Deconv(32, kernel_size=(4, 4), strides=1, padding='same', activation='relu')(x)
x = UpSampling2D((2, 2))(x)
x = Deconv(16, kernel_size=(4, 4), strides=1, padding='same', activation='relu')(x)
x = UpSampling2D((2, 2))(x)
x = Conv2D(1, kernel_size=(5, 5), strides=1, padding='same')(x)
model = Model(inputs=base_model.input, outputs=x)

Answered By - Marcin Możejko