seq2seq-chatbot

A sequence2sequence chatbot implementation with TensorFlow.

Chatting with a trained model

For console chat:

• Run chatconsolebestweightstraining.bat or chatconsolebestweightsvalidation.bat

For web chat:

• Run chatwebbestweightstraining.bat or chatwebbestweightsvalidation.bat

• Open a browser to the URL indicated by the server console, followed by /chatui.html. This is typically: http://localhost:8080/chatui.html

To chat with a trained model from a python console:

• Set console working directory to the seq2seq-chatbot directory. This directory should have the models and datasets directories directly within it.

• Run chat.py with the model checkpoint path:

run chat.py models\cornellmoviedialog\trainedmodel\bestweights_training.ckpt

Training a model

To train a model from a python console:

• To train a new model, run train.py with the dataset path:

run train.py --datasetdir=datasets\dataset_name

Or to resume training an existing model, run train.py with the model checkpoint path:

run train.py --checkpointfile=models\datasetname\modelname\checkpoint.ckpt

Visualizing a model in TensorBoard

To start TensorBoard from a terminal:

tensorboard --logdir=model_dir

Dependencies

The following python packages are used in seq2seq-chatbot: (excluding packages that come with Anaconda)

• TensorFlow

pip install --upgrade tensorflow

For GPU support: (See here for full GPU install instructions including CUDA and cuDNN)

pip install --upgrade tensorflow-gpu

• jsonpickle

pip install --upgrade jsonpickle

• flask 0.12.4 and flask-restful (required to run the web interface)

pip install flask==0.12.4
    pip install --upgrade flask-restful