nihui
/
ncnn
Not watched
1
0
Fork 0
Code
Releases 46 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: f1bdc87478
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'f1bdc87478'
${ noResults }
ncnn/tools/pnnx/tests/test_nn_Upsample.py

165 lines
7.1 kB
Raw Blame History 

  1. # Tencent is pleased to support the open source community by making ncnn available.

  2. #

  3. # Copyright (C) 2021 THL A29 Limited, a Tencent company. All rights reserved.

  4. #

  5. # Licensed under the BSD 3-Clause License (the "License"); you may not use this file except

  6. # in compliance with the License. You may obtain a copy of the License at

  7. #

  8. # https://opensource.org/licenses/BSD-3-Clause

  9. #

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

  11. # under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR

  12. # CONDITIONS OF ANY KIND, either express or implied. See the License for the

  13. # specific language governing permissions and limitations under the License.

  14. 

  15. import torch

  16. import torch.nn as nn

  17. import torch.nn.functional as F

  18. from packaging import version

  19. 

  20. class Model(nn.Module):

  21.     def __init__(self):

  22.         super(Model, self).__init__()

  23. 

  24.         self.up_1d_0_0 = nn.Upsample(size=16)

  25.         self.up_1d_0_1 = nn.Upsample(scale_factor=2, mode='nearest')

  26.         self.up_1d_0_2 = nn.Upsample(size=(20), mode='nearest')

  27.         self.up_1d_0_3 = nn.Upsample(scale_factor=(4), mode='nearest')

  28.         self.up_1d_1_0 = nn.Upsample(size=16, mode='linear')

  29.         self.up_1d_1_1 = nn.Upsample(scale_factor=2, mode='linear')

  30.         self.up_1d_1_2 = nn.Upsample(size=(24), mode='linear', align_corners=True)

  31.         if version.parse(torch.__version__) >= version.parse('1.11'):

  32.             self.up_1d_1_3 = nn.Upsample(scale_factor=(3.1), mode='linear', align_corners=True, recompute_scale_factor=True)

  33.         else:

  34.             self.up_1d_1_3 = nn.Upsample(scale_factor=(3.1), mode='linear', align_corners=True)

  35. 

  36.         self.up_2d_0_0 = nn.Upsample(size=16)

  37.         self.up_2d_0_1 = nn.Upsample(scale_factor=2, mode='nearest')

  38.         self.up_2d_0_2 = nn.Upsample(size=(20,20), mode='nearest')

  39.         if version.parse(torch.__version__) >= version.parse('1.11'):

  40.             self.up_2d_0_3 = nn.Upsample(scale_factor=(4,4), mode='nearest', recompute_scale_factor=True)

  41.         else:

  42.             self.up_2d_0_3 = nn.Upsample(scale_factor=(4,4), mode='nearest')

  43.         self.up_2d_0_4 = nn.Upsample(size=(16,24), mode='nearest')

  44.         self.up_2d_0_5 = nn.Upsample(scale_factor=(2,3), mode='nearest')

  45.         self.up_2d_1_0 = nn.Upsample(size=16, mode='bilinear')

  46.         self.up_2d_1_1 = nn.Upsample(scale_factor=2, mode='bilinear')

  47.         self.up_2d_1_2 = nn.Upsample(size=(20,20), mode='bilinear', align_corners=False)

  48.         if version.parse(torch.__version__) >= version.parse('1.11'):

  49.             self.up_2d_1_3 = nn.Upsample(scale_factor=(4,4), mode='bilinear', align_corners=False, recompute_scale_factor=True)

  50.         else:

  51.             self.up_2d_1_3 = nn.Upsample(scale_factor=(4,4), mode='bilinear', align_corners=False)

  52.         self.up_2d_1_4 = nn.Upsample(size=(16,24), mode='bilinear', align_corners=True)

  53.         self.up_2d_1_5 = nn.Upsample(scale_factor=(2,3), mode='bilinear', align_corners=True)

  54.         self.up_2d_2_0 = nn.Upsample(size=16, mode='bicubic')

  55.         self.up_2d_2_1 = nn.Upsample(scale_factor=2, mode='bicubic')

  56.         self.up_2d_2_2 = nn.Upsample(size=(20,20), mode='bicubic', align_corners=False)

  57.         self.up_2d_2_3 = nn.Upsample(scale_factor=(4,4), mode='bicubic', align_corners=False)

  58.         self.up_2d_2_4 = nn.Upsample(size=(16,24), mode='bicubic', align_corners=True)

  59.         if version.parse(torch.__version__) >= version.parse('1.11'):

  60.             self.up_2d_2_5 = nn.Upsample(scale_factor=(2,3.11), mode='bicubic', align_corners=True, recompute_scale_factor=True)

  61.         else:

  62.             self.up_2d_2_5 = nn.Upsample(scale_factor=(2,3.11), mode='bicubic', align_corners=True)

  63. 

  64.         self.up_3d_0_0 = nn.Upsample(size=16)

  65.         self.up_3d_0_1 = nn.Upsample(scale_factor=2, mode='nearest')

  66.         self.up_3d_0_2 = nn.Upsample(size=(20,20,20), mode='nearest')

  67.         if version.parse(torch.__version__) >= version.parse('1.11'):

  68.             self.up_3d_0_3 = nn.Upsample(scale_factor=(4,4,4), mode='nearest', recompute_scale_factor=True)

  69.         else:

  70.             self.up_3d_0_3 = nn.Upsample(scale_factor=(4,4,4), mode='nearest')

  71.         self.up_3d_0_4 = nn.Upsample(size=(16,24,20), mode='nearest')

  72.         self.up_3d_0_5 = nn.Upsample(scale_factor=(2,3,4), mode='nearest')

  73.         self.up_3d_1_0 = nn.Upsample(size=16, mode='trilinear')

  74.         self.up_3d_1_1 = nn.Upsample(scale_factor=2, mode='trilinear')

  75.         self.up_3d_1_2 = nn.Upsample(size=(20,20,20), mode='trilinear', align_corners=False)

  76.         self.up_3d_1_3 = nn.Upsample(scale_factor=(4,4,4), mode='trilinear', align_corners=False)

  77.         self.up_3d_1_4 = nn.Upsample(size=(16,24,20), mode='trilinear', align_corners=True)

  78.         if version.parse(torch.__version__) >= version.parse('1.11'):

  79.             self.up_3d_1_5 = nn.Upsample(scale_factor=(2,3,4.11), mode='trilinear', align_corners=True, recompute_scale_factor=True)

  80.         else:

  81.             self.up_3d_1_5 = nn.Upsample(scale_factor=(2,3,4.11), mode='trilinear', align_corners=True)

  82. 

  83.         self.up_w = nn.Upsample(scale_factor=(1.499,1.499), mode='nearest')

  84. 

  85.     def forward(self, x, y, z, w):

  86.         x = self.up_1d_0_0(x)

  87.         x = self.up_1d_0_1(x)

  88.         x = self.up_1d_0_2(x)

  89.         x = self.up_1d_0_3(x)

  90.         x = self.up_1d_1_0(x)

  91.         x = self.up_1d_1_1(x)

  92.         x = self.up_1d_1_2(x)

  93.         x = self.up_1d_1_3(x)

  94. 

  95.         y = self.up_2d_0_0(y)

  96.         y = self.up_2d_0_1(y)

  97.         y = self.up_2d_0_2(y)

  98.         y = self.up_2d_0_3(y)

  99.         y = self.up_2d_0_4(y)

  100.         y = self.up_2d_0_5(y)

  101.         y = self.up_2d_1_0(y)

  102.         y = self.up_2d_1_1(y)

  103.         y = self.up_2d_1_2(y)

  104.         y = self.up_2d_1_3(y)

  105.         y = self.up_2d_1_4(y)

  106.         y = self.up_2d_1_5(y)

  107.         y = self.up_2d_2_0(y)

  108.         y = self.up_2d_2_1(y)

  109.         y = self.up_2d_2_2(y)

  110.         y = self.up_2d_2_3(y)

  111.         y = self.up_2d_2_4(y)

  112.         y = self.up_2d_2_5(y)

  113. 

  114.         z = self.up_3d_0_0(z)

  115.         z = self.up_3d_0_1(z)

  116.         z = self.up_3d_0_2(z)

  117.         z = self.up_3d_0_3(z)

  118.         z = self.up_3d_0_4(z)

  119.         z = self.up_3d_0_5(z)

  120.         z = self.up_3d_1_0(z)

  121.         z = self.up_3d_1_1(z)

  122.         z = self.up_3d_1_2(z)

  123.         z = self.up_3d_1_3(z)

  124.         z = self.up_3d_1_4(z)

  125.         z = self.up_3d_1_5(z)

  126. 

  127.         w = self.up_w(w)

  128. 

  129.         return x, y, z, w

  130. 

  131. def test():

  132.     net = Model()

  133.     net.eval()

  134. 

  135.     torch.manual_seed(0)

  136.     x = torch.rand(1, 3, 32)

  137.     y = torch.rand(1, 3, 32, 32)

  138.     z = torch.rand(1, 3, 32, 32, 32)

  139.     w = torch.rand(1, 8, 12, 12)

  140. 

  141.     a = net(x, y, z, w)

  142. 

  143.     # export torchscript

  144.     mod = torch.jit.trace(net, (x, y, z, w))

  145.     mod.save("test_nn_Upsample.pt")

  146. 

  147.     # torchscript to pnnx

  148.     import os

  149.     os.system("../src/pnnx test_nn_Upsample.pt inputshape=[1,3,32],[1,3,32,32],[1,3,32,32,32],[1,8,12,12]")

  150. 

  151.     # pnnx inference

  152.     import test_nn_Upsample_pnnx

  153.     b = test_nn_Upsample_pnnx.test_inference()

  154. 

  155.     for a0, b0 in zip(a, b):

  156.         if not torch.equal(a0, b0):

  157.             return False

  158.     return True

  159. 

  160. if __name__ == "__main__":

  161.     if test():

  162.         exit(0)

  163.     else:

  164.         exit(1)