Is DDP training and nn.SyncBatchNorm.convert_sync_batchnorm supported ? - Githubissues

mit-han-lab / torchsparse

[MICRO'23, MLSys'22] TorchSparse: Efficient Training and Inference Framework for Sparse Convolution on GPUs.

https://torchsparse.mit.edu

MIT License

1.22k stars 143 forks source link

Is DDP training and nn.SyncBatchNorm.convert_sync_batchnorm supported ? #332

Open barzanisar opened 1 month ago

barzanisar commented 1 month ago

Hey,

Is DDP training and nn.SyncBatchNorm.convert_sync_batchnorm supported ?

NoOne-AndYou commented 1 week ago

I also want to know this