From 9f41a9e8b78d8774334b48aeb7d0707f7c459de2 Mon Sep 17 00:00:00 2001 From: William Falcon Date: Fri, 12 Jul 2019 16:35:20 -0400 Subject: [PATCH] fixed nccl init --- pytorch_lightning/models/trainer.py | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) diff --git a/pytorch_lightning/models/trainer.py b/pytorch_lightning/models/trainer.py index 9c58c05b..c7bbc1a6 100644 --- a/pytorch_lightning/models/trainer.py +++ b/pytorch_lightning/models/trainer.py @@ -378,11 +378,7 @@ class Trainer(TrainerIO): port = 12910 os.environ['MASTER_PORT'] = f'{port}' - print('-'*100) - print(f'PORT: {port}') - print('-'*100) - - sleep(self.proc_rank * 2) + sleep(self.proc_rank*0.5) root_node = os.environ['SLURM_NODELIST'].split(' ')[0] os.environ['MASTER_ADDR'] = root_node