• Search by Module
• Search by Words
• Search Projects

• Java
• Python
• JavaScript
• TypeScript
• C++
• Scala
• Blog

Project: fancy-cnn (GitHub Link)

• fancy-cnn-master
  • textclf
    • wordvectors
      • char.py
      • misc.py
      • glove.py
      • __init__.py
    • nn
      • timedistributed.py
      • embeddings.py
      • __init__.py
      • train_neural.py
    • util
      • misc.py
      • __init__.py
      • language.py
    • __init__.py
  • datasets
    • aclImdb
      • README.md
      • download.sh
    • __init__.py
    • reuters21578
      • data
        • all-places-strings.lc.txt
        • all-topics-strings.lc.txt
        • README.txt
        • lewis.dtd
        • all-orgs-strings.lc.txt
        • cat-descriptions_120396.txt
        • all-exchanges-strings.lc.txt
        • all-people-strings.lc.txt
      • __init__.py
      • parser.py
    • yelp
      • data_handling.py
      • data_funny_binary_balanced
      • yelp_w2v.py
      • __init__.py
  • prepare_imdb_as_text.py
  • prepare_yelp_char.py
  • examples
    • yelp
      • test_birnn.py
      • test_cnn.py
      • test_rcnn.py
      • test_rcnnchar.py
      • test_rnn.py
      • outputs
        • useful
          • log-model-cnn-1
          • log-model-cnn-2
          • log-model-yelp-wvs-cnn-1
          • log-model-birnn-1
          • log-model-rnn-2
          • log-model-yelp-wvs-rnn-1
          • log-model-gru-1
          • log-model-birnn-2
          • log-model-yelp-wvs-birnn-1
          • log-model-birnn-2~
        • cool
          • log-model-cnn-1
          • log-model-yelp-wvs-cnn-1
          • log-model-birnn-1
          • log-model-yelp-wvs-rnn-1
          • log-model-yelp-wvs-birnn-1
          • log-model-rnn-1
        • funny
          • log-model-rcnn-2
          • log-model-cnn-1
          • log-model-yelp-wvs-cnn-1
          • log-model-birnn-1
          • log-model-yelp-wvs-rnn-1
          • log-model-gru-1
          • log-model-yelp-wvs-birnn-1
          • log-model-rcnn-1
      • test_meanwvs.py
    • imdb
      • test_vggcnn.py
      • test_birnn.py
      • test_cnn.py
      • test_rcnn.py
      • test_rcnnchar.py
      • test_rnn.py
      • outputs
        • log-model-cnn-1
        • log-model-multigru-3
        • log-model-multigru-2
        • log-model-bigru-1
        • log-model-gru
        • log-model-multigru-1
        • log-model-multigru-4
      • test_multi_rnn.py
  • embeddings
    • README.md
    • download.sh
  • LICENSE
  • .old
    • model-yelp.py
    • cnn
      • layers
        • version.py
        • convolutions.py
        • embeddings.py
        • __init__.py
      • __init__.py
    • multichannel-imdb.py
  • prepare_imdb_sentences.py
  • prepare_imdb_char.py
  • prepare_yelp_as_text.py
  • README.md
  • prepare_yelp_sentences.py
  • .gitignore

fancy-cnn

Multiparadigm Sequential Convolutional Neural Networks for text classification.

General Considerations

First off, this class of sequential convolutional networks is quite GEMM intensive, and really isn't suited to a CPU. You also really should use CuDNN when training.

Bleeding edge installations of Theano and Keras are required. The TensorFlow backend has not been tested with the fancier TimeDistributed architectures. When considering your hardware, note that...

• On a 2.6 GHz Intel Core i7, one epoch of IMDB training takes ~3.5 days
• On a GRID K520 without CuDNN, one epoch of IMDB training takes ~1 hour
• On a GRID K520 with CuDNN, one epoch of IMDB training takes ~30 minutes
• On a GTX Titan X with CuDNN, one epoch of IMDB training takes ~11 minutes

If you want to use CuDNN, you really should also put

[dnn]
conv.algo_fwd = time_on_shape_change
conv.algo_bwd = time_on_shape_change

to your .theanorc. This will let Theano (at compilation time) pick the fastest convolution implementation given your input and kernel size.

Setup

1. Go to datasets/aclImdb/
2. ./download.sh
3. Go to embeddings/
4. ./download.sh

This gets all the embeddings from S3 and sets up the IMDB dataset for processing. Yelp instructions forthcoming.

IMDB Dataset

Look at (and run) prepare_imdb*.py to prepare your data! Look at this to see how to train the IMDB model.

Yelp Humor Dataset

Look at (and run) prepare_yelp*.py to prepare your data! Look at this to see how to train a model on Yelp.