diff --git a/aten/src/ATen/native/cpu/UnaryOpsKernel.cpp b/aten/src/ATen/native/cpu/UnaryOpsKernel.cpp
index 32ebaf7752f7..6ed4b3af6f23 100644
--- a/aten/src/ATen/native/cpu/UnaryOpsKernel.cpp
+++ b/aten/src/ATen/native/cpu/UnaryOpsKernel.cpp
@@ -336,7 +336,7 @@ static void cosh_kernel(TensorIterator& iter) {
 }
 
 static void acosh_kernel(TensorIterator& iter) {
-    AT_DISPATCH_FLOATING_TYPES(iter.dtype(), "acosh_cpu", [&]() {
+    AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(iter.dtype(), "acosh_cpu", [&]() {
       cpu_kernel(
         iter,
         [=](scalar_t a) -> scalar_t { return std::acosh(a); });
@@ -344,7 +344,7 @@ static void acosh_kernel(TensorIterator& iter) {
 }
 
 static void asinh_kernel(TensorIterator& iter) {
-    AT_DISPATCH_FLOATING_TYPES(iter.dtype(), "asinh_cpu", [&]() {
+    AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(iter.dtype(), "asinh_cpu", [&]() {
       cpu_kernel(
         iter,
         [=](scalar_t a) -> scalar_t { return std::asinh(a); });
@@ -352,7 +352,7 @@ static void asinh_kernel(TensorIterator& iter) {
 }
 
 static void atanh_kernel(TensorIterator& iter) {
-    AT_DISPATCH_FLOATING_TYPES(iter.dtype(), "atanh_cpu", [&]() {
+    AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(iter.dtype(), "atanh_cpu", [&]() {
       cpu_kernel(
         iter,
         [=](scalar_t a) -> scalar_t { return std::atanh(a); });
diff --git a/aten/src/ATen/native/cuda/UnaryGeometricKernels.cu b/aten/src/ATen/native/cuda/UnaryGeometricKernels.cu
index 867855217092..bac3a05439d2 100644
--- a/aten/src/ATen/native/cuda/UnaryGeometricKernels.cu
+++ b/aten/src/ATen/native/cuda/UnaryGeometricKernels.cu
@@ -75,7 +75,7 @@ void tanh_kernel_cuda(TensorIterator& iter) {
 }
 
 void acosh_kernel_cuda(TensorIterator& iter) {
-  AT_DISPATCH_FLOATING_TYPES_AND2(ScalarType::Half, ScalarType::BFloat16, iter.common_dtype(), "acosh_cuda", [&]() {
+  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES_AND2(ScalarType::Half, ScalarType::BFloat16, iter.common_dtype(), "acosh_cuda", [&]() {
     gpu_kernel(iter, []GPU_LAMBDA(scalar_t a) -> scalar_t {
             return ::acosh(a);
     });
@@ -83,7 +83,7 @@ void acosh_kernel_cuda(TensorIterator& iter) {
 }
 
 void asinh_kernel_cuda(TensorIterator& iter) {
-  AT_DISPATCH_FLOATING_TYPES_AND2(ScalarType::Half, ScalarType::BFloat16, iter.common_dtype(), "asinh_cuda", [&]() {
+  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES_AND2(ScalarType::Half, ScalarType::BFloat16, iter.common_dtype(), "asinh_cuda", [&]() {
     gpu_kernel(iter, []GPU_LAMBDA(scalar_t a) -> scalar_t {
             return ::asinh(a);
     });
@@ -91,7 +91,7 @@ void asinh_kernel_cuda(TensorIterator& iter) {
 }
 
 void atanh_kernel_cuda(TensorIterator& iter) {
-  AT_DISPATCH_FLOATING_TYPES_AND2(ScalarType::Half, ScalarType::BFloat16, iter.common_dtype(), "atanh_cuda", [&]() {
+  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES_AND2(ScalarType::Half, ScalarType::BFloat16, iter.common_dtype(), "atanh_cuda", [&]() {
     gpu_kernel(iter, []GPU_LAMBDA(scalar_t a) -> scalar_t {
             return ::atanh(a);
     });
diff --git a/tools/autograd/gen_variable_type.py b/tools/autograd/gen_variable_type.py
index c25c724559f6..08b6672d96fd 100644
--- a/tools/autograd/gen_variable_type.py
+++ b/tools/autograd/gen_variable_type.py
@@ -80,7 +80,7 @@
     'unbind', 'split', 'split_with_sizes', 'unsafe_split', 'split_with_sizes_backward',
     'dot', 'vdot', 'cholesky', 'triangular_solve', 'mm', '_unsafe_view', 'mv', 'ger',
     'bmm', 'diagonal', 'alias', 'atan', 'log', 'log10', 'log1p', 'log2', 'reciprocal',
-    'tan', 'pow', 'rsqrt', 'tanh', 'tanh_backward', 'asinh', 'acosh', 'take', 'fill_',
+    'tan', 'pow', 'rsqrt', 'tanh', 'tanh_backward', 'asinh', 'acosh', 'atanh', 'take', 'fill_',
     'exp', 'nonzero', 'mean', 'inverse', 'solve', 'linalg_cholesky', 'addcmul', 'addcdiv',
     'matrix_exp', 'linalg_eigh', 'cholesky_solve', 'linalg_qr', '_svd_helper', '_fft_c2c', '_fft_r2c',
     'linalg_solve', 'sqrt', 'stack', 'gather', 'index_select', 'index_add_', 'linalg_inv',
diff --git a/torch/testing/_internal/common_methods_invocations.py b/torch/testing/_internal/common_methods_invocations.py
index 6c983abffba0..52a2c75da97e 100644
--- a/torch/testing/_internal/common_methods_invocations.py
+++ b/torch/testing/_internal/common_methods_invocations.py
@@ -773,9 +773,9 @@ def sample_inputs_fliplr_flipud(op_info, device, dtype, requires_grad):
     UnaryUfuncInfo('acosh',
                    ref=np.arccosh,
                    domain=(1, float('inf')),
-                   dtypes=all_types_and(torch.bool),
-                   dtypesIfCPU=all_types_and(torch.bool),
-                   dtypesIfCUDA=all_types_and(torch.bool, torch.half, torch.bfloat16),
+                   dtypes=all_types_and_complex_and(torch.bool),
+                   dtypesIfCPU=all_types_and_complex_and(torch.bool),
+                   dtypesIfCUDA=all_types_and_complex_and(torch.bool, torch.half, torch.bfloat16),
                    promotes_integers_to_float=True,
                    decorators=(precisionOverride({torch.bfloat16: 5e-2}),),
                    test_inplace_grad=False,
@@ -783,6 +783,11 @@ def sample_inputs_fliplr_flipud(op_info, device, dtype, requires_grad):
                        # RuntimeError: "rsqrt_cuda" not implemented for 'BFloat16'
                        SkipInfo('TestCommon', 'test_variant_consistency_jit',
                                 device_type='cuda', dtypes=[torch.bfloat16]),
+                       SkipInfo('TestUnaryUfuncs', 'test_reference_numerics',
+                                device_type='cpu', dtypes=[torch.cfloat, torch.cdouble]),
+                       SkipInfo('TestUnaryUfuncs', 'test_reference_numerics',
+                                device_type='cuda', dtypes=[torch.cfloat, torch.cdouble],
+                                active_if=IS_WINDOWS),
                    )),
     OpInfo('addmm',
            dtypes=floating_types(),
@@ -827,9 +832,9 @@ def sample_inputs_fliplr_flipud(op_info, device, dtype, requires_grad):
     # NOTE: derivative for inplace asinh is not implemented
     UnaryUfuncInfo('asinh',
                    ref=np.arcsinh,
-                   dtypes=all_types_and(torch.bool),
-                   dtypesIfCPU=all_types_and(torch.bool),
-                   dtypesIfCUDA=all_types_and(torch.bool, torch.half, torch.bfloat16),
+                   dtypes=all_types_and_complex_and(torch.bool),
+                   dtypesIfCPU=all_types_and_complex_and(torch.bool),
+                   dtypesIfCUDA=all_types_and_complex_and(torch.bool, torch.half, torch.bfloat16),
                    promotes_integers_to_float=True,
                    decorators=(precisionOverride({torch.bfloat16: 5e-2}),),
                    test_inplace_grad=False,
@@ -837,6 +842,11 @@ def sample_inputs_fliplr_flipud(op_info, device, dtype, requires_grad):
                        # RuntimeError: "rsqrt_cuda" not implemented for 'BFloat16'
                        SkipInfo('TestCommon', 'test_variant_consistency_jit',
                                 device_type='cuda', dtypes=[torch.bfloat16]),
+                       SkipInfo('TestUnaryUfuncs', 'test_reference_numerics',
+                                device_type='cpu', dtypes=[torch.cfloat, torch.cdouble]),
+                       SkipInfo('TestUnaryUfuncs', 'test_reference_numerics',
+                                device_type='cuda', dtypes=[torch.cfloat, torch.cdouble],
+                                active_if=IS_WINDOWS),
                    )),
     UnaryUfuncInfo('atan',
                    ref=np.arctan,
@@ -857,12 +867,19 @@ def sample_inputs_fliplr_flipud(op_info, device, dtype, requires_grad):
     UnaryUfuncInfo('atanh',
                    ref=np.arctanh,
                    domain=(-1, 1),
-                   dtypes=all_types_and(torch.bool),
-                   dtypesIfCPU=all_types_and(torch.bool),
-                   dtypesIfCUDA=all_types_and(torch.bool, torch.half, torch.bfloat16),
+                   dtypes=all_types_and_complex_and(torch.bool),
+                   dtypesIfCPU=all_types_and_complex_and(torch.bool),
+                   dtypesIfCUDA=all_types_and_complex_and(torch.bool, torch.half, torch.bfloat16),
                    promotes_integers_to_float=True,
                    decorators=(precisionOverride({torch.bfloat16: 1e-2}),),
-                   test_inplace_grad=False),
+                   test_inplace_grad=False,
+                   skips=(
+                       SkipInfo('TestUnaryUfuncs', 'test_reference_numerics',
+                                device_type='cpu', dtypes=[torch.cfloat, torch.cdouble]),
+                       SkipInfo('TestUnaryUfuncs', 'test_reference_numerics',
+                                device_type='cuda', dtypes=[torch.cfloat, torch.cdouble],
+                                active_if=IS_WINDOWS),
+                   )),
     OpInfo('broadcast_to',
            dtypes=all_types_and_complex_and(torch.bool, torch.float16, torch.bfloat16),
            supports_tensor_out=False,