Update handler.py

handler.py

@@ -33,7 +33,7 @@ class EndpointHandler():
               truncation=True,
               max_length=512,
           )
-          token_ids, token_mask = tokens.input_ids.cuda(), tokens.attention_mask.cuda()
+          token_ids, token_mask = tokenized.input_ids.cuda(), tokenized.attention_mask.cuda()
           with torch.no_grad():
               out = model(token_ids, attention_mask=token_mask)
           meaned_logits = (out.logits * token_mask.unsqueeze(-1)).sum(1) / token_mask.sum(