Sentence classification on the Sentihood dataset with a couple CRF models.
Code is tested with torch 1.0.0, pyro 0.3.0, and torchtext 0.4.0, and spacy 2.0.18.
Data was obtained from jack.
Run the models with the following command:
python main.py --flat-data --model {lstmfinal | crflstmdiag | crflstmlstm | crfemblstm} {--save} {--devid #}
Add the --save option to save checkpoints along with their valid and test performance.
Be sure to create the save/{modelname} directory before saving.
Models are saved based on validation accuracy.
Use the --devid # option to utilize a GPU.