Track: Track1; Team name: topolab; Model: GREAD

2026_tdl_challenge/run_evaluation.ipynb

@@ -104,7 +104,7 @@
    "outputs": [],
    "source": [
     "# Your model configuration (e.g., \"graph/gcn\", \"graph/gin\", \"graph/gat\")\n",
-    "MODEL_CONFIG = \"graph/gin\""
+    "MODEL_CONFIG = \"graph/gread\""
    ]
   },
   {

configs/model/graph/gread.yaml

@@ -0,0 +1,50 @@
+_target_: topobench.model.TBModel
+
+model_name: gread
+model_domain: graph
+
+feature_encoder:
+  _target_: topobench.nn.encoders.${model.feature_encoder.encoder_name}
+  encoder_name: AllCellFeatureEncoder
+  in_channels: ${infer_in_channels:${dataset},${oc.select:transforms,null}}
+  out_channels: 64
+  proj_dropout: 0.0
+
+backbone:
+  _target_: topobench.nn.backbones.GREAD
+  in_channels: ${model.feature_encoder.out_channels}
+  hidden_channels: ${model.feature_encoder.out_channels}
+  reaction_term: bspm  # GREAD-BS. Options: bspm, fisher, allen-cahn, zeldovich, st, fb, fb3, none
+  time: 3.0
+  step_size: 1.0
+  # Per-channel reaction gate (the "(VC)" variants of the paper); every
+  # tuned dataset configuration in the reference implementation's
+  # src/gread_params.py sets beta_diag: true.
+  beta_diag: true
+  add_source: false
+  data_norm: rw
+  self_loop_weight: 1.0
+  # Terminal ReLU pinned explicitly for the submitted benchmark runs
+  # (the constructor default is false, matching the reference CLI default).
+  xn_activation: true
+  input_dropout: 0.0
+  dropout: 0.0
+
+backbone_wrapper:
+  _target_: topobench.nn.wrappers.GNNWrapper
+  _partial_: true
+  wrapper_name: GNNWrapper
+  out_channels: ${model.feature_encoder.out_channels}
+  num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}}
+
+readout:
+  _target_: topobench.nn.readouts.${model.readout.readout_name}
+  readout_name: NoReadOut #  Use <NoReadOut> in case readout is not needed Options: PropagateSignalDown
+  num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} # The highest order of cell dimensions to consider
+  hidden_dim: ${model.feature_encoder.out_channels}
+  out_channels: ${dataset.parameters.num_classes}
+  task_level: ${define_task_level:${dataset.parameters.task_level},${dataset.split_params.learning_setting}} # Handles the edge case of node-inductive task
+  pooling_type: sum
+
+# compile model for faster training with pytorch 2.0
+compile: false

configs/model/graph/gread_allen_cahn.yaml

@@ -0,0 +1,47 @@
+_target_: topobench.model.TBModel
+
+model_name: gread_allen_cahn
+model_domain: graph
+
+feature_encoder:
+  _target_: topobench.nn.encoders.${model.feature_encoder.encoder_name}
+  encoder_name: AllCellFeatureEncoder
+  in_channels: ${infer_in_channels:${dataset},${oc.select:transforms,null}}
+  out_channels: 64
+  proj_dropout: 0.0
+
+backbone:
+  _target_: topobench.nn.backbones.GREAD
+  in_channels: ${model.feature_encoder.out_channels}
+  hidden_channels: ${model.feature_encoder.out_channels}
+  reaction_term: allen-cahn
+  time: 3.0
+  step_size: 1.0
+  beta_diag: false
+  add_source: false
+  data_norm: rw
+  self_loop_weight: 1.0
+  # Terminal ReLU pinned explicitly for the submitted benchmark runs
+  # (the constructor default is false, matching the reference CLI default).
+  xn_activation: true
+  input_dropout: 0.0
+  dropout: 0.0
+
+backbone_wrapper:
+  _target_: topobench.nn.wrappers.GNNWrapper
+  _partial_: true
+  wrapper_name: GNNWrapper
+  out_channels: ${model.feature_encoder.out_channels}
+  num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}}
+
+readout:
+  _target_: topobench.nn.readouts.${model.readout.readout_name}
+  readout_name: NoReadOut #  Use <NoReadOut> in case readout is not needed Options: PropagateSignalDown
+  num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} # The highest order of cell dimensions to consider
+  hidden_dim: ${model.feature_encoder.out_channels}
+  out_channels: ${dataset.parameters.num_classes}
+  task_level: ${define_task_level:${dataset.parameters.task_level},${dataset.split_params.learning_setting}} # Handles the edge case of node-inductive task
+  pooling_type: sum
+
+# compile model for faster training with pytorch 2.0
+compile: false

configs/model/graph/gread_fisher.yaml

@@ -0,0 +1,47 @@
+_target_: topobench.model.TBModel
+
+model_name: gread_fisher
+model_domain: graph
+
+feature_encoder:
+  _target_: topobench.nn.encoders.${model.feature_encoder.encoder_name}
+  encoder_name: AllCellFeatureEncoder
+  in_channels: ${infer_in_channels:${dataset},${oc.select:transforms,null}}
+  out_channels: 64
+  proj_dropout: 0.0
+
+backbone:
+  _target_: topobench.nn.backbones.GREAD
+  in_channels: ${model.feature_encoder.out_channels}
+  hidden_channels: ${model.feature_encoder.out_channels}
+  reaction_term: fisher
+  time: 3.0
+  step_size: 1.0
+  beta_diag: false
+  add_source: false
+  data_norm: rw
+  self_loop_weight: 1.0
+  # Terminal ReLU pinned explicitly for the submitted benchmark runs
+  # (the constructor default is false, matching the reference CLI default).
+  xn_activation: true
+  input_dropout: 0.0
+  dropout: 0.0
+
+backbone_wrapper:
+  _target_: topobench.nn.wrappers.GNNWrapper
+  _partial_: true
+  wrapper_name: GNNWrapper
+  out_channels: ${model.feature_encoder.out_channels}
+  num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}}
+
+readout:
+  _target_: topobench.nn.readouts.${model.readout.readout_name}
+  readout_name: NoReadOut #  Use <NoReadOut> in case readout is not needed Options: PropagateSignalDown
+  num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} # The highest order of cell dimensions to consider
+  hidden_dim: ${model.feature_encoder.out_channels}
+  out_channels: ${dataset.parameters.num_classes}
+  task_level: ${define_task_level:${dataset.parameters.task_level},${dataset.split_params.learning_setting}} # Handles the edge case of node-inductive task
+  pooling_type: sum
+
+# compile model for faster training with pytorch 2.0
+compile: false

configs/model/graph/gread_source_term.yaml

@@ -0,0 +1,47 @@
+_target_: topobench.model.TBModel
+
+model_name: gread_source_term
+model_domain: graph
+
+feature_encoder:
+  _target_: topobench.nn.encoders.${model.feature_encoder.encoder_name}
+  encoder_name: AllCellFeatureEncoder
+  in_channels: ${infer_in_channels:${dataset},${oc.select:transforms,null}}
+  out_channels: 64
+  proj_dropout: 0.0
+
+backbone:
+  _target_: topobench.nn.backbones.GREAD
+  in_channels: ${model.feature_encoder.out_channels}
+  hidden_channels: ${model.feature_encoder.out_channels}
+  reaction_term: st
+  time: 3.0
+  step_size: 1.0
+  beta_diag: false
+  add_source: false
+  data_norm: rw
+  self_loop_weight: 1.0
+  # Terminal ReLU pinned explicitly for the submitted benchmark runs
+  # (the constructor default is false, matching the reference CLI default).
+  xn_activation: true
+  input_dropout: 0.0
+  dropout: 0.0
+
+backbone_wrapper:
+  _target_: topobench.nn.wrappers.GNNWrapper
+  _partial_: true
+  wrapper_name: GNNWrapper
+  out_channels: ${model.feature_encoder.out_channels}
+  num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}}
+
+readout:
+  _target_: topobench.nn.readouts.${model.readout.readout_name}
+  readout_name: NoReadOut #  Use <NoReadOut> in case readout is not needed Options: PropagateSignalDown
+  num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} # The highest order of cell dimensions to consider
+  hidden_dim: ${model.feature_encoder.out_channels}
+  out_channels: ${dataset.parameters.num_classes}
+  task_level: ${define_task_level:${dataset.parameters.task_level},${dataset.split_params.learning_setting}} # Handles the edge case of node-inductive task
+  pooling_type: sum
+
+# compile model for faster training with pytorch 2.0
+compile: false

configs/model/graph/gread_zeldovich.yaml

@@ -0,0 +1,47 @@
+_target_: topobench.model.TBModel
+
+model_name: gread_zeldovich
+model_domain: graph
+
+feature_encoder:
+  _target_: topobench.nn.encoders.${model.feature_encoder.encoder_name}
+  encoder_name: AllCellFeatureEncoder
+  in_channels: ${infer_in_channels:${dataset},${oc.select:transforms,null}}
+  out_channels: 64
+  proj_dropout: 0.0
+
+backbone:
+  _target_: topobench.nn.backbones.GREAD
+  in_channels: ${model.feature_encoder.out_channels}
+  hidden_channels: ${model.feature_encoder.out_channels}
+  reaction_term: zeldovich
+  time: 3.0
+  step_size: 1.0
+  beta_diag: false
+  add_source: false
+  data_norm: rw
+  self_loop_weight: 1.0
+  # Terminal ReLU pinned explicitly for the submitted benchmark runs
+  # (the constructor default is false, matching the reference CLI default).
+  xn_activation: true
+  input_dropout: 0.0
+  dropout: 0.0
+
+backbone_wrapper:
+  _target_: topobench.nn.wrappers.GNNWrapper
+  _partial_: true
+  wrapper_name: GNNWrapper
+  out_channels: ${model.feature_encoder.out_channels}
+  num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}}
+
+readout:
+  _target_: topobench.nn.readouts.${model.readout.readout_name}
+  readout_name: NoReadOut #  Use <NoReadOut> in case readout is not needed Options: PropagateSignalDown
+  num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} # The highest order of cell dimensions to consider
+  hidden_dim: ${model.feature_encoder.out_channels}
+  out_channels: ${dataset.parameters.num_classes}
+  task_level: ${define_task_level:${dataset.parameters.task_level},${dataset.split_params.learning_setting}} # Handles the edge case of node-inductive task
+  pooling_type: sum
+
+# compile model for faster training with pytorch 2.0
+compile: false