Make release 1.12.0 candidate rc4 for testing (#4267)

* Use op name rather than hard-coding Hann in window op doc strings. (#4248) And strip trailing spaces. Signed-off-by: Gary Miguel <garymiguel@microsoft.com> Signed-off-by: Chun-Wei Chen <jacky82226@gmail.com> * Fix: Window functions generate incorrect shape for with symmetric attribute (#4256) * fix symmetric function definition, and add shape inference test Signed-off-by: Sheil Kumar <smk2007@gmail.com> * whitespace... Signed-off-by: Sheil Kumar <smk2007@gmail.com> * whitespace... Signed-off-by: Sheil Kumar <smk2007@gmail.com> * flake8 whitespace error Signed-off-by: Sheil Kumar <smk2007@gmail.com> * fix stft test Signed-off-by: Sheil Kumar <smk2007@gmail.com> Signed-off-by: Chun-Wei Chen <jacky82226@gmail.com> * support truncation & fix big-endian support (#4238) Signed-off-by: Ian Bearman <ianb@microsoft.com> Signed-off-by: Chun-Wei Chen <jacky82226@gmail.com> * Fix layer normalization's reference outputs (#4263) Signed-off-by: Chun-Wei Chen <jacky82226@gmail.com> * handle raw types correctly in helper.make_tensor (#4262) Signed-off-by: Chun-Wei Chen <jacky82226@gmail.com> * bump to 1.12.0rc4 for cherry-picks Signed-off-by: Chun-Wei Chen <jacky82226@gmail.com> * Fix sub-graph generation for LN (#4268) Signed-off-by: Wei-Sheng Chin <wschin@outlook.com> Signed-off-by: Chun-Wei Chen <jacky82226@gmail.com> Co-authored-by: Gary Miguel <garymiguel@microsoft.com> Co-authored-by: Sheil Kumar <smk2007@gmail.com> Co-authored-by: Ian Bearman <ianb@microsoft.com> Co-authored-by: Wei-Sheng Chin <wschin@outlook.com>
onnx · Jun 10, 2022 · b2845aa · b2845aa
1 parent 19944f1
commit b2845aa
Show file tree

Hide file tree

Showing 46 changed files with 144 additions and 90 deletions.
diff --git a/VERSION_NUMBER b/VERSION_NUMBER
@@ -1 +1 @@
-1.12.0rc3
+1.12.0rc4
diff --git a/docs/Changelog.md b/docs/Changelog.md
@@ -20845,7 +20845,7 @@ This version of the operator has been available since version 17 of the default
 
 <dl>
 <dt><tt>output</tt> (non-differentiable) : T2</dt>
-<dd>A Hann window with length: size. The output has the shape: [size].</dd>
+<dd>A Blackman window with length: size. The output has the shape: [size].</dd>
 </dl>
 
 #### Type Constraints
@@ -20929,7 +20929,7 @@ This version of the operator has been available since version 17 of the default
 
 <dl>
 <dt><tt>output</tt> (non-differentiable) : T2</dt>
-<dd>A Hann window with length: size. The output has the shape: [size].</dd>
+<dd>A Hamming window with length: size. The output has the shape: [size].</dd>
 </dl>
 
 #### Type Constraints

diff --git a/docs/Operators.md b/docs/Operators.md
@@ -2417,7 +2417,7 @@ This version of the operator has been available since version 17 of the default
 
 <dl>
 <dt><tt>output</tt> (non-differentiable) : T2</dt>
-<dd>A Hann window with length: size. The output has the shape: [size].</dd>
+<dd>A Blackman window with length: size. The output has the shape: [size].</dd>
 </dl>
 
 #### Type Constraints
@@ -8232,7 +8232,7 @@ This version of the operator has been available since version 17 of the default
 
 <dl>
 <dt><tt>output</tt> (non-differentiable) : T2</dt>
-<dd>A Hann window with length: size. The output has the shape: [size].</dd>
+<dd>A Hamming window with length: size. The output has the shape: [size].</dd>
 </dl>
 
 #### Type Constraints
@@ -9796,19 +9796,20 @@ for i in range(len(X.shape)):
 <summary>d_epsilon</summary>
 
 ```python
+epsilon = 1e-1
 X = np.random.randn(2, 3, 5).astype(np.float32)
 
 def case(axis: int) -> None:
     normalized_shape = calculate_normalized_shape(X.shape, axis)
     W = np.random.randn(*normalized_shape).astype(np.float32)
     B = np.random.randn(*normalized_shape).astype(np.float32)
-    Y, mean, inv_std_dev = _layer_normalization(X, W, B, axis)
+    Y, mean, inv_std_dev = _layer_normalization(X, W, B, axis, epsilon)
     node = onnx.helper.make_node(
         'LayerNormalization',
         inputs=['X', 'W', 'B'],
         outputs=['Y', 'Mean', 'InvStdDev'],
         axis=axis,
-        epsilon=1e-1
+        epsilon=epsilon
     )
 
     if axis < 0:
@@ -9848,7 +9849,7 @@ node = onnx.helper.make_node(
 )
 
 expect(node, inputs=[X, W, B], outputs=[Y, mean, inv_std_dev],
-        name='test_layer_normalization_default_axis')
+       name='test_layer_normalization_default_axis')
 ```
 
 </details>
@@ -19263,8 +19264,8 @@ nstfts = ((signal.shape[1] - length) // step) + 1
 output = np.empty([1, nstfts, onesided_length, 2], dtype=np.float32)
 for i in range(nstfts):
     start = i * step
-    stop = i * step + onesided_length
-    complex_out = np.fft.fft(signal[0, start:stop, 0])
+    stop = i * step + length
+    complex_out = np.fft.fft(signal[0, start:stop, 0])[0:onesided_length]
     output[0, i] = np.stack((complex_out.real, complex_out.imag), axis=1)
 
 expect(node, inputs=[signal, step, length], outputs=[output],

diff --git a/docs/TestCoverage.md b/docs/TestCoverage.md
@@ -6017,19 +6017,20 @@ for i in range(len(X.shape)):
 <summary>d_epsilon</summary>
 
 ```python
+epsilon = 1e-1
 X = np.random.randn(2, 3, 5).astype(np.float32)
 
 def case(axis: int) -> None:
     normalized_shape = calculate_normalized_shape(X.shape, axis)
     W = np.random.randn(*normalized_shape).astype(np.float32)
     B = np.random.randn(*normalized_shape).astype(np.float32)
-    Y, mean, inv_std_dev = _layer_normalization(X, W, B, axis)
+    Y, mean, inv_std_dev = _layer_normalization(X, W, B, axis, epsilon)
     node = onnx.helper.make_node(
         'LayerNormalization',
         inputs=['X', 'W', 'B'],
         outputs=['Y', 'Mean', 'InvStdDev'],
         axis=axis,
-        epsilon=1e-1
+        epsilon=epsilon
     )
 
     if axis < 0:
@@ -6067,7 +6068,7 @@ node = onnx.helper.make_node(
 )
 
 expect(node, inputs=[X, W, B], outputs=[Y, mean, inv_std_dev],
-        name='test_layer_normalization_default_axis')
+       name='test_layer_normalization_default_axis')
 ```
 
 </details>
@@ -12138,8 +12139,8 @@ nstfts = ((signal.shape[1] - length) // step) + 1
 output = np.empty([1, nstfts, onesided_length, 2], dtype=np.float32)
 for i in range(nstfts):
     start = i * step
-    stop = i * step + onesided_length
-    complex_out = np.fft.fft(signal[0, start:stop, 0])
+    stop = i * step + length
+    complex_out = np.fft.fft(signal[0, start:stop, 0])[0:onesided_length]
     output[0, i] = np.stack((complex_out.real, complex_out.imag), axis=1)
 
 expect(node, inputs=[signal, step, length], outputs=[output],

diff --git a/onnx/backend/test/case/node/layernormalization.py b/onnx/backend/test/case/node/layernormalization.py
@@ -1,8 +1,8 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import numpy as np  # type: ignore
-
 import onnx
+
 from ..base import Base
 from . import expect
 
@@ -112,7 +112,7 @@ def export_default_axis() -> None:
         )
 
         expect(node, inputs=[X, W, B], outputs=[Y, mean, inv_std_dev],
-                name='test_layer_normalization_default_axis')
+               name='test_layer_normalization_default_axis')
 
     @staticmethod
     def export2d() -> None:
@@ -145,19 +145,20 @@ def case(axis: int) -> None:
 
     @staticmethod
     def export3d_epsilon() -> None:
+        epsilon = 1e-1
         X = np.random.randn(2, 3, 5).astype(np.float32)
 
         def case(axis: int) -> None:
             normalized_shape = calculate_normalized_shape(X.shape, axis)
             W = np.random.randn(*normalized_shape).astype(np.float32)
             B = np.random.randn(*normalized_shape).astype(np.float32)
-            Y, mean, inv_std_dev = _layer_normalization(X, W, B, axis)
+            Y, mean, inv_std_dev = _layer_normalization(X, W, B, axis, epsilon)
             node = onnx.helper.make_node(
                 'LayerNormalization',
                 inputs=['X', 'W', 'B'],
                 outputs=['Y', 'Mean', 'InvStdDev'],
                 axis=axis,
-                epsilon=1e-1
+                epsilon=epsilon
             )
 
             if axis < 0:

diff --git a/onnx/backend/test/case/node/stft.py b/onnx/backend/test/case/node/stft.py
@@ -29,8 +29,8 @@ def export() -> None:
         output = np.empty([1, nstfts, onesided_length, 2], dtype=np.float32)
         for i in range(nstfts):
             start = i * step
-            stop = i * step + onesided_length
-            complex_out = np.fft.fft(signal[0, start:stop, 0])
+            stop = i * step + length
+            complex_out = np.fft.fft(signal[0, start:stop, 0])[0:onesided_length]
             output[0, i] = np.stack((complex_out.real, complex_out.imag), axis=1)
 
         expect(node, inputs=[signal, step, length], outputs=[output],

diff --git a/onnx/backend/test/data/node/test_blackmanwindow_expanded/model.onnx b/onnx/backend/test/data/node/test_blackmanwindow_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_blackmanwindow_symmetric_expanded/model.onnx b/onnx/backend/test/data/node/test_blackmanwindow_symmetric_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_hammingwindow_expanded/model.onnx b/onnx/backend/test/data/node/test_hammingwindow_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_hammingwindow_symmetric_expanded/model.onnx b/onnx/backend/test/data/node/test_hammingwindow_symmetric_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_hannwindow_expanded/model.onnx b/onnx/backend/test/data/node/test_hannwindow_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_hannwindow_symmetric_expanded/model.onnx b/onnx/backend/test/data/node/test_hannwindow_symmetric_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_layer_normalization_2d_axis0_expanded/model.onnx b/onnx/backend/test/data/node/test_layer_normalization_2d_axis0_expanded/model.onnx
diff --git a/...kend/test/data/node/test_layer_normalization_3d_axis0_epsilon/test_data_set_0/output_0.pb b/...kend/test/data/node/test_layer_normalization_3d_axis0_epsilon/test_data_set_0/output_0.pb
diff --git a/...kend/test/data/node/test_layer_normalization_3d_axis0_epsilon/test_data_set_0/output_2.pb b/...kend/test/data/node/test_layer_normalization_3d_axis0_epsilon/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJX�l?
+B	InvStdDevJp!c?
diff --git a/onnx/backend/test/data/node/test_layer_normalization_3d_axis0_epsilon_expanded/model.onnx b/onnx/backend/test/data/node/test_layer_normalization_3d_axis0_epsilon_expanded/model.onnx
diff --git a/.../data/node/test_layer_normalization_3d_axis0_epsilon_expanded/test_data_set_0/output_0.pb b/.../data/node/test_layer_normalization_3d_axis0_epsilon_expanded/test_data_set_0/output_0.pb
diff --git a/.../data/node/test_layer_normalization_3d_axis0_epsilon_expanded/test_data_set_0/output_2.pb b/.../data/node/test_layer_normalization_3d_axis0_epsilon_expanded/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJX�l?
+B	InvStdDevJp!c?
diff --git a/...kend/test/data/node/test_layer_normalization_3d_axis1_epsilon/test_data_set_0/output_0.pb b/...kend/test/data/node/test_layer_normalization_3d_axis1_epsilon/test_data_set_0/output_0.pb
@@ -1 +1 @@
-BYJx��A?�ʿ����-���F7>��@�+?w;�?f��>�c�MT�?�1t?�*�����>==V?lN?$k���:�?�o�>)�@$N<?pQ�>�|h>b��l�?���J�|�n����ܟ�
+BYJxF?C?��ſĢ���t�[I@>4x	@l�%?�u�?N��>�Vf�k{?ӀW?������@{�>�LV?>{??�g��䉎?|�>w�@4�8?���>�&L>��߿�?�?8����fz�� �����
diff --git a/...kend/test/data/node/test_layer_normalization_3d_axis1_epsilon/test_data_set_0/output_2.pb b/...kend/test/data/node/test_layer_normalization_3d_axis1_epsilon/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJ6�?5P?
+B	InvStdDevJ�>�?K�I?
diff --git a/.../data/node/test_layer_normalization_3d_axis1_epsilon_expanded/test_data_set_0/output_0.pb b/.../data/node/test_layer_normalization_3d_axis1_epsilon_expanded/test_data_set_0/output_0.pb
@@ -1 +1 @@
-BYJx��A?�ʿ����-���F7>��@�+?w;�?f��>�c�MT�?�1t?�*�����>==V?lN?$k���:�?�o�>)�@$N<?pQ�>�|h>b��l�?���J�|�n����ܟ�
+BYJxF?C?��ſĢ���t�[I@>4x	@l�%?�u�?N��>�Vf�k{?ӀW?������@{�>�LV?>{??�g��䉎?|�>w�@4�8?���>�&L>��߿�?�?8����fz�� �����
diff --git a/.../data/node/test_layer_normalization_3d_axis1_epsilon_expanded/test_data_set_0/output_2.pb b/.../data/node/test_layer_normalization_3d_axis1_epsilon_expanded/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJ6�?5P?
+B	InvStdDevJ�>�?K�I?
diff --git a/...kend/test/data/node/test_layer_normalization_3d_axis2_epsilon/test_data_set_0/output_0.pb b/...kend/test/data/node/test_layer_normalization_3d_axis2_epsilon/test_data_set_0/output_0.pb
diff --git a/...kend/test/data/node/test_layer_normalization_3d_axis2_epsilon/test_data_set_0/output_2.pb b/...kend/test/data/node/test_layer_normalization_3d_axis2_epsilon/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJ���?���?��@ ҥ?�?�d?
+B	InvStdDevJ�t�?X�?���?�q�?�?\?
diff --git a/.../data/node/test_layer_normalization_3d_axis2_epsilon_expanded/test_data_set_0/output_0.pb b/.../data/node/test_layer_normalization_3d_axis2_epsilon_expanded/test_data_set_0/output_0.pb
diff --git a/.../data/node/test_layer_normalization_3d_axis2_epsilon_expanded/test_data_set_0/output_2.pb b/.../data/node/test_layer_normalization_3d_axis2_epsilon_expanded/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJ���?���?��@ ҥ?�?�d?
+B	InvStdDevJ�t�?X�?���?�q�?�?\?
diff --git a/...data/node/test_layer_normalization_3d_axis_negative_1_epsilon/test_data_set_0/output_0.pb b/...data/node/test_layer_normalization_3d_axis_negative_1_epsilon/test_data_set_0/output_0.pb
diff --git a/...data/node/test_layer_normalization_3d_axis_negative_1_epsilon/test_data_set_0/output_2.pb b/...data/node/test_layer_normalization_3d_axis_negative_1_epsilon/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJ���?���?��@ ҥ?�?�d?
+B	InvStdDevJ�t�?X�?���?�q�?�?\?
diff --git a/.../test_layer_normalization_3d_axis_negative_1_epsilon_expanded/test_data_set_0/output_0.pb b/.../test_layer_normalization_3d_axis_negative_1_epsilon_expanded/test_data_set_0/output_0.pb
diff --git a/.../test_layer_normalization_3d_axis_negative_1_epsilon_expanded/test_data_set_0/output_2.pb b/.../test_layer_normalization_3d_axis_negative_1_epsilon_expanded/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJ���?���?��@ ҥ?�?�d?
+B	InvStdDevJ�t�?X�?���?�q�?�?\?
diff --git a/...data/node/test_layer_normalization_3d_axis_negative_2_epsilon/test_data_set_0/output_0.pb b/...data/node/test_layer_normalization_3d_axis_negative_2_epsilon/test_data_set_0/output_0.pb
@@ -1 +1,2 @@
-BYJxc�����?	�?���G�@"�2@J����??�	>ءH��#`>��=�~)�?L'�ktW�EA-��w�?Vl�?�M?�M@d�%�*_�?�O�?�NO@K�=��k���9�8�|�l��>
+BYJx�a�3�?�f?��ֿ�
+@3&@�����?<�?�&*>m<V��AP>c�=�mÁ?qJu�c3V�k�!����?�A�?LkQ?͚F@v�"��A�?��?�J@��=�jg��9���r����>
diff --git a/...data/node/test_layer_normalization_3d_axis_negative_2_epsilon/test_data_set_0/output_2.pb b/...data/node/test_layer_normalization_3d_axis_negative_2_epsilon/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJ6�?5P?
+B	InvStdDevJ�>�?K�I?
diff --git a/.../test_layer_normalization_3d_axis_negative_2_epsilon_expanded/test_data_set_0/output_0.pb b/.../test_layer_normalization_3d_axis_negative_2_epsilon_expanded/test_data_set_0/output_0.pb
@@ -1 +1,2 @@
-BYJxc�����?	�?���G�@"�2@J����??�	>ءH��#`>��=�~)�?L'�ktW�EA-��w�?Vl�?�M?�M@d�%�*_�?�O�?�NO@K�=��k���9�8�|�l��>
+BYJx�a�3�?�f?��ֿ�
+@3&@�����?<�?�&*>m<V��AP>c�=�mÁ?qJu�c3V�k�!����?�A�?LkQ?͚F@v�"��A�?��?�J@��=�jg��9���r����>
diff --git a/.../test_layer_normalization_3d_axis_negative_2_epsilon_expanded/test_data_set_0/output_2.pb b/.../test_layer_normalization_3d_axis_negative_2_epsilon_expanded/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJ6�?5P?
+B	InvStdDevJ�>�?K�I?
diff --git a/...data/node/test_layer_normalization_3d_axis_negative_3_epsilon/test_data_set_0/output_0.pb b/...data/node/test_layer_normalization_3d_axis_negative_3_epsilon/test_data_set_0/output_0.pb
@@ -1 +1 @@
-BYJx�y��/撿��>��<@z�f�@���,GܾLL^?6�?�{�>� ���8��$��P��[#����>��2>+u>G)>��7��6���?��,�܅����?��쾰X�����?F~2�ׯ�>
+BYJx��Ľ�����e�>ԙ8@`�y����aܾbh?�]?ں�>t����+�9���Fz�|#����>�eC>o�b>��2>��2�D���l�?;0�vį�tp�?�о�����?��8����>
diff --git a/...data/node/test_layer_normalization_3d_axis_negative_3_epsilon/test_data_set_0/output_2.pb b/...data/node/test_layer_normalization_3d_axis_negative_3_epsilon/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJX�l?
+B	InvStdDevJp!c?
diff --git a/.../test_layer_normalization_3d_axis_negative_3_epsilon_expanded/test_data_set_0/output_0.pb b/.../test_layer_normalization_3d_axis_negative_3_epsilon_expanded/test_data_set_0/output_0.pb
@@ -1 +1 @@
-BYJx�y��/撿��>��<@z�f�@���,GܾLL^?6�?�{�>� ���8��$��P��[#����>��2>+u>G)>��7��6���?��,�܅����?��쾰X�����?F~2�ׯ�>
+BYJx��Ľ�����e�>ԙ8@`�y����aܾbh?�]?ں�>t����+�9���Fz�|#����>�eC>o�b>��2>��2�D���l�?;0�vį�tp�?�о�����?��8����>
diff --git a/.../test_layer_normalization_3d_axis_negative_3_epsilon_expanded/test_data_set_0/output_2.pb b/.../test_layer_normalization_3d_axis_negative_3_epsilon_expanded/test_data_set_0/output_2.pb
@@ -1 +1 @@
-B	InvStdDevJX�l?
+B	InvStdDevJp!c?
diff --git a/onnx/backend/test/data/node/test_layer_normalization_4d_axis0_expanded/model.onnx b/onnx/backend/test/data/node/test_layer_normalization_4d_axis0_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_stft/test_data_set_0/output_0.pb b/onnx/backend/test/data/node/test_stft/test_data_set_0/output_0.pb
diff --git a/onnx/defs/math/defs.cc b/onnx/defs/math/defs.cc
@@ -757,7 +757,7 @@ ONNX_OPERATOR_SET_SCHEMA(
         .TypeAndShapeInferenceFunction(propagateShapeAndTypeFromFirstInput)
         .FunctionBody(R"ONNX(
           {
-            HS_X = HardSigmoid<alpha = 0.16666667163372, beta = 0.5>(X) 
+            HS_X = HardSigmoid<alpha = 0.16666667163372, beta = 0.5>(X)
             Y = Mul (X, HS_X)
           }
         )ONNX"));
@@ -1708,14 +1708,14 @@ ONNX_OPERATOR_SET_SCHEMA(
 
 static const char* QLinearMatMul_ver10_doc = R"DOC(
 Matrix product that behaves like numpy.matmul: https://docs.scipy.org/doc/numpy-1.13.0/reference/generated/numpy.matmul.html.
-It consumes two quantized input tensors, their scales and zero points, scale and zero point of output, 
-and computes the quantized output. The quantization formula is y = saturate((x / y_scale) + y_zero_point). 
-For (x / y_scale), it is rounding to nearest ties to even. Refer to https://en.wikipedia.org/wiki/Rounding for details. 
-Scale and zero point must have same shape. They must be either scalar (per tensor) or N-D tensor 
-(per row for 'a' and per column for 'b'). Scalar refers to per tensor quantization whereas N-D refers to per row 
-or per column quantization. If the input is 2D of shape [M, K] then zero point and scale tensor may be 
-an M element vector [v_1, v_2, ..., v_M] for per row quantization and K element vector of shape [v_1, v_2, ..., v_K] 
-for per column quantization. If the input is N-D tensor with shape [D1, D2, M, K] then zero point and scale tensor may 
+It consumes two quantized input tensors, their scales and zero points, scale and zero point of output,
+and computes the quantized output. The quantization formula is y = saturate((x / y_scale) + y_zero_point).
+For (x / y_scale), it is rounding to nearest ties to even. Refer to https://en.wikipedia.org/wiki/Rounding for details.
+Scale and zero point must have same shape. They must be either scalar (per tensor) or N-D tensor
+(per row for 'a' and per column for 'b'). Scalar refers to per tensor quantization whereas N-D refers to per row
+or per column quantization. If the input is 2D of shape [M, K] then zero point and scale tensor may be
+an M element vector [v_1, v_2, ..., v_M] for per row quantization and K element vector of shape [v_1, v_2, ..., v_K]
+for per column quantization. If the input is N-D tensor with shape [D1, D2, M, K] then zero point and scale tensor may
 have shape [D1, D2, M, 1] for per row quantization and shape [D1, D2, 1, K] for per column quantization.
 Production must never overflow, and accumulation may overflow if and only if in 32 bits.
 )DOC";
@@ -2858,16 +2858,9 @@ Generates a {name} window as described in the paper https://ieeexplore.ieee.org/
         true,
         1,
         OpSchema::NonDifferentiable);
-    schema.Output(
-        0,
-        "output",
-        "A Hann window with length: size. "
-        "The output has the shape: [size].",
-        "T2",
-        OpSchema::Single,
-        true,
-        1,
-        OpSchema::NonDifferentiable);
+    std::string output_doc("A {name} window with length: size. The output has the shape: [size].");
+    ReplaceAll(output_doc, "{name}", name);
+    schema.Output(0, "output", output_doc, "T2", OpSchema::Single, true, 1, OpSchema::NonDifferentiable);
     schema.TypeAndShapeInferenceFunction([](ONNX_NAMESPACE::InferenceContext& ctx) {
       // Update the output data type to the output_datatype
       auto output_datatype = getAttribute(ctx, "output_datatype", static_cast<int64_t>(TensorProto_DataType_FLOAT));
@@ -2925,7 +2918,7 @@ ONNX_OPERATOR_SET_SCHEMA(
           Symmetric_Component = Mul(Symmetric_Size_FP, IsSymmetric_FP)
           Size_FP = Add(Periodic_Component, Symmetric_Component)
           AngularIncrement = Div (Tau, Size_FP)
-          Range = Range (Zero, Size_FP, One)
+          Range = Range (Zero, Periodic_Size_FP, One)
           RangeAngular = Mul (Range, AngularIncrement)
           TwoRangeAngular = Mul (RangeAngular, Two)
           CosTwoRangeAngular = Cos (TwoRangeAngular)
@@ -2963,7 +2956,7 @@ ONNX_OPERATOR_SET_SCHEMA(
           Symmetric_Component = Mul(Symmetric_Size_FP, IsSymmetric_FP)
           Size_FP = Add(Periodic_Component, Symmetric_Component)
           AngularIncrement = Div (Tau, Size_FP)
-          Range = Range (Zero, Size_FP, One)
+          Range = Range (Zero, Periodic_Size_FP, One)
           RangeAngular = Mul (Range, AngularIncrement)
           TwoRangeAngular = Mul (RangeAngular, Two)
           CosTwoRangeAngular = Cos (TwoRangeAngular)
@@ -3001,7 +2994,7 @@ ONNX_OPERATOR_SET_SCHEMA(
           Symmetric_Component = Mul(Symmetric_Size_FP, IsSymmetric_FP)
           Size_FP = Add(Periodic_Component, Symmetric_Component)
           AngularIncrement = Div (Tau, Size_FP)
-          Range = Range (Zero, Size_FP, One)
+          Range = Range (Zero, Periodic_Size_FP, One)
           RangeAngular = Mul (Range, AngularIncrement)
           TwoRangeAngular = Mul (RangeAngular, Two)
           CosTwoRangeAngular = Cos (TwoRangeAngular)
@@ -3017,7 +3010,7 @@ ONNX_OPERATOR_SET_SCHEMA(
 static const char* MelWeightMatrix_ver17_doc = R"DOC(
 Generate a MelWeightMatrix that can be used to re-weight a Tensor containing a linearly sampled frequency spectra (from DFT or STFT) into num_mel_bins frequency information based on the [lower_edge_hertz, upper_edge_hertz] range on the mel scale.
 This function defines the mel scale in terms of a frequency in hertz according to the following formula:
-    
+
     mel(f) = 2595 * log10(1 + f/700)
 
 In the returned matrix, all the triangles (filterbanks) have a peak value of 1.0.

diff --git a/onnx/defs/nn/defs.cc b/onnx/defs/nn/defs.cc
@@ -2353,7 +2353,7 @@ ONNX_OPERATOR_SET_SCHEMA(
               .Add("Axis1D = Constant()", "value", mktensor(axis)) // [axis] : 1D tensor
               .Add("PrefixShape = Slice (XShape, Zero1D, Axis1D)") // [d[0], ..., d[axis-1]]
               .Add(
-                  axis > 0 // number of axes that are reduced =
+                  axis >= 0 // number of axes that are reduced =
                       ? "NumReducedAxes = Sub (Rank, Axis1D)" // [rank - axis]: 1D tensor
                       : "NumReducedAxes = Neg (Axis1D)") // [-axis] : 1D tensor
               .Add(