Need guidance on training xtts_v2 model from scratch in Hindi

[lalit-mohanani]

I've been attempting to train the xtts_v2 model on my custom dataset in Hindi without using the xtts checkpoint, but I haven't been successful in getting any meaningful results. It seems I might be missing some crucial steps or configurations. Could someone please provide guidance on how to train the xtts_v2 model from scratch in Hindi?
I would greatly appreciate any assistance or pointers on the necessary preprocessing steps, model architecture adjustments, hyperparameter tuning, or any other relevant details that could help me achieve success in training the xtts_v2 model for Hindi.

[avii-km]

were you able to train model for hindi ?

[lalit-mohanani]

yes, but not from scrach but finetuned it.

[avii-km]

kindly share implementation for fine tuning for hindi lang dataset

[lalit-mohanani]

You can refer https://docs.coqui.ai/en/latest/models/xtts.html

[VivekGangurde94]

File /opt/conda/lib/python3.10/site-packages/IPython/core/ultratb.py:1131, in VerboseTB.get_records(self, etb, number_of_lines_of_context, tb_offset)
1129 while cf is not None:
1130 try:
-> 1131 mod = inspect.getmodule(cf.tb_frame)
1132 if mod is not None:
1133 mod_name = mod.name

AttributeError: 'tuple' object has no attribute 'tb_frame'
did u got this error while fine tunning

[lalit-mohanani]

Nope

[Aml-Hassan-Abd-El-hamid]

File /opt/conda/lib/python3.10/site-packages/IPython/core/ultratb.py:1131, in VerboseTB.get_records(self, etb, number_of_lines_of_context, tb_offset) 1129 while cf is not None: 1130 try: -> 1131 mod = inspect.getmodule(cf.tb_frame) 1132 if mod is not None: 1133 mod_name = mod.name

AttributeError: 'tuple' object has no attribute 'tb_frame' did u got this error while fine tunning

I had the same issue trying to fine-tune for the Egyptian dialect, did you find any solution? @VivekGangurde94