Hello!
I get this message when I try to generate features on my audio keyword recognition project for school: /home/create_features.sh: line 3: 7 Killed node “/app/node/windowing/build/window-time-series.js” “/home/input.json” “/home/output.json”
Application exited with code 137 (OOMKilled)
Im guessing this means out of memory?
I have not reached the limit for total audio length, but I have a lot of 1s files. Is there some workaround for this?
Project id: 85711
Thanks!
/Jonas
Hi @Instant_exit ,
From your project, I can see the feature generation and training steps went through.
Clinton
Yes, I took away data from the training side until it worked, so It seems the project is at its maximum capacity as is? It would be nice If I could add some more varieties of my keyword if possible.
cheers,
/j
Hi @Instant_exit ,
The free developer version is only limited to 20 mins of training time and 4GB of RAM. That’s why your feature generation or training crashes when you have a large set of training examples.
Regards,
Aha, ok! Thanks. I am well under the limit for audio total length so I figured I might ask. Im guessing a lot of small files eat up ram faster than fewer and longer ones. Anyway, thanks for replying to me, and for providing a great service for free.
1 Like
@Instant_exit thanks for reporting, have pinged infra on this too.
1 Like
Hi @Instant_exit - we’ve upped the memory limits for all jobs to be less stringent (they can go over memory limits without being killed immediately) and this should resolve all OOMKilled issues. We’re monitoring actively to see if any others happen and can tweak the limits if that’s the case.