diff --git a/bert_score/utils.py b/bert_score/utils.py
index 3ec0a7c..455f5b8 100644
--- a/bert_score/utils.py
+++ b/bert_score/utils.py
@@ -107,19 +107,27 @@ def sent_encode(tokenizer, sent):
         # for RoBERTa and GPT-2
         import transformers
 
-        if LooseVersion(transformers.__version__) >= LooseVersion("3.0.0"):
+        if LooseVersion(transformers.__version__) >= LooseVersion("4.0.0"):
             return tokenizer.encode(
                 sent, add_special_tokens=True, add_prefix_space=True, max_length=tokenizer.model_max_length,
                 truncation=True
             )
+        elif LooseVersion(transformers.__version__) >= LooseVersion("3.0.0"):
+            return tokenizer.encode(
+                sent, add_special_tokens=True, add_prefix_space=True, max_length=tokenizer.max_len,
+                truncation=True
+            )
         else:
             return tokenizer.encode(sent, add_special_tokens=True, add_prefix_space=True, max_length=tokenizer.max_len)
     else:
         import transformers
 
-        if LooseVersion(transformers.__version__) >= LooseVersion("3.0.0"):
+        if LooseVersion(transformers.__version__) >= LooseVersion("4.0.0"):
             return tokenizer.encode(sent, add_special_tokens=True, max_length=tokenizer.model_max_length,
                                     truncation=True)
+        elif LooseVersion(transformers.__version__) >= LooseVersion("3.0.0"):
+            return tokenizer.encode(sent, add_special_tokens=True, max_length=tokenizer.max_len,
+                                    truncation=True)
         else:
             return tokenizer.encode(sent, add_special_tokens=True, max_length=tokenizer.max_len)