Empty Audio Issue in Fine-Tuning
#3
by
Shanos76
- opened
Hi,
I'm trying to fine-tune the model for a single speaker, but after training, it only generates empty audio with some noise.
I'm using Flan-t5-Large as a tokenizer, and I noticed that the config.json file includes a path to the decoder. Could this be something I'm missing? Your assistance would be greatly appreciated!