in train.py, if config.inference['normalize'] is True, transform estimates after demix

ZFTurbo / Music-Source-Separation-Training

Repository for training models for music source separation.

MIT License

462 stars 63 forks source link

in train.py, if config.inference['normalize'] is True, transform estimates after demix #68

Open hunterhogan opened 2 months ago

hunterhogan commented 2 months ago

When the new parameter in config.inference, normalize, was added in this commit, logic was added to valid.py and inference.py to remove the normalization transformation from estimates. In train.py, the logic to transform mix was added, however, nothing was added for estimates.

https://github.com/ZFTurbo/Music-Source-Separation-Training/blob/54bd7259dca5864bd38c7962ec36d05579aba7e7/train.py#L182-L212

Would this affect the SDR calculation in sdr_val = sdr(references, estimates)[0]?

ZFTurbo commented 2 months ago

I added normalize only to support one model (SCNet) for inference. I didn't change training to support normalization.

Kenus23 commented 3 weeks ago

how use training for stem separation?