master
[v2.0][LICENSE] Port #20493 (#20608)
Automatic Layout Management (#20718)
[Master] Clang-formatter: only src/ directory (#20571)
Improve AMP, bf16 support. Support oneDNN ops in AMP (#20753)
Get rid of warnings (#21099)
broadcast_like CPU optimization (#21004)
Add oneDNN support for reduce operators (#20669)
Unify all names used to refer to oneDNN library in logs and docs to oneDNN (#20719)
[operator] Integrate oneDNN matmul primitive to mxnet dot operator (#20911)
Refactor SupportDNNL functions (#21032)
[FEATURE] Add quantization for npi_add with oneDNN (#21041)
[FEATURE] Dnnl sum primitive path (#21132)
Fix broadcast ops descriptions (#21087)
[master][clang-format] Re-format cc. .h. .cu files; cond. (#20704)
[FEATURE] Add _npi_power_scalar and _npi_multiply_scalar fuse (#20976)
[master] Remove dnnl_ops-inl.h file (#20997)
Improve bf16 support (#21002)
[2.0] [BACKPORT] of [1.x][FEATURE] CUDA graphs support (#19142) (#20324)
Optimize 'take' operator for CPU (#20745)
Add size threshold for few oneDNN operators (#21106)
[NumPy] Wrap unravel_index backend implementation instead of fallback (#20730)
[master][style-fix] Clang-format comment style fix (#20744)