A question on the number of parameter

[who7384]

Hi Akara,

Thanks for your excellent project! But I am confused about the number of parameters. The paper said the parameter number is about 1.3M, but when I added some extra code to calculate parameter numbers, such as:

num_params = 0
        for variable in tf.trainable_variables():
            shape = variable.get_shape()
            num_params += reduce(mul, [dim.value for dim in shape], 1)
print(num_params)

The output was:

cnn/conv1d_1/conv2d/kernel:0 6400
cnn/bn_1/batch_normalization/gamma:0 128
cnn/bn_1/batch_normalization/beta:0 128
cnn/conv1d_2_1/conv2d/kernel:0 131072
cnn/bn_2_1/batch_normalization/gamma:0 128
cnn/bn_2_1/batch_normalization/beta:0 128
cnn/conv1d_2_2/conv2d/kernel:0 131072
cnn/bn_2_2/batch_normalization/gamma:0 128
cnn/bn_2_2/batch_normalization/beta:0 128
cnn/conv1d_2_3/conv2d/kernel:0 131072
cnn/bn_2_3/batch_normalization/gamma:0 128
cnn/bn_2_3/batch_normalization/beta:0 128
rnn/rnn/multi_rnn_cell/cell_0/lstm_cell/kernel:0 1114112
rnn/rnn/multi_rnn_cell/cell_0/lstm_cell/bias:0 512
rnn/rnn/multi_rnn_cell/cell_0/lstm_cell/w_f_diag:0 128
rnn/rnn/multi_rnn_cell/cell_0/lstm_cell/w_i_diag:0 128
rnn/rnn/multi_rnn_cell/cell_0/lstm_cell/w_o_diag:0 128
1515648

Would you please tell me why?