延長訓練schedule steps，合成雜音多

請問我這樣修改對嘛?
訓練完還是一堆雜音，語料大概161小時。204個語者。有沒有辦法預估這樣的語料量需要訓練幾個step?


The issue names no files, tests, or entry points; it only includes an image of a training-schedule change. Start by reviewing that change and the training configuration it refers to, then reproduce the noisy synthesis with the reported 161 hours and 204 speakers. Done means documenting whether the step count explains the noise and what training guidance is supported.

Written by the indexing model from the issue text.