nnAudio-CPU memory growth

[will-wiki]

Hello, I am using nnAudio to extract the CQT features of the audio in the following way

from nnAudio import features
cqt_layer = features.CQT2010v2(sr=16000).to(torch.device('cuda'))
data, sr = librosa.load(audio_file, sr=None)
 with torch.no_grad():
        data = torch.tensor(data, dtype=torch.float32).to(torch.device('cuda'))
        cqt = cqt_layer(data)[0].cpu().numpy()

However, the problem is that when the amount of data requested is fixed, increasing concurrency causes memory to keep growing

[gudgud96]

How do you increase concurrency? Using torch.multiprocessing? Maybe you can share your code with concurrency included.

[migperfer]

Hi! @will-wiki , what do you mean by "fixed amount of data"?