Add permute and pow ops and fix attribute value issue.

onnxscript/function_libs/torch_aten/ops/core.py

@@ -3708,10 +3708,11 @@ def aten_pdist(self: TensorType, p: float = 2) -> TensorType:
     raise NotImplementedError()
 
 
-def aten_permute(self: TensorType, dims: Sequence[int]) -> TensorType:
+@torch_op("aten::permute")
+def aten_permute(self: TTensor, dims: Sequence[int]) -> TTensor:
     # permute(Tensor(a) self, int[] dims) -> Tensor(a)
 
-    raise NotImplementedError()
+    return op.Transpose(self, perm=dims)
 
 
 def aten_permute_copy(self: TensorType, dims: Sequence[int]) -> TensorType:
@@ -3781,6 +3782,13 @@ def aten_positive(self: TensorType) -> TensorType:
     raise NotImplementedError()
 
 
+@torch_op("aten::pow")
+def aten_pow(self: TReal, exponent: TTensor) -> TReal:
+    # pow(Tensor self, Tensor exponent) -> Tensor
+
+    return op.Pow(self, exponent)
+
+
 def aten_prelu(self: TensorType, weight: TensorType) -> TensorType:
     # prelu(Tensor self, Tensor weight) -> Tensor
 

onnxscript/test/function_libs/torch_aten/ops_correctness_test.py

@@ -315,6 +315,8 @@ def _topk_input_wrangler(
     "nonzero": core_ops.aten_nonzero,
     "ones_like": core_ops.aten_ones_like,
     "ones": core_ops.aten_ones,
+    "permute": core_ops.aten_permute,
+    "pow": core_ops.aten_pow,
     "reciprocal": core_ops.aten_reciprocal,
     "remainder": core_ops.aten_remainder,
     "repeat": core_ops.aten_repeat,
@@ -452,6 +454,16 @@ def _topk_input_wrangler(
         matcher=lambda sample: "scale_factor" in sample.kwargs,
         reason="fixme: the scale_factor tests",
     ),
+    skip(
+        "permute",
+        matcher=lambda sample: len(list(filter(lambda v: v < 0, sample.args[0]))) > 0,
+        reason="Negative value in perm is not supported",
+    ),
+    skip(
+        "permute",
+        matcher=lambda sample: len(sample.args[0]) == 0,
+        reason="Empty perm is not supported",
+    ),
     skip(
         "slice",
         # kwargs {dim, start, end, step} is empty, we cannot give the default value

onnxscript/values.py

@@ -133,10 +133,19 @@ def adapt_kwargs(self, kwargs):
                 )
         return kwargs, closure
 
+    def _convert_kwargs_to_numpy(self, kwargs: dict[str, Any]) -> dict[str, Any]:
+        new_kwargs = {}
+        for k, v in kwargs.items():
+            new_kwargs[k] = v
+            if isinstance(v, tensor.Tensor):
+                new_kwargs[k] = v.value
+
+        return new_kwargs
+
     def __call__(self, *args, **kwargs):
         from onnxscript import evaluator  # pylint: disable=import-outside-toplevel
 
-        return evaluator.eval(self.opschema, args, kwargs)
+        return evaluator.eval(self.opschema, args, self._convert_kwargs_to_numpy(kwargs))
 
 
 @dataclasses.dataclass(repr=False, eq=False)