/whisper-ctc

Whisper Encoder (extracted from pretrained) with a Linear on top and solve using CTC criterion

Primary LanguagePython

Whisper Encoder + CTC

Install dependencies

apt install sox
apt-get install espeak-ng -y
pip install -r requirements.txt + sox

Usage

python main.py

for more configuration, please refer to https://hydra.cc/docs/intro/

python finetuning.py \
    data_cfg.dataloader.batch_size=16 \
    experiment_cfg.train=True