out of memory when inference

[xiaolongwu0713]

Hi, I was following the example in the github about how to use the pre-trained model at : https://github.com/kan-bayashi/ParallelWaveGAN#how-to-use-pretrained-models ;
I used this model: csmsc_parallel_wavegan.v1;
But when I run the last step: $ parallel-wavegan-decode
--checkpoint pretrain_model/<pretrain_model_tag>/checkpoint-400000steps.pkl
--dumpdir dump/sample/norm
--outdir sample
I have the OOM error:

What should I do now?

[xiaolongwu0713]

It still OOM even though I used a very short audio, like 5 seconds. It is the same when I force the program to use the CPU.