Commits · 915afc26b978d37d7f42cd972b3b3f32f58dc49f · Max Ehrlich / cdcnn

26 Dec, 2018 1 commit
- Final relu training results · 915afc26
  Max Ehrlich authored Dec 26, 2018
  
  915afc26
19 Dec, 2018 7 commits
- Fix typo · 6bee207f
  Max Ehrlich authored Dec 19, 2018
  
  6bee207f
- Final throughput results · 061a87c8
  Max Ehrlich authored Dec 19, 2018
  
  061a87c8
- Clean up experiments section · ceb14e53
  Max Ehrlich authored Dec 19, 2018
  
  ceb14e53
- Tweaks to tensor notation · 88877e12
  Max Ehrlich authored Dec 19, 2018
  
  88877e12
- Add plot for relu training · dfd67e2c
  Max Ehrlich authored Dec 19, 2018
  
  dfd67e2c
- Fix sizing and x-axis tics for relu · 1e027406
  Max Ehrlich authored Dec 19, 2018
  
  1e027406
- Add training accuracies · ea730558
  Max Ehrlich authored Dec 19, 2018
  
  ea730558
18 Dec, 2018 2 commits
- Initial throughput results and plot · 7e3ea638
  Max Ehrlich authored Dec 18, 2018
  
  7e3ea638
- Fix throughput experiments · f75864f2
  Max Ehrlich authored Dec 18, 2018
  
  f75864f2
17 Dec, 2018 2 commits
- Clean up plots · 35351a27
  Max Ehrlich authored Dec 17, 2018
  
  35351a27
- Shorten paper and move things to supplement · af6ca03d
  Max Ehrlich authored Dec 17, 2018
  
  af6ca03d
14 Dec, 2018 2 commits
- Updates to spatial throughput test · 8d890409
  Max Ehrlich authored Dec 14, 2018
  
  8d890409
- Add throughput tests · 81bf349d
  Max Ehrlich authored Dec 14, 2018
  
  81bf349d
13 Dec, 2018 2 commits
- Add relu training experiment · d4208bd3
  Max Ehrlich authored Dec 13, 2018
  
  d4208bd3
- Correct the batch norm forward training algorithm · 5b29b727
  Max Ehrlich authored Dec 13, 2018
  
  5b29b727
11 Dec, 2018 1 commit
- Use cloned weights so that a converted model can be trained separately · 40cef28c
  Max Ehrlich authored Dec 11, 2018
  
  40cef28c
07 Dec, 2018 7 commits
- Factor out more common plot settings · 925df178
  Max Ehrlich authored Dec 07, 2018
  
  925df178
- Add cifar 100 results · b20a6dd2
  Max Ehrlich authored Dec 07, 2018
  
  b20a6dd2
- Move styles to common settings · 0a2be12b
  Max Ehrlich authored Dec 07, 2018
  
  0a2be12b
- Factor out common settings · c3243659
  Max Ehrlich authored Dec 07, 2018
  
  c3243659
- Add plot for relu accuracy · 8b03dc99
  Max Ehrlich authored Dec 07, 2018
  
  8b03dc99
- Add relu accuracy for MNIST and CIFAR10 · 09f7b5b0
  Max Ehrlich authored Dec 07, 2018
  
  09f7b5b0
- Simplify relu blocks plot script · bd20f96b
  Max Ehrlich authored Dec 07, 2018
  
  bd20f96b
06 Dec, 2018 9 commits
- Use apxrelu from the layers package · f3bf1bec
  Max Ehrlich authored Dec 06, 2018
  
  f3bf1bec
- Add relu accuracy experiment · e249d6e7
  Max Ehrlich authored Dec 06, 2018
  
  e249d6e7
- Add relu layer using direct approximation · df707d83
  Max Ehrlich authored Dec 06, 2018
  
  df707d83
- Slight change to relu blocks output data format · 46fb2b06
  Max Ehrlich authored Dec 06, 2018
  
  46fb2b06
- Formatting changes for the relu blocks plot · 71631865
  Max Ehrlich authored Dec 06, 2018
  
  71631865
- Corrected relu blocks results · de6a8362
  Max Ehrlich authored Dec 06, 2018
  
  de6a8362
- Minor fixes to relu blocks test · 4851be26
  Max Ehrlich authored Dec 06, 2018
  
  4851be26
- Preliminary plot for relu_blocks error · 30d1c887
  Max Ehrlich authored Dec 06, 2018
  
  30d1c887
- Add relu blocks result · 409b93b9
  Max Ehrlich authored Dec 06, 2018
  
  409b93b9
05 Dec, 2018 7 commits
- Simple training experiment · 65a729a5
  Max Ehrlich authored Dec 05, 2018
  
  65a729a5
- Batch norm training mode first attempt · cc5094a1
  Max Ehrlich authored Dec 05, 2018
  
  cc5094a1
- Remove double import · 79824871
  Max Ehrlich authored Dec 05, 2018
  
  79824871
- Make sure conv weight is registered as a param · b36dfadd
  Max Ehrlich authored Dec 05, 2018
  
  b36dfadd
- Dont retain the graph, it wastes memory · 81d5c18c
  Max Ehrlich authored Dec 05, 2018
  
  81d5c18c
- Fix shape broadcasting bug · a99367a8
  Max Ehrlich authored Dec 05, 2018
  
  a99367a8
- Remove scipy · 90171997
  Max Ehrlich authored Dec 05, 2018
  
  90171997