From 4696e126410f71058d36eceb9564c783f5375207 Mon Sep 17 00:00:00 2001 From: William Falcon Date: Fri, 12 Jul 2019 15:53:45 -0400 Subject: [PATCH] fixed nccl init --- pytorch_lightning/models/trainer.py | 1 + 1 file changed, 1 insertion(+) diff --git a/pytorch_lightning/models/trainer.py b/pytorch_lightning/models/trainer.py index da8f5559..a6d8e1e8 100644 --- a/pytorch_lightning/models/trainer.py +++ b/pytorch_lightning/models/trainer.py @@ -379,6 +379,7 @@ class Trainer(TrainerIO): :param tries: :return: """ + sleep(self.proc_rank*2) root_node = os.environ['SLURM_NODELIST'].split(' ')[0] os.environ['MASTER_ADDR'] = root_node os.environ['MASTER_PORT'] = f'{port}'