diff --git a/train.py b/train.py
index a13433ca..5a229907 100644
--- a/train.py
+++ b/train.py
@@ -199,9 +199,11 @@ def train():
 
     # Dataloader
     batch_size = min(batch_size, len(dataset))
+    nw = min([os.cpu_count(), batch_size if batch_size > 1 else 0, 16])
+    print('Using %g dataloader workers' % nw)
     dataloader = torch.utils.data.DataLoader(dataset,
                                              batch_size=batch_size,
-                                             num_workers=min([os.cpu_count(), batch_size if batch_size > 1 else 0, 16]),
+                                             num_workers=nw,
                                              shuffle=not opt.rect,  # Shuffle=True unless rectangular training is used
                                              pin_memory=True,
                                              collate_fn=dataset.collate_fn)