zzy34407230
/
mindspore2022
Not watched
1
0
Fork 0
Code
Releases 22 Wiki evaluate Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain HPC
You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
Tree: 8d00a8d803
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '8d00a8d803'
${ noResults }
mindspore2022/tests/st/ops/gpu/test_scatter_func_op.py

721 lines
27 kB
Raw Blame History 

  1. # Copyright 2020-2021 Huawei Technologies Co., Ltd

  2. #

  3. # Licensed under the Apache License, Version 2.0 (the "License");

  4. # you may not use this file except in compliance with the License.

  5. # You may obtain a copy of the License at

  6. #

  7. # http://www.apache.org/licenses/LICENSE-2.0

  8. #

  9. # Unless required by applicable law or agreed to in writing, software

  10. # distributed under the License is distributed on an "AS IS" BASIS,

  11. # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

  12. # See the License for the specific language governing permissions and

  13. # limitations under the License.

  14. # ============================================================================

  15. 

  16. import numpy as np

  17. import pytest

  18. import mindspore.context as context

  19. import mindspore.nn as nn

  20. from mindspore import Tensor, Parameter

  21. from mindspore.ops import operations as P

  22. from mindspore.ops.operations import _inner_ops as inner

  23. 

  24. context.set_context(mode=context.GRAPH_MODE, device_target="GPU")

  25. # all cases tested against dchip

  26. 

  27. func_map = {

  28.     "update": P.ScatterUpdate,

  29.     "add": P.ScatterAdd,

  30.     "sub": P.ScatterSub,

  31. }

  32. 

  33. 

  34. class TestScatterFuncNet(nn.Cell):

  35.     def __init__(self, func, lock, inputx, indices, updates):

  36.         super(TestScatterFuncNet, self).__init__()

  37. 

  38.         self.scatter_func = func_map[func](use_locking=lock)

  39.         self.inputx = Parameter(inputx, name="inputx")

  40.         self.indices = Parameter(indices, name="indices")

  41.         self.updates = Parameter(updates, name="updates")

  42. 

  43.     def construct(self):

  44.         out = self.scatter_func(self.inputx, self.indices, self.updates)

  45.         return out

  46. 

  47. 

  48. def scatter_func_net(func, inputx, indices, updates):

  49.     lock = True

  50.     net = TestScatterFuncNet(func, lock, inputx, indices, updates)

  51.     return net()

  52. 

  53. 

  54. def scatter_func_use_locking_false_net(func, inputx, indices, updates):

  55.     lock = False

  56.     net = TestScatterFuncNet(func, lock, inputx, indices, updates)

  57.     return net()

  58. 

  59. 

  60. class TestScatterFuncDynamicNet(nn.Cell):

  61.     def __init__(self, func, inputx, indices, updates):

  62.         super(TestScatterFuncDynamicNet, self).__init__()

  63.         self.scatter_func = func_map[func]()

  64.         self.test_dynamic = inner.GpuConvertToDynamicShape()

  65.         self.inputx = Parameter(inputx, name="inputx")

  66.         self.indices = Parameter(indices, name="indices")

  67.         self.updates = Parameter(updates, name="updates")

  68. 

  69.     def construct(self):

  70.         indices = self.test_dynamic(self.indices)

  71.         updates = self.test_dynamic(self.updates)

  72.         out = self.scatter_func(self.inputx, indices, updates)

  73.         return out

  74. 

  75. 

  76. def scatter_func_d_net(func, inputx, indices, updates):

  77.     context.set_context(mode=context.GRAPH_MODE, device_target="GPU")

  78.     net = TestScatterFuncDynamicNet(func, inputx, indices, updates)

  79.     return net()

  80. 

  81. 

  82. class TestScatterFuncDynamicNet2(nn.Cell):

  83.     def __init__(self, func, inputx):

  84.         super(TestScatterFuncDynamicNet2, self).__init__()

  85.         self.scatter_func = func_map[func]()

  86.         self.test_dynamic = inner.GpuConvertToDynamicShape()

  87.         self.inputx = Parameter(inputx, name="inputx")

  88. 

  89.     def construct(self, indices, updates):

  90.         indices = self.test_dynamic(indices)

  91.         updates = self.test_dynamic(updates)

  92.         out = self.scatter_func(self.inputx, indices, updates)

  93.         return out

  94. 

  95. 

  96. def scatter_func_d2_net(func, inputx, indices_1, updates_1, indices_2, updates_2):

  97.     context.set_context(mode=context.GRAPH_MODE, device_target="GPU")

  98.     net = TestScatterFuncDynamicNet2(func, inputx)

  99.     out1 = net(indices_1, updates_1)

  100.     out2 = net(indices_2, updates_2)

  101.     return (out1, out2)

  102. 

  103. 

  104. @pytest.mark.level0

  105. @pytest.mark.platform_x86_gpu_training

  106. @pytest.mark.env_onecard

  107. def test_scatter_func_small_float32():

  108.     inputx = Tensor(np.zeros((2, 3)).astype(np.float32))

  109.     indices = Tensor(np.array([[0, 1], [0, 1]]).astype(np.int32))

  110.     updates = Tensor(np.arange(12).reshape((2, 2, 3)).astype(np.float32))

  111. 

  112.     # update

  113.     output = scatter_func_net("update", inputx, indices, updates)

  114.     expected = np.array([[0.0, 1.0, 2.0], [3.0, 4.0, 5.0]])

  115.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  116. 

  117.     # add

  118.     output = scatter_func_net("add", inputx, indices, updates)

  119.     expected = np.array([[6.0, 8.0, 10.0], [12.0, 14.0, 16.0]])

  120.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  121. 

  122.     # sub

  123.     output = scatter_func_net("sub", inputx, indices, updates)

  124.     expected = np.array([[-6.0, -8.0, -10.0], [-12.0, -14.0, -16.0]])

  125.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  126. 

  127. 

  128. @pytest.mark.level0

  129. @pytest.mark.platform_x86_gpu_training

  130. @pytest.mark.env_onecard

  131. def test_scatter_func_input_updated():

  132.     inputx = Tensor(np.zeros((2, 3)).astype(np.float32))

  133.     indices = Tensor(np.array([[0, 1], [0, 1]]).astype(np.int32))

  134.     updates = Tensor(np.arange(12).reshape((2, 2, 3)).astype(np.float32))

  135.     lock = True

  136. 

  137.     # update

  138.     net = TestScatterFuncNet("update", lock, inputx, indices, updates)

  139.     net()

  140.     expected = np.array([[0.0, 1.0, 2.0], [3.0, 4.0, 5.0]])

  141.     np.testing.assert_array_almost_equal(net.inputx.asnumpy(), expected)

  142. 

  143.     # add

  144.     net = TestScatterFuncNet("add", lock, inputx, indices, updates)

  145.     net()

  146.     expected = np.array([[6.0, 8.0, 10.0], [12.0, 14.0, 16.0]])

  147.     np.testing.assert_array_almost_equal(net.inputx.asnumpy(), expected)

  148. 

  149.     # sub

  150.     net = TestScatterFuncNet("sub", lock, inputx, indices, updates)

  151.     net()

  152.     expected = np.array([[-6.0, -8.0, -10.0], [-12.0, -14.0, -16.0]])

  153.     np.testing.assert_array_almost_equal(net.inputx.asnumpy(), expected)

  154. 

  155. 

  156. @pytest.mark.level0

  157. @pytest.mark.platform_x86_gpu_training

  158. @pytest.mark.env_onecard

  159. def test_scatter_func_large_shape_float32():

  160.     inputx = Tensor(np.ones((4, 2, 3, 4)).astype(np.float32))

  161.     indices = Tensor(np.array([[0, 2], [3, 1]]).astype(np.int32))

  162.     updates = Tensor(np.arange(96).reshape((2, 2, 2, 3, 4)).astype(np.float32))

  163. 

  164.     # update

  165.     output = scatter_func_net("update", inputx, indices, updates)

  166.     expected = np.array(

  167.         [

  168.             [

  169.                 [[0.0, 1.0, 2.0, 3.0], [4.0, 5.0, 6.0, 7.0], [8.0, 9.0, 10.0, 11.0]],

  170.                 [[12.0, 13.0, 14.0, 15.0], [16.0, 17.0, 18.0, 19.0], [20.0, 21.0, 22.0, 23.0]],

  171.             ],

  172.             [

  173.                 [[72.0, 73.0, 74.0, 75.0], [76.0, 77.0, 78.0, 79.0], [80.0, 81.0, 82.0, 83.0]],

  174.                 [[84.0, 85.0, 86.0, 87.0], [88.0, 89.0, 90.0, 91.0], [92.0, 93.0, 94.0, 95.0]],

  175.             ],

  176.             [

  177.                 [[24.0, 25.0, 26.0, 27.0], [28.0, 29.0, 30.0, 31.0], [32.0, 33.0, 34.0, 35.0]],

  178.                 [[36.0, 37.0, 38.0, 39.0], [40.0, 41.0, 42.0, 43.0], [44.0, 45.0, 46.0, 47.0]],

  179.             ],

  180.             [

  181.                 [[48.0, 49.0, 50.0, 51.0], [52.0, 53.0, 54.0, 55.0], [56.0, 57.0, 58.0, 59.0]],

  182.                 [[60.0, 61.0, 62.0, 63.0], [64.0, 65.0, 66.0, 67.0], [68.0, 69.0, 70.0, 71.0]],

  183.             ],

  184.         ]

  185.     )

  186.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  187. 

  188.     # add

  189.     output = scatter_func_net("add", inputx, indices, updates)

  190.     expected = np.array(

  191.         [

  192.             [

  193.                 [[1.0, 2.0, 3.0, 4.0], [5.0, 6.0, 7.0, 8.0], [9.0, 10.0, 11.0, 12.0]],

  194.                 [[13.0, 14.0, 15.0, 16.0], [17.0, 18.0, 19.0, 20.0], [21.0, 22.0, 23.0, 24.0]],

  195.             ],

  196.             [

  197.                 [[73.0, 74.0, 75.0, 76.0], [77.0, 78.0, 79.0, 80.0], [81.0, 82.0, 83.0, 84.0]],

  198.                 [[85.0, 86.0, 87.0, 88.0], [89.0, 90.0, 91.0, 92.0], [93.0, 94.0, 95.0, 96.0]],

  199.             ],

  200.             [

  201.                 [[25.0, 26.0, 27.0, 28.0], [29.0, 30.0, 31.0, 32.0], [33.0, 34.0, 35.0, 36.0]],

  202.                 [[37.0, 38.0, 39.0, 40.0], [41.0, 42.0, 43.0, 44.0], [45.0, 46.0, 47.0, 48.0]],

  203.             ],

  204.             [

  205.                 [[49.0, 50.0, 51.0, 52.0], [53.0, 54.0, 55.0, 56.0], [57.0, 58.0, 59.0, 60.0]],

  206.                 [[61.0, 62.0, 63.0, 64.0], [65.0, 66.0, 67.0, 68.0], [69.0, 70.0, 71.0, 72.0]],

  207.             ],

  208.         ]

  209.     )

  210.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  211. 

  212.     # sub

  213.     output = scatter_func_net("sub", inputx, indices, updates)

  214.     expected = np.array(

  215.         [

  216.             [

  217.                 [[1.0, 0.0, -1.0, -2.0], [-3.0, -4.0, -5.0, -6.0], [-7.0, -8.0, -9.0, -10.0]],

  218.                 [

  219.                     [-11.0, -12.0, -13.0, -14.0],

  220.                     [-15.0, -16.0, -17.0, -18.0],

  221.                     [-19.0, -20.0, -21.0, -22.0],

  222.                 ],

  223.             ],

  224.             [

  225.                 [

  226.                     [-71.0, -72.0, -73.0, -74.0],

  227.                     [-75.0, -76.0, -77.0, -78.0],

  228.                     [-79.0, -80.0, -81.0, -82.0],

  229.                 ],

  230.                 [

  231.                     [-83.0, -84.0, -85.0, -86.0],

  232.                     [-87.0, -88.0, -89.0, -90.0],

  233.                     [-91.0, -92.0, -93.0, -94.0],

  234.                 ],

  235.             ],

  236.             [

  237.                 [

  238.                     [-23.0, -24.0, -25.0, -26.0],

  239.                     [-27.0, -28.0, -29.0, -30.0],

  240.                     [-31.0, -32.0, -33.0, -34.0],

  241.                 ],

  242.                 [

  243.                     [-35.0, -36.0, -37.0, -38.0],

  244.                     [-39.0, -40.0, -41.0, -42.0],

  245.                     [-43.0, -44.0, -45.0, -46.0],

  246.                 ],

  247.             ],

  248.             [

  249.                 [

  250.                     [-47.0, -48.0, -49.0, -50.0],

  251.                     [-51.0, -52.0, -53.0, -54.0],

  252.                     [-55.0, -56.0, -57.0, -58.0],

  253.                 ],

  254.                 [

  255.                     [-59.0, -60.0, -61.0, -62.0],

  256.                     [-63.0, -64.0, -65.0, -66.0],

  257.                     [-67.0, -68.0, -69.0, -70.0],

  258.                 ],

  259.             ],

  260.         ]

  261.     )

  262.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  263. 

  264. 

  265. @pytest.mark.level0

  266. @pytest.mark.platform_x86_gpu_training

  267. @pytest.mark.env_onecard

  268. def test_scatter_func_small_float32_use_locking_false():

  269.     inputx = Tensor(np.zeros((2, 3)).astype(np.float32))

  270.     indices = Tensor(np.array([1, 0]).astype(np.int32))

  271.     updates = Tensor(np.arange(6).reshape((2, 3)).astype(np.float32))

  272. 

  273.     # update

  274.     output = scatter_func_use_locking_false_net("update", inputx, indices, updates)

  275.     expected = np.array([[3.0, 4.0, 5.0], [0.0, 1.0, 2.0]])

  276.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  277. 

  278.     # add

  279.     output = scatter_func_use_locking_false_net("add", inputx, indices, updates)

  280.     expected = np.array([[3.0, 4.0, 5.0], [0.0, 1.0, 2.0]])

  281.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  282. 

  283.     # sub

  284.     output = scatter_func_use_locking_false_net("sub", inputx, indices, updates)

  285.     expected = np.array([[-3.0, -4.0, -5.0], [0.0, -1.0, -2.0]])

  286.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  287. 

  288. 

  289. @pytest.mark.level0

  290. @pytest.mark.platform_x86_gpu_training

  291. @pytest.mark.env_onecard

  292. def test_scatter_func_input_less_than_1_float32():

  293.     inputx = Tensor(

  294.         np.array(

  295.             [

  296.                 [0.214141, 0.415151, 0.51516],

  297.                 [0.876542, 0.451611, 0.55112],

  298.                 [0.111244, 0.633333, 0.34444],

  299.             ]

  300.         ).astype(np.float32)

  301.     )

  302.     indices = Tensor(np.array([[[1, 0, 2], [2, 2, 0]], [[1, 0, 1], [2, 1, 2]]]).astype(np.int32))

  303.     updates = Tensor(np.arange(34, 70).reshape((2, 2, 3, 3)).astype(np.float32))

  304. 

  305.     # update

  306.     output = scatter_func_net("update", inputx, indices, updates)

  307.     expected = np.array(

  308.         [[37.0, 38.0, 39.0], [34.0, 35.0, 66.0], [67.0, 68.0, 69.0],], dtype=np.float32,

  309.     )

  310.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  311. 

  312.     # add

  313.     output = scatter_func_net("add", inputx, indices, updates)

  314.     expected = np.array(

  315.         [

  316.             [141.21414, 144.41515, 147.51517],

  317.             [208.87654, 212.45161, 216.55112],

  318.             [257.11124, 262.63333, 267.34442],

  319.         ],

  320.         dtype=np.float32,

  321.     )

  322.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  323. 

  324.     # sub

  325.     output = scatter_func_net("sub", inputx, indices, updates)

  326.     expected = np.array(

  327.         [

  328.             [-140.78586, -143.58485, -146.48483],

  329.             [-207.12346, -211.54839, -215.44888],

  330.             [-256.88876, -261.36667, -266.65558],

  331.         ],

  332.         dtype=np.float32,

  333.     )

  334.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  335. 

  336. 

  337. @pytest.mark.level0

  338. @pytest.mark.platform_x86_gpu_training

  339. @pytest.mark.env_onecard

  340. def test_scatter_func_float16():

  341.     inputx = Tensor(np.zeros((2, 3)).astype(np.float16))

  342.     indices = Tensor(np.array([[0, 1], [0, 1]]).astype(np.int32))

  343.     updates = Tensor(np.arange(12).reshape((2, 2, 3)).astype(np.float16))

  344. 

  345.     # update

  346.     output = scatter_func_net("update", inputx, indices, updates)

  347.     expected = np.array([[0.0, 1.0, 2.0], [3.0, 4.0, 5.0]])

  348.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  349. 

  350.     # add

  351.     output = scatter_func_net("add", inputx, indices, updates)

  352.     expected = np.array([[6.0, 8.0, 10.0], [12.0, 14.0, 16.0]])

  353.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  354. 

  355.     # sub

  356.     output = scatter_func_net("sub", inputx, indices, updates)

  357.     expected = np.array([[-6.0, -8.0, -10.0], [-12.0, -14.0, -16.0]])

  358.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  359. 

  360. 

  361. @pytest.mark.level0

  362. @pytest.mark.platform_x86_gpu_training

  363. @pytest.mark.env_onecard

  364. def test_scatter_func_large_float16():

  365.     inputx = Tensor(np.zeros((2, 3, 4)).astype(np.float16))

  366.     indices = Tensor(np.array([[0, 0], [1, 1]]).astype(np.int32))

  367.     updates = Tensor(np.arange(63, 111).reshape((2, 2, 3, 4)).astype(np.float16))

  368. 

  369.     # update

  370.     output = scatter_func_net("update", inputx, indices, updates)

  371.     expected = np.array(

  372.         [

  373.             [[63.0, 64.0, 65.0, 66.0], [67.0, 68.0, 69.0, 70.0], [71.0, 72.0, 73.0, 74.0],],

  374.             [[99.0, 100.0, 101.0, 102.0], [103.0, 104.0, 105.0, 106.0], [95.0, 96.0, 97.0, 98.0],],

  375.         ]

  376.     )

  377.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  378. 

  379.     # add

  380.     output = scatter_func_net("add", inputx, indices, updates)

  381.     expected = np.array(

  382.         [

  383.             [

  384.                 [138.0, 140.0, 142.0, 144.0],

  385.                 [146.0, 148.0, 150.0, 152.0],

  386.                 [154.0, 156.0, 158.0, 160.0],

  387.             ],

  388.             [

  389.                 [186.0, 188.0, 190.0, 192.0],

  390.                 [194.0, 196.0, 198.0, 200.0],

  391.                 [202.0, 204.0, 206.0, 208.0],

  392.             ],

  393.         ]

  394.     )

  395.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  396. 

  397.     # sub

  398.     output = scatter_func_net("sub", inputx, indices, updates)

  399.     expected = np.array(

  400.         [

  401.             [

  402.                 [-138.0, -140.0, -142.0, -144.0],

  403.                 [-146.0, -148.0, -150.0, -152.0],

  404.                 [-154.0, -156.0, -158.0, -160.0],

  405.             ],

  406.             [

  407.                 [-186.0, -188.0, -190.0, -192.0],

  408.                 [-194.0, -196.0, -198.0, -200.0],

  409.                 [-202.0, -204.0, -206.0, -208.0],

  410.             ],

  411.         ]

  412.     )

  413.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  414. 

  415. 

  416. @pytest.mark.level0

  417. @pytest.mark.platform_x86_gpu_training

  418. @pytest.mark.env_onecard

  419. def test_scatter_func_disordered_float16():

  420.     inputx = Tensor(np.flip(np.arange(34, 46).reshape(3, 4).astype(np.float16)))

  421.     indices = Tensor(np.array([[[0, 1, 2], [2, 1, 0]], [[0, 0, 0], [2, 2, 2]]]).astype(np.int32))

  422.     updates = Tensor(np.arange(63, 111).reshape((2, 2, 3, 4)).astype(np.float16))

  423. 

  424.     # update

  425.     output = scatter_func_net("update", inputx, indices, updates)

  426.     expected = np.array(

  427.         [[95.0, 96.0, 97.0, 98.0], [67.0, 68.0, 69.0, 70.0], [99.0, 100.0, 101.0, 102.0]]

  428.     )

  429.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  430. 

  431.     # add

  432.     output = scatter_func_net("add", inputx, indices, updates)

  433.     expected = np.array(

  434.         [[464.0, 468.0, 472.0, 476.0], [187.0, 188.0, 189.0, 190.0], [492.0, 496.0, 500.0, 504.0]]

  435.     )

  436.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  437. 

  438.     # sub

  439.     output = scatter_func_net("sub", inputx, indices, updates)

  440.     expected = np.array(

  441.         [

  442.             [-374.0, -380.0, -386.0, -392.0],

  443.             [-105.0, -108.0, -111.0, -114.0],

  444.             [-418.0, -424.0, -430.0, -436.0],

  445.         ]

  446.     )

  447.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  448. 

  449. 

  450. @pytest.mark.level0

  451. @pytest.mark.platform_x86_gpu_training

  452. @pytest.mark.env_onecard

  453. def test_scatter_func_large_int32():

  454.     inputx = Tensor(np.zeros((2, 3, 4)).astype(np.int32))

  455.     indices = Tensor(np.array([[0, 0], [1, 1]]).astype(np.int32))

  456.     updates = Tensor(np.arange(63, 111).reshape((2, 2, 3, 4)).astype(np.int32))

  457. 

  458.     # update

  459.     output = scatter_func_net("update", inputx, indices, updates)

  460.     expected = np.array(

  461.         [

  462.             [[63.0, 64.0, 65.0, 66.0], [67.0, 68.0, 69.0, 70.0], [71.0, 72.0, 73.0, 74.0],],

  463.             [[99.0, 100.0, 101.0, 102.0], [103.0, 104.0, 105.0, 106.0], [95.0, 96.0, 97.0, 98.0],],

  464.         ]

  465.     ).astype(np.int32)

  466.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  467. 

  468.     # add

  469.     output = scatter_func_net("add", inputx, indices, updates)

  470.     expected = np.array(

  471.         [

  472.             [

  473.                 [138.0, 140.0, 142.0, 144.0],

  474.                 [146.0, 148.0, 150.0, 152.0],

  475.                 [154.0, 156.0, 158.0, 160.0],

  476.             ],

  477.             [

  478.                 [186.0, 188.0, 190.0, 192.0],

  479.                 [194.0, 196.0, 198.0, 200.0],

  480.                 [202.0, 204.0, 206.0, 208.0],

  481.             ],

  482.         ]

  483.     ).astype(np.int32)

  484.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  485. 

  486.     # sub

  487.     output = scatter_func_net("sub", inputx, indices, updates)

  488.     expected = np.array(

  489.         [

  490.             [

  491.                 [-138.0, -140.0, -142.0, -144.0],

  492.                 [-146.0, -148.0, -150.0, -152.0],

  493.                 [-154.0, -156.0, -158.0, -160.0],

  494.             ],

  495.             [

  496.                 [-186.0, -188.0, -190.0, -192.0],

  497.                 [-194.0, -196.0, -198.0, -200.0],

  498.                 [-202.0, -204.0, -206.0, -208.0],

  499.             ],

  500.         ]

  501.     ).astype(np.int32)

  502.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  503. 

  504. 

  505. @pytest.mark.level0

  506. @pytest.mark.platform_x86_gpu_training

  507. @pytest.mark.env_onecard

  508. def test_scatter_func_disordered_int32():

  509.     inputx = Tensor(np.flip(np.arange(34, 46).reshape(3, 4).astype(np.int32)))

  510.     indices = Tensor(np.array([[[0, 1, 2], [2, 1, 0]], [[0, 0, 0], [2, 2, 2]]]).astype(np.int32))

  511.     updates = Tensor(np.arange(63, 111).reshape((2, 2, 3, 4)).astype(np.int32))

  512. 

  513.     # update

  514.     output = scatter_func_net("update", inputx, indices, updates)

  515.     expected = np.array(

  516.         [[95.0, 96.0, 97.0, 98.0], [67.0, 68.0, 69.0, 70.0], [99.0, 100.0, 101.0, 102.0]]

  517.     ).astype(np.int32)

  518.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  519. 

  520.     # add

  521.     output = scatter_func_net("add", inputx, indices, updates)

  522.     expected = np.array(

  523.         [[464.0, 468.0, 472.0, 476.0], [187.0, 188.0, 189.0, 190.0], [492.0, 496.0, 500.0, 504.0]]

  524.     ).astype(np.int32)

  525.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  526. 

  527.     # sub

  528.     output = scatter_func_net("sub", inputx, indices, updates)

  529.     expected = np.array(

  530.         [

  531.             [-374.0, -380.0, -386.0, -392.0],

  532.             [-105.0, -108.0, -111.0, -114.0],

  533.             [-418.0, -424.0, -430.0, -436.0],

  534.         ]

  535.     ).astype(np.int32)

  536.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  537. 

  538. 

  539. @pytest.mark.level0

  540. @pytest.mark.platform_x86_gpu_training

  541. @pytest.mark.env_onecard

  542. def test_scatter_func_disordered_dynamic_int32():

  543.     inputx = Tensor(np.flip(np.arange(34, 46).reshape(3, 4).astype(np.int32)))

  544.     indices = Tensor(np.array([[[0, 1, 2], [2, 1, 0]], [[0, 0, 0], [2, 2, 2]]]).astype(np.int32))

  545.     updates = Tensor(np.arange(63, 111).reshape((2, 2, 3, 4)).astype(np.int32))

  546. 

  547.     # update

  548.     output = scatter_func_d_net("update", inputx, indices, updates)

  549.     expected = np.array(

  550.         [[95.0, 96.0, 97.0, 98.0], [67.0, 68.0, 69.0, 70.0], [99.0, 100.0, 101.0, 102.0]]

  551.     ).astype(np.int32)

  552.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  553. 

  554.     # add

  555.     output = scatter_func_d_net("add", inputx, indices, updates)

  556.     expected = np.array(

  557.         [[464.0, 468.0, 472.0, 476.0], [187.0, 188.0, 189.0, 190.0], [492.0, 496.0, 500.0, 504.0]]

  558.     ).astype(np.int32)

  559.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  560. 

  561.     # sub

  562.     output = scatter_func_d_net("sub", inputx, indices, updates)

  563.     expected = np.array(

  564.         [

  565.             [-374.0, -380.0, -386.0, -392.0],

  566.             [-105.0, -108.0, -111.0, -114.0],

  567.             [-418.0, -424.0, -430.0, -436.0],

  568.         ]

  569.     ).astype(np.int32)

  570.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  571. 

  572. 

  573. @pytest.mark.level0

  574. @pytest.mark.platform_x86_gpu_training

  575. @pytest.mark.env_onecard

  576. def test_scatter_func_disordered_dynamic_int8():

  577.     inputx = Tensor(np.flip(np.arange(34, 46).reshape(3, 4).astype(np.int8)))

  578.     indices = Tensor(np.array([[[0, 1, 2], [2, 1, 0]], [[0, 0, 0], [2, 2, 2]]]).astype(np.int32))

  579.     updates = Tensor(np.arange(63, 111).reshape((2, 2, 3, 4)).astype(np.int8))

  580. 

  581.     # update

  582.     output = scatter_func_d_net("update", inputx, indices, updates)

  583.     expected = np.array(

  584.         [[95.0, 96.0, 97.0, 98.0], [67.0, 68.0, 69.0, 70.0], [99.0, 100.0, 101.0, 102.0]]

  585.     ).astype(np.int8)

  586.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  587. 

  588.     # add

  589.     output = scatter_func_d_net("add", inputx, indices, updates)

  590.     expected = np.array(

  591.         [[464.0, 468.0, 472.0, 476.0], [187.0, 188.0, 189.0, 190.0], [492.0, 496.0, 500.0, 504.0]]

  592.     ).astype(np.int8)

  593.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  594. 

  595.     # sub

  596.     output = scatter_func_d_net("sub", inputx, indices, updates)

  597.     expected = np.array(

  598.         [

  599.             [-118.0, -124.0, 126.0, 120.0],

  600.             [-105.0, -108.0, -111.0, -114.0],

  601.             [94.0, 88.0, 82.0, 76.0],

  602.         ]

  603.     ).astype(np.int8)

  604.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  605. 

  606. 

  607. @pytest.mark.level0

  608. @pytest.mark.platform_x86_gpu_training

  609. @pytest.mark.env_onecard

  610. def test_scatter_func_disordered_dynamic_uint8():

  611.     inputx = Tensor(np.flip(np.arange(34, 46).reshape(3, 4).astype(np.uint8)))

  612.     indices = Tensor(np.array([[[0, 1, 2], [2, 1, 0]], [[0, 0, 0], [2, 2, 2]]]).astype(np.int32))

  613.     updates = Tensor(np.arange(63, 111).reshape((2, 2, 3, 4)).astype(np.uint8))

  614. 

  615.     # update

  616.     output = scatter_func_d_net("update", inputx, indices, updates)

  617.     expected = np.array(

  618.         [[95.0, 96.0, 97.0, 98.0], [67.0, 68.0, 69.0, 70.0], [99.0, 100.0, 101.0, 102.0]]

  619.     ).astype(np.uint8)

  620.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  621. 

  622.     # add

  623.     output = scatter_func_d_net("add", inputx, indices, updates)

  624.     expected = np.array(

  625.         [[464.0, 468.0, 472.0, 476.0], [187.0, 188.0, 189.0, 190.0], [492.0, 496.0, 500.0, 504.0]]

  626.     ).astype(np.uint8)

  627.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  628. 

  629.     # sub

  630.     output = scatter_func_d_net("sub", inputx, indices, updates)

  631.     expected = np.array(

  632.         [[138.0, 132.0, 126.0, 120.0], [151.0, 148.0, 145.0, 142.0], [94.0, 88.0, 82.0, 76.0]]

  633.     ).astype(np.uint8)

  634.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  635. 

  636. 

  637. @pytest.mark.level0

  638. @pytest.mark.platform_x86_gpu_training

  639. @pytest.mark.env_onecard

  640. def test_scatter_func_input_less_than_1_dynamic_float32():

  641.     inputx = Tensor(

  642.         np.array(

  643.             [

  644.                 [0.214141, 0.415151, 0.51516],

  645.                 [0.876542, 0.451611, 0.55112],

  646.                 [0.111244, 0.633333, 0.34444],

  647.             ]

  648.         ).astype(np.float32)

  649.     )

  650.     indices = Tensor(np.array([[[1, 0, 2], [2, 2, 0]], [[1, 0, 1], [2, 1, 2]]]).astype(np.int32))

  651.     updates = Tensor(np.arange(34, 70).reshape((2, 2, 3, 3)).astype(np.float32))

  652. 

  653.     # update

  654.     output = scatter_func_d_net("update", inputx, indices, updates)

  655.     expected = np.array(

  656.         [[37.0, 38.0, 39.0], [34.0, 35.0, 66.0], [67.0, 68.0, 69.0],], dtype=np.float32,

  657.     )

  658.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  659. 

  660.     # add

  661.     output = scatter_func_d_net("add", inputx, indices, updates)

  662.     expected = np.array(

  663.         [

  664.             [141.21414, 144.41515, 147.51517],

  665.             [208.87654, 212.45161, 216.55112],

  666.             [257.11124, 262.63333, 267.34442],

  667.         ],

  668.         dtype=np.float32,

  669.     )

  670.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  671. 

  672.     # sub

  673.     output = scatter_func_d_net("sub", inputx, indices, updates)

  674.     expected = np.array(

  675.         [

  676.             [-140.78586, -143.58485, -146.48483],

  677.             [-207.12346, -211.54839, -215.44888],

  678.             [-256.88876, -261.36667, -266.65558],

  679.         ],

  680.         dtype=np.float32,

  681.     )

  682.     np.testing.assert_array_almost_equal(output.asnumpy(), expected)

  683. 

  684. 

  685. @pytest.mark.level0

  686. @pytest.mark.platform_x86_gpu_training

  687. @pytest.mark.env_onecard

  688. def test_scatter_func_dynamic_two_inputs():

  689.     inputx = Tensor(np.zeros((2, 3)).astype(np.float32))

  690.     indices_1 = Tensor(np.array([[0, 1], [0, 1]]).astype(np.int32))

  691.     updates_1 = Tensor(np.arange(12).reshape((2, 2, 3)).astype(np.float32))

  692.     indices_2 = Tensor(np.array([[0, 0], [1, 1], [1, 0]]).astype(np.int32))

  693.     updates_2 = Tensor(np.flip(np.arange(18).reshape((3, 2, 3)).astype(np.float32)))

  694. 

  695.     # update

  696.     output_1, output_2 = scatter_func_d2_net(

  697.         "update", inputx, indices_1, updates_1, indices_2, updates_2

  698.     )

  699.     expected_1 = np.array([[0.0, 1.0, 2.0], [3.0, 4.0, 5.0]])

  700.     expected_2 = np.array([[17.0, 16.0, 15.0], [11.0, 10.0, 9.0]])

  701.     np.testing.assert_array_almost_equal(output_1.asnumpy(), expected_1)

  702.     np.testing.assert_array_almost_equal(output_2.asnumpy(), expected_2)

  703. 

  704.     # add

  705.     output_1, output_2 = scatter_func_d2_net(

  706.         "add", inputx, indices_1, updates_1, indices_2, updates_2

  707.     )

  708.     expected_1 = np.array([[6.0, 8.0, 10.0], [12.0, 14.0, 16.0]])

  709.     expected_2 = np.array([[39.0, 38.0, 37.0], [36.0, 35.0, 34.0]])

  710.     np.testing.assert_array_almost_equal(output_1.asnumpy(), expected_1)

  711.     np.testing.assert_array_almost_equal(output_2.asnumpy(), expected_2)

  712. 

  713.     # sub

  714.     output_1, output_2 = scatter_func_d2_net(

  715.         "sub", inputx, indices_1, updates_1, indices_2, updates_2

  716.     )

  717.     expected_1 = np.array([[-6.0, -8.0, -10.0], [-12.0, -14.0, -16.0]])

  718.     expected_2 = np.array([[-39.0, -38.0, -37.0], [-36.0, -35.0, -34.0]])

  719.     np.testing.assert_array_almost_equal(output_1.asnumpy(), expected_1)

  720.     np.testing.assert_array_almost_equal(output_2.asnumpy(), expected_2)