search

qq456cvb / Point-Transformers

Point Transformers
MIT License
609 stars 102 forks source link

add parallelism to the classification training. #32

Open exx8 opened 1 year ago

exx8 commented 1 year ago

add parallelism to the classification training.

qq456cvb commented 1 year ago

Hi, have you tested if the performance keeps the same or similar, since dataparallel may affect the batchnorm behavior? See https://discuss.pytorch.org/t/how-does-dataparallel-handels-batch-norm/14040