IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an email to s dot adaszewski at gmail dot com. User accounts are meant only to report issues and/or generate pull requests. This is a purpose-specific Git hosting for ADARED projects. Thank you for your understanding!
Você não pode selecionar mais de 25 tópicos Os tópicos devem começar com uma letra ou um número, podem incluir traços ('-') e podem ter até 35 caracteres.

test_declayer.py 3.1KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889
  1. #
  2. # Copyright (C) Stanislaw Adaszewski, 2020
  3. # License: GPLv3
  4. #
  5. from icosagon.input import OneHotInputLayer
  6. from icosagon.convlayer import DecagonLayer
  7. from icosagon.declayer import DecodeLayer
  8. from icosagon.decode import DEDICOMDecoder
  9. from icosagon.data import Data
  10. import torch
  11. def test_decode_layer_01():
  12. d = Data()
  13. d.add_node_type('Dummy', 100)
  14. d.add_relation_type('Dummy Relation 1', 0, 0,
  15. torch.rand((100, 100), dtype=torch.float32).round().to_sparse())
  16. in_layer = OneHotInputLayer(d)
  17. d_layer = DecagonLayer(in_layer.output_dim, 32, d)
  18. seq = torch.nn.Sequential(in_layer, d_layer)
  19. last_layer_repr = seq(None)
  20. dec = DecodeLayer(input_dim=d_layer.output_dim, data=d, keep_prob=1.,
  21. decoder_class=DEDICOMDecoder, activation=lambda x: x)
  22. pred_adj_matrices = dec(last_layer_repr)
  23. assert isinstance(pred_adj_matrices, dict)
  24. assert len(pred_adj_matrices) == 1
  25. assert isinstance(pred_adj_matrices[0, 0], list)
  26. assert len(pred_adj_matrices[0, 0]) == 1
  27. def test_decode_layer_02():
  28. d = Data()
  29. d.add_node_type('Dummy', 100)
  30. d.add_relation_type('Dummy Relation 1', 0, 0,
  31. torch.rand((100, 100), dtype=torch.float32).round().to_sparse())
  32. in_layer = OneHotInputLayer(d)
  33. d_layer = DecagonLayer(in_layer.output_dim, 32, d)
  34. dec_layer = DecodeLayer(input_dim=d_layer.output_dim, data=d, keep_prob=1.,
  35. decoder_class=DEDICOMDecoder, activation=lambda x: x)
  36. seq = torch.nn.Sequential(in_layer, d_layer, dec_layer)
  37. pred_adj_matrices = seq(None)
  38. assert isinstance(pred_adj_matrices, dict)
  39. assert len(pred_adj_matrices) == 1
  40. assert isinstance(pred_adj_matrices[0, 0], list)
  41. assert len(pred_adj_matrices[0, 0]) == 1
  42. def test_decode_layer_03():
  43. d = Data()
  44. d.add_node_type('Dummy 1', 100)
  45. d.add_node_type('Dummy 2', 100)
  46. d.add_relation_type('Dummy Relation 1', 0, 1,
  47. torch.rand((100, 100), dtype=torch.float32).round().to_sparse())
  48. in_layer = OneHotInputLayer(d)
  49. d_layer = DecagonLayer(in_layer.output_dim, 32, d)
  50. dec_layer = DecodeLayer(input_dim=d_layer.output_dim, data=d, keep_prob=1.,
  51. decoder_class={(0, 1): DEDICOMDecoder}, activation=lambda x: x)
  52. seq = torch.nn.Sequential(in_layer, d_layer, dec_layer)
  53. pred_adj_matrices = seq(None)
  54. assert isinstance(pred_adj_matrices, dict)
  55. assert len(pred_adj_matrices) == 2
  56. assert isinstance(pred_adj_matrices[0, 1], list)
  57. assert isinstance(pred_adj_matrices[1, 0], list)
  58. assert len(pred_adj_matrices[0, 1]) == 1
  59. assert len(pred_adj_matrices[1, 0]) == 1
  60. def test_decode_layer_04():
  61. d = Data()
  62. d.add_node_type('Dummy', 100)
  63. assert len(d.relation_types[0, 0]) == 0
  64. in_layer = OneHotInputLayer(d)
  65. d_layer = DecagonLayer(in_layer.output_dim, 32, d)
  66. dec_layer = DecodeLayer(input_dim=d_layer.output_dim, data=d, keep_prob=1.,
  67. decoder_class=DEDICOMDecoder, activation=lambda x: x)
  68. seq = torch.nn.Sequential(in_layer, d_layer, dec_layer)
  69. pred_adj_matrices = seq(None)
  70. assert isinstance(pred_adj_matrices, dict)
  71. assert len(pred_adj_matrices) == 0