diff --git a/pytorch_lightning/models/trainer.py b/pytorch_lightning/models/trainer.py index f2252186..f1e1881d 100644 --- a/pytorch_lightning/models/trainer.py +++ b/pytorch_lightning/models/trainer.py @@ -289,7 +289,11 @@ class Trainer(TrainerIO): :return: """ # node rank using relative slurm id - node_rank = int(os.environ['SLURM_NODEID']) + # otherwise default to node rank 0 + try: + node_rank = int(os.environ['SLURM_NODEID']) + except KeyError as e: + node_rank = 0 # recover original exp before went into process self.experiment = self.experiment.get_non_ddp_exp()