search

yl4579 / StyleTTS2

StyleTTS 2: Towards Human-Level Text-to-Speech through Style Diffusion and Adversarial Training with Large Speech Language Models
MIT License
4.98k stars 422 forks source link

In 2nd stage training AttributeError: 'AudioDiffusionConditional' object has no attribute 'module' #263

Open SandyPanda-MLDL opened 4 months ago

SandyPanda-MLDL commented 4 months ago

I am unable to debug this specific issue which is appearing in 2nd stage training as follows:

Traceback (most recent call last): File "/hdd2/Sandipan/SDhar-Projects/StyleTTS2-new/train_second.py", line 829, in main() File "/hdd2/Sandipan/envs/stylett2new/lib/python3.9/site-packages/click/core.py", line 1157, in call return self.main(args, kwargs) File "/hdd2/Sandipan/envs/stylett2new/lib/python3.9/site-packages/click/core.py", line 1078, in main rv = self.invoke(ctx) File "/hdd2/Sandipan/envs/stylett2new/lib/python3.9/site-packages/click/core.py", line 1434, in invoke return ctx.invoke(self.callback, ctx.params) File "/hdd2/Sandipan/envs/stylett2new/lib/python3.9/site-packages/click/core.py", line 783, in invoke return __callback(args, **kwargs) File "/hdd2/Sandipan/SDhar-Projects/StyleTTS2-new/train_second.py", line 353, in main model.diffusion.module.diffusion.sigma_data = s_trg.std(axis=-1).mean().item() # batch-wise std estimation File "/hdd2/Sandipan/envs/stylett2new/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1614, in getattr raise AttributeError("'{}' object has no attribute '{}'".format( AttributeError: 'AudioDiffusionConditional' object has no attribute 'module'