Getting segmentaiton fault on my docker machine. It has low RAM, 2GB, and i'm wondering if this is supported? I didnt find anything in the docs.
INFO:spleeter:Downloading model archive https://github.com/deezer/spleeter/releases/download/v1.4.0/2stems.tar.gz
INFO:spleeter:Validating archive checksum
INFO:spleeter:Extracting downloaded 2stems archive
INFO:spleeter:2stems model file(s) extracted
OMP: Info #212: KMP_AFFINITY: decoding x2APIC ids.
OMP: Info #210: KMP_AFFINITY: Affinity capable, using global cpuid leaf 11 info
OMP: Info #154: KMP_AFFINITY: Initial OS proc set respected: 0
OMP: Info #156: KMP_AFFINITY: 1 available OS procs
OMP: Info #157: KMP_AFFINITY: Uniform topology
OMP: Info #159: KMP_AFFINITY: 1 packages x 1 cores/pkg x 1 threads/core (1 total cores)
OMP: Info #214: KMP_AFFINITY: OS proc to physical thread map:
OMP: Info #171: KMP_AFFINITY: OS proc 0 maps to package 0
OMP: Info #250: KMP_AFFINITY: pid 202 tid 202 thread 0 bound to OS proc set 0
OMP: Info #250: KMP_AFFINITY: pid 202 tid 224 thread 1 bound to OS proc set 0
OMP: Info #250: KMP_AFFINITY: pid 202 tid 225 thread 2 bound to OS proc set 0
/entrypoint.sh: line 20: 202 Segmentation fault (core dumped) spleeter separate -i $f -o /music/split -p spleeter:2stems-16kHz
OS | Ubuntu
Installation type | conda in custom docker image
RAM available | 2GB
Hardware spec | Intel CPU
Hi @peerjuettner
It may be link to #229. In any case 2Gb is probably too low to have spleeter process complete audio files.
Thanks for the reply. As the processing speed doesn't really matter to me, I will try upping the swap space. I'll report if this helps.
Edit: Sometimes, the application also gets killed with this message:
INFO:spleeter:Downloading model archive https://github.com/deezer/spleeter/releases/download/v1.4.0/2stems.tar.gz
INFO:spleeter:Validating archive checksum
INFO:spleeter:Extracting downloaded 2stems archive
INFO:spleeter:2stems model file(s) extracted
OMP: Info #212: KMP_AFFINITY: decoding x2APIC ids.
OMP: Info #210: KMP_AFFINITY: Affinity capable, using global cpuid leaf 11 info
OMP: Info #154: KMP_AFFINITY: Initial OS proc set respected: 0
OMP: Info #156: KMP_AFFINITY: 1 available OS procs
OMP: Info #157: KMP_AFFINITY: Uniform topology
OMP: Info #159: KMP_AFFINITY: 1 packages x 1 cores/pkg x 1 threads/core (1 total cores)
OMP: Info #214: KMP_AFFINITY: OS proc to physical thread map:
OMP: Info #171: KMP_AFFINITY: OS proc 0 maps to package 0
OMP: Info #250: KMP_AFFINITY: pid 83 tid 83 thread 0 bound to OS proc set 0
OMP: Info #250: KMP_AFFINITY: pid 83 tid 105 thread 1 bound to OS proc set 0
OMP: Info #250: KMP_AFFINITY: pid 83 tid 106 thread 2 bound to OS proc set 0
/entrypoint.sh: line 20: 83 Killed spleeter separate -i $f -o /music/split -p spleeter:2stems-16kHz
i have read that someone have execute spleeter with good result in a 4gb machine with 2stems, 4 and 5 stems are for more than 8gb ( i have to buy more ddr ;o) )
Hi everyone.
We will soon release a new version (1.5) with a small change allowing to use librosa implementation of the STFT instead of the tensorflow one that appears to be very memory consuming. The PR is #282 so feel free to test it and/or join the conversation there.
Hi @peerjuettner
We've released spleeter 1.5, could you check if this new version solves your problem ?
With some 10 minutes songs I have OOMKill even I have 16GB of RAM ;)
@mmoussallam I'll check the latest version!
@mmoussallam is this already on conda? It keeps downloading model 1.4 for me.
Hi @peerjuettner,
We seem to have an issue with the conda package on windows, so we're stuck on 1.4. We're looking into it.
@mmoussallam Thanks for the heads up. Could I build from source and try again or is there another way to test the newest release? I'm eager to try it.
Hi @peerjuettner
the issus was solved and you should probably have no issue anymore.
Most helpful comment
Hi everyone.
We will soon release a new version (1.5) with a small change allowing to use
librosaimplementation of the STFT instead of the tensorflow one that appears to be very memory consuming. The PR is #282 so feel free to test it and/or join the conversation there.