Additional Sequence to Sequence Features for TensorFlow

Hey Guys, TensorFlow is great, but there are some seq2seq features that could be added. The goal is to add them as research in this field unfolds.

That's why there's this additional python package. This is meant to work in conjunction with TensorFlow. Simply clone this and import as:

sys.path.insert(0, os.environ['HOME']) #add the dir that you cloned to
from Seq2Seq_Upgrade_TensorFlow.seq2seq_upgrade import seq2seq_enhanced, rnn_cell_enhanced

Main Features Include:

• Different RNN layers on different GPU's
• Regularizing RNNs by Stabilizing Activations -- Krueger's paper
• GRU Mutants -- Jozefowicz's paper
• Identity RNN's -- Le's paper
• Averaging Hidden States During Decoding
• Temperature Sampling During Decoding

Currently Working On:

• Unitary RNN (will take at least 2 weeks) --Arjovsky's paper
• Adversarial Seq2Seq Training (will take at least 4 or 5 weeks)

Other Features That Might Happen If There's Time:

• Async Loading Data during training
• Curriculum Learning
• Removing Biases From GRU and LSTM (some reports that it improves performance)

You will find many, many comments in the code. Just like you guys, I'm still very much learning and it helps me to comment as much as possible.

Lastly, I'm still have much to learn, so I apologize for mistakes in advance. I welcome pull requests and any help.

##Using Seq2Seq Upgrade

Import All RNN Materials from Seq2Seq_Upgrade_TensorFlow_Tensorflow Only!

If you are using Seq2Seq_Upgrade_Tensorflow, please do not import:

• rnn.py
• rnn_cell.py
• seq2seq.py

from the regular tensorflow. Instead import:

• rnn_enhanced.py
• rnn_cell_enhanced.py
• seq2seq_enhanced.py

from Seq2Seq_Upgrade_Tensorflow. Otherwise class inheritance will be thrown off, and you will get an isinstance error!

You can also try to install seq2seq_upgrade as a python package

sudo pip install git+ssh://github.com/LeavesBreathe/Seq2Seq_Upgrade_TensorFlow.git

or a bit longer version in case the previous one didn't work

git clone git@github.com:LeavesBreathe/Seq2Seq_Upgrade_TensorFlow.git
cd Seq2Seq_Upgrade_TensorFlow
sudo python setup.py build & sudo python setup.py install

After that you hopefully be able to simply write import seq2seq_upgrade

##Different RNN Layers on Multiple GPU's -- Working

To call in GRU for gpu 0, simply call in the class

rnn_cell_enhanced.GRUCell(size, gpu_number = 0)

from Seq2Seq_Upgrade_TensorFlow.seq2seq_upgrade import rnn_cell_enhanced

#assuming you're using two gpu's
first_layer = rnn_cell_enhanced.GRUCell(size, gpu_for_layer = 0)
second_layer = rnn_cell_enhanced.GRUCell(size, gpu_for_layer = 1)
cell = rnn_cell.MultiRNNCell(([first_layer]*(num_layers/2)) + ([second_layer]*(num_layers/2)))

#notice that we put consecutive layers on the same gpu. Also notice that you need to use an even number of layers.

You can apply dropout on a specific GPU as well:

from Seq2Seq_Upgrade_TensorFlow.seq2seq_upgrade import rnn_cell_enhanced

first_layer = rnn_cell_enhanced.GRUCell(size, gpu_for_layer = 1)
dropout_first_layer = rnn_cell_enhanced.DropoutWrapper(first_layer, output_keep_prob = 0.80, gpu_for_layer = 1)

##Norm Regularize Hidden States And Outputs -- Testing Krueger's paper

Adds an additional cost to regularize hidden state activations and/or output activations (logits).

By default this feature is set to off.

To use:

from Seq2Seq_Upgrade_TensorFlow.seq2seq_upgrade import seq2seq_enhanced
#to regularize both
seq2seq_enhanced.model_with_buckets(...norm_regularize_hidden_states = True, 
										norm_regularize_logits = True, norm_regularize_factor = 50)
#to regularize one
seq2seq_enhanced.model_with_buckets(...norm_regularize_hidden_states = True, 
										norm_regularize_logits = False, norm_regularize_factor = 50)

norm_regularizer_factor: The factor required to apply norm stabilization. Keep in mind that a larger factor will allow you to achieve a lower loss, but it will take many more epochs to do so!

##GRU Mutants -- Working Jozefowicz's paper

These mutants do better in some seq2seq tasks. Memory wise, they approximately the same as GRU.

To use this simply:

from Seq2Seq_Upgrade_TensorFlow.seq2seq_upgrade import rnn_cell_enhanced

first_layer = rnn_cell_enhanced.JZS1Cell(size)

Mutants are called in by:

1. rnn_cell_enhanced.JZS1Cell(size, gpu_for_layer = 0)
2. rnn_cell_enhanced.JZS2Cell(size, gpu for layer = 1)
3. rnn_cell_enhanced.JZS3Cell(size, gpu_for_layer = 2)

*Gpu arguments are not necessary.

##Identity RNN -- Almost Ready for Testing Allows you to run an IRNN (on specified gpu of your choice). To call in:

from Seq2Seq_Upgrade_TensorFlow.seq2seq_upgrade import rnn_cell_enhanced

first_layer = rnn_cell_enhanced.IdentityRNNCell(size, gpu_for_layer = 0)

##Averaging Hidden States -- Testing Allows you to set ratio of last hidden state to mean hidden state

Simply call in:

from Seq2Seq_Upgrade_TensorFlow.seq2seq_upgrade import seq2seq_enhanced as sse

seq2seq_model = sse.embedding_attention_seq2seq(....,average_states = True, average_hidden_state_influence = 0.5)

Note that average_hidden_state_influence should be any real number between 0 and 1. The higher the number, the higher the percentage that the inputted hidden state will be influenced by the average hidden state.

For example, an average_hidden_state_influence = .75 means 75% of the hidden state will be the average hidden state and the remainder 25% of the hidden state will be the last hidden state. This allows you to choose how much you want the inputted hidden state to be affected by the previous timestep.

##Temperature Sampling During Decoding -- Working Allows you to use a temperature to alter the output of your softmax.

Higher temperature will result in more diversity in output, but more errors. Do not use this during training! Use this only while decoding after a model has been trained.

Note: This does not affect the internal decoding process at all. Rather, this is meant to replace the argmax function that is commonly used in greedy decoding.

To use:

from Seq2Seq_Upgrade_TensorFlow.seq2seq_upgrade import decoding_enhanced

for each_temperature in [0.2, 0.4, 0.6, 0.8, 1.0, 1.5]:
            outputs = []
            for logit in output_logits:
              select_word_number = int(decoding_enhanced.sample_with_temperature(logit[0], each_temperature))
              outputs.append(select_word_number)


            # If there is an EOS symbol in outputs, cut them at that point.
            if data_utils.EOS_ID in outputs:
              outputs = outputs[:outputs.index(data_utils.EOS_ID)]

            print() #put extra space between input and output
            print('--------Temperature is: ', each_temperature)
            print("Output Sentence", sentence_num, "Below")
            print(" ".join([rev_fr_vocab[output] for output in outputs])) #place space inbetween output here
            print()