Can I use the same parameters for Speech Overlap for the segmentation checkpoint of branches @Interspeech2021 and @main in hf?

[guilhermehge]

I am using the Speech Overlap Pipeline and I see that the config.yaml refers to a branch in HF called @Interspeech2021. I see that there is an updated version of the segmentation model at the @main branch.

Question is, can I use the same hyperparameters that are in the config.yaml file that refers to the Interspeech2021 model for the @ main model?

Hyperparameters in the config.yaml file
min_duration_off: 0.09791355693027545 min_duration_on: 0.05537587440407595 offset: 0.4806866463041527 onset: 0.8104268538848918
These two checkpoints, while using the same hyperparameters, yield different results, I reckon that the older checkpoint tends to detect sounds like "hmm" or "hmpf" when detecting overlap, while the newer one, with the same parameters, tend to ignore those, or even the other way arount can also happen, this is why I am opening this discussion.

I do think that the newer model is better.

Thanks!