Reproduce the paper on MIR_ST500 or MDB-stem-synth

Hi, I'm new here and I wanted to reproduce the result from the paper. Currently I can reproduce the result on MIR-1K, the default config matches the meta data of MIR-1K. However, I have troubles with the other two datasets: 1. Now I want to try MDB-stem-synth. As far as I know, the hop size of the annotation is `~2.8ms`, but your config is `20ms`. What did you do to get a frame of `20ms` for pitch note for this dataset? 2. The same question for MIR_ST500 which is a real dataset, not synthetic one. Thanks for the contribution.