Skip to content

Repository files navigation

legality_check_model

This is the tiramisu legality check model. Setting it up is similar to the cost model. However, the utility files have been updated and will work differently.

pikle.py generates a dataset from the actual data.

generate_dataset.py generates data for the benchmark.

balance.py helps balance the legal and illegal schedules in the dataset to get better accuracy in the f1 score.

train_model.py for training and evaluate_model.py for evaluating.

You can set the configuration in the config.yaml

About

research project using deep learning to predict the legality of compiler transformation schedule

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages