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!
Vous ne pouvez pas sélectionner plus de 25 sujets Les noms de sujets doivent commencer par une lettre ou un nombre, peuvent contenir des tirets ('-') et peuvent comporter jusqu'à 35 caractères.

89 lignes
2.8KB

  1. from icosagon.fastconv import _sparse_diag_cat, \
  2. _cat, \
  3. FastGraphConv
  4. from icosagon.data import _equal
  5. import torch
  6. import pdb
  7. import time
  8. def test_sparse_diag_cat_01():
  9. matrices = [ torch.rand(5, 10).round() for _ in range(7) ]
  10. ground_truth = torch.zeros(35, 70)
  11. ground_truth[0:5, 0:10] = matrices[0]
  12. ground_truth[5:10, 10:20] = matrices[1]
  13. ground_truth[10:15, 20:30] = matrices[2]
  14. ground_truth[15:20, 30:40] = matrices[3]
  15. ground_truth[20:25, 40:50] = matrices[4]
  16. ground_truth[25:30, 50:60] = matrices[5]
  17. ground_truth[30:35, 60:70] = matrices[6]
  18. res = _sparse_diag_cat([ m.to_sparse() for m in matrices ])
  19. res = res.to_dense()
  20. assert torch.all(res == ground_truth)
  21. def test_sparse_diag_cat_02():
  22. x = [ torch.rand(5, 10).round() for _ in range(7) ]
  23. a = [ m.to_sparse() for m in x ]
  24. a = _sparse_diag_cat(a)
  25. b = torch.rand(70, 64)
  26. res = torch.sparse.mm(a, b)
  27. ground_truth = torch.zeros(35, 64)
  28. ground_truth[0:5, :] = torch.mm(x[0], b[0:10])
  29. ground_truth[5:10, :] = torch.mm(x[1], b[10:20])
  30. ground_truth[10:15, :] = torch.mm(x[2], b[20:30])
  31. ground_truth[15:20, :] = torch.mm(x[3], b[30:40])
  32. ground_truth[20:25, :] = torch.mm(x[4], b[40:50])
  33. ground_truth[25:30, :] = torch.mm(x[5], b[50:60])
  34. ground_truth[30:35, :] = torch.mm(x[6], b[60:70])
  35. assert torch.all(res == ground_truth)
  36. def test_cat_01():
  37. matrices = [ torch.rand(5, 10) for _ in range(7) ]
  38. res = _cat(matrices)
  39. assert res.shape == (35, 10)
  40. assert not res.is_sparse
  41. ground_truth = torch.zeros(35, 10)
  42. for i in range(7):
  43. ground_truth[i*5:(i+1)*5, :] = matrices[i]
  44. assert torch.all(res == ground_truth)
  45. def test_cat_02():
  46. matrices = [ torch.rand(5, 10) for _ in range(7) ]
  47. ground_truth = torch.zeros(35, 10)
  48. for i in range(7):
  49. ground_truth[i*5:(i+1)*5, :] = matrices[i]
  50. res = _cat([ m.to_sparse() for m in matrices ])
  51. assert res.shape == (35, 10)
  52. assert res.is_sparse
  53. assert torch.all(res.to_dense() == ground_truth)
  54. def test_fast_graph_conv_01():
  55. # pdb.set_trace()
  56. adj_mats = [ torch.rand(10, 15).round().to_sparse() \
  57. for _ in range(23) ]
  58. fgc = FastGraphConv(32, 64, adj_mats)
  59. in_repr = torch.rand(15, 32)
  60. _ = fgc(in_repr)
  61. def test_fast_graph_conv_02():
  62. t = time.time()
  63. m = (torch.rand(2000, 2000) < .001).to(torch.float32).to_sparse()
  64. adj_mats = [ m for _ in range(1300) ]
  65. print('Generating adj_mats took:', time.time() - t)
  66. t = time.time()
  67. fgc = FastGraphConv(32, 64, adj_mats)
  68. print('FGC constructor took:', time.time() - t)
  69. in_repr = torch.rand(2000, 32)
  70. for _ in range(3):
  71. t = time.time()
  72. _ = fgc(in_repr)
  73. print('FGC forward pass took:', time.time() - t)