Closed wichtounet closed 7 years ago
in train.py try replacing the line
state = model.initial_state.eval()
with
model.initial_state = tf.convert_to_tensor(model.initial_state) state = model.initial_state.eval()
It worked for me but I'm not sure how it would affect performance. Interestingly, I only experienced this problem when I compiled TF for use of my gpu.
I am using Ubuntu/Linux 64-bit CPU Only, Python 2.7 Tensorflow-0.11 and this fixed my issue
I fixed this issue using sess.run which seems to make more sense in my opinion.
Hi,
I tried this implementation of char-rnn and I have an issue with the train script:
I'm using the last version of tensorflow and Python 3.4.
Thanks