updates

train.py

@@ -140,11 +140,12 @@ def train(
     # plt.savefig('LR.png', dpi=300)
 
     # Dataset
+    rectangular_training = False
     dataset = LoadImagesAndLabels(train_path,
                                   img_size,
                                   batch_size,
                                   augment=True,
-                                  rect=False)
+                                  rect=rectangular_training)
 
     # Initialize distributed training
     if torch.cuda.device_count() > 1:
@@ -156,7 +157,7 @@ def train(
     dataloader = DataLoader(dataset,
                             batch_size=batch_size,
                             num_workers=opt.num_workers,
-                            shuffle=True,  # disable rectangular training if True
+                            shuffle=not rectangular_training,  # Shuffle=True unless rectangular training is used
                             pin_memory=True,
                             collate_fn=dataset.collate_fn)