update configs and examples

molmod · Jul 20, 2024 · d4f4bcc · d4f4bcc
1 parent b061b87
commit d4f4bcc
Show file tree

Hide file tree

Showing 6 changed files with 89 additions and 8 deletions.
diff --git a/configs/hortense.yaml b/configs/hortense.yaml
@@ -28,15 +28,15 @@ ModelTraining:
     walltime: "12:00:00"
     scheduler_options: "#SBATCH --clusters=dodrio\n#SBATCH --gpus=1\n"
 CP2K:
-  cores_per_worker: 64
+  cores_per_worker: 128
   max_evaluation_time: 30
   launch_command: 'apptainer exec -e --no-init oras://ghcr.io/molmod/cp2k:2023.2 /opt/entry.sh mpirun -np 32 --bind-to core --map-by core -x OMP_NUM_THREADS=1 cp2k.psmp -i cp2k.inp'
   slurm:
     partition: "cpu_rome"
     account: "2024_079"
     nodes_per_block: 1
-    cores_per_node: 64
-    max_blocks: 2
+    cores_per_node: 32
+    max_blocks: 25
     walltime: "06:00:00"
     scheduler_options: "#SBATCH --clusters=dodrio\n"
 ...
diff --git a/configs/lumi.yaml b/configs/lumi.yaml
@@ -29,11 +29,11 @@ ModelTraining:
   cores_per_worker: 7
   gpu: true
   slurm:
-    partition: "small-g"
+    partition: "standard-g"
     account: "project_465001125"
     nodes_per_block: 1
-    cores_per_node: 7
+    cores_per_node: 56
     walltime: "01:00:00"
-    scheduler_options: "#SBATCH --gres=gpu:1\n"
+    scheduler_options: "#SBATCH --gres=gpu:8\n"
     worker_init: "ml LUMI/23.09 && ml partition/G && ml rocm/5.6\n"
 ...
diff --git a/examples/hortense.yaml b/examples/hortense.yaml
@@ -0,0 +1,42 @@
+---
+parsl_log_level: DEBUG
+container_engine: 'apptainer'
+container_uri: 'oras://ghcr.io/molmod/psiflow:main_cu118'
+default_threads: 8
+ModelEvaluation:
+  cores_per_worker: 12
+  gpu: True
+  max_simulation_time: 20
+  slurm:
+    partition: "gpu_rome_a100"
+    account: "2023_070"
+    nodes_per_block: 1
+    cores_per_node: 48
+    max_blocks: 1
+    walltime: "12:00:00"
+    scheduler_options: "#SBATCH --clusters=dodrio\n#SBATCH --gpus=4\n"
+ModelTraining:
+  cores_per_worker: 12
+  gpu: true
+  max_training_time: 40
+  slurm:
+    partition: "gpu_rome_a100"
+    account: "2023_070"
+    nodes_per_block: 1
+    cores_per_node: 12
+    max_blocks: 1
+    walltime: "12:00:00"
+    scheduler_options: "#SBATCH --clusters=dodrio\n#SBATCH --gpus=1\n"
+CP2K:
+  cores_per_worker: 64
+  max_evaluation_time: 30
+  launch_command: 'apptainer exec -e --no-init oras://ghcr.io/molmod/cp2k:2023.2 /opt/entry.sh mpirun -np 32 --bind-to core --map-by core -x OMP_NUM_THREADS=1 cp2k.psmp -i cp2k.inp'
+  slurm:
+    partition: "cpu_rome"
+    account: "2024_079"
+    nodes_per_block: 1
+    cores_per_node: 64
+    max_blocks: 2
+    walltime: "06:00:00"
+    scheduler_options: "#SBATCH --clusters=dodrio\n"
+...
diff --git a/examples/lumi.yaml b/examples/lumi.yaml
@@ -0,0 +1,39 @@
+---
+parsl_log_level: DEBUG
+container_engine: 'singularity'
+container_uri: 'oras://ghcr.io/molmod/psiflow:main_rocm5.6'
+default_threads: 8
+CP2K:
+  cores_per_worker: 32
+  max_evaluation_time: 20
+  launch_command: 'singularity exec -e --no-init oras://ghcr.io/molmod/cp2k:2023.2 /opt/entry.sh mpirun -np 32 -x OMP_NUM_THREADS=1 cp2k.psmp -i cp2k.inp'
+  slurm:
+    partition: "standard"
+    account: "project_465001125"
+    nodes_per_block: 1
+    cores_per_node: 128
+    max_blocks: 10
+    walltime: "01:00:00"
+ModelEvaluation:
+  cores_per_worker: 7
+  gpu: True
+  slurm:
+    partition: "standard-g"
+    account: "project_465001125"
+    nodes_per_block: 1
+    cores_per_node: 56
+    walltime: "01:00:00"
+    scheduler_options: "#SBATCH --gres=gpu:8\n"
+    worker_init: "ml LUMI/23.09 && ml partition/G && ml rocm/5.6\n"
+ModelTraining:
+  cores_per_worker: 7
+  gpu: true
+  slurm:
+    partition: "standard-g"
+    account: "project_465001125"
+    nodes_per_block: 1
+    cores_per_node: 56
+    walltime: "01:00:00"
+    scheduler_options: "#SBATCH --gres=gpu:8\n"
+    worker_init: "ml LUMI/23.09 && ml partition/G && ml rocm/5.6\n"
+...
diff --git a/examples/submit_hortense.sh b/examples/submit_hortense.sh
@@ -11,7 +11,7 @@ files=(
     "alanine_replica_exchange.py"
 )
 
-curl -O https://raw.githubusercontent.com/molmod/psiflow/main/configs/hortense.yaml
+curl -O https://raw.githubusercontent.com/molmod/psiflow/main/examples/hortense.yaml
 
 run_dir=$(pwd)/run_examples
 mkdir $run_dir && cp hortense.yaml $run_dir && cd $run_dir

diff --git a/examples/submit_lumi.sh b/examples/submit_lumi.sh
@@ -11,7 +11,7 @@ files=(
     "alanine_replica_exchange.py"
 )
 
-curl -O https://raw.githubusercontent.com/molmod/psiflow/main/configs/lumi.yaml
+curl -O https://raw.githubusercontent.com/molmod/psiflow/main/examples/lumi.yaml
 
 run_dir=$(pwd)/run_examples
 mkdir $run_dir && cp lumi.yaml $run_dir && cd $run_dir