Fix an autoquant bug in flatten/unflatten

test/integration/test_integration.py

@@ -777,6 +777,18 @@ def test_aq_float8_weight_only_quant_subclass(self, device, dtype):
             AQFloat8WeightOnlyQuantizedLinearWeight.from_float, device, 30, test_dtype=dtype
         )
 
+    def test_autoquantizable_flatten_unflatten(self):
+        from torchao.quantization import DEFAULT_AUTOQUANT_CLASS_LIST
+        weight = torch.randn(16, 32)
+        qtensor_class_list = DEFAULT_AUTOQUANT_CLASS_LIST
+        aqw = AutoQuantizableLinearWeight.from_float(weight, qtensor_class_list)
+        tensor_data_name_dict, tensor_attributes = aqw.__tensor_flatten__()
+        tensor_data_dict = {name: getattr(aqw, name) for name in tensor_data_name_dict}
+        outer_size = aqw.size()
+        outer_stride = aqw.stride()
+        reconstructed = type(aqw).__tensor_unflatten__(tensor_data_dict, tensor_attributes, outer_size, outer_stride)
+
+
     @parameterized.expand(COMMON_DEVICE_DTYPE)
     @unittest.skipIf(not TORCH_VERSION_AT_LEAST_2_5, "autoquant+aqt needs newer pytorch")
     @unittest.skipIf(not is_H100, "Need H100 to run")

torchao/quantization/autoquant.py

@@ -217,7 +217,7 @@ def __tensor_unflatten__(
         cls, tensor_data_dict, tensor_attributes, outer_size=None, outer_stride=None
     ):
         weight = tensor_data_dict["weight"]
-        qtensor_class_list, mode, dtype, shape = tensor_attributes[0]
+        qtensor_class_list, mode, dtype, shape = tensor_attributes
         return cls(
             weight,
             qtensor_class_list,