Add op test for torch.unique_consecutive

tests/function_libs/torch_lib/extra_opinfo.py

@@ -2270,6 +2270,16 @@ def sample_inputs_unique_dim(op_info, device, dtype, requires_grad, **kwargs):
             yield sample
 
 
+def sample_inputs_unique_consecutive(op_info, device, dtype, requires_grad, **kwargs):
+    for sample in common_methods_invocations.sample_inputs_unique(
+        op_info, device, dtype, requires_grad, **kwargs
+    ):
+        # unique_consecutive only supports dim=None or (dim=0 with rank=1)
+        # So filter out samples with dim != None
+        if sample.kwargs.get("dim") is None:
+            yield sample
+
+
 def sample_inputs_upsample_trilinear3d_vec(op_info, device, dtype, requires_grad, **kwargs):
     del op_info
     del kwargs
@@ -2878,6 +2888,14 @@ def __init__(self):
         supports_out=False,
         supports_autograd=False,
     ),
+    opinfo_core.OpInfo(
+        "ops.aten.unique_consecutive",
+        aten_name="unique_consecutive",
+        dtypes=common_dtype.integral_types(),
+        sample_inputs_func=sample_inputs_unique_consecutive,
+        supports_out=False,
+        supports_autograd=False,
+    ),
     opinfo_core.OpInfo(
         "ops.aten.upsample_bicubic2d.default",
         aten_name="upsample_bicubic2d",

tests/function_libs/torch_lib/ops_test_data.py

@@ -1865,6 +1865,7 @@ def _where_input_wrangler(
             "Our implementation is based on that for CUDA"
         ),
     ),
+    TorchLibOpInfo("ops.aten.unique_consecutive", core_ops.aten_unique_consecutive),
     TorchLibOpInfo("ops.prims.broadcast_in_dim.default", prims_ops.prims_broadcast_in_dim),
     TorchLibOpInfo(
         "ops.prims.var.default", prims_ops.prims_var, tolerance={torch.float16: (1e-3, 5e-2)}