Deep Learning with TensorFlow

Running the example

The PTB model can support small, medium, and large dataset configurations. The small model should be able to reach perplexity below 120 on the test set and the large one below 80, though it might take several hours to train.

We executed the model on the small size dataset. To do this, simply type in your command prompt the following:

    python ptb_word_lm --data_path=/tmp/simple-examples/data/ --model small 

This is where you extracted the previously downloaded PTB dataset in /tmp/simple-examples/data/.

The preceding perplexity value after eight hours of training and 13 training epochs are as follows:

Epoch: 1 Learning rate: 1.000 0.004 perplexity: 5263.762 speed: 391 wps 0.104 perplexity: 837.607 speed: 429 wps 0.204 ...