Update README.md

README.md

@@ -8,7 +8,7 @@ per_device_train_batch_size=16
 gradient_accumulation_steps=1
 total train batch size (w. parallel, distributed & accumulation) = 16
 learning_rate=3e-5
-num_train_epochs=2
+num_train_epochs=3
 ```
 
 ## Evaluation results