IndexError: index 1 is out of bounds for axis 0 with size 1

[tonyatml]

IndexError Traceback (most recent call last)
in ()
29 {input_text: dyn_input, initial_state: prev_state})
30
---> 31 pred_word = pick_word(probabilities[dyn_seq_length-1], int_to_vocab)
32
33 gen_sentences.append(pred_word)

IndexError: index 1 is out of bounds for axis 0 with size 1

[YangWangGit]

Try to replace the line
pred_word = pick_word(probabilities[dyn_seq_length-1], int_to_vocab)
with
pred_word = pick_word(probabilities[0,dyn_seq_length-1,:], int_to_vocab)
This fix seems to work for me.

By the way, I finished the training, but did not get the same, reasonable quality of predicted text as zackthoutt demonstrated.

[zackthoutt]

@tonyatml, @YangWangGit are you using Tensorflow 1.0.0? @YangWangGit, the output of the network should not have a depth of three. That is likely why you are not getting good results.

Also keep in mind that training networks like this is tricky and involves a lot of tinkering. When I trained the network I ran several instances at once with different combinations of hyperparameters. I also downloaded all of the networks every half hour or so while they were training and tested how well they generated text.

I think that the model I actually ended up generating the text with used slightly different hyperparameters than what I have posted in this repo. I think my best one had a larger word embedding layer and a longer sequence length. Also note that I did not train the full number of epochs defined in the code. I stopped all of the models training well before reaching the max value. Training too much will either force your weights into a local minima or overfit, both of which lead to worse results.

[tonyatml]

@zackthoutt ，I am using version 1.3.0 installed via 'pip install' with GPU support. The python version is 2.7.12.

[zackthoutt]

You need to use Tensorflow 1.0.0. There are updates that will break the code in 1.3.0.

[vinothkumarsvv]

Traceback (most recent call last):
File "train.py", line 249, in
main()
File "train.py", line 246, in main
history = parallel_model.fit_generator(train_data_generator, epochs=FLAGS.max_epochs, steps_per_epoch=train_samples_count/FLAGS.batch_size, use_multiprocessing=False, callbacks=callbacks, verbose=1)
File "C:\Users\VINOTH KUMAR S\Anaconda3\lib\site-packages\keras\legacy\interfaces.py", line 91, in wrapper
return func(*args, **kwargs)
File "C:\Users\VINOTH KUMAR S\Anaconda3\lib\site-packages\keras\engine\training.py", line 1418, in fit_generator
initial_epoch=initial_epoch)
File "C:\Users\VINOTH KUMAR S\Anaconda3\lib\site-packages\keras\engine\training_generator.py", line 251, in fit_generator
callbacks.on_epoch_end(epoch, epoch_logs)
File "C:\Users\VINOTH KUMAR S\Anaconda3\lib\site-packages\keras\callbacks.py", line 79, in on_epoch_end
callback.on_epoch_end(epoch, logs)
File "train.py", line 104, in on_epoch_end
input_image_summary = make_image_summary(((data[0][0][i] + 1) * 127.5).astype('uint8'))
IndexError: index 1 is out of bounds for axis 0 with size 1