diff --git a/alpha_automl/wrapper_primitives/huggingface_text.py b/alpha_automl/wrapper_primitives/huggingface_text.py
index 383d304f..c6e0be41 100644
--- a/alpha_automl/wrapper_primitives/huggingface_text.py
+++ b/alpha_automl/wrapper_primitives/huggingface_text.py
@@ -13,9 +13,10 @@
 
 class HuggingfaceTextTransformer(BasePrimitive):
 
-    def __init__(self, name, tokenizer=None):
+    def __init__(self, name, tokenizer=None, max_length=512):
         self.name = name
         self.tokenizer = tokenizer if tokenizer else name
+        self.max_length = max_length
 
     def fit(self, X, y=None):
         return self
@@ -44,10 +45,8 @@ def transform(self, texts):
                 batch_texts = list_texts[start: start + batch_size]
 
 #             batch_texts = [' '.join(line.split()) if str(line)!='nan' else '' for line in batch_texts]
-            ids = tokenizer(batch_texts, padding=True, return_tensors="pt")
-            ids['input_ids'] = ids['input_ids'][:, :512]
-            ids['token_type_ids'] = ids['token_type_ids'][:, :512]
-            ids['attention_mask'] = ids['attention_mask'][:, :512]
+            ids = tokenizer(batch_texts, padding=True, return_tensors="pt", max_length=self.max_length)
+
             device = 'cuda' if torch.cuda.is_available() else 'cpu'
             model.to(device)
             ids = ids.to(device)