Hi, I understand I can use @rank_zero_only
to wrap a whole function, but is it possible to only run part of the function on rank 0 and the rest on all ranks? like:
def on_train_batch_start(self, batch, batch_idx):
with rank_zero_only:
...
# on all ranks
self.examples = batch