i

entry.py

@@ -41,11 +41,11 @@ def get_batch(keys, value_dict, N, device="cuda"):
                 .to(device)
                 .repeat(*N, 1)
             )
-            batch_uc["original_size_as_tuple"] = (
+            # batch_uc["original_size_as_tuple"] = (
-                torch.tensor([value_dict["orig_height"], value_dict["orig_width"]])
+            #     torch.tensor([value_dict["orig_height"], value_dict["orig_width"]])
-                .to(device)
+            #     .to(device)
-                .repeat(*N, 1) / 2
+            #     .repeat(*N, 1) / 2
-            )
+            # )
         elif key == "crop_coords_top_left":
             batch["crop_coords_top_left"] = (
                 torch.tensor(
@@ -70,11 +70,11 @@ def get_batch(keys, value_dict, N, device="cuda"):
                 .to(device)
                 .repeat(*N, 1)
             )
-            batch_uc["target_size_as_tuple"] = (
+            # batch_uc["target_size_as_tuple"] = (
-                torch.tensor([value_dict["target_height"], value_dict["target_width"]])
+            #     torch.tensor([value_dict["target_height"], value_dict["target_width"]])
-                .to(device)
+            #     .to(device)
-                .repeat(*N, 1) / 2.0
+            #     .repeat(*N, 1) / 2.0
-            )
+            # )
         else:
             batch[key] = value_dict[key]
 
@@ -100,7 +100,7 @@ sampler = EulerAncestralSampler(
     verbose=True,
 )
 
-torch.manual_seed(123)
+torch.manual_seed(12345)
 
 config_path = './sd_xl_base.yaml'
 config = OmegaConf.load(config_path)