1. c8922fe Python string formatting (#21136) by hankaj · 8 days ago master
  2. bd6405b Add quantized batch norm operator fused with ReLU (#21137) by hankaj · 9 days ago
  3. f803641 [DOC] Add custom strategy script to quantization with INC example (#21134) by Andrzej Kotłowski · 2 weeks ago
  4. 3a19f0e [FEATURE] Dnnl sum primitive path (#21132) by Kacper Pietkun · 3 weeks ago
  5. 8d933fd Add proper link to scripts in quantization with INC example (#21133) by Andrzej Kotłowski · 4 weeks ago
  6. 2d72ce4 [DOC] Add tutotrial about improving accuracy of quantization with oneDNN (#21127) by Andrzej Kotłowski · 4 weeks ago
  7. 1a418e4 [FEATURE] Add query_keys transformer version without split (#21115) by AdamGrabowski · 5 weeks ago
  8. daac02c Fix fused resnet low accuracy (#21122) by hankaj · 5 weeks ago
  9. 7748ae7 docs: Fix a few typos (#21094) by Tim Gates · 6 weeks ago
  10. 1058369 [BUGFIX] _npi_repeats with swap (#21112) by Kacper Pietkun · 6 weeks ago
  11. 736313f Add support for bool data type for condition in where operator (#21103) by bgawrych · 6 weeks ago
  12. 6d1fbe3 Add size threshold for few oneDNN operators (#21106) by bgawrych · 7 weeks ago
  13. 9975ab4 [BUGFIX] Reenable fwd conv engine 5 on test_group_conv2d_16c (#21104) by Dick Carter · 7 weeks ago
  14. 97e25cf [submodule] Upgrade oneDNN to v2.6.1 (#21108) by bartekkuncer · 7 weeks ago
  15. 0b4ecdb [BUGFIX] Fix threadsafety and shutdown issues with threaded_engine_perdevice (#21110) by Dick Carter · 8 weeks ago
  16. dedb8c9 [WIP] [BUGFIX] Fix flakey TemporaryDirectory() cleanup on Windows (#21107) by Dick Carter · 8 weeks ago
  17. db39bb1 Fix multivariate normal bug (#21105) by hankaj · 8 weeks ago
  18. 5e5e0e3 [BUGFIX] Fix SupportDNNL for multiple inputs (#21102) by AdamGrabowski · 9 weeks ago
  19. ecb5026 Comments formatting fix (#21101) by DominikaJedynak · 9 weeks ago
  20. 183e012 Get rid of warnings (#21099) by hankaj · 10 weeks ago
  21. 7b1daf9 Requantize scale fix (#21100) by DominikaJedynak · 10 weeks ago
  22. cca8f4e Reduce overhead in sg_onednn_fully_connected for floats (#21092) by bgawrych · 10 weeks ago
  23. ef0415d [BUGFIX] Fix floor divide (#21096) by Kacper Pietkun · 10 weeks ago
  24. cf15e0a [BUGFIX] Fix remove Cast fuse (#21086) by bartekkuncer · 2 months ago
  25. f6d1ed1 Improve bf16 support (#21002) by Paweł Głomski · 2 months ago
  26. ded6096 [FEATURE] Add pytest with benchmarking operator (#21088) by AdamGrabowski · 2 months ago
  27. e522bea [BUGFIX] Fix Gluon2.0 guide (#21090) by bartekkuncer · 2 months ago
  28. e36c9f0 Refactor fc_sum_fuse (#21077) by bartekkuncer · 3 months ago
  29. 26243ee Fix broadcast ops descriptions (#21087) by bartekkuncer · 3 months ago
  30. 84b1626 oneDNN FullyConnected weight caching & refactor (#21047) by bgawrych · 3 months ago
  31. b713dc5 [BUGFIX] Fix DNNL requantize operator overflow error (#21079) by Andrzej Kotłowski · 3 months ago
  32. 5abdc77 [FEATURE] Add _npi_power_scalar and _npi_multiply_scalar fuse (#20976) by bartekkuncer · 3 months ago
  33. 9745d36 Improve masked_softmax performance with temperature parameter (#21082) by AdamGrabowski · 3 months ago
  34. d8872c8 Fix building of master website due to removed blog page. (#21083) by Joe Evans · 3 months ago
  35. cdffaf0 [FEATURE] Add tanh approximation for GeLU activation (#21034) by bgawrych · 3 months ago
  36. c486a0e [master] Node elimination graph pass (#21046) by PiotrWolinski - Intel · 3 months ago
  37. afbef15 Type fix for FullyConnected with sum (#21043) by DominikaJedynak · 3 months ago
  38. b4aca83 Use requested mem in dot op to reduce memory usage (#21067) by bartekkuncer · 3 months ago
  39. ef2be51 Refactor SupportDNNL functions (#21032) by AdamGrabowski · 3 months ago
  40. 63aea9e [FEATURE] Add quantization for npi_add with oneDNN (#21041) by Andrzej Kotłowski · 3 months ago
  41. 1ad198d [FEATURE] Refactor SwapAxis operator. (#21024) by AdamGrabowski · 3 months ago
  42. 1dba769 Diversify default RNG seed (#21058) by bartekkuncer · 3 months ago
  43. 08f578b Fix test_bf16_binary_broadcast_elemwise_mixed_input (#20986) by AdamGrabowski · 3 months ago
  44. b322bee [FEATURE] Add property removing duplicate Cast operations (#21020) by bartekkuncer · 3 months ago
  45. 36f4f58 Fix pip package description (#21064) by bartekkuncer · 3 months ago
  46. 6b2d4d4 Merge #21055 from v1.9.x - website changes. (#21068) by Joe Evans · 3 months ago
  47. 9491442 Update oneDNN quantization tutorial (#21060) by Andrzej Kotłowski · 3 months ago
  48. 3263aee fix linkcheck (#21057) by bgawrych · 3 months ago
  49. 4e161d0 update readme for mx1.9.1 (#21050) by waytrue17 · 4 months ago
  50. 15cacdb broadcast_like CPU optimization (#21004) by bgawrych · 4 months ago
  51. 102388a [master] Remove dnnl_ops-inl.h file (#20997) by PiotrWolinski - Intel · 4 months ago
  52. 9152b90 Fix linkcheck (#21045) by AdamGrabowski · 4 months ago
  53. ab30c56 [master] Fix issue with fc_eltwise fusing (#20958) by PiotrWolinski - Intel · 4 months ago
  54. 1687b44 [website] Update website for mxnet 1.9.1 (#21038) by waytrue17 · 4 months ago
  55. 9ca3b27 [master] Enabled tests using the whole batch for calibration (#21008) by DominikaJedynak · 4 months ago
  56. 0497efd fix transformer optimization for gpt-2 (#21007) by bgawrych · 4 months ago
  57. a9574d2 Fix quantized_elemwise_add (#21031) by Andrzej Kotłowski · 4 months ago
  58. 3098af4 AMP improvements + enable bf16 input for quantize_v2 (#20983) by Paweł Głomski · 4 months ago
  59. 0208063 Port website fixes (#20993, #20995, #21009) from v1.9.x branch (#21015) by Joe Evans · 4 months ago
  60. eb8982d Fix CI in master (#21026) by Joe Evans · 4 months ago
  61. 024e791 Fix bitwise_right_shift with uint input (#20974) by AdamGrabowski · 5 months ago
  62. bb91978 [master] Unified oneDNN implementation calls for quantized operators (#20987) by PiotrWolinski - Intel · 5 months ago
  63. e43c1e0 [BUGFIX] Fix and cleanup _contrib_quantized_elemwise_add (#20999) by Andrzej Kotłowski · 5 months ago
  64. c70ea50 [FEATURE] Integrate oneDNN support for scalar power() operator (#20904) by AdamGrabowski · 5 months ago
  65. 90d9054 [submodule] Upgrade oneDNN to v2.6 (#20982) by bartekkuncer · 5 months ago
  66. dadb5b7 Fix CI on master branch (#21001) by bartekkuncer · 5 months ago
  67. ea6c91a Restore quantized RNN to master (#20952) by bgawrych · 6 months ago
  68. 08737b2 [Master] Decouple OneDNN data structures in MXNet C++ API (#20624) by mozga · 6 months ago
  69. f16c3c7 Post quantize property improvement (#20929) by DominikaJedynak · 6 months ago
  70. e8ff13c Improve AMP, bf16 support. Support oneDNN ops in AMP (#20753) by Paweł Głomski · 6 months ago
  71. 4672ca1 [FEATURE] Add oneDNN support for npi: exp, square and sqrt (#20981) by bartekkuncer · 6 months ago
  72. 52d8d48 fix onednn install header location (#20893) by RafLit · 6 months ago
  73. 6b9213c Support bfloat and float mixed input in elemwise broadcast ops (#20932) by AdamGrabowski · 6 months ago
  74. db730cc [submodule] Upgrade oneDNN to v2.5.4 (#20980) by bartekkuncer · 6 months ago
  75. e165b35 Enable possibility of disabling particular graph pass (#20961) by Andrzej Kotłowski · 6 months ago
  76. b03d717 [BUGFIX] Fix tests/python/dnnl/subgraphs/test_conv_subgraph.py (#20970) by bartekkuncer · 6 months ago
  77. ef378e4 [operator] Integrate oneDNN matmul primitive to mxnet dot operator (#20911) by bartekkuncer · 6 months ago
  78. 54d153c [operator] Integrate oneDNN eltwise_tanh into _npi_tanh operator (#20972) by bartekkuncer · 6 months ago
  79. 11423bc Fix subgraph tests (#20969) by Andrzej Kotłowski · 6 months ago
  80. 83c875d Rename SetMKLMem func to SetDNNLMem to avoid confusion (#20948) by bartekkuncer · 6 months ago
  81. 51e5204 Add support for up 12 dims for oneDNN tensors in MXNet (#20913) by bartekkuncer · 6 months ago
  82. 5ab7f64 [2.0] [BACKPORT] of [1.x][FEATURE] CUDA graphs support (#19142) (#20324) by Dick Carter · 6 months ago
  83. 95a6a54 [v2.0] [BUGFIX] Port #20940 (Upgrade numpy to <1.20.0 ...) from v1.9.x (#20957) by Dick Carter · 6 months ago
  84. e972556 [BUGFIX] Fix mxnet.numpy.eye() handling of extreme k parameter values (#20965) by Dick Carter · 6 months ago
  85. 8f199c2 Fix broken web link (#20966) by Dick Carter · 6 months ago
  86. ded0400 [submodule] Upgrade oneDNN to v2.5.3 (#20946) by bartekkuncer · 6 months ago
  87. a03dc1e port #20926 to master (#20949) by RafLit · 6 months ago
  88. 927d68b [PERFORMANCE] Improve shuffle implementation (#20928) by Andrzej Kotłowski · 6 months ago
  89. fc54fab [FEATURE] Add g5 instance to CI (#20876) by Dick Carter · 7 months ago 2.0.0.beta1 2.0.0.beta1.rc1
  90. 13b8690 Avoid modifying loaded library map while iterating in lib_close() (#20941) by Joe Evans · 7 months ago
  91. 7a49008 [BUGFIX] Type fix for large tensors (#20922) by DominikaJedynak · 7 months ago
  92. 588f541 [master] 2022.00 MKL' version, update (#20865) by mozga · 7 months ago
  93. 54e122c quantized transpose operator (#20817) by RafLit · 7 months ago
  94. 6c2b3dc Remove first_quantization_pass FC property (#20908) by Andrzej Kotłowski · 7 months ago
  95. 6e25c88 Add oneDNN support for "where" operator (#20862) by bgawrych · 7 months ago
  96. 93aa9e3 add Bartłomiej as committer (#20896) by Sheng Zha · 7 months ago
  97. 8069a18 [master] Fix issue with even number of channels in BatchNorm (#20907) by PiotrWolinski - Intel · 7 months ago
  98. 188d7b6 [master] Implemented oneDNN Backward Adaptive Pooling kernel (#20825) by PiotrWolinski - Intel · 7 months ago
  99. 9f47730 fix python docs ci (#20903) by RafLit · 7 months ago zero_sharding
  100. f6266f0 Reduce after quantization memory usage (#20894) by bgawrych · 7 months ago