diff --git a/libs/infinity_emb/infinity_emb/transformer/classifier/torch.py b/libs/infinity_emb/infinity_emb/transformer/classifier/torch.py
index d503c27a..a251c586 100644
--- a/libs/infinity_emb/infinity_emb/transformer/classifier/torch.py
+++ b/libs/infinity_emb/infinity_emb/transformer/classifier/torch.py
@@ -35,6 +35,7 @@ def __init__(
 
         self._infinity_tokenizer = AutoTokenizer.from_pretrained(
             engine_args.model_name_or_path,
+            revision=engine_args.revision,
             trust_remote_code=engine_args.trust_remote_code,
         )
 
diff --git a/libs/infinity_emb/infinity_emb/transformer/crossencoder/optimum.py b/libs/infinity_emb/infinity_emb/transformer/crossencoder/optimum.py
index e0f2cf80..6992d30e 100644
--- a/libs/infinity_emb/infinity_emb/transformer/crossencoder/optimum.py
+++ b/libs/infinity_emb/infinity_emb/transformer/crossencoder/optimum.py
@@ -40,14 +40,18 @@ def __init__(self, *, engine_args: EngineArgs):
             file_name=onnx_file.as_posix(),
             optimize_model=not os.environ.get("INFINITY_ONNX_DISABLE_OPTIMIZE", False),
             model_class=ORTModelForSequenceClassification,
+            revision=engine_args.revision,
+            trust_remote_code=engine_args.trust_remote_code,
         )
         self.model.use_io_binding = False
         self.tokenizer = AutoTokenizer.from_pretrained(
             engine_args.model_name_or_path,
+            revision=engine_args.revision,
             trust_remote_code=engine_args.trust_remote_code,
         )
         self.config = AutoConfig.from_pretrained(
             engine_args.model_name_or_path,
+            revision=engine_args.revision,
             trust_remote_code=engine_args.trust_remote_code,
         )
         self._infinity_tokenizer = copy.deepcopy(self.tokenizer)
diff --git a/libs/infinity_emb/infinity_emb/transformer/embedder/neuron.py b/libs/infinity_emb/infinity_emb/transformer/embedder/neuron.py
index 95e9e6d7..7781ca25 100644
--- a/libs/infinity_emb/infinity_emb/transformer/embedder/neuron.py
+++ b/libs/infinity_emb/infinity_emb/transformer/embedder/neuron.py
@@ -82,8 +82,16 @@ def __init__(self, *, engine_args: EngineArgs):
             else cls_token_pooling
         )
 
-        self.tokenizer = AutoTokenizer.from_pretrained(engine_args.model_name_or_path)
-        self.config = AutoConfig.from_pretrained(engine_args.model_name_or_path)
+        self.tokenizer = AutoTokenizer.from_pretrained(
+            engine_args.model_name_or_path,
+            revision=engine_args.revision,
+            trust_remote_code=engine_args.trust_remote_code,
+        )
+        self.config = AutoConfig.from_pretrained(
+            engine_args.model_name_or_path,
+            revision=engine_args.revision,
+            trust_remote_code=engine_args.trust_remote_code,
+        )
         self._infinity_tokenizer = copy.deepcopy(self.tokenizer)
 
         compiler_args = {"num_cores": get_nc_count(), "auto_cast_type": "fp16"}
@@ -97,6 +105,8 @@ def __init__(self, *, engine_args: EngineArgs):
         }
         self.model = NeuronModelForFeatureExtraction.from_pretrained(
             model_id=engine_args.model_name_or_path,
+            revision=engine_args.revision,
+            trust_remote_code=engine_args.trust_remote_code,
             export=True,
             **compiler_args,
             **input_shapes,
diff --git a/libs/infinity_emb/infinity_emb/transformer/embedder/optimum.py b/libs/infinity_emb/infinity_emb/transformer/embedder/optimum.py
index d80d3d52..fdf31d2d 100644
--- a/libs/infinity_emb/infinity_emb/transformer/embedder/optimum.py
+++ b/libs/infinity_emb/infinity_emb/transformer/embedder/optimum.py
@@ -50,6 +50,8 @@ def __init__(self, *, engine_args: EngineArgs):
 
         self.model = optimize_model(
             model_name_or_path=engine_args.model_name_or_path,
+            revision=engine_args.revision,
+            trust_remote_code=engine_args.trust_remote_code,
             execution_provider=provider,
             file_name=onnx_file.as_posix(),
             optimize_model=not os.environ.get("INFINITY_ONNX_DISABLE_OPTIMIZE", False),
@@ -59,10 +61,12 @@ def __init__(self, *, engine_args: EngineArgs):
 
         self.tokenizer = AutoTokenizer.from_pretrained(
             engine_args.model_name_or_path,
+            revision=engine_args.revision,
             trust_remote_code=engine_args.trust_remote_code,
         )
         self.config = AutoConfig.from_pretrained(
             engine_args.model_name_or_path,
+            revision=engine_args.revision,
             trust_remote_code=engine_args.trust_remote_code,
         )
         self._infinity_tokenizer = copy.deepcopy(self.tokenizer)
diff --git a/libs/infinity_emb/infinity_emb/transformer/utils_optimum.py b/libs/infinity_emb/infinity_emb/transformer/utils_optimum.py
index c044339e..4021fd4b 100644
--- a/libs/infinity_emb/infinity_emb/transformer/utils_optimum.py
+++ b/libs/infinity_emb/infinity_emb/transformer/utils_optimum.py
@@ -70,6 +70,7 @@ def optimize_model(
     file_name: str,
     optimize_model=False,
     revision: Optional[str] = None,
+    trust_remote_code: bool = True,
 ):
     CHECK_ONNXRUNTIME.mark_required()
     path_folder = (
@@ -82,6 +83,7 @@ def optimize_model(
         return model_class.from_pretrained(
             model_name_or_path,
             revision=revision,
+            trust_remote_code=trust_remote_code,
             provider=execution_provider,
             file_name=file_name,
             provider_options={
@@ -100,15 +102,17 @@ def optimize_model(
         return model_class.from_pretrained(
             file_optimized.parent.as_posix(),
             revision=revision,
+            trust_remote_code=trust_remote_code,
             provider=execution_provider,
             file_name=file_optimized.name,
         )
 
     unoptimized_model_path = model_class.from_pretrained(
         model_name_or_path,
+        revision=revision,
+        trust_remote_code=trust_remote_code,
         provider=execution_provider,
         file_name=file_name,
-        revision=revision,
     )
     if not optimize_model or execution_provider == "TensorrtExecutionProvider":
         return unoptimized_model_path
@@ -137,6 +141,8 @@ def optimize_model(
 
         model = model_class.from_pretrained(
             optimized_model_path,
+            revision=revision,
+            trust_remote_code=trust_remote_code,
             provider=execution_provider,
             file_name=Path(file_name).name.replace(".onnx", "_optimized.onnx"),
         )