zackthoutt / got-book-6 Goto Github PK

Nice project.

However, you uploaded all five books of ASOIAF illegally, and you better remove it before you get hit with a DMCA.

Is this a vanilla RNN or an LSTM? I can't find the actual code.

Also take a look at the newer text generation papers. In particular, there was this one paper (from Baidu research i believe?) on Chinese poetry generation that did incredibly well. I would in particular try GANs and autoregressive CNN models.

Can you also do a separate write up about how you come up with the model?

1.Wiil it work if I put no GOT book?
2.Is there any way to start training and when I dont wanna train i can pause it and
resume it from place that I stoped?
3.How can I generate book using trained model without starting training from begining?

Hope you will anserw my questions :)

Your project is so amazing 👍

No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without permission in writing from the publisher.

This passage (line 27 /data/got1.txt) is stylistically different from George's usual prose, and may have some impact on the quality of the data that can be generated.

Hi, I'm wondering if it is possible to reload the model from a saved checkpoint. I've tried to implement this myself, but I don't really know much about tensorflow and run into errors.

Hi. that's a great project there.. i loved it.. Could you please tell me about your hardware, I have MacBook Air with 8gb ram and it took almost 13 hours and then it crashed.. I was thinking to run it on a server so if you could just give me an idea about the memory.. i would really appreciate it... Thanks!

Greenbeard was actually a character (although I've forgotten him)
I was initially perplexed thinking the network came up with a word outside the vocab.

http://awoiaf.westeros.org/index.php/Greenbeard

Just clearing it up for anybody who might be thinking along similar lines.

In order to enable open-source contributions, it would be formative to consider the specifics of a license agreement.

Hello,
has anybody successfully ported to TensorFlow 1.4 ?
I am encountering problems, and would appreciate help.

1. crash here

38     # Build RNN
---> 39     outputs, final_state = tf.nn.dynamic_rnn(cell, embed, dtype=tf.float32)

if embed_dim is different from rnn_size.
For the time being, I leave the 2 values equal.

2. crash here

---> 34         pred_word = pick_word(probabilities[dyn_seq_length-1], int_to_vocab)
IndexError: index 2 is out of bounds for axis 0 with size 1

same problem as #15

Any ideas ?

In the notebook, the last train_loss is approximately 2.4.
Since I am training a language generating model like yours, i am very curious about the final train_loss you get.
Thanks!

Greetings,

Nice project... With the default parameters you've gave, it takes days not hours, even with powerful GPUs on AWS.

num_epochs = 10000; batch_size = 512; rnn_size = 512; num_layers = 3; keep_prob = 0.7; embed_dim = 512; seq_length = 30; learning_rate = 0.001;

Also I have to cut the batch_size to 256 (as I think you also did) since I've got a GPU memory overflow. I'm curious to know if there is less greedy hyperparameters. In your Notebook, the executable shows more than 24 days.

I'm also curious to compare with a simple Markov chain.

Anyway thanks for the nice project that allows us to dream with tales.

Claude Coulombe

Hodor looked at them bellowing, "which road you should be home."

I think a special case needs handled here ;-D
Awesome job

Nice project!
Is there any chance you could provide the trained model, so we can play a bit with it without going through the trouble of training a new one from scratch?

Hi Buddy,

There is an error when I run your program via 'ipython notebook' with following tips

NFO:tensorflow:Restoring parameters from ./save

ValueError Traceback (most recent call last)
in ()
30
31 #pred_word = pick_word(probabilities[dyn_seq_length-1], int_to_vocab)
---> 32 pred_word = pick_word(probabilities[0], int_to_vocab)
33
34 gen_sentences.append(pred_word)

in pick_word(probabilities, int_to_vocab)
6 :return: String of the predicted word
7 """
----> 8 return np.random.choice(list(int_to_vocab.values()), 1, p=probabilities)[0]

mtrand.pyx in mtrand.RandomState.choice (numpy/random/mtrand/mtrand.c:17602)()

ValueError: object too deep for desired array

How long did it take to train your model?
Best flo

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