search

rhasspy / piper

A fast, local neural text to speech system
https://rhasspy.github.io/piper-samples/
MIT License
6.24k stars 457 forks source link

Floating point exception (core dumped) #7

Closed xuezhongfei2008 closed 1 year ago

xuezhongfei2008 commented 1 year ago

turn onnx

synesthesiam commented 1 year ago

Can you provide more information please?

xuezhongfei2008 commented 1 year ago

Can you provide more information please?

hello, my environment: Linux ubuntu516 4.15.0-112-generic #113-Ubuntu SMP Thu Jul 9 23:41:39 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux Python 3.8.13 torch 1.12.1 torchaudio 0.12.1 torchinfo 1.7.1 torchmetrics 0.9.0 torchvision 0.13.1 cudatoolkit 11.3.1
onnxruntime 1.13.1

when usesdp=False G.pth convert to onnx ,is ok when usesdp=True G.pth convert to onnx Floating point exception (core dumped)

synesthesiam commented 1 year ago

The model itself comes from here: https://github.com/jaywalnut310/vits/ I've never had this problem when exporting a model with use_sdp=True. I would either try a different version of PyTorch or lower OPSET_VERSION in export_onnx.py

tuannvhust commented 1 year ago

@xuezhongfei2008 i got the same issue, did you solve this bug?

d-r-a-b commented 1 year ago

Although there are no details in this issue, I believe I experienced the same bug while trying to run python -m larynx_train.export_onnyx. For my own case, I was able to fix it by editing requirements.txt to use newer version of torch and onnxruntime. For reference, I am on Python 3.10.9, using Arch Linux through WSL.

requirements.txt:

cython>=0.29.0,<1
espeak-phonemizer>=1.1.0,<2
librosa>=0.9.2,<1
numpy>=1.19.0
onnxruntime~=1.13.0
pytorch-lightning~=1.7.0
torch~=1.13.0

Then I just activate venv again and then pip3 install -r requirements.txt.

Does this fix things for others?