git profiling tools

root_gnn_dgl/configs/stats_100K/ttH_CP_even_vs_odd_batch_size_2048.yaml

@@ -0,0 +1,57 @@
+Training_Name: ttH_CP_even_vs_odd_batch_size_2048
+Training_Directory: trainings/stats_100K/ttH_CP_even_vs_odd_batch_size_2048
+Model:
+  module: models.GCN
+  class: Edge_Network
+  args:
+    hid_size: 64
+    in_size: 7
+    out_size: 1
+    n_layers: 4
+    n_proc_steps: 4
+    dropout: 0
+Training:
+  epochs: 500
+  batch_size: 2048
+  learning_rate: 0.0001
+  gamma: 0.99
+Datasets:
+  ttH_CP_even: &dataset_defn
+    module: root_gnn_base.dataset
+    class: LazyDataset
+    shuffle_chunks: 3
+    batch_size: 2048
+    padding_mode: NONE #one of STEPS, FIXED, or NONE
+    args: &dataset_args
+      name: ttH_CP_even
+      label: 0
+      # weight_var: weight
+      chunks: 3
+      buffer_size: 2
+      file_names: ttH_NLO.root
+      tree_name: output
+      fold_var: Number
+      raw_dir: /global/cfs/projectdirs/trn007/lbl_atlas/data/stats_100K/
+      save_dir: /global/cfs/projectdirs/trn007/lbl_atlas/data/processed_graphs/stats_100K/ttH_CP_even_vs_odd_batch_size_2048/
+      node_branch_names:
+        - [jet_pt, ele_pt, mu_pt, ph_pt, MET_met]
+        - [jet_eta, ele_eta, mu_eta, ph_eta, 0]
+        - [jet_phi, ele_phi, mu_phi, ph_phi, MET_phi]
+        - CALC_E
+        - [jet_btag, 0, 0, 0, 0]
+        - [0, ele_charge, mu_charge, 0, 0]
+        - NODE_TYPE
+      node_branch_types: [vector, vector, vector, vector, single]
+      node_feature_scales: [1e-1, 1, 1, 1e-1, 1, 1, 1]
+    folding:
+      n_folds: 4
+      test: [0]
+      # validation: 1
+      train: [1, 2, 3]
+  ttH_CP_odd:
+    <<: *dataset_defn
+    args:
+      <<: *dataset_args
+      name: ttH_CP_odd
+      label: 1
+      file_names: ttH_CPodd.root

root_gnn_dgl/configs/stats_100K/ttH_CP_even_vs_odd_batch_size_4096.yaml

@@ -0,0 +1,57 @@
+Training_Name: ttH_CP_even_vs_odd_batch_size_4096
+Training_Directory: trainings/stats_100K/ttH_CP_even_vs_odd_batch_size_4096
+Model:
+  module: models.GCN
+  class: Edge_Network
+  args:
+    hid_size: 64
+    in_size: 7
+    out_size: 1
+    n_layers: 4
+    n_proc_steps: 4
+    dropout: 0
+Training:
+  epochs: 500
+  batch_size: 1024
+  learning_rate: 0.0001
+  gamma: 0.99
+Datasets:
+  ttH_CP_even: &dataset_defn
+    module: root_gnn_base.dataset
+    class: LazyDataset
+    shuffle_chunks: 3
+    batch_size: 4096
+    padding_mode: NONE #one of STEPS, FIXED, or NONE
+    args: &dataset_args
+      name: ttH_CP_even
+      label: 0
+      # weight_var: weight
+      chunks: 3
+      buffer_size: 2
+      file_names: ttH_NLO.root
+      tree_name: output
+      fold_var: Number
+      raw_dir: /global/cfs/projectdirs/trn007/lbl_atlas/data/stats_100K/
+      save_dir: /global/cfs/projectdirs/trn007/lbl_atlas/data/processed_graphs/stats_100K/ttH_CP_even_vs_odd_batch_size_4096/
+      node_branch_names:
+        - [jet_pt, ele_pt, mu_pt, ph_pt, MET_met]
+        - [jet_eta, ele_eta, mu_eta, ph_eta, 0]
+        - [jet_phi, ele_phi, mu_phi, ph_phi, MET_phi]
+        - CALC_E
+        - [jet_btag, 0, 0, 0, 0]
+        - [0, ele_charge, mu_charge, 0, 0]
+        - NODE_TYPE
+      node_branch_types: [vector, vector, vector, vector, single]
+      node_feature_scales: [1e-1, 1, 1, 1e-1, 1, 1, 1]
+    folding:
+      n_folds: 4
+      test: [0]
+      # validation: 1
+      train: [1, 2, 3]
+  ttH_CP_odd:
+    <<: *dataset_defn
+    args:
+      <<: *dataset_args
+      name: ttH_CP_odd
+      label: 1
+      file_names: ttH_CPodd.root

root_gnn_dgl/configs/stats_100K/ttH_CP_even_vs_odd_batch_size_8192.yaml

@@ -0,0 +1,57 @@
+Training_Name: ttH_CP_even_vs_odd_batch_size_8192
+Training_Directory: trainings/stats_100K/ttH_CP_even_vs_odd_batch_size_8192
+Model:
+  module: models.GCN
+  class: Edge_Network
+  args:
+    hid_size: 64
+    in_size: 7
+    out_size: 1
+    n_layers: 4
+    n_proc_steps: 4
+    dropout: 0
+Training:
+  epochs: 500
+  batch_size: 2048
+  learning_rate: 0.0001
+  gamma: 0.99
+Datasets:
+  ttH_CP_even: &dataset_defn
+    module: root_gnn_base.dataset
+    class: LazyDataset
+    shuffle_chunks: 3
+    batch_size: 2048
+    padding_mode: NONE #one of STEPS, FIXED, or NONE
+    args: &dataset_args
+      name: ttH_CP_even
+      label: 0
+      # weight_var: weight
+      chunks: 3
+      buffer_size: 2
+      file_names: ttH_NLO.root
+      tree_name: output
+      fold_var: Number
+      raw_dir: /global/cfs/projectdirs/trn007/lbl_atlas/data/stats_100K/
+      save_dir: /global/cfs/projectdirs/trn007/lbl_atlas/data/processed_graphs/stats_100K/ttH_CP_even_vs_odd_batch_size_8192/
+      node_branch_names:
+        - [jet_pt, ele_pt, mu_pt, ph_pt, MET_met]
+        - [jet_eta, ele_eta, mu_eta, ph_eta, 0]
+        - [jet_phi, ele_phi, mu_phi, ph_phi, MET_phi]
+        - CALC_E
+        - [jet_btag, 0, 0, 0, 0]
+        - [0, ele_charge, mu_charge, 0, 0]
+        - NODE_TYPE
+      node_branch_types: [vector, vector, vector, vector, single]
+      node_feature_scales: [1e-1, 1, 1, 1e-1, 1, 1, 1]
+    folding:
+      n_folds: 4
+      test: [0]
+      # validation: 1
+      train: [1, 2, 3]
+  ttH_CP_odd:
+    <<: *dataset_defn
+    args:
+      <<: *dataset_args
+      name: ttH_CP_odd
+      label: 1
+      file_names: ttH_CPodd.root

root_gnn_dgl/profile.sh

@@ -0,0 +1,35 @@
+nsys profile \
+  -o /pscratch/sd/j/joshuaho/my_profile_report_1_gpu_batch_size_1028 \
+  --capture-range=cudaProfilerApi \
+  --capture-range-end=stop-shutdown \
+  --force-overwrite true \
+  --trace=nvtx \
+  --cudabacktrace=all \
+  python scripts/training_script.py --config configs/stats_100K/ttH_CP_even_vs_odd.yaml --preshuffle --nocompile --lazy --restart --profile
+
+nsys profile \
+  -o /pscratch/sd/j/joshuaho/my_profile_report_1_gpu_batch_size_2048 \
+  --capture-range=cudaProfilerApi \
+  --capture-range-end=stop-shutdown \
+  --force-overwrite true \
+  --trace=nvtx \
+  --cudabacktrace=all \
+  python scripts/training_script.py --config configs/stats_100K/ttH_CP_even_vs_odd_batch_size_2048.yaml --preshuffle --nocompile --lazy --restart --profile
+
+nsys profile \
+  -o /pscratch/sd/j/joshuaho/my_profile_report_1_gpu_batch_size_4096 \
+  --capture-range=cudaProfilerApi \
+  --capture-range-end=stop-shutdown \
+  --force-overwrite true \
+  --trace=nvtx \
+  --cudabacktrace=all \
+  python scripts/training_script.py --config configs/stats_100K/ttH_CP_even_vs_odd_batch_size_4096.yaml --preshuffle --nocompile --lazy --restart --profile
+
+nsys profile \
+  -o /pscratch/sd/j/joshuaho/my_profile_report_1_gpu_batch_size_8192 \
+  --capture-range=cudaProfilerApi \
+  --capture-range-end=stop-shutdown \
+  --force-overwrite true \
+  --trace=nvtx \
+  --cudabacktrace=all \
+  python scripts/training_script.py --config configs/stats_100K/ttH_CP_even_vs_odd_batch_size_8192.yaml --preshuffle --nocompile --lazy --restart --profile