成功的训练版本，跑的epoch数太少，学习率过大，效果不好。

config/config.yaml

@@ -3,8 +3,12 @@ project: deblur_gan
 experiment_desc: fpn
 
 train:
-  files_a: &FILES_A /datasets/my_dataset/**/*.jpg
+#  files_a: &FILES_A /datasets/my_dataset/**/*.jpg
+  #可以使用绝对路径
+  files_a: &FILES_A D:/deblur/goprol_large/**/*.png
   files_b: *FILES_A
+#  files_a: &FILES_A ./dataset1/blur/*.png
+#  files_b: &FILES_B ./dataset1/sharp/*.png
   size: &SIZE 256
   crop: random
   preload: &PRELOAD false
@@ -30,6 +34,8 @@ train:
 val:
   files_a: *FILES_A
   files_b: *FILES_A
+#  files_a: &FILES_A
+#  files_b: &FILES_B
   size: *SIZE
   scope: geometric
   crop: center
@@ -52,15 +58,15 @@ model:
   norm_layer: instance
   dropout: True
 
-num_epochs: 200
+num_epochs: 10
 train_batches_per_epoch: 1000
 val_batches_per_epoch: 100
 batch_size: 1
 image_size: [256, 256]
 
 optimizer:
   name: adam
-  lr: 0.0001
+  lr: 0.01
 scheduler:
   name: linear
   start_epoch: 50

predict.py

@@ -15,7 +15,7 @@
 
 class Predictor:
     def __init__(self, weights_path: str, model_name: str = ''):
-        with open('config/config.yaml') as cfg:
+        with open('config/config.yaml',encoding='utf-8') as cfg:
             config = yaml.load(cfg, Loader=yaml.FullLoader)
         model = get_generator(model_name or config['model'])
         model.load_state_dict(torch.load(weights_path)['model'])
@@ -122,4 +122,4 @@ def sorted_glob(pattern):
 
 if __name__ == '__main__':
   #  Fire(main)
-    main('submit/t6.jpg')
+    main('test_img/000027.png')

train.py

@@ -171,13 +171,16 @@ def _init_params(self):
 
 
 def main(config_path='config/config.yaml'):
-    with open(config_path, 'r') as f:
+    with open(config_path, 'r',encoding='utf-8') as f:
         config = yaml.load(f, Loader=yaml.SafeLoader)
 
     batch_size = config.pop('batch_size')
+    # get_dataloader = partial(DataLoader,
+    #                          batch_size=batch_size,
+    #                          num_workers=0 if os.environ.get('DEBUG') else cpu_count(),
+    #                          shuffle=True, drop_last=True)
     get_dataloader = partial(DataLoader,
                              batch_size=batch_size,
-                             num_workers=0 if os.environ.get('DEBUG') else cpu_count(),
                              shuffle=True, drop_last=True)
 
     datasets = map(config.pop, ('train', 'val'))