Drylab Progress 2020
TODO List:
- Consider torch.cuda. I wrote my code to be run on a CPU, but it would be much more efficient to train on a GPU in the future. Luckily, I don't think the necessary changes are that large...hopefully (Alston)
- If necessary, consider using in-place operations to save memory (Alston).
- Work out how to split the meltome dataset so that it provides a good training and testing environment. Considering 3 partitions, train, validation, and test. Perhaps 80/10/10? or 60/20/20? or 60/30/10? (Ryan)
- Develop a system of easy training of future classifiers (Ryan)
Datasets in the drive: https://drive.google.com/drive/folders/1jCpfrNbQasSjeHxNEaMxZtv0VBX690ky?usp=sharing
- UniRep vectors for Meltome dataset
- UniRep vectors for GFP dataset