kalihale/costa_w_qa

COSTA for QA

By Kali Hale and Setareh Najafi Khoshnoo

Run instructions

To run as-is, run run.sh.

A number of learning rates can be passed with the -lr option.

The number of epochs (an int) should be passed with -e.

run.sh defines the number of epochs and the learning rates that were used in our project.

Metrics output

The loss is output to the /runs folder using Tensorboard and SummaryWriter. The information can be viewed live by running the command tensorboard --logdir=./runs in the local directory, or run.sh will run the command when the training and testing is done.

The test results are printed to the console after testing of each model is complete.

The GitHub repository is at https://github.com/kalihale/COSTA_w_QA