From a315cebd6c46e087e20d60a066a3a2c09bbe947c Mon Sep 17 00:00:00 2001 From: niravshah241 Date: Thu, 21 Dec 2023 12:38:45 +0000 Subject: [PATCH] thermomech gpu in work --- .../dlrbnicsx_thermal_multigpu.py | 4 ++-- demo/thermomechanical_dlrbnicsx_multigpu/test_sbatch_gpu | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/demo/thermomechanical_dlrbnicsx_multigpu/dlrbnicsx_thermal_multigpu.py b/demo/thermomechanical_dlrbnicsx_multigpu/dlrbnicsx_thermal_multigpu.py index 8208878..40388c9 100644 --- a/demo/thermomechanical_dlrbnicsx_multigpu/dlrbnicsx_thermal_multigpu.py +++ b/demo/thermomechanical_dlrbnicsx_multigpu/dlrbnicsx_thermal_multigpu.py @@ -490,7 +490,7 @@ def generate_ann_output_set(problem, reduced_problem, input_set, print("\n") - thermal_gpu_group0_procs = world_comm.group.Incl([0, 1, 2, 3]) + thermal_gpu_group0_procs = world_comm.group.Incl([0, 1, 2]) thermal_gpu_group0_comm = world_comm.Create_group(thermal_gpu_group0_procs) # ANN model @@ -499,7 +499,7 @@ def generate_ann_output_set(problem, reduced_problem, input_set, if thermal_gpu_group0_comm != MPI.COMM_NULL: - cuda_rank_list = [0, 1, 2, 3] + cuda_rank_list = [0, 1, 2] init_gpu_process_group(thermal_gpu_group0_comm) thermal_training_set_indices_gpu = np.arange(thermal_gpu_group0_comm.rank, diff --git a/demo/thermomechanical_dlrbnicsx_multigpu/test_sbatch_gpu b/demo/thermomechanical_dlrbnicsx_multigpu/test_sbatch_gpu index 03cbdc8..47b9a02 100644 --- a/demo/thermomechanical_dlrbnicsx_multigpu/test_sbatch_gpu +++ b/demo/thermomechanical_dlrbnicsx_multigpu/test_sbatch_gpu @@ -5,10 +5,10 @@ #SBATCH --error=test_job_gpu_%A_%a.err #SBATCH --nodes=1 -#SBATCH --gres=gpu:4 -#SBATCH --ntasks=8 +#SBATCH --gres=gpu:3 +#SBATCH --ntasks=16 #SBATCH --time=02:00:00 -#SBATCH --mem=7000mb +#SBATCH --mem=3500mb #SBATCH -p ampere