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.

105 Zeilen
3.6KB

  1. from icosagon.data import Data, \
  2. _equal
  3. from icosagon.model import Model
  4. from icosagon.trainprep import PreparedData, \
  5. PreparedRelationFamily, \
  6. PreparedRelationType, \
  7. TrainValTest, \
  8. norm_adj_mat_one_node_type
  9. import torch
  10. from icosagon.input import OneHotInputLayer
  11. from icosagon.convlayer import DecagonLayer
  12. from icosagon.declayer import DecodeLayer
  13. def _is_identity_function(f):
  14. for x in range(-100, 101):
  15. if f(x) != x:
  16. return False
  17. return True
  18. def test_model_01():
  19. d = Data()
  20. d.add_node_type('Dummy', 10)
  21. fam = d.add_relation_family('Dummy-Dummy', 0, 0, False)
  22. fam.add_relation_type('Dummy Rel', torch.rand(10, 10).round())
  23. m = Model(d)
  24. assert m.data == d
  25. assert m.layer_dimensions == [32, 64]
  26. assert (m.ratios.train, m.ratios.val, m.ratios.test) == (.8, .1, .1)
  27. assert m.keep_prob == 1.
  28. assert _is_identity_function(m.rel_activation)
  29. assert m.layer_activation == torch.nn.functional.relu
  30. assert _is_identity_function(m.dec_activation)
  31. assert m.lr == 0.001
  32. assert m.loss == torch.nn.functional.binary_cross_entropy_with_logits
  33. assert m.batch_size == 100
  34. assert isinstance(m.prep_d, PreparedData)
  35. assert isinstance(m.seq, torch.nn.Sequential)
  36. assert isinstance(m.opt, torch.optim.Optimizer)
  37. def test_model_02():
  38. d = Data()
  39. d.add_node_type('Dummy', 10)
  40. fam = d.add_relation_family('Dummy-Dummy', 0, 0, False)
  41. mat = torch.rand(10, 10).round().to_sparse()
  42. fam.add_relation_type('Dummy Rel', mat)
  43. m = Model(d, ratios=TrainValTest(1., 0., 0.))
  44. assert isinstance(m.prep_d, PreparedData)
  45. assert isinstance(m.prep_d.relation_families, list)
  46. assert len(m.prep_d.relation_families) == 1
  47. assert isinstance(m.prep_d.relation_families[0], PreparedRelationFamily)
  48. assert len(m.prep_d.relation_families[0].relation_types) == 1
  49. assert isinstance(m.prep_d.relation_families[0].relation_types[0], PreparedRelationType)
  50. assert m.prep_d.relation_families[0].relation_types[0].adjacency_matrix_backward is None
  51. assert torch.all(_equal(m.prep_d.relation_families[0].relation_types[0].adjacency_matrix,
  52. norm_adj_mat_one_node_type(mat)))
  53. assert isinstance(m.seq[0], OneHotInputLayer)
  54. assert isinstance(m.seq[1], DecagonLayer)
  55. assert isinstance(m.seq[2], DecagonLayer)
  56. assert isinstance(m.seq[3], DecodeLayer)
  57. assert len(m.seq) == 4
  58. def test_model_03():
  59. d = Data()
  60. d.add_node_type('Dummy', 10)
  61. fam = d.add_relation_family('Dummy-Dummy', 0, 0, False)
  62. mat = torch.rand(10, 10).round().to_sparse()
  63. fam.add_relation_type('Dummy Rel', mat)
  64. m = Model(d, ratios=TrainValTest(1., 0., 0.))
  65. state_dict = m.opt.state_dict()
  66. assert isinstance(state_dict, dict)
  67. # print(state_dict['param_groups'])
  68. # print(list(m.seq.parameters()))
  69. assert len(list(m.seq[0].parameters())) == 1
  70. assert len(list(m.seq[1].parameters())) == 1
  71. assert len(list(m.seq[2].parameters())) == 1
  72. assert len(list(m.seq[3].parameters())) == 2
  73. # print(list(m.seq[1].parameters()))
  74. def test_model_04():
  75. d = Data()
  76. d.add_node_type('Dummy', 10)
  77. fam = d.add_relation_family('Dummy-Dummy', 0, 0, False)
  78. mat = torch.rand(10, 10).round().to_sparse()
  79. fam.add_relation_type('Dummy Rel 1', mat)
  80. fam.add_relation_type('Dummy Rel 2', mat.clone())
  81. m = Model(d, ratios=TrainValTest(1., 0., 0.))
  82. assert len(list(m.seq[0].parameters())) == 1
  83. assert len(list(m.seq[1].parameters())) == 2
  84. assert len(list(m.seq[2].parameters())) == 2
  85. assert len(list(m.seq[3].parameters())) == 3