How to let the model randomly choose a voice timbre in command line inference?

[Uitwaaien1123]

I want the model randomly choose a voice timbre in command line inference, I have seen the note "If you plan to let the model randomly choose a voice timbre, you can skip this step." in the step1 (Generate prompt from voice).
But I have no idea about what to do in the Step 2(Generate semantic tokens from text).
1)when I run the command line like
"python tools/llama/generate.py
--text "The text you want to convert"
--checkpoint-path "checkpoints/fish-speech-1.4"
--num-samples 2 "
it turns out "get extra arguments"
2)when I use an empty file "fake.npy", it turns out something like "feature is empty".
So what can I do?