diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 9ce558495..6f22b7cb9 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -7,7 +7,7 @@ repos:
       - id: trailing-whitespace
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.0.291
+    rev: v0.0.290
     hooks:
       - id: ruff
         args: [ --fix ]
diff --git a/text/chinese_bert.py b/text/chinese_bert.py
index 8159425df..a7607198f 100644
--- a/text/chinese_bert.py
+++ b/text/chinese_bert.py
@@ -4,8 +4,6 @@
 
 tokenizer = AutoTokenizer.from_pretrained("./bert/chinese-roberta-wwm-ext-large")
 
-models = dict()
-
 
 def get_bert_feature(text, word2ph, device=None):
     if (
@@ -16,15 +14,14 @@ def get_bert_feature(text, word2ph, device=None):
         device = "mps"
     if not device:
         device = "cuda"
-    if device not in models.keys():
-        models[device] = AutoModelForMaskedLM.from_pretrained(
-            "./bert/chinese-roberta-wwm-ext-large"
-        ).to(device)
+    model = AutoModelForMaskedLM.from_pretrained(
+        "./bert/chinese-roberta-wwm-ext-large"
+    ).to(device)
     with torch.no_grad():
         inputs = tokenizer(text, return_tensors="pt")
         for i in inputs:
             inputs[i] = inputs[i].to(device)
-        res = models[device](**inputs, output_hidden_states=True)
+        res = model(**inputs, output_hidden_states=True)
         res = torch.cat(res["hidden_states"][-3:-2], -1)[0].cpu()
 
     assert len(word2ph) == len(text) + 2
diff --git a/text/japanese_bert.py b/text/japanese_bert.py
index 69004002d..0953b1bd3 100644
--- a/text/japanese_bert.py
+++ b/text/japanese_bert.py
@@ -9,8 +9,6 @@
 # default value is 3(untested)
 BERT_LAYER = 3
 
-models = dict()
-
 
 def get_bert_feature(text, word2ph, device=None):
     if (
@@ -21,10 +19,7 @@ def get_bert_feature(text, word2ph, device=None):
         device = "mps"
     if not device:
         device = "cuda"
-    if device not in models.keys():
-        models[device] = AutoModelForMaskedLM.from_pretrained(
-            "./bert/bert-base-japanese-v3"
-        ).to(device)
+    model = AutoModelForMaskedLM.from_pretrained(BERT).to(device)
     with torch.no_grad():
         inputs = tokenizer(text, return_tensors="pt")
         for i in inputs: