diff --git a/src/llmcompressor/modifiers/quantization/gptq/base.py b/src/llmcompressor/modifiers/quantization/gptq/base.py
index 8c66c272b..0c3c96d34 100644
--- a/src/llmcompressor/modifiers/quantization/gptq/base.py
+++ b/src/llmcompressor/modifiers/quantization/gptq/base.py
@@ -120,7 +120,7 @@ class GPTQModifier(Modifier, HooksMixin):
     disable_quantization_observer_epoch: Optional[float] = None
 
     # private variables
-    _quantization_modifier: Optional[QuantizationModifier] = PrivateAttr()
+    _quantization_modifier: Optional[QuantizationModifier] = PrivateAttr(default=None)
     _hessians: Dict[torch.nn.Module, torch.Tensor] = PrivateAttr(default_factory=dict)
     _num_samples: Dict[torch.nn.Module, int] = PrivateAttr(default_factory=dict)
     _update_size: Optional[int] = PrivateAttr(default=None)
diff --git a/tests/llmcompressor/modifiers/quantization/gptq/utils/test_gptq_wrapper.py b/tests/llmcompressor/modifiers/quantization/gptq/utils/test_gptq_wrapper.py
deleted file mode 100644
index 203d1fe03..000000000
--- a/tests/llmcompressor/modifiers/quantization/gptq/utils/test_gptq_wrapper.py
+++ /dev/null
@@ -1,41 +0,0 @@
-from collections import OrderedDict
-
-import torch
-from compressed_tensors.quantization.lifecycle.apply import apply_quantization_config
-from compressed_tensors.quantization.quant_config import QuantizationConfig
-from compressed_tensors.quantization.quant_scheme import preset_name_to_scheme
-from loguru import logger
-
-from llmcompressor.modifiers.quantization.gptq.utils.gptq_wrapper import GPTQWrapper
-
-
-def test_ignore():
-    model = torch.nn.Sequential(
-        OrderedDict(
-            [
-                ("first_layer", torch.nn.Linear(2, 3)),
-                ("second_layer", torch.nn.Linear(3, 5)),
-            ]
-        )
-    )
-
-    config = QuantizationConfig(
-        config_groups={"group_0": preset_name_to_scheme("W8A8", targets=["Linear"])},
-        ignore=["first_layer"],
-    )
-    apply_quantization_config(model, config)
-
-    messages = []
-    logger.add(lambda m: messages.append(m))
-
-    with torch.no_grad():
-        first_compressor = GPTQWrapper("first_layer", model.first_layer)
-        first_compressor.add_batch(torch.ones(2), None)
-        first_compressor.compress()
-
-        second_compressor = GPTQWrapper("second_layer", model.second_layer)
-        second_compressor.add_batch(torch.ones(3), None)
-        second_compressor.compress()
-
-    assert sum("Skipping unquantized layer first_layer" in m for m in messages) == 1
-    assert sum("Skipping unquantized layer second_layer" in m for m in messages) == 0