pyannote-audio runs with --cpu flag even if --gpu is passed to ./run_ALICE.sh
fix:
on ./run_ALICE.sh line 32 should be:
GPU="gpu";
and line 47 should read:
bash $THISDIR/voice-type-classifier/apply.sh $THISDIR/tmp_data/ "MAL FEM" --device=$GPU |& sed '/^Took/d'
--device=gpu argument must be passed instead of --gpu
I think this bug was already in last version since my execution time for a 2 hour audio dropped from 70min to 10min
Hmm that's weird. We processed all ACLEW data on a computing cluster specifically using GPUs, but this was with an older version of pyannote. Maybe the syntax has changed. Anyway, fixed!
pyannote-audio
runs with--cpu
flag even if--gpu
is passed to ./run_ALICE.shfix: on ./run_ALICE.sh line 32 should be:
GPU="gpu";
and line 47 should read:bash $THISDIR/voice-type-classifier/apply.sh $THISDIR/tmp_data/ "MAL FEM" --device=$GPU |& sed '/^Took/d'
--device=gpu argument must be passed instead of --gpu
I think this bug was already in last version since my execution time for a 2 hour audio dropped from 70min to 10min