EveryVoice
Optimize fs2 preprocess by maximizing parallelization benefits
#40

Merged

Commits

perf: use joblib.Parallel instead of pathos.multiprocessing for faster results

joanise committed 1 year ago
refactor: make function process_one_audio in prep for parallelization

joanise committed 1 year ago
perf: parallelize audio processing too

joanise committed 1 year ago
ci: fix CI by using pytorch==2.0.1 and torchaudio==2.0.2 explicitly

joanise committed 1 year ago
fix: make counters compatible with multi-processing

joanise committed 1 year ago
style: make torch and rich package use explicit

joanise committed 1 year ago
fix(tests): failing tests is an error, fix dataloader tests to pass

joanise committed 1 year ago
perf: gather stats in parallel too, plus some style improvements

joanise committed 1 year ago
fix(deps): remove pathos dependency, no longer used

joanise committed 1 year ago
ci: fix the installation of torch and friends

joanise committed 1 year ago