search

yizhilll / MERT

Official implementation of the paper "Acoustic Music Understanding Model with Large-Scale Self-supervised Training".
Apache License 2.0
301 stars 18 forks source link

What is the sampling rate during training? Is it okay to upsample 16k to 24k during testing? #9

Closed xuanjihe closed 11 months ago

xuanjihe commented 12 months ago

What is the sampling rate during training? Is it okay to upsample 16k to 24k during testing?

yizhilll commented 11 months ago

In the released results, all the audios are resampled to the frequency same as in training.

However, we've tried testing the models with the non-training frequency and the model still can comprehend the audios (but necessarily perform the same).