start mlp on gpu

core/model_finetuning/mlp.py

@@ -52,28 +52,6 @@ def __len__(self):
     def __getitem__(self, idx):
         return self.embeddings[idx], self.labels[idx]
 
-class MLP(nn.Module):
-    def __init__(self, input_size, hidden_size, num_classes):
-        super(MLP, self).__init__()
-        self.fc1 = nn.Linear(input_size, hidden_size)
-        self.fc2 = nn.Linear(hidden_size, hidden_size)
-        self.fc3 = nn.Linear(hidden_size, num_classes)
-        self.relu = nn.ReLU()
-        self.dropout = nn.Dropout(p=0.2)
-        self.bn1 = nn.BatchNorm1d(hidden_size)
-        self.bn2 = nn.BatchNorm1d(hidden_size)
-
-    def forward(self, x):
-        out = self.fc1(x)
-        out = self.bn1(out)
-        out = self.relu(out)
-        out = self.dropout(out)
-        out = self.fc2(out)
-        out = self.bn2(out)
-        out = self.relu(out)
-        out = self.dropout(out)
-        out = self.fc3(out)
-        return out
 
 def init_weights(m):
     if isinstance(m, nn.Linear):
@@ -86,7 +64,7 @@ def parse_args() -> argparse.Namespace:
     r"""Parses the command line arguments."""
     parser = argparse.ArgumentParser(description='GraphGym')
     parser.add_argument('--data', dest='data', type=str, required=True,
-                        default='pubmed',
+                        default='cora',
                         help='data name')
     parser.add_argument('--device', dest='device', required=False, 
                         help='device id')
@@ -101,9 +79,7 @@ def parse_args() -> argparse.Namespace:
                         help='word embedding method')
     parser.add_argument('--score', dest='score', type=str, required=False, default='mlp_score',
                         help='decoder name')
-    parser.add_argument('--decoder', dest='decoder', type=str, required=False, default='MLP',
-                        help='decoder name')
-    parser.add_argument('--repeat', type=int, default=3,
+    parser.add_argument('--repeat', type=int, default=5,
                         help='The number of repeated jobs.')
     parser.add_argument('opts', default=None, nargs=argparse.REMAINDER,
                         help='See graphgym/config.py for remaining options.')
@@ -132,7 +108,6 @@ def project_main():
     custom_set_out_dir(cfg, args.cfg_file, cfg.wandb.name_tag)
     # torch.set_num_threads(20)
     loggers = create_logger(args.repeat)
-
     for run_id, seed, split_index in zip(*run_loop_settings(cfg, args)):
         print(f'run id : {run_id}')
         # Set configurations for each run TODO clean code here 
@@ -152,7 +127,7 @@ def project_main():
             clf = RidgeClassifier(tol=1e-2, max_iter=10000, solver="sparse_cg")
             clf.fit(train_dataset, train_labels)
         elif args.decoder == 'MLP':
-            clf = MLPClassifier(random_state=run_id, max_iter=10000).fit(train_dataset, train_labels)  
+            clf = MLPClassifier(random_state=run_id, max_iter=100).fit(train_dataset, train_labels)  
 
         test_pred = clf.predict(test_dataset)
         test_acc = sum(np.asarray(test_labels) == test_pred ) / len(test_labels)

core/model_finetuning/mlp_tfidf.py

@@ -59,7 +59,7 @@ def parse_args() -> argparse.Namespace:
                         help='word embedding method')
     parser.add_argument('--score', dest='score', type=str, required=False, default='mlp_score',
                         help='decoder name')
-    parser.add_argument('--repeat', type=int, default=3,
+    parser.add_argument('--repeat', type=int, default=5,
                         help='The number of repeated jobs.')
     parser.add_argument('opts', default=None, nargs=argparse.REMAINDER,
                         help='See graphgym/config.py for remaining options.')
@@ -77,7 +77,23 @@ def __len__(self):
     def __getitem__(self, idx):
         return self.embeddings[idx], self.labels[idx]
 
-
+
+
+class MLP(nn.Module):
+    def __init__(self, input_dim, hidden_dim, output_dim):
+        super(MLP, self).__init__()
+        self.model = nn.Sequential(
+            nn.Linear(input_dim, hidden_dim),
+            nn.ReLU(),
+            nn.Linear(hidden_dim, hidden_dim),
+            nn.ReLU(),
+            nn.Linear(hidden_dim, output_dim)
+        )
+
+    def forward(self, x):
+        return self.model(x)
+
+
 def project_main():  # sourcery skip: avoid-builtin-shadow, hoist-statement-from-loop
 
     # process params
@@ -130,13 +146,16 @@ def project_main():  # sourcery skip: avoid-builtin-shadow, hoist-statement-from
         )
 
         dump_cfg(cfg)
-        in_channels = train_dataset.shape[1]
-
-        model = mlp_model(in_channels, 
-                            cfg.decoder.hidden_channels, 
-                            cfg.decoder.out_channels, 
-                            cfg.decoder.num_layers,
-                            cfg.decoder.dropout).to(cfg.device)
+        # model = mlp_model(in_channels, 
+        #                     cfg.decoder.hidden_channels, 
+        #                     cfg.decoder.out_channels, 
+        #                     cfg.decoder.num_layers,
+        #                     cfg.decoder.dropout).to(cfg.device)
+        input_dim = train_dataset.shape[1]
+        hidden_dim = 128
+        output_dim = len(np.unique(train_labels))
+        model = MLP(input_dim, hidden_dim, output_dim).to(cfg.device)
+
         model = model.to(cfg.device)
 
         print_logger.info(f"{model} on {next(model.parameters()).device}" )

core/model_finetuning/scripts/create_data.sh

@@ -1,7 +1,7 @@
 #!/bin/sh
 #SBATCH --time=3-00:00:00
 #SBATCH --partition=cpuonly
-#SBATCH --job-name=gnn_wb
+#SBATCH --job-name=tfidf
 
 
 

core/model_finetuning/horeka_sh/cross_encoder.sh → core/model_finetuning/scripts/mlp.sh

@@ -1,15 +1,12 @@
 #!/bin/sh
-#SBATCH --time=02:00:00
-#SBATCH --nodes=1
-#SBATCH --ntasks=1
-#SBATCH --partition=accelerated
-#SBATCH --job-name=cross_encoder
-#SBATCH --mem=50160mb
-#BATCH  --cpu-per-gpu=38
+#SBATCH --time=3-00:00:00
+#SBATCH --partition=cpuonly
+#SBATCH --job-name=gnn_wb
+
+
+
 #SBATCH --output=log/TAG_Benchmark_%j.output
 #SBATCH --error=error/TAG_Benchmark_%j.error
-#SBATCH --gres=gpu:4
-#SBATCH --account=hk-project-test-p0022257  # specify the project group
 
 
 #SBATCH --chdir=/hkfs/work/workspace/scratch/cc7738-benchmark_tag/TAPE_chen/batch
@@ -18,10 +15,10 @@
 #SBATCH --mail-type=ALL
 #SBATCH --mail-user=cc7738@kit.edu
 
-source /hkfs/home/project/hk-project-test-p0021478/cc7738/anaconda3/etc/profile.d/conda.sh
+source /hkfs/home/haicore/aifb/cc7738/anaconda3/etc/profile.d/conda.sh
 
 conda activate base
-conda activate TAG_LP
+conda activate ss
 # <<< conda initialize <<<
 module purge
 module load devel/cmake/3.18
@@ -32,4 +29,9 @@ module load compiler/gnu/12
 cd /hkfs/work/workspace/scratch/cc7738-benchmark_tag/TAPE_chen/core/model_finetuning
 
 
-python cross_encoder.py
+for data in  cora arxiv_2023 pubmed ogbn-arxiv ogbn-products; do
+    python mlp.py --data $data --decoder MLP
+done
+for data in cora  arxiv_2023 pubmed ogbn-arxiv ogbn-products; do
+    python mlp.py --data $data --decoder Ridge
+done