- 28 Feb, 2018 5 commits
-
-
Robert Kimball authored
* add support for both command line arg and env var to set NGRAPH_WARINGS_AS_ERRORS'
-
Robert Kimball authored
* make json lib an external project * add env var to turn warnings to errors
-
L.S. Cook authored
* Improve the readability of imported doxygenclass api doc style from being so small (WIP). Also fixed the not op to not import abs doxygenclass. * Make sure no large files get added to git * Other doxygen files to exclude from git * WIP phase two styling code to save and test * Cleanup for PR
-
Fenglei authored
Fix bugs and enable dot in GPU
-
fenglei.tian authored
-
- 27 Feb, 2018 14 commits
-
-
fenglei.tian authored
-
Fenglei authored
Add nvrtc support for CUDA
-
fenglei.tian authored
Merge branch 'tfl/gpu_add_nvrtc' of github.com:NervanaSystems/private-ngraph-cpp into tfl/gpu_add_nvrtc
-
fenglei.tian authored
-
Scott Cyphers authored
-
sharathns93 authored
* add TF serialzed graphs
-
fenglei.tian authored
-
fenglei.tian authored
-
Fenglei authored
-
Scott Cyphers authored
* Replace using aliases with actual classes
-
Jayaram Bobba authored
Elementwise Add mkldnn support in cpu emitter
-
pthoreho authored
-
pthoreho authored
-
pthoreho authored
- support layout assignment pass to mkldnn add, to insert reorder if the input1 and input0 format are different
-
- 26 Feb, 2018 21 commits
-
-
Fenglei authored
-
fenglei.tian authored
-
fenglei.tian authored
-
fenglei.tian authored
-
Fenglei Tian authored
Merge branch 'tfl/gpu_add_nvrtc' of github.com:NervanaSystems/private-ngraph-cpp into tfl/gpu_add_nvrtc
-
Fenglei Tian authored
-
Jayaram Bobba authored
-
pthoreho authored
-
pthoreho authored
- select mkldnn add only if the input_tensor size is >= 1
-
Yixing Lao authored
* initial support for hybrid transformer * add broadcast_vector_rowwise_reversed for hybrid test * headerc * get function placement fix * conv ref test generator graph node in labmda fuction * rename map_parameter_to_source_node * type change map_parameter_to_source_node * use interpreter for numerical derivative * better comments
-
Jayaram Bobba authored
MKLDNN optimized layouts: convolution bprop, avg_pool fprop and bprop
-
pthoreho authored
-
pthoreho authored
- modified mkldnn add kernel to use mkldnn_utils to build primitive and stream execution
-
fenglei.tian authored
-
Fenglei Tian authored
-
pthoreho authored
-
Fenglei Tian authored
-
Jayaram Bobba authored
-
Robert Kimball authored
Add simple method to add default values for missing deserialized fields. Provided to ease backwards compatibility. (#547)
-
fenglei.tian authored
-
Robert Kimball authored
* working * cleanup
-