This is an implementation of a LSTM model to solve the IMDB sentiment classification task.
The model script is included with the neon repo examples (link)
The trained weights file can be downloaded from AWS using the following link: [imdb_lstm.p][S3_WEIGHTS_FILE]. [S3_WEIGHTS_FILE]: https://s3-us-west-1.amazonaws.com/nervana-modelzoo/imdb_lstm/imdb_lstm.p
The model weight file above has been generated using neon version tag [v1.4.0]((https://github.com/NervanaSystems/neon/releases/tag/v1.4.0). It may not work with other versions.
This model is acheiving 94.7% and 84.8% accuracy on the training and validation set, respectively.
Commands should be run from the neon installation root directory. If neon was installed into a virtualenv, make sure that it is active first.
To train the model from scratch for 2 epochs, use the command:
python examples/imdb_lstm.py -b gpu -e 2 -eval 1 -r 0 -s imdb_lstm.p
When Are Tree Structures Necessary for Deep Learning of Representations? Jiwei Li, Dan Jurafsky and Eduard Hovy http://arxiv.org/pdf/1503.00185v1.pdf