search

NVIDIA / tacotron2

Tacotron 2 - PyTorch implementation with faster-than-realtime inference
BSD 3-Clause "New" or "Revised" License
5.12k stars 1.39k forks source link

AssertionError: Default process group is not initialized #519

Closed jinny1208 closed 3 years ago

jinny1208 commented 3 years ago

I am trying to do inference based on a model trained with 2 GPUs, and with distributed_run=True and fp16_run=False set in hparams.py. The command used to train the model is python -m multiproc train.py -o [output path] -l [log path].

Training works well, but when I load the model for inference, the following error message occurs. I do not know what the problem is. (The same problem occurs when I do inference on a model that was trained with only one GPU.)

8-23

I tried looking for different ways to solve this, but was not able to. I would appreciate it if someone could suggest possible solutions. Thanks in advance.

jinny1208 commented 3 years ago

Deleting the ipynb checkpoints and reloading the inference file worked. Still not sure what exactly the problem was though.