How to store printed values from tf.print into a text file

[dia...@gmail.com]

Hello guys, I'm trying to print the values of an LSTM forget gate in a text file:

When I do:

f = tf.Print(f, [f], message="This is f: ", summarize=120)

... I receive by console the result in numbers

However when I try to store this data using:

with open("test.txt", "a") as myfile:
                myfile.write(str(tf.Print(f, [f], message="This is f: ", summarize=120)))

In the text file I just receive per processed data,

Tensor("RNN/LayerNormalizedLSTMCell_5/Print_1:0", shape=(?, 10), dtype=float32)

Could you tell me please how solve this?

Thanks

[Jordi]

Hello Dialgop,

From the tensorflow documentation (https://www.tensorflow.org/versions/r0.10/api_docs/python/control_flow_ops.html#Print) :

- "This is an identity op with the SIDE EFFECT of printing data when evaluating."

I guess that it will write it on the "cout" channel, so you could try to redirect the output of the execution to a file using the console. If not, you can not use that functions to print into a file.

[Dmytro Filatov]

Hi DialGop,

Did you found a good solution for this? Have you tried to experiment with tf.py_func?

This may provide some insights: https://github.com/tensorflow/tensorflow/issues/1702

Please let me know if you will find the good way to do this.

Thanks

On Monday, 29 August 2016 08:12:29 UTC-7, dia...@gmail.com wrote: