DeepICF & NAIS issue with loss

[theconsolelogger]

DeepICF and NAIS models generates the following error on line 215 for DeepICF:

NeuRec/model/general_recommender/DeepICF.py

Line 215 in f8eb375

loss, _ = self.sess.run((self.loss, self.optimizer), feed_dict=feed_dict)

and line 216 for NAIS:

NeuRec/model/general_recommender/NAIS.py

Line 216 in f8eb375

loss, _ = self.sess.run((self.loss, self.optimizer), feed_dict=feed_dict)

ValueError: setting an array element with a sequence.

The properties for DeepICF is set to the following:

[DEFAULT]
data.input.path=neurec
data.input.dataset=ml-100k.rating
data.column.format=UIRT
data.convert.separator='\t'
data.convert.binarize.threshold=0
data.splitter=loo
data.splitterratio=[0.8,0.2]|float
rec.number.thread=20
rec.evaluate.neg=0
recommender=deepicf
topk=[10, 20, 30, 40]|int
gpu_id=0

pretrain_file=None

#Interval of evaluation
verbose=1

#adagrad,rmsprop,adam
learner=adam

#user: generate batches by user, fixed:batch_size: generate batches by batch size
batch_size=256

#Number of epochs.
epochs=300
#Size of each layer
layers=[64,32,16]|int

weight_size=16

embedding_size=16

#Index of coefficient of embedding vector
data_alpha=0

#Regularization for user and item embeddings.[1e-7,1e-7,0]
regs=[0.000001,0.000001,0.00001]|float
#L_2 regularization on each layer weights.
regw=[10,10]|int
#Index of coefficient of embedding vector
alpha=0

#Index of coefficient of sum of exp(A)
beta=0.5

#Number of negative instances to pair with a positive instance.
num_neg=4

learning_rate=0.001

#Activation for ReLU, sigmoid, tanh.
activation=Relu

#0 for deepICF_prod, 1 for deepicf_concat
algorithm=1

#Whether to perform batch norm (0 or 1)
batch_norm=1

#tnormal: truncated_normal_initializer, uniform: random_uniform_initializer,
#normal: random_normal_initializer, xavier_normal, xavier_uniform, 
#he_normal, he_uniform. Defualt: tnormal
embed_init_method=tnormal
weight_init_method=he_normal
bias_init_method=he_normal
stddev=0.01

The properties for NAIS is set to the following:

[DEFAULT]
data.input.path=neurec
data.input.dataset=ml-100k.rating
data.column.format=UIRT
data.convert.separator='\t'
data.convert.binarize.threshold=0
data.splitter=loo
data.splitterratio=[0.8,0.2]|float
rec.number.thread=20
rec.evaluate.neg=0
recommender=nais
topk=[10, 20, 30, 40]|int
gpu_id=0

pretrain=1

#Interval of evaluation
verbose=1

#adagrad,rmsprop,adam
learner=adam
#user: generate batches by user, fixed:batch_size: generate batches by batch size
batch_size=256

#Number of epochs.
epochs=100

weight_size=16

embedding_size=16

#Index of coefficient of embedding vector
data_alpha=0

#Regularization for user and item embeddings.
regs=[1e-7,1e-7,1e-5]|float

#Index of coefficient of embedding vector
alpha=0

#Index of coefficient of sum of exp(A)
beta=0.5

#Number of negative instances to pair with a positive instance.
num_neg=4

learning_rate=0.001

#Activation for ReLU, sigmoid, tanh.
activation=Relu

#0 for NAIS_prod, 1 for NAIS_concat
algorithm=0

is_pairwise=false
#pairwise: BPR(BPRFPMC),hinge,square
#pointwise: cross_entropy,square
loss_function=cross_entropy
#tnormal: truncated_normal_initializer, uniform: random_uniform_initializer,
#normal: random_normal_initializer, xavier_normal, xavier_uniform, 
#he_normal, he_uniform. Defualt: tnormal
embed_init_method=tnormal
weight_init_method=he_normal
stddev=0.01
pretrain_file=None