deeper drop edge network

MartinXPN · Feb 6, 2021 · e6ff617 · e6ff617
1 parent d6b383d
commit e6ff617
Show file tree

Hide file tree

Showing 3 changed files with 7 additions and 6 deletions.
diff --git a/abcde/data.py b/abcde/data.py
@@ -31,7 +31,7 @@ def __post_init__(self, save_path):
         workers: int = max(os.cpu_count(), 1)
 
         if save_path is not None and save_path.exists():
-            print('Found existing dataset at', save_path, 'loading it...')
+            print(f'Found existing dataset at {save_path} => loading it...')
             self.graphs = torch.load(save_path)
             return
 

diff --git a/abcde/dropout.py b/abcde/dropout.py
@@ -11,7 +11,7 @@ def dropout_adj(edge_index, edge_attr=None, p=0.5, force_undirected=False,
         raise ValueError('Dropout probability has to be between 0 and 1, '
                          'but got {}'.format(p))
 
-    if not training:
+    if not training or p == 0:
         return edge_index, edge_attr
 
     N = maybe_num_nodes(edge_index, num_nodes)

diff --git a/train.py b/train.py
@@ -29,9 +29,9 @@
         AimLogger(experiment=experiment.name),
     ]
     # Previous best: nb_gcn_cycles=(4, 4, 6, 6, 8), conv_sizes=(64, 64, 32, 32, 16), drops=(0, 0, 0, 0, 0)
-    model = ABCDE(nb_gcn_cycles=(4, 4, 6, 6, 8),
-                  conv_sizes=(64, 64, 32, 32, 16),
-                  drops=(0, 0, 0, 0, 0),
+    model = ABCDE(nb_gcn_cycles=(4, 4, 6, 6, 8, 8),
+                  conv_sizes=(64, 48, 32, 32, 24, 24),
+                  drops=(0.4, 0.3, 0.2, 0.2, 0.1, 0.1),
                   lr_reduce_patience=2, dropout=0.1)
     data = GraphDataModule(min_nodes=4000, max_nodes=5000, nb_train_graphs=160, nb_valid_graphs=240,
                            batch_size=16, graph_type='powerlaw', repeats=8, regenerate_epoch_interval=10,
@@ -41,8 +41,9 @@
                       max_epochs=100, terminate_on_nan=True, enable_pl_optimizer=True,
                       reload_dataloaders_every_epoch=True,
                       callbacks=[
-                          EarlyStopping(monitor='val_kendal', patience=7, verbose=True, mode='max'),
+                          EarlyStopping(monitor='val_kendal', patience=5, verbose=True, mode='max'),
                           ModelCheckpoint(dirpath=experiment.model_save_path, filename='drop-{epoch:02d}-{val_kendal:.2f}', monitor='val_kendal', save_top_k=5, verbose=True, mode='max'),
                           LearningRateMonitor(logging_interval='epoch'),
                       ])
+    trainer.fit = telegram_sender(token='1653878275:AAEIr-mLt9-SSAyYPon1n-CgFQpINjUWHDw', chat_id=695404691)(trainer.fit)
     trainer.fit(model, datamodule=data)