training updates

train.py

@@ -209,7 +209,7 @@ def train():
 
     # Start training
     nb = len(dataloader)  # number of batches
-    n_burn = max(3 * nb, 500)  # burn-in iterations, max(3 epochs, 300 iterations)
+    n_burn = max(3 * nb, 500)  # burn-in iterations, max(3 epochs, 500 iterations)
     maps = np.zeros(nc)  # mAP per class
     # torch.autograd.set_detect_anomaly(True)
     results = (0, 0, 0, 0, 0, 0, 0)  # 'P', 'R', 'mAP', 'F1', 'val GIoU', 'val Objectness', 'val Classification'