diff --git a/forge/forge/op/eval/forge/transpose.py b/forge/forge/op/eval/forge/transpose.py
index 46641d429..bfb7cb544 100644
--- a/forge/forge/op/eval/forge/transpose.py
+++ b/forge/forge/op/eval/forge/transpose.py
@@ -63,67 +63,6 @@ def decompose(self, dc, inputs):
                 )
             )
 
-    def decompose_post_optimize(self, dc, inputs):
-        orig_shape = inputs[0].shape
-        if (
-            len(orig_shape) > 2
-            and self.dim0 == -3
-            and self.dim1 == -1
-            and ((len(orig_shape) == 4 and orig_shape[-4] == 1) or len(orig_shape) < 4)
-        ):
-            # XZ transpose
-            result = inputs[0]
-            use_sparse_mm = True
-
-            result = inputs[0]
-            result = dc.op("pad_tile", [result], (-2, orig_shape[-2]))
-            result = dc.op("pad_tile", [result], (-1, orig_shape[-1]))
-            result = dc.op(TransposeTM.create(-2, -1), [result])
-
-            if result.shape[-3] > 1:
-                result = dc.op("vstack", [result], (orig_shape[-3],))
-            i_spm = sparse_utils.create_sparse_interleave_picker_matrix(
-                result.shape[-2], orig_shape[-1], orig_shape[-3]
-            )
-            result = picker_matmul(use_sparse_mm, dc, i_spm, result)
-
-            if orig_shape[-1] > 1:
-                result = dc.op("vslice", [result], (orig_shape[-1],))
-            result = dc.op(TransposeTM.create(-2, -1), [result])
-
-            result = dc.op("narrow", [result], (-2, 0, orig_shape[-2], result.shape[-2]))
-            result = dc.op("narrow", [result], (-1, 0, orig_shape[-3], result.shape[-1]))
-
-            dc.fuse(result)
-
-        elif (
-            self.dim0 == -3
-            and self.dim1 == -2
-            and ((len(orig_shape) == 4 and orig_shape[0] == 1) or len(orig_shape) == 3)
-        ):
-            # YZ transpose
-            result = inputs[0]
-            use_sparse_mm = True
-
-            result = inputs[0]
-            result = dc.op("pad_tile", [result], (-2, orig_shape[-2]))
-            result = dc.op("pad_tile", [result], (-1, orig_shape[-1]))
-
-            if result.shape[-3] > 1:
-                result = dc.op("vstack", [result], (orig_shape[-3],))
-            i_spm = sparse_utils.create_sparse_interleave_picker_matrix(
-                result.shape[-2], orig_shape[-2], orig_shape[-3]
-            )
-            result = picker_matmul(use_sparse_mm, dc, i_spm, result)
-
-            if orig_shape[-2] > 1:
-                result = dc.op("vslice", [result], (orig_shape[-2],))
-
-            result = dc.op("narrow", [result], (-2, 0, orig_shape[-3], result.shape[-2]))
-            result = dc.op("narrow", [result], (-1, 0, orig_shape[-1], result.shape[-1]))
-
-            dc.fuse(result)
-
 
 def picker_matmul(use_sparse_mm, dc, s, result):
     if use_sparse_mm:
diff --git a/forge/test/mlir/test_ops.py b/forge/test/mlir/test_ops.py
index 48dc50257..37e7a8e47 100644
--- a/forge/test/mlir/test_ops.py
+++ b/forge/test/mlir/test_ops.py
@@ -275,6 +275,9 @@ def forward(self, a, b):
         ((32, 128), (0, 1)),
         ((18, 65), (1, 0)),
         ((6, 33, 34), (-1, 1)),
+        ((1, 32, 64), (-2, -3)),
+        ((6, 33, 34), (-1, -3)),
+        ((32, 128, 24), (1, -3)),
     ],
 )
 def test_transpose(params):
@@ -293,6 +296,10 @@ def forward(self, a):
     fw_out = framework_model(*inputs)
 
     compiled_model = forge.compile(framework_model, sample_inputs=inputs)
+
+    if params[1][1] == -3:
+        pytest.xfail("Currently the lowering to TTNN is not supported for -3 dim")
+
     co_out = compiled_model(*inputs)
 
     co_out = [co.to("cpu") for co in co_out]