Trying to fine-tune Tacotron2-DDC checkpoint but coarse_decoder layers are missing

[serg06]

Hello,

I want to finetune the tts_models/en/ljspeech/tacotron2-DDC model, but the checkpoint is missing coarse_decoder layers. Is there a way to get the full checkpoint that has those layers? @erogol I notice you mentioned that such a checkpoint exists?

[erogol]

Unfortunately, I don't keep the full checkpoint.

But you can initiate training from that model and start with a newly initialized coarse decoder. That'd not make a big difference.