checkpoint only on rank=0 now

This commit is contained in:
William Falcon
2019-07-03 18:14:34 -04:00
parent 75e32daad4
commit f101152650
+2 -1
View File
@@ -588,4 +588,5 @@ class Trainer(TrainerIO):
# model checkpointing
print('save callback...')
self.checkpoint_callback.on_epoch_end(epoch=self.current_epoch, logs=self.__tng_tqdm_dic)
if self.proc_rank == 0:
self.checkpoint_callback.on_epoch_end(epoch=self.current_epoch, logs=self.__tng_tqdm_dic)