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!
Du kannst nicht mehr als 25 Themen auswählen Themen müssen entweder mit einem Buchstaben oder einer Ziffer beginnen. Sie können Bindestriche („-“) enthalten und bis zu 35 Zeichen lang sein.

test_trainprep.py 6.2KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153
  1. #
  2. # Copyright (C) Stanislaw Adaszewski, 2020
  3. # License: GPLv3
  4. #
  5. from icosagon.trainprep import TrainValTest, \
  6. train_val_test_split_edges, \
  7. get_edges_and_degrees, \
  8. prepare_adj_mat, \
  9. prepare_relation_type, \
  10. prep_rel_one_node_type
  11. import torch
  12. import pytest
  13. import numpy as np
  14. from itertools import chain
  15. from icosagon.data import RelationType
  16. def test_train_val_test_split_edges_01():
  17. edges = torch.randint(0, 10, (10, 2))
  18. with pytest.raises(ValueError):
  19. _ = train_val_test_split_edges(edges, TrainValTest(.5, .5, .5))
  20. with pytest.raises(ValueError):
  21. _ = train_val_test_split_edges(edges, TrainValTest(.2, .2, .2))
  22. with pytest.raises(ValueError):
  23. _ = train_val_test_split_edges(edges, None)
  24. with pytest.raises(ValueError):
  25. _ = train_val_test_split_edges(edges, (.8, .1, .1))
  26. with pytest.raises(ValueError):
  27. _ = train_val_test_split_edges(np.random.randint(0, 10, (10, 2)), TrainValTest(.8, .1, .1))
  28. with pytest.raises(ValueError):
  29. _ = train_val_test_split_edges(torch.randint(0, 10, (10, 3)), TrainValTest(.8, .1, .1))
  30. with pytest.raises(ValueError):
  31. _ = train_val_test_split_edges(torch.randint(0, 10, (10, 2, 1)), TrainValTest(.8, .1, .1))
  32. with pytest.raises(ValueError):
  33. _ = train_val_test_split_edges(None, TrainValTest(.8, .2, .2))
  34. res = train_val_test_split_edges(edges, TrainValTest(.8, .1, .1))
  35. assert res.train.shape == (8, 2) and res.val.shape == (1, 2) and \
  36. res.test.shape == (1, 2)
  37. res = train_val_test_split_edges(edges, TrainValTest(.8, .0, .2))
  38. assert res.train.shape == (8, 2) and res.val.shape == (0, 2) and \
  39. res.test.shape == (2, 2)
  40. res = train_val_test_split_edges(edges, TrainValTest(.8, .2, .0))
  41. assert res.train.shape == (8, 2) and res.val.shape == (2, 2) and \
  42. res.test.shape == (0, 2)
  43. res = train_val_test_split_edges(edges, TrainValTest(.0, .5, .5))
  44. assert res.train.shape == (0, 2) and res.val.shape == (5, 2) and \
  45. res.test.shape == (5, 2)
  46. res = train_val_test_split_edges(edges, TrainValTest(.0, .0, 1.))
  47. assert res.train.shape == (0, 2) and res.val.shape == (0, 2) and \
  48. res.test.shape == (10, 2)
  49. res = train_val_test_split_edges(edges, TrainValTest(.0, 1., .0))
  50. assert res.train.shape == (0, 2) and res.val.shape == (10, 2) and \
  51. res.test.shape == (0, 2)
  52. def test_train_val_test_split_edges_02():
  53. edges = torch.randint(0, 30, (30, 2))
  54. ratios = TrainValTest(.8, .1, .1)
  55. res = train_val_test_split_edges(edges, ratios)
  56. edges = [ tuple(a) for a in edges ]
  57. res = [ tuple(a) for a in chain(res.train, res.val, res.test) ]
  58. assert all([ a in edges for a in res ])
  59. def test_get_edges_and_degrees_01():
  60. adj_mat_dense = (torch.rand((10, 10)) > .5)
  61. adj_mat_sparse = adj_mat_dense.to_sparse()
  62. edges_dense, degrees_dense = get_edges_and_degrees(adj_mat_dense)
  63. edges_sparse, degrees_sparse = get_edges_and_degrees(adj_mat_sparse)
  64. assert torch.all(degrees_dense == degrees_sparse)
  65. edges_dense = [ tuple(a) for a in edges_dense ]
  66. edges_sparse = [ tuple(a) for a in edges_dense ]
  67. assert len(edges_dense) == len(edges_sparse)
  68. assert all([ a in edges_dense for a in edges_sparse ])
  69. assert all([ a in edges_sparse for a in edges_dense ])
  70. # assert torch.all(edges_dense == edges_sparse)
  71. def test_prepare_adj_mat_01():
  72. adj_mat = (torch.rand((10, 10)) > .5)
  73. adj_mat = adj_mat.to_sparse()
  74. ratios = TrainValTest(.8, .1, .1)
  75. _ = prepare_adj_mat(adj_mat, ratios)
  76. def test_prepare_adj_mat_02():
  77. adj_mat = (torch.rand((10, 10)) > .5)
  78. adj_mat = adj_mat.to_sparse()
  79. ratios = TrainValTest(.8, .1, .1)
  80. (adj_mat_train, edges_pos, edges_neg) = prepare_adj_mat(adj_mat, ratios)
  81. assert isinstance(adj_mat_train, torch.Tensor)
  82. assert adj_mat_train.is_sparse
  83. assert adj_mat_train.shape == adj_mat.shape
  84. assert adj_mat_train.dtype == adj_mat.dtype
  85. assert isinstance(edges_pos, TrainValTest)
  86. assert isinstance(edges_neg, TrainValTest)
  87. for a in ['train', 'val', 'test']:
  88. for b in [edges_pos, edges_neg]:
  89. edges = getattr(b, a)
  90. assert isinstance(edges, torch.Tensor)
  91. assert len(edges.shape) == 2
  92. assert edges.shape[1] == 2
  93. def test_prepare_relation_type_01():
  94. adj_mat = (torch.rand((10, 10)) > .5)
  95. r = RelationType('Test', 0, 0, adj_mat, True)
  96. ratios = TrainValTest(.8, .1, .1)
  97. _ = prepare_relation_type(r, ratios, False)
  98. def test_prep_rel_one_node_type_01():
  99. adj_mat = torch.zeros(100)
  100. perm = torch.randperm(100)
  101. adj_mat[perm[:10]] = 1
  102. adj_mat = adj_mat.view(10, 10)
  103. rel = RelationType('Dummy Relation', 0, 0, adj_mat, None)
  104. ratios = TrainValTest(.8, .1, .1)
  105. prep_rel = prep_rel_one_node_type(rel, ratios)
  106. assert prep_rel.name == rel.name
  107. assert prep_rel.node_type_row == rel.node_type_row
  108. assert prep_rel.node_type_column == rel.node_type_column
  109. assert prep_rel.adjacency_matrix.shape == rel.adjacency_matrix.shape
  110. assert prep_rel.adjacency_matrix_backward is None
  111. assert len(prep_rel.edges_pos.train) == 8
  112. assert len(prep_rel.edges_pos.val) == 1
  113. assert len(prep_rel.edges_pos.test) == 1
  114. assert len(prep_rel.edges_neg.train) == 8
  115. assert len(prep_rel.edges_neg.val) == 1
  116. assert len(prep_rel.edges_neg.test) == 1
  117. assert len(prep_rel.edges_back_pos.train) == 0
  118. assert len(prep_rel.edges_back_pos.val) == 0
  119. assert len(prep_rel.edges_back_pos.test) == 0
  120. assert len(prep_rel.edges_back_neg.train) == 0
  121. assert len(prep_rel.edges_back_neg.val) == 0
  122. assert len(prep_rel.edges_back_neg.test) == 0
  123. # def prepare_relation(r, ratios):
  124. # adj_mat = r.adjacency_matrix
  125. # adj_mat_train, edges_pos, edges_neg = prepare_adj_mat(adj_mat)
  126. #
  127. # if r.node_type_row == r.node_type_column:
  128. # adj_mat_train = norm_adj_mat_one_node_type(adj_mat_train)
  129. # else:
  130. # adj_mat_train = norm_adj_mat_two_node_types(adj_mat_train)
  131. #
  132. # return PreparedRelation(r.name, r.node_type_row, r.node_type_column,
  133. # adj_mat_train, edges_pos, edges_neg)