belka_gate_fusions: trainable version

src/models/dual_encoder.py

@@ -23,7 +23,6 @@ import torch.nn.functional as F
 
 # residual fusions exp
 from .residual_fusions import ResidualGateType, GatedFusionResidual
-# from residual_fusions import ResidualGateType, GatedFusionResidual
 
 
 class ProjectionHead(nn.Module):

src/models/residual_fusions.py

@@ -142,3 +142,13 @@ class GatedFusionResidual(nn.Module):
             fused_vec = self.FuseARGFCat(img_feat=img_feat, text_feat=text_feat)
 
         return fused_vec
+    
+    @property
+    def gate_value(self) -> float:
+        """Current gate value (image weight). 1.0 = text ignored."""
+        if self.baseline_mode:
+            return 1.0
+        return torch.sigmoid(self.alpha).item()
+    
+
+    

src/training/train.py

@@ -56,8 +56,8 @@ class TrainConfig:
         self,
         train_query_file: str = "Index/train_query.txt",
         val_query_file: str = "Index/test_query.txt",
-        data_root: str = "/mnt/data1tb/cvgl_datasets/UAV-GeoLoc",
+        data_root: str = "/media/servml/SSD_2_TB/datasets/cvgl_datasets/UAV-GeoLoc",
-        output_dir: str = "out/caption_test",
+        output_dir: str = "out/caption_test_exp_gate_SRGF",
         epochs: int = 10,
         batch_size: int = 128,
         num_workers: int = 4,