FIX Int8 check for torchao v0.7.0 (#2284)

At one point, we need to perform a check for the quantization dtype.
This used to rely on the layout_tensor attribute, which was renamed to
tensor_impl. The code now checks both attributes.

src/peft/tuners/lora/torchao.py

@@ -43,7 +43,14 @@ def _check_dtype_supported(self):
         # TODO: Not required once int4_weight_only is properly supported by torchao
         base_layer = self.get_base_layer()
         weight = base_layer.weight
-        if hasattr(weight, "layout_tensor") and (weight.layout_tensor.data.dtype != torch.int8):
+        # pytest tests/test_gpu_examples.py::PeftTorchaoGPUTests::test_causal_lm_training_single_gpu_torchao_0_int8_weight_only
+        if (
+            # torchao 0.7.0+
+            (hasattr(weight, "tensor_impl") and (weight.tensor_impl.data.dtype != torch.int8))
+            or
+            # torchao < 0.7.0
+            (hasattr(weight, "layout_tensor") and (weight.layout_tensor.data.dtype != torch.int8))
+        ):
             raise ValueError(f"{type(self).__name__} only supports int8 weights for now.")
 
     def merge(self, safe_merge: bool = False, adapter_names: Optional[list[str]] = None) -> None: