Add Alpha and Beta parameters in Backward method
Compare changes
Solve the issue !272 for CUDA backend.
The CUDA kernel for the following operators is modified: ILayerNorm, ShiftGELU, ShiftMax.
Add Alpha=1 and Beta=1 parameters to Backward() method so as to ADD a contribution to the upstream gradient tensor.
Copyright © Eclipse Foundation, Inc. All Rights Reserved. Privacy Policy | Terms of Use | Copyright Agent