Add Scatter op to ONNX (onnx#1517)

* Adding Scatter op to ONNX

* fixing build onnx#6598 break

* fixing build onnx#6601 break

* Addressing PR Feedback onnx#1

* fixing build onnx#6601 break

* removing renamed/redundant test folders from commit

* updating testcoverage.md

* Addressing PR Feedback onnx#2

* updating testcoverage file

* Update onnx/defs/tensor/defs.cc

Co-Authored-By: pk-g <[email protected]>

* resolving merge conflicts

* resolving merge conflicts onnx#2

* Applying PR Feedback

* applying PR feedback and resolving merge conflicts

Author: pk-g

docs/Changelog.md

@@ -9263,6 +9263,80 @@ This version of the operator has been available since version 9 of the default O
 <dd>All Tensor types</dd>
 </dl>
 
+### <a name="Scatter-9"></a>**Scatter-9**</a>
+
+  Given `data`, `updates` and `indices` input tensors of rank r >= 1, write the values provided by `updates` 
+  into the first input, `data`, along `axis` dimension of `data` (by default outer-most one as axis=0) at corresponding `indices`. 
+  For each entry in `updates`, the target index in `data` is specified by corresponding entry in `indices`
+  for dimension = axis, and index in source for dimension != axis. For instance, in a 2-D tensor case,
+  data[indices[i][j]][j] = updates[i][j] if axis = 0, or data[i][indices[i][j]] = updates[i][j] if axis = 1,
+  where i and j are loop counters from 0 up to the respective size in `updates` - 1.
+  
+  Example 1:
+    data = [
+        [0.0, 0.0, 0.0],
+        [0.0, 0.0, 0.0],
+        [0.0, 0.0, 0.0],
+    ]
+    indices = [
+        [1, 0, 2],
+        [0, 2, 1],
+    ]
+    updates = [
+        [1.0, 1.1, 1.2],
+        [2.0, 2.1, 2.2],
+    ]
+    output = [
+        [2.0, 1.1, 0.0]
+        [1.0, 0.0, 2.2]
+        [0.0, 2.1, 1.2]
+    ]
+  
+  Example 2:
+    data = [[1.0, 2.0, 3.0, 4.0, 5.0]]
+    indices = [[1, 3]]
+    updates = [[1.1, 2.1]]
+    axis = 1
+    output = [[1.0, 1.1, 3.0, 2.1, 5.0]]
+
+#### Version
+
+This version of the operator has been available since version 9 of the default ONNX operator set.
+
+#### Attributes
+
+<dl>
+<dt><tt>axis</tt> : int (default is 0)</dt>
+<dd>Which axis to scatter on. Negative value means counting dimensions from the back. Accepted range in [-r, r-1]</dd>
+</dl>
+
+#### Inputs
+
+<dl>
+<dt><tt>data</tt> : T</dt>
+<dd>Tensor of rank r >= 1.</dd>
+<dt><tt>indices</tt> : Tind</dt>
+<dd>Tensor of int32/int64 indices, of r >= 1 (same rank as input).</dd>
+<dt><tt>updates</tt> : T</dt>
+<dd>Tensor of rank r >=1 (same rank and shape as indices)</dd>
+</dl>
+
+#### Outputs
+
+<dl>
+<dt><tt>output</tt> : T</dt>
+<dd>Tensor of rank r >= 1 (same rank as input).</dd>
+</dl>
+
+#### Type Constraints
+
+<dl>
+<dt><tt>T</tt> : tensor(uint8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(int8), tensor(int16), tensor(int32), tensor(int64), tensor(float16), tensor(float), tensor(double), tensor(string), tensor(bool), tensor(complex64), tensor(complex128)</dt>
+<dd>Input and output types can be of any tensor type.</dd>
+<dt><tt>Tind</tt> : tensor(int32), tensor(int64)</dt>
+<dd>Constrain indices to integer types</dd>
+</dl>
+
 ### <a name="Sign-9"></a>**Sign-9**</a>
 
   Calculate the sign of the given input tensor element-wise.

docs/Operators.md

@@ -96,6 +96,7 @@
   * <a href="#Relu">Relu</a>
   * <a href="#Reshape">Reshape</a>
   * <a href="#Scan">Scan</a>
+  * <a href="#Scatter">Scatter</a>
   * <a href="#Selu">Selu</a>
   * <a href="#Shape">Shape</a>
   * <a href="#Sigmoid">Sigmoid</a>
@@ -9879,6 +9880,132 @@ expect(node, inputs=[initial, x], outputs=[y, z],
 </details>
 
 
+### <a name="Scatter"></a><a name="scatter">**Scatter**</a>
+
+  Given `data`, `updates` and `indices` input tensors of rank r >= 1, write the values provided by `updates` 
+  into the first input, `data`, along `axis` dimension of `data` (by default outer-most one as axis=0) at corresponding `indices`. 
+  For each entry in `updates`, the target index in `data` is specified by corresponding entry in `indices`
+  for dimension = axis, and index in source for dimension != axis. For instance, in a 2-D tensor case,
+  data[indices[i][j]][j] = updates[i][j] if axis = 0, or data[i][indices[i][j]] = updates[i][j] if axis = 1,
+  where i and j are loop counters from 0 up to the respective size in `updates` - 1.
+  
+  Example 1:
+    data = [
+        [0.0, 0.0, 0.0],
+        [0.0, 0.0, 0.0],
+        [0.0, 0.0, 0.0],
+    ]
+    indices = [
+        [1, 0, 2],
+        [0, 2, 1],
+    ]
+    updates = [
+        [1.0, 1.1, 1.2],
+        [2.0, 2.1, 2.2],
+    ]
+    output = [
+        [2.0, 1.1, 0.0]
+        [1.0, 0.0, 2.2]
+        [0.0, 2.1, 1.2]
+    ]
+  
+  Example 2:
+    data = [[1.0, 2.0, 3.0, 4.0, 5.0]]
+    indices = [[1, 3]]
+    updates = [[1.1, 2.1]]
+    axis = 1
+    output = [[1.0, 1.1, 3.0, 2.1, 5.0]]
+
+#### Version
+
+This version of the operator has been available since version 9 of the default ONNX operator set.
+
+#### Attributes
+
+<dl>
+<dt><tt>axis</tt> : int (default is 0)</dt>
+<dd>Which axis to scatter on. Negative value means counting dimensions from the back. Accepted range in [-r, r-1]</dd>
+</dl>
+
+#### Inputs
+
+<dl>
+<dt><tt>data</tt> : T</dt>
+<dd>Tensor of rank r >= 1.</dd>
+<dt><tt>indices</tt> : Tind</dt>
+<dd>Tensor of int32/int64 indices, of r >= 1 (same rank as input).</dd>
+<dt><tt>updates</tt> : T</dt>
+<dd>Tensor of rank r >=1 (same rank and shape as indices)</dd>
+</dl>
+
+#### Outputs
+
+<dl>
+<dt><tt>output</tt> : T</dt>
+<dd>Tensor of rank r >= 1 (same rank as input).</dd>
+</dl>
+
+#### Type Constraints
+
+<dl>
+<dt><tt>T</tt> : tensor(uint8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(int8), tensor(int16), tensor(int32), tensor(int64), tensor(float16), tensor(float), tensor(double), tensor(string), tensor(bool), tensor(complex64), tensor(complex128)</dt>
+<dd>Input and output types can be of any tensor type.</dd>
+<dt><tt>Tind</tt> : tensor(int32), tensor(int64)</dt>
+<dd>Constrain indices to integer types</dd>
+</dl>
+
+
+#### Examples
+
+<details>
+<summary>scatter_with_axis</summary>
+
+```python
+node = onnx.helper.make_node(
+    'Scatter',
+    inputs=['data', 'indices', 'updates'],
+    outputs=['y'],
+    axis=1,
+)
+data = np.array([[1.0, 2.0, 3.0, 4.0, 5.0]], dtype=np.float32)
+indices = np.array([[1, 3]], dtype=np.int64)
+updates = np.array([[1.1, 2.1]], dtype=np.float32)
+
+y = np.array([[1.0, 1.1, 3.0, 2.1, 5.0]], dtype=np.float32)
+
+expect(node, inputs=[data, indices, updates], outputs=[y],
+       name='test_scatter_with_axis')
+```
+
+</details>
+
+
+<details>
+<summary>scatter_without_axis</summary>
+
+```python
+node = onnx.helper.make_node(
+    'Scatter',
+    inputs=['data', 'indices', 'updates'],
+    outputs=['y'],
+)
+data = np.zeros((3, 3), dtype=np.float32)
+indices = np.array([[1, 0, 2], [0, 2, 1]], dtype=np.int64)
+updates = np.array([[1.0, 1.1, 1.2], [2.0, 2.1, 2.2]], dtype=np.float32)
+
+y = np.array([
+    [2.0, 1.1, 0.0],
+    [1.0, 0.0, 2.2],
+    [0.0, 2.1, 1.2]
+], dtype=np.float32)
+
+expect(node, inputs=[data, indices, updates], outputs=[y],
+       name='test_scatter_without_axis')
+```
+
+</details>
+
+
 ### <a name="Selu"></a><a name="selu">**Selu**</a>
 
   Selu takes one input data (Tensor<T>) and produces one output data

docs/TestCoverage.md

@@ -5,7 +5,7 @@
 * [Overall Test Coverage](#overall-test-coverage)
 # Node Test Coverage
 ## Summary
-Node tests have covered 105/112 (93.75%, 5 generators excluded) common operators.
+Node tests have covered 106/113 (93.81%, 5 generators excluded) common operators.
 
 Node tests have covered 2/12 (16.67%, 0 generators excluded) experimental operators.
 
@@ -5184,6 +5184,55 @@ expect(node, inputs=[initial, x], outputs=[y, z],
 </details>
 
 
+### Scatter
+There are 2 test cases, listed as following:
+<details>
+<summary>scatter_with_axis</summary>
+
+```python
+node = onnx.helper.make_node(
+    'Scatter',
+    inputs=['data', 'indices', 'updates'],
+    outputs=['y'],
+    axis=1,
+)
+data = np.array([[1.0, 2.0, 3.0, 4.0, 5.0]], dtype=np.float32)
+indices = np.array([[1, 3]], dtype=np.int64)
+updates = np.array([[1.1, 2.1]], dtype=np.float32)
+
+y = np.array([[1.0, 1.1, 3.0, 2.1, 5.0]], dtype=np.float32)
+
+expect(node, inputs=[data, indices, updates], outputs=[y],
+       name='test_scatter_with_axis')
+```
+
+</details>
+<details>
+<summary>scatter_without_axis</summary>
+
+```python
+node = onnx.helper.make_node(
+    'Scatter',
+    inputs=['data', 'indices', 'updates'],
+    outputs=['y'],
+)
+data = np.zeros((3, 3), dtype=np.float32)
+indices = np.array([[1, 0, 2], [0, 2, 1]], dtype=np.int64)
+updates = np.array([[1.0, 1.1, 1.2], [2.0, 2.1, 2.2]], dtype=np.float32)
+
+y = np.array([
+    [2.0, 1.1, 0.0],
+    [1.0, 0.0, 2.2],
+    [0.0, 2.1, 1.2]
+], dtype=np.float32)
+
+expect(node, inputs=[data, indices, updates], outputs=[y],
+       name='test_scatter_without_axis')
+```
+
+</details>
+
+
 ### Selu
 There are 2 test cases, listed as following:
 <details>

onnx/backend/test/case/node/scatter.py

@@ -0,0 +1,50 @@
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+from __future__ import unicode_literals
+
+import numpy as np  # type: ignore
+
+import onnx
+from ..base import Base
+from . import expect
+
+
+class Scatter(Base):
+
+    @staticmethod
+    def export_scatter_without_axis():  # type: () -> None
+        node = onnx.helper.make_node(
+            'Scatter',
+            inputs=['data', 'indices', 'updates'],
+            outputs=['y'],
+        )
+        data = np.zeros((3, 3), dtype=np.float32)
+        indices = np.array([[1, 0, 2], [0, 2, 1]], dtype=np.int64)
+        updates = np.array([[1.0, 1.1, 1.2], [2.0, 2.1, 2.2]], dtype=np.float32)
+
+        y = np.array([
+            [2.0, 1.1, 0.0],
+            [1.0, 0.0, 2.2],
+            [0.0, 2.1, 1.2]
+        ], dtype=np.float32)
+
+        expect(node, inputs=[data, indices, updates], outputs=[y],
+               name='test_scatter_without_axis')
+
+    @staticmethod
+    def export_scatter_with_axis():  # type: () -> None
+        node = onnx.helper.make_node(
+            'Scatter',
+            inputs=['data', 'indices', 'updates'],
+            outputs=['y'],
+            axis=1,
+        )
+        data = np.array([[1.0, 2.0, 3.0, 4.0, 5.0]], dtype=np.float32)
+        indices = np.array([[1, 3]], dtype=np.int64)
+        updates = np.array([[1.1, 2.1]], dtype=np.float32)
+
+        y = np.array([[1.0, 1.1, 3.0, 2.1, 5.0]], dtype=np.float32)
+
+        expect(node, inputs=[data, indices, updates], outputs=[y],
+               name='test_scatter_with_axis')

onnx/backend/test/data/node/test_scatter_with_axis/model.onnx

@@ -0,0 +1,22 @@
+backend-test:�

+1
+data
+indices
+updates
y"Scatter*
+axis
�
test_scatter_with_axisZ
+data
+

+

+Z
+indices
+
+

+Z
+updates
+

+

+b
+
y
+

+

+B	

onnx/backend/test/data/node/test_scatter_with_axis/test_data_set_0/input_0.pb

@@ -0,0 +1 @@
+

BupdatesJ�̌?ff@

onnx/backend/test/data/node/test_scatter_with_axis/test_data_set_0/input_1.pb

@@ -0,0 +1,21 @@
+backend-test:�

+$
+data
+indices
+updates
y"Scattertest_scatter_without_axisZ
+data
+

+
+Z
+indices
+
+
+Z
+updates
+

+
+b
+
y
+

+
+B