diff --git a/src/app/visual_bge/eva_clip/hf_model.py b/src/app/visual_bge/eva_clip/hf_model.py
index c4b9fd8..51bcff3 100644
--- a/src/app/visual_bge/eva_clip/hf_model.py
+++ b/src/app/visual_bge/eva_clip/hf_model.py
@@ -101,8 +101,7 @@ def __init__(
             else:
                 create_func, model_args = (AutoModel.from_pretrained, model_name_or_path) if pretrained else (
                     AutoModel.from_config, self.config)
-            # TODO: do all model configs have this attribute? PretrainedConfig does so yes??
-            if hasattr(self.config, "is_encoder_decoder") and self.config.is_encoder_decoder:
+            if getattr(self.config, "is_encoder_decoder", False):
                 self.transformer = create_func(model_args)
                 self.transformer = self.transformer.encoder
             else:
diff --git a/src/tests/test_hf_model.py b/src/tests/test_hf_model.py
new file mode 100644
index 0000000..d2d82d7
--- /dev/null
+++ b/src/tests/test_hf_model.py
@@ -0,0 +1,87 @@
+import unittest
+from unittest.mock import MagicMock, patch
+import torch.nn as nn
+
+from app.visual_bge.eva_clip.hf_model import HFTextEncoder
+
+
+class TestHFTextEncoderConfig(unittest.TestCase):
+    @patch("app.visual_bge.eva_clip.hf_model.AutoTokenizer")
+    @patch("app.visual_bge.eva_clip.hf_model.AutoModel")
+    @patch("app.visual_bge.eva_clip.hf_model.AutoConfig")
+    def test_encoder_decoder_true(
+        self, mock_autoconfig, mock_automodel, mock_autotokenizer
+    ):
+        mock_config = MagicMock()
+        mock_config.is_encoder_decoder = True
+        mock_config.model_type = "bert"
+        mock_config.hidden_size = 768
+        mock_config.vocab_size = 30522
+        mock_autoconfig.from_pretrained.return_value = mock_config
+
+        mock_model_instance = MagicMock()
+        mock_encoder = MagicMock(spec=nn.Module)
+        mock_model_instance.encoder = mock_encoder
+        mock_automodel.from_pretrained.return_value = mock_model_instance
+
+        encoder = HFTextEncoder(
+            model_name_or_path="dummy-model",
+            output_dim=768,
+            tokenizer_name="dummy-model",
+        )
+
+        mock_automodel.from_pretrained.assert_called_once_with("dummy-model")
+        self.assertEqual(encoder.transformer, mock_encoder)
+
+    @patch("app.visual_bge.eva_clip.hf_model.AutoTokenizer")
+    @patch("app.visual_bge.eva_clip.hf_model.AutoModel")
+    @patch("app.visual_bge.eva_clip.hf_model.AutoConfig")
+    def test_encoder_decoder_false(
+        self, mock_autoconfig, mock_automodel, mock_autotokenizer
+    ):
+        mock_config = MagicMock()
+        mock_config.is_encoder_decoder = False
+        mock_config.model_type = "bert"
+        mock_config.hidden_size = 768
+        mock_config.vocab_size = 30522
+        mock_autoconfig.from_pretrained.return_value = mock_config
+
+        mock_model_instance = MagicMock(spec=nn.Module)
+        mock_automodel.from_pretrained.return_value = mock_model_instance
+
+        encoder = HFTextEncoder(
+            model_name_or_path="dummy-model",
+            output_dim=768,
+            tokenizer_name="dummy-model",
+        )
+
+        mock_automodel.from_pretrained.assert_called_once_with(
+            "dummy-model", add_pooling_layer=False
+        )
+        self.assertEqual(encoder.transformer, mock_model_instance)
+
+    @patch("app.visual_bge.eva_clip.hf_model.AutoTokenizer")
+    @patch("app.visual_bge.eva_clip.hf_model.AutoModel")
+    @patch("app.visual_bge.eva_clip.hf_model.AutoConfig")
+    def test_encoder_decoder_missing_attribute(
+        self, mock_autoconfig, mock_automodel, mock_autotokenizer
+    ):
+        mock_config = MagicMock(spec=["model_type", "hidden_size", "vocab_size"])
+        mock_config.model_type = "bert"
+        mock_config.hidden_size = 768
+        mock_config.vocab_size = 30522
+        mock_autoconfig.from_pretrained.return_value = mock_config
+
+        mock_model_instance = MagicMock(spec=nn.Module)
+        mock_automodel.from_pretrained.return_value = mock_model_instance
+
+        encoder = HFTextEncoder(
+            model_name_or_path="dummy-model",
+            output_dim=768,
+            tokenizer_name="dummy-model",
+        )
+
+        mock_automodel.from_pretrained.assert_called_once_with(
+            "dummy-model", add_pooling_layer=False
+        )
+        self.assertEqual(encoder.transformer, mock_model_instance)