update

src/common/validation.py

@@ -7,7 +7,7 @@ from PIL import Image
 
 # @ray.remote(num_cpus=1, num_gpus=0.3)
 def val_image_pair(model, hr_image, lr_image, output_image_path=None):            
-    with torch.no_grad():
+    with torch.inference_mode():
         # prepare lr_image
         lr_image = torch.tensor(lr_image).type(torch.float32).permute(2,0,1)
         lr_image = lr_image.unsqueeze(0).cuda()

src/scripts/image_demo.py

@@ -57,6 +57,7 @@ image_gt = cv2.imread(str(config.hr_image_path))[:,:,::-1].copy()
 
 input_image = torch.tensor(lr_image).type(torch.float32).permute(2,0,1)[None,...].cuda()
 
+with torch.inference_mode():
     net_prediction = net_model(input_image).cpu().type(torch.uint8).squeeze().permute(1,2,0).numpy().copy()
     lut_prediction = lut_model(input_image).cpu().type(torch.uint8).squeeze().permute(1,2,0).numpy().copy()
 

src/scripts/transfer_to_lut.py

@@ -70,8 +70,14 @@ if __name__ == "__main__":
     lut_model_size = np.sum([x.nelement()*x.element_size() for x in lut_model.parameters()])
     print("Saved to", lut_path, f"{lut_model_size/(2**20):.3f} MB")
 
-    Path(config.models_dir / f"last_transfered_net.pth").symlink_to(config.model_path.resolve())
-    Path(config.models_dir / f"last_transfered_lut.pth").symlink_to(lut_path.resolve())
+    link = Path(config.models_dir / f"last_transfered_net.pth")
+    if link.exists():
+        link.unlink()
+        link.symlink_to(config.model_path.resolve())
+    link = Path(config.models_dir / f"last_transfered_lut.pth")
+    if link.exists():
+        link.unlink()
+        link.symlink_to(lut_path.resolve())
     print("Updated link", config.models_dir / f"last_transfered_net.pth")
     print("Updated link", config.models_dir / f"last_transfered_lut.pth")