[BUG🐛] Title: Incompatibility Between XTTS_v2 and DVAE Decoder Leads to Inference Errors

[elvinzade]

Title: Incompatibility Between XTTS_v2 and DVAE Decoder Leads to Inference Errors

Description:

I encountered an issue while training the XTTS_v2 model using the DVAE Decoder. The provided solution on GitHub is configured to work with the Hifi-GAN Encoder. When running inference with the DVAE Decoder setup, the following error occurs:

Error Output:

Unexpected keys:

Unexpected key(s) in state_dict: "dvae.codebook.cluster_size", "dvae.codebook.embed", "dvae.codebook.embed_avg"

Missing keys:

Missing key(s) in state_dict: "mel_stats", "conditioning_encoder.init.weight", "conditioning_encoder.init.bias"

[mlinmg]

We'll add support for the VAE vocoder in future releases, atm only the gan vocoder is supported