out-of-memory

[Z-Times]

Hi, when i use the instruction python synthesize_fuse.py -S data/may -M output/may_talkingface --use_train --audio data/may/aud.npy, the error comes: slurmstepd: error: Detected 1 oom-kill event(s) in StepId=3618241.batch cgroup. Some of your processes may have been killed by the cgroup out-of-memory handler.
So i am wondering how much memory should i have to work it. And how large the batch_size is? Thank you.

[Fictionarry]

Hi, it should take memory less than the training process. If you successfully run the training, you can run the final inference as well with the same device. To fully load the data into the memory, it takes about 60GB data for the sample of May. However, most of these data are not in use in synthesize_fuse.py. You can manually code the dataloader to prune them.

[Z-Times]

Thank you, I tried several times and finally it worked.

[Ahmed-Aboalasaad]

Thank you, I tried several times and finally it worked.

Hello, Z-Times! I'm still stuck in the training process (python data_utils/process.py data//.mp4).
I have an RTX 3060 Ti Desktop with 8GB of vRAM and 64GB of RAM.
Is this enough? And should I have some special configurations so that it works properly?