diff --git a/pytorch_lightning/models/trainer.py b/pytorch_lightning/models/trainer.py index ba5e7272..79d2c2dc 100644 --- a/pytorch_lightning/models/trainer.py +++ b/pytorch_lightning/models/trainer.py @@ -394,6 +394,7 @@ class Trainer(TrainerIO): # whenever we have the correct number of tasks, we let slurm manage processes # otherwise we launch the required number of processes nb_requested_gpus = len(self.data_parallel_device_ids) * self.nb_gpu_nodes + nb_slurm_tasks = 0 try: nb_slurm_tasks = int(os.environ['SLURM_NTASKS']) is_slurm_managing_tasks = nb_slurm_tasks == nb_requested_gpus