Error when running csnn_pc_mnist.py

[pedrombmachado]

Dear Peter and dear Matthew,
I would like to congratulate you both for the high-quality paper that you have published. I am trying to run your code and I am getting the following error:
$python csnn_pc_mnist.py

Optional argument values:

• post_pre : postpre
• conv_features : 50
• random_seed : 42
• conv_size : 16
• weight_dependence : no_weight_dependence
• weight_sharing : no_weight_sharing
• reduced_dataset : False
• conv_stride : 4
• sort_euclidean : False
• do_plot : False
• connectivity : none
• mode : train
• num_examples : 10000
• random_lattice_prob : 0.0
• examples_per_class : 100
• random_inhibition_prob : 0.0
• top_percent : 10
• num_classes : 10
• lattice_structure : 4

Traceback (most recent call last):
File "csnn_pc_mnist.py", line 970, in
elif test_mode:
NameError: name 'test_mode' is not defined

Can you please advise on how to solve it?
BW
Pedro

[pedrombmachado]

I was able to pass the error by defining the test_mode=True after the import statements.

[djsaunde]

Hi Pedro,

This repo contains code extended from that of the paper by Peter Diehl and Matthew Cook. To see the original code, check out this repo.

[pedrombmachado]

Hi @djsaunde ,
Many thanks for the update. BTW there is another bug here:
Classification performance (all) [ 13. 59. 70. 77. 73. 71. 72. 82. 82. 73. 67. 78. 74. 80. 79. 77. 84. 83. 84. 84. 82. 83. 82. 82. 82. 81. 80. 81. 83. 72. 76. 82. 85. 73. 85. 76. 81. 83. 78. 81.
82. 82. 80. 81. 82. 80. 79. 77. 78. 79. 85. 79. 80. 77. 85. 82. 80. 71. 72. 83. 89. 89. 78. 90. 74. 87. 85. 83. 78. 77. 72. 76. 74. 81. 83. 78. 69. 74. 75. 75.
86. 74. 69. 83. 70. 86. 77. 73. 78. 77. 82. 83. 87. 71. 81. 78. 87. 75. 86.]
Average performance: 78.3232323232

Classification performance (top_percent) [ 13. 65. 65. 76. 67. 71. 75. 78. 73. 77. 66. 77. 71. 78. 76. 77. 78. 84. 80. 75. 80. 80. 83. 77. 80. 80. 77. 79. 78. 69. 74. 79. 78. 71. 86. 77. 82. 80. 77. 84.
81. 83. 81. 80. 85. 82. 76. 79. 82. 81. 75. 79. 77. 77. 85. 80. 80. 72. 81. 82. 86. 88. 78. 88. 74. 84. 85. 78. 76. 73. 73. 70. 72. 77. 86. 77. 64. 67. 73. 75.
88. 70. 60. 81. 69. 81. 71. 73. 79. 75. 77. 83. 86. 66. 79. 63. 84. 70. 85.]
Average performance: 76.5151515152

runs done: 9910 of 10000 (time taken for past 10 runs: 7.21255397797)
runs done: 9920 of 10000 (time taken for past 10 runs: 6.59374213219)
runs done: 9930 of 10000 (time taken for past 10 runs: 6.44509911537)
runs done: 9940 of 10000 (time taken for past 10 runs: 5.79141306877)
runs done: 9950 of 10000 (time taken for past 10 runs: 6.67625212669)
runs done: 9960 of 10000 (time taken for past 10 runs: 6.6978521347)
runs done: 9970 of 10000 (time taken for past 10 runs: 5.98725914955)
runs done: 9980 of 10000 (time taken for past 10 runs: 6.86167407036)
runs done: 9990 of 10000 (time taken for past 10 runs: 6.04048490524)

...Saving results
Traceback (most recent call last):
File "csnn_pc_mnist.py", line 1176, in
save_results()
File "csnn_pc_mnist.py", line 849, in save_results
save_connections(weights_dir, connections, input_connections, ending)
TypeError: save_connections() takes exactly 5 arguments (4 given)

I hope it helps!
BW
Pedro

[pedrombmachado]

Hi @djsaunde ,
I created a pull request for correcting the bug.
BW
Pedro