Update README.md

README.md

@@ -41,7 +41,7 @@ input_ids = tokenizer.encode('Tinha uma pedra no meio do caminho.', return_tenso
 
 with torch.no_grad():
     outs = model(input_ids)
-    encoded = outs.last_hidden_state[0, 1:-1]  # Ignore [CLS] and [SEP] special tokens
+    encoded = outs.last_hidden_state[0, 0]  # Take [CLS] special token representation
 ```
 
 ## Citation