CSE481N NLP Capstone Project (Project Webpage)
This project uses Python 3.
First, create a Python 3 virtual environment using python -m venv venv.
Next, activate the virtual environment with source venv/bin/activate.
Dependencies are listed in requirements.txt. They can be installed with pip
by using pip install -r requirements.txt.
Data is stored in the data directory and is not under version control.
There are scripts to download the various datasets.
SSH into nlpgpu01.cs.washington.edu and clone the repository.
The setup_gpu_machine.sh script will take care of:
CUDA requires some environment variables. These can be set using source cuda_path.sh.
Run source disable_tf_logging.sh.
The experiment.py script is a utility for experimenting with different model
parameters. It accepts multiple values for each parameter and tries all
possible combinations of them.
For example, the following command will try batch sizes of 16, 32, and 64, with question types of entity and location. It will always run for 10 epochs.
python experiment.py experiment_name -b 16 32 64 -q entity location -e 10Results are dumped to the tensorboard_models/ directory and can be viewed
by using tensorboard --logdir tensorboard_models/.
The name of each experiment includes the parameters used.
The error_analysis.py script takes the development dataset used, references, and one or more candidates.
It prints HTML which can be piped to a file.
This file must be in the error_analysis/ directory for the CSS to work.
python error_analysis.py datasets/msmarco/dev/location.json references/location.json candidates/attention-batch_size=128-epochs=50-hidden_size=100-keep_prob=0.3-learning_rate=0.01-question_type=location.json candidates/baseline-batch_size=1024-epochs=50-hidden_size=100-keep_prob=0.3-learning_rate=0.01-question_type=location.json > error_analysis/index.html