weights of keras model are nan

Question

I'm using below model for regression purpose; with input of size 2 and output of size 28.

from keras.models import Sequential
from keras.layers import Dense
from keras.optimizers import SGD

model = Sequential()
model.add(Dense(16, input_dim=2, activation='relu'))
model.add(Dense(16, activation='relu'))
model.add(Dense(28, activation='linear'))
sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='mean_squared_error',optimizer=sgd)

In training all going well, but when I save and reload the model; I'm getting weights as a nan.

from keras.models import model_from_json

model_json = model.to_json()
with open('/models/model_ar.json', "w") as json_file:
     json_file.write(model_json)
model.save_weights('/models/model_wt.h5')

json_file = open('/models/model_ar.json', 'r')
loaded_model_json = json_file.read()
json_file.close()
new_model = model_from_json(loaded_model_json)
# load weights into new model
new_model.load_weights('/models/model_wt.h5')

getting weights as 'nan'. What is reason to get the all weights as nan

    new_model.get_weights()
[array([[nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan, nan],
        [nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan, nan]], dtype=float32),
 array([nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
        nan, nan, nan], dtype=float32),
 array([[nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan, nan],
        [nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan, nan],
        [nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan, nan],
        [nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan, nan],
        [nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan, nan],
        [nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan, nan]], dtype=float32),
 array([nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
        nan, nan, nan], dtype=float32),
 array([[nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan],
        [nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan],
        [nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
         nan, nan]], dtype=float32),
 array([nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
        nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
        nan, nan], dtype=float32)]

Answer 1

Try

sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True, clipvalue=0.5)

from https://www.dlology.com/blog/how-to-deal-with-vanishingexploding-gradients-in-keras/
You can also try clipnorm=1. or try one of those parameters with even smaller values.

This limits how much the weights can change in each step of the gradient descent. It worked for me when I had the same problem and I hope it helps!

Answer 2

In my case, there were NaN rows in the dataset that I didn't notice. So please check your dataset and values. Check the answer here