Tensorflow not outputting any log information in Spyder?

Question

What it says on the tin. This issue:

https://github.com/spyder-ide/spyder/issues/3777

Has been open for years now. Surely someone has gotten tensorflow to output information to the Spyder console?

For the record, my Tensorflow works. I ran the simple example here:

a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3], name='a')
b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2], name='b')
c = tf.matmul(a, b)
sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))
print(sess.run(c))

And it just gives me the answer, not the logging information I want. My actual code that I'm trying to run does the same thing (answer, no log info), but since it takes so long I would really like the console output.

But I have no idea how to actually implement the code snippet it has there.

Answer 1

I've had no trouble with Spyder printing logging information. You may need to enable logging in your application:

tf.logging.set_verbosity(tf.logging.INFO)

A lot of developers set the TF_CPP_MIN_LOG_LEVEL environment variable to 3 to prevent TensorFlow from whining about the device's supported capabilities. I'd check that.

If you run the following code, will it print anything?

tf.logging.info('Hi there!')