Tensorflow gpu starter (tensorflow-gpu 1.12 with CUDA 10.0)

tf-starter.py

import os
import tensorflow as tf;

# Just disables the warning, doesn't enable AVX/FMA
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

def main():
    tf.enable_eager_execution();
    result = tf.reduce_sum(tf.random_normal([1000,1000]))
    print(result)
    print("Setup successful")

if __name__ == "__main__":
    main()
    

We can disable the warning log level because the standard version is supposed to support as many CPU as possible, but we do not have to care when using tensorflow with gpu. All credits go to https://stackoverflow.com/questions/47068709/your-cpu-supports-instructions-that-this-tensorflow-binary-was-not-compiled-to-u