diff --git a/optimum/intel/neural_compressor/modeling_base.py b/optimum/intel/neural_compressor/modeling_base.py
index 0a6521b211..63110bf305 100644
--- a/optimum/intel/neural_compressor/modeling_base.py
+++ b/optimum/intel/neural_compressor/modeling_base.py
@@ -186,7 +186,7 @@ def _from_pretrained(
                 **kwargs,
             )
 
-            return model
+            return cls(model, config=config, model_save_dir=None, **kwargs).model
 
         if use_auth_token is not None:
             warnings.warn(
diff --git a/setup.py b/setup.py
index 13c1b6796b..d61d5c1858 100644
--- a/setup.py
+++ b/setup.py
@@ -59,7 +59,7 @@
 QUALITY_REQUIRE = ["black~=23.1", "ruff==0.4.4"]
 
 EXTRAS_REQUIRE = {
-    "neural-compressor": ["neural-compressor>3.0", "accelerate", "transformers<4.43"],
+    "neural-compressor": ["neural-compressor>3.0", "accelerate", "transformers<=4.43.2"],
     "openvino": ["openvino>=2023.3", "nncf>=2.11.0", "openvino-tokenizers[transformers]"],
     "nncf": ["nncf>=2.11.0"],
     "ipex": ["intel-extension-for-pytorch", "transformers>=4.39,<4.45"],