Speech-to-text models

Pytorch implementation and comparison speech-to-text (STT) models.

References:

• Deep Speech: Scaling up end-to-end speech recognition
• Wav2Letter (WIP)
• Jasper: An End-to-End Convolutional Neural Acoustic Model (WIP)

Train on CPU/GPU

python main.py --num-workers 0 --batch-size 32 --train-data-urls train-clean-100 train-clean-360 --num-epochs 15 --window-stride 20 --optimizer adam --learning-rate 3e-4 --log-steps 100 --checkpoint test

• trained on train-clean-100 train-clean-360.
• WER on dev-clean (9 epochs): 0.33
• pre-trained weights: https://github.com/discort/stt_models/releases/tag/0.1

ToDo:

• Parallelize model training
• write checkpointing and tensorboard logger
• training using mixed precision