add sampling paradigm

.gitignore

@@ -0,0 +1,11 @@
+__pycache__/
+
+
+*.egg-info/
+
+build/
+dist/
+
+logs/
+
+configs/

examples/GDA/configs/FLAG.yml

@@ -0,0 +1,57 @@
+dataset:
+ classname: "Ogbn"
+ name: "arxiv"
+ root: "/home/ssq/test_data/"
+model:
+ gnn_type: 'gcn'
+ hidden_dim: 256
+ dropout: 0.5
+ n_layers: 3
+ step_size: 0.001
+ augM: 3
+ batch_norm: True
+task:
+ lr: 0.01
+ seed: 12345
+ epochs: 500
+ patience: 80
+ weight_decay: 0
+# dataset:
+#  classname: "Ogbn"
+#  name: "arxiv"
+#  root: "/home/ssq/test_data/"
+# model:
+#  gnn_type: 'sage'
+#  hidden_dim: 256
+#  dropout: 0.5
+#  n_layers: 3
+#  step_size: 0.001
+#  augM: 3
+#  batch_norm: True
+# task:
+#  lr: 0.01
+#  seed: 12345
+#  epochs: 500
+#  patience: 80
+#  weight_decay: 0
+# dataset:
+#  classname: "Ogbn"
+#  name: "arxiv"
+#  root: "/home/ssq/test_data/"
+# model:
+#  gnn_type: 'gat'
+#  hidden_dim: 256
+#  n_heads: '8,8,1'
+#  dropout: 0.6
+#  attn_dropout: 0.6
+#  n_layers: 3
+#  step_size: 0.001
+#  augM: 3
+#  amp: 2
+#  batch_norm: True
+# task:
+#  lr: 0.002
+#  seed: 12345
+#  epochs: 500
+#  patience: 100
+#  weight_decay: 0

examples/GDA/configs/GAugM.yml

@@ -0,0 +1,59 @@
+dataset:
+ classname: "Planetoid"
+ name: "cora"
+ root: "/home/ssq/test_data/"
+model:
+ model_name: 'GAugM'
+ gnn_type: 'gcn'
+ feat_norm: 'row'
+ hidden_dim: 256
+ dropout: 0.5
+ n_layers: 2
+ choose_idx: 5
+ rm_pct: 2
+ add_pct: 57
+task:
+ lr: 0.01
+ seed: 42
+ epochs: 200
+ weight_decay: 0.0005
+# dataset:
+#  classname: "Planetoid"
+#  name: "cora"
+#  root: "/home/ssq/test_data/"
+# model:
+#  model_name: 'GAugM'
+#  gnn_type: 'gsage'
+#  feat_norm: 'row'
+#  normalize: True
+#  hidden_dim: 128
+#  dropout: 0.5
+#  n_layers: 2
+#  choose_idx: 2
+#  rm_pct: 1
+#  add_pct: 80
+# task:
+#  lr: 0.01
+#  seed: 42
+#  epochs: 200
+#  weight_decay: 0.0005
+# dataset:
+#   classname: "Planetoid"
+#   name: "cora"
+#   root: "/home/ssq/test_data/"
+# model:
+#   model_name: 'GAugM'
+#   gnn_type: 'gat'
+#   feat_norm: 'row'
+#   activation: 'elu'
+#   hidden_dim: 128
+#   dropout: 0.5
+#   n_layers: 2
+#   choose_idx: 2
+#   rm_pct: 1
+#   add_pct: 68
+# task:
+#   lr: 0.01
+#   seed: 42
+#   epochs: 200
+#   weight_decay: 0.0005

examples/GDA/configs/GAugO.yml

@@ -0,0 +1,80 @@
+# dataset:
+#  classname: "Planetoid"
+#  name: "cora"
+#  root: "/home/ssq/test_data/"
+# model:
+#  model_name: 'GAugO'
+#  gnn_type: 'gcn'
+#  alpha: 1.0
+#  temperature: 1.2
+#  hidden_dim: 256
+#  emb_size: 32
+#  dropout: 0.5
+#  n_layers: 2
+#  gae: true
+#  feat_norm: 'row'
+#  sample_type: 'add_sample'
+# task:
+#  lr: 0.01
+#  seed: 42
+#  warmup: 0
+#  beta: 0.8
+#  epochs: 200
+#  weight_decay: 0.0005
+#  pretrain_ep: 160
+#  pretrain_nc: 30
+#  max_patience: 50
+# dataset:
+#   classname: "Planetoid"
+#   name: "cora"
+#   root: "/home/ssq/test_data/"
+# model:
+#   model_name: 'GAugO'
+#   gnn_type: 'gsage'
+#   alpha: 0.13
+#   temperature: 1.0
+#   hidden_dim: 256
+#   emb_size: 32
+#   dropout: 0.5
+#   n_layers: 2
+#   gae: true
+#   feat_norm: 'row'
+#   normalize: True
+#   sample_type: 'add_sample'
+# task:
+#   lr: 0.01
+#   seed: 42
+#   warmup: 2
+#   beta: 3.2
+#   epochs: 200
+#   weight_decay: 0.0005
+#   pretrain_ep: 195
+#   pretrain_nc: 35
+#   max_patience: 50
+dataset:
+ classname: "Planetoid"
+ name: "cora"
+ root: "/home/ssq/test_data/"
+model:
+  model_name: 'GAugO'
+  gnn_type: 'gat'
+  alpha: 0.02
+  temperature: 1.7
+  hidden_dim: 128
+  emb_size: 32
+  dropout: 0.6
+  n_layers: 2
+  activation: "elu"
+  gae: true
+  feat_norm: 'row'
+  sample_type: 'add_sample'
+task:
+  lr: 0.01
+  seed: 42
+  warmup: 1
+  beta: 3.2
+  epochs: 200
+  weight_decay: 0.0005
+  pretrain_ep: 175
+  pretrain_nc: 45
+  max_patience: 50

examples/GDA/configs/GAugOMini.yml

@@ -0,0 +1,30 @@
+dataset:
+ classname: "Planetoid"
+ name: "pubmed"
+ root: "/home/ssq/test_data/"
+model:
+ model_name: 'GAugO'
+ gnn_type: 'gcn'
+ alpha: 0.79
+ temperature: 1.6
+ hidden_dim: 256
+ emb_size: 64
+ dropout: 0.5
+ n_layers: 2
+ gae: true
+ feat_norm: 'row'
+ sample_type: 'add_sample'
+ minibatch: True
+task:
+ lr: 0.01
+ ep_lr: 0.006
+ seed: 42
+ warmup: 7
+ beta: 3.3
+ epochs: 200
+ weight_decay: 0.0005
+ pretrain_ep: 10
+ pretrain_nc: 40
+ max_patience: 50
+ train_batch_size: 8192
+ pretrain_batch_size: 8192

examples/GDA/configs/Mixup.yml

@@ -0,0 +1,20 @@
+dataset:
+  classname: "Planetoid"
+  name: "pubmed"
+  root: "/home/ssq/test_data/"
+  split: "full"
+model:
+  gnn_type: 'sage'
+  hidden_dim: 256
+  dropout: 0.5
+  alpha: 4
+  beta: 4
+  n_layers: 3
+  feat_norm: "none"
+  batch_norm: True
+task:
+  lr: 0.01
+  seed: 12345
+  epochs: 300
+  patience: 30
+  weight_decay: 0

examples/GDA/configs/SampleFLAG.yml

@@ -0,0 +1,78 @@
+# dataset:
+#   classname: "Ogbn"
+#   name: "products"
+#   root: "/home/ssq/test_data/"
+# sampler:
+#   training:
+#     name: "NeighborSampler"
+#     layer_sizes: "15,10,5"
+#     prob_type: "normalize"
+#     replace: False
+#   eval:
+#     name: "NeighborSampler"
+#     layer_sizes: "-1"
+#     replace: False
+# model:
+#   gnn_type: 'sage'
+#   hidden_dim: 256
+#   dropout: 0.5
+#   n_layers: 3
+#   step_size: 0.008
+#   augM: 3
+#   amp: 2
+#   batch_norm: False
+# task:
+#   name: "NodeClassification_Sampling"
+#   lr: 0.003
+#   seed: 12345
+#   epochs: 20
+#   patience: 10
+#   weight_decay: 0
+#   train_batch_size: 1024
+#   eval_batch_size: 4096
+#   train_num_workers: 12
+#   eval_num_workers: 12
+#   eval_together: True
+#   eval_freq: 2
+#   eval_start: 10
+#   loss_fn: "nll_loss"
+dataset:
+  classname: "Ogbn"
+  name: "arxiv"
+  root: "/home/ssq/test_data/"
+sampler:
+  training:
+    name: "NeighborSampler"
+    layer_sizes: "15,10,5"
+    prob_type: "normalize"
+    replace: False
+  eval:
+    name: "NeighborSampler"
+    layer_sizes: "-1"
+    replace: False
+model:
+  gnn_type: 'gat'
+  hidden_dim: 64
+  n_heads: "8,8,1"
+  dropout: 0.5
+  n_layers: 3
+  step_size: 0.008
+  augM: 3
+  amp: 2
+  batch_norm: False
+task:
+  name: "NodeClassification_Sampling"
+  lr: 0.002
+  seed: 12345
+  epochs: 500
+  patience: 50
+  weight_decay: 0
+  train_batch_size: 1024
+  eval_batch_size: 4096
+  train_num_workers: 12
+  eval_num_workers: 12
+  eval_together: True
+  pin_memory: True
+  eval_freq: 2
+  eval_start: 1
+  loss_fn: "nll_loss"

examples/GDA/configs/SampleMixup.yml

@@ -0,0 +1,36 @@
+dataset:
+  classname: "Ogbn"
+  name: "arxiv"
+  root: "/home/ssq/test_data/"
+sampler:
+  training:
+    name: "NeighborSampler"
+    layer_sizes: "15, 10, 5"
+    prob_type: "normalize"
+    replace: False 
+  eval:
+    name: "NeighborSampler"
+    layer_sizes: "-1"
+    replace: False
+model:
+  gnn_type: 'sage'
+  hidden_dim: 256
+  dropout: 0.5
+  alpha: 4
+  beta: 4
+  n_layers: 3
+  feat_norm: "none"
+  batch_norm: True
+task:
+  name: "NodeClassification_Sampling"
+  lr: 0.003
+  seed: 12345
+  epochs: 300
+  patience: 30
+  weight_decay: 0
+  train_batch_size: 1024
+  eval_batch_size: 4096 
+  train_num_workers: 12 
+  eval_num_workers: 12
+  eval_together: True 
+  eval_freq: 2