diff --git a/fastai_contrib/utils.py b/fastai_contrib/utils.py
index 0cd855e..3017ffc 100644
--- a/fastai_contrib/utils.py
+++ b/fastai_contrib/utils.py
@@ -97,7 +97,7 @@ def get_sentencepiece(path:PathOrStr, trn_path:Path, name:str, rules:ListRules=N
     vocab = Vocab(pickle.load(open(path / 'models' / f'itos_{name}.pkl', 'rb')))
     # We cannot use lambdas or local methods here, since `tok_func` needs to be
     # pickle-able in order to be called in subprocesses when multithread tokenizing
-    tokenizer = Tokenizer(tok_func=SentencepieceTokenizer, lang=str(path / 'models'), pre_rules=rules, post_rules=[])
+    tokenizer = Tokenizer(tok_func=SentencepieceTokenizer, lang=str(path / 'models'), rules=rules)
     
     clear_cache_directory(path, cache_name)
 
diff --git a/ulmfit/pretrain_lm.py b/ulmfit/pretrain_lm.py
index fe2d4b7..64b2da2 100644
--- a/ulmfit/pretrain_lm.py
+++ b/ulmfit/pretrain_lm.py
@@ -72,7 +72,7 @@ def pretrain_lm(dir_path, lang='en', cuda_id=0, qrnn=True, subword=False, max_vo
 
         sp = get_sentencepiece(dir_path, trn_path, name, vocab_size=max_vocab)
 
-        data_lm = TextLMDataBunch.from_csv(dir_path, **sp)
+        data_lm = TextLMDataBunch.from_csv(dir_path, 'train.csv', **sp)
         itos = data_lm.train_ds.vocab.itos
         stoi = data_lm.train_ds.vocab.stoi
     else: