[Wait for #2579][unittest] Add mixed precision unit test#2589

Closed

DonghakPark wants to merge 14 commits intonnstreamer:mainfrom DonghakPark:mixed_pre_2

+1,699-319

Commits on May 17, 2024

[ Weight ] Add Var32 Tensor in Weight.

jijoongmoon
authored and
DonghakPark
committed
[ Mixed ] Create weight with var32 tensor

jijoongmoon
authored and
DonghakPark
committed
[ Layers ] Update Layers to support FP16

jijoongmoon
authored and
DonghakPark
committed
[ Test ] Mixed Precision Test Case

jijoongmoon
authored and
DonghakPark
committed
[ Optimizer ] Update Optimizer / Adam to support Mixed training

jijoongmoon
authored and
DonghakPark
committed
[ Tensor ] add is_NaN check in Tensor

jijoongmoon
authored and
DonghakPark
committed
[ Context ] Add loss scale in Context & using mse loss

jijoongmoon
authored and
DonghakPark
committed
[ Mixed Precision ] Enable Mixed Precision

jijoongmoon
authored and
DonghakPark
committed
[ Tensor ] Add inifinity check in Tensor

jijoongmoon
authored and
DonghakPark
committed
[ MSE ] Fix for better MSE loss precision

jijoongmoon
authored and
DonghakPark
committed
[Test] Add conv2d test for fp16

jihochu
authored and
DonghakPark
committed
[Layers] Modify layers for data type

jihochu
authored and
DonghakPark
committed
[Mixed] Mixed Precision Layer update
DonghakPark
committed
[unittest] Add mixed precision unit test
DonghakPark
committed