diff --git a/src/base/elu.h b/src/base/elu.h
index 0d942f97f..9d47fd49e 100644
--- a/src/base/elu.h
+++ b/src/base/elu.h
@@ -7,6 +7,21 @@ namespace infini::ops {
 
 class Elu : public Operator<Elu> {
  public:
+  Elu(const Tensor input, const double alpha, Tensor out)
+      : input_shape_{input.shape()},
+        input_strides_{input.strides()},
+        input_type_{input.dtype()},
+        out_shape_{out.shape()},
+        out_strides_{out.strides()},
+        out_type_{out.dtype()},
+        alpha_{alpha},
+        scale_{1.0},
+        input_scale_{1.0},
+        device_index_{out.device().index()} {}
+
+  /// \deprecated Use `Elu(input, alpha, out)`. This constructor will be
+  /// removed in a future release.
+  [[deprecated("Use the `(input, alpha, out)` overload instead.")]]
   Elu(const Tensor input, const double alpha, const double scale,
       const double input_scale, Tensor out)
       : input_shape_{input.shape()},
@@ -20,6 +35,13 @@ class Elu : public Operator<Elu> {
         input_scale_{input_scale},
         device_index_{out.device().index()} {}
 
+  void operator()(const Tensor input, const double alpha, Tensor out) const {
+    return operator()(input, alpha, 1.0, 1.0, out);
+  }
+
+  /// \deprecated Use `operator()(input, alpha, out)`. This overload will be
+  /// removed in a future release.
+  [[deprecated("Use `operator()(input, alpha, out)` instead.")]]
   virtual void operator()(const Tensor input, const double alpha,
                           const double scale, const double input_scale,
                           Tensor out) const = 0;
diff --git a/tests/test_torch_ops.py b/tests/test_torch_ops.py
index 99021467d..2bfd8c50d 100644
--- a/tests/test_torch_ops.py
+++ b/tests/test_torch_ops.py
@@ -96,8 +96,8 @@
     ("softplus", "beta"): 1.0,
     ("softplus", "threshold"): 20.0,
     ("elu", "alpha"): 1.0,
-    ("elu", "scale"): 1.0,
-    ("elu", "input_scale"): 1.0,
+    ("elu", "scale"): 1.25,
+    ("elu", "input_scale"): 0.75,
     ("sub", "alpha"): 1.0,
     ("addcmul", "value"): 1.0,
     ("addcdiv", "value"): 1.0,
@@ -530,8 +530,13 @@ def test_op(op_meta, shape, dtype, device, rtol, atol):
         clone_strided(x) if isinstance(x, torch.Tensor) else x for x in inputs
     ]
 
+    if aten_name == "elu":
+        ref_func = torch.ops.aten.elu.default
+    else:
+        ref_func = _torch_func(aten_name)
+
     try:
-        ref = _torch_func(aten_name)(*ref_inputs)
+        ref = ref_func(*ref_inputs)
     except (
         RuntimeError,
         TypeError,
@@ -595,3 +600,11 @@ def test_op(op_meta, shape, dtype, device, rtol, atol):
 
     for actual, expected in zip(outs, ref_outs):
         _assert_close(actual, expected, rtol, atol)
+
+    if aten_name == "elu":
+        input, alpha = inputs[:2]
+        alpha_out = torch.empty_like(ref_outs[0])
+        _call_infini(op_name, input, alpha, alpha_out)
+        _assert_close(
+            alpha_out, torch.nn.functional.elu(input, alpha=alpha), rtol, atol
+        )