diff --git a/src/cellmap_models/pytorch/cosem/load_model.py b/src/cellmap_models/pytorch/cosem/load_model.py
index 6aff03b..1294f95 100755
--- a/src/cellmap_models/pytorch/cosem/load_model.py
+++ b/src/cellmap_models/pytorch/cosem/load_model.py
@@ -78,12 +78,27 @@ def load_model(checkpoint_name: str) -> torch.nn.Module:
     new_checkpoint = deepcopy(checkpoint)
     for key in checkpoint["model"].keys():
         if "chain" in key:
+            # Verify that the chain tensors are redundant and can be removed
+            if key.replace("chain.0", "architecture") in checkpoint["model"].keys():
+                assert (
+                    checkpoint["model"][key]
+                    == checkpoint["model"][key.replace("chain.0", "architecture")]
+                ).all(), f"Chain {key} does not match architecture {key.replace('chain.0', 'architecture')}"
+            elif (
+                key.replace("chain.1", "prediction_head") in checkpoint["model"].keys()
+            ):
+                assert (
+                    checkpoint["model"][key]
+                    == checkpoint["model"][key.replace("chain.1", "prediction_head")]
+                ).all(), f"Chain {key} does not match prediction_head {key.replace('chain.1', 'prediction_head')}"
+            else:
+                raise ValueError(f"No match found for key {key} in checkpoint")
             new_checkpoint["model"].pop(key)
             continue
         new_key = key.replace("architecture.", "")
         new_key = new_key.replace("unet.", "backbone.")
         new_checkpoint["model"][new_key] = new_checkpoint["model"].pop(key)
-    model.load_state_dict(new_checkpoint["model"])
+    model.load_state_dict(new_checkpoint["model"], strict=True)
     model.eval()
 
     return model