Code Implementation for the paper Self-Supervised Representation Learning for Basecalling.
This implementation is built over SACall. Hence, for supervised training and finetuning it is necessary to install ctc-decode.
To perform Self-Supervised pre-training, move to the corresponding directory and run the train_simclr.py or train_wav2vec2.py.
We also provide the pre-trained models if you wish to fine-tune them with your own data.
We are working on improving the data loader for it to be more memory efficent.