denispeskoff / qbasr_generatedata

QBASR_GenerateData

The following document allows you to Slurm documents. There are two parts.

Part 1: Audio Generation

This includes submit.sh, template.sh, QBASR_GenerateData.py

1. Add file in proper json format from https://sites.google.com/view/qanta/datasets?authuser=0
2. Update QBASR_GenerateData to save in the right place
3. Update template.sh to refer to the proper anaconda path, file, etc.
4. Run "submit.sh 48" (or appropriate amount of partitions)

Part 2: Decoding

This includes submit_decode.sh, template_decode.sh, QBASR_GenerateFileList, QBASR_DecodeAudio.py Depends on kaldi installation and ffmpeg (to make .wav files from .mp3s)

1. Run QBASR_GenerateFileList on the proper folder e.g. "/dev/*.mp3"
2. Update template.sh to refer to proper file and DecodeAudio to write to correct location
3. Run submit_decode.sh