feat(arg_parse): added dropout_prot and pro_emb_dim

src/train_test/distributed.py

@@ -56,6 +56,8 @@ def dtrain(args):
     print(f"-------------- HYPERPARAMETERS -----------")
     print(f"            Learning rate: {args.learning_rate}")
     print(f"                  Dropout: {args.dropout}")
+    print(f"             dropout_prot: {args.dropout_prot}")
+    print(f"              pro_emb_dim: {args.pro_emb_dim}")
     print(f"               Num epochs: {args.num_epochs}\n")
 
     print(f"----------------- DISTRIBUTED ARGS -----------------")
@@ -87,7 +89,10 @@ def dtrain(args):
     # ==== Load model ====
     # args.gpu is the local rank for this process
     model = Loader.init_model(model=MODEL, pro_feature=FEATURE, pro_edge=EDGEW, 
-                              dropout=args.dropout).cuda(args.gpu)
+                              dropout=args.dropout, 
+                              dropout_prot=args.dropout_prot, 
+                              pro_emb_dim=args.pro_emb_dim).cuda(args.gpu)
+
     cp_saver = CheckpointSaver(model=model, save_path=f'{cfg.MODEL_SAVE_DIR}/{MODEL_KEY}.model',
                             train_all=False,
                             patience=50, min_delta=0.2,

src/utils/arg_parse.py

@@ -95,6 +95,17 @@ def add_hyperparam_args(parser: argparse.ArgumentParser):
         action='store', type=float, default=0.4,
         help='Dropout rate for training (default: 0.4)'
     )
+    parser.add_argument('-dop',
+        '--dropout_prot',
+        action='store', type=float, default=0.4,
+        help='Dropout rate for protein GCN branch for training (default: 0.4)'
+    )
+    parser.add_argument('-embP',
+        '--pro_emb_dim',
+        action='stor', type=int, default=128,
+        help='Embedding dimension for protein GCN branch for training (default: 128)'
+    )
+
     parser.add_argument('-ne',
         '--num_epochs',
         action='store', type=int, default=2000,