Segmentation fault on training

[nicolagambini]

Hi,

whenever I try to run train.py, I get a segmentation fault error. After some debugging, I found out that it crashes on the following instruction (line 136):
_, gs = sess.run([g.train_op, g.global_step])

What could be the cause? Did anyone experience the same error before?
Please let me know if you need more detail.

Python interpreter version: 2.7, 3.6.1 (same outcome with both versions)
Tensorflow version: 1.14.0

[nicolagambini]

Update: preprocessing the data seems to resolve the issue.

I would then assume that it's mandatory to do so, even if the documentation does not specify it. Can someone motivate this?