LSTM Based Essay Grading

Using LSTMs to automatically grade Kaggle's ASAP Essays.

Models

Pre-trained models can be downloaded from here.

Training

1. Download GloVe's 6B.100d word vectors from here. You can also train on 840B.300d for better results.
2. Set your desired config in config.py and simply run the training using train.py.
3. You can serve you own trained models by modifying models.config inside the /models directory.

Usage

In 1st Terminal

git clone https://github.com/tanmaypandey7/essay-grading
cd essay-grading
pip install -r requirements.txt
python app.py

In 2nd Terminal

sudo service docker start
# For GPU
docker run -t --rm --gpus all --name=tf-serving -p  8501:8501     \
-v "$(pwd)/models:/models/" tensorflow/serving:latest-gpu   \
     --model_config_file=/models/models.config
# For CPU
docker run -p  8501:8501 \
-v "$(pwd)/models:/models/" tensorflow/serving:latest   \
     --model_config_file=/models/models.config