-
Notifications
You must be signed in to change notification settings - Fork 90
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[SYCLomatic] Enable migration of CUBLASLT_EPILOGUE_GELU & GELU_AUX #2436
Conversation
cublasLtEpilogue_t e; | ||
e = CUBLASLT_EPILOGUE_DEFAULT; | ||
e = CUBLASLT_EPILOGUE_RELU; | ||
e = CUBLASLT_EPILOGUE_GELU; | ||
e = CUBLASLT_EPILOGUE_GELU_AUX; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
GELU_AUX is not as same as GELU.
According to CUBLAS doc, GELU_AUX is
Apply GELU point-wise transform to the results (x := GELU(x)). This epilogue mode outputs GELU input as a separate matrix (useful for training). See CUBLASLT_MATMUL_DESC_EPILOGUE_AUX_POINTER of [cublasLtMatmulDescAttributes_t](https://docs.nvidia.com/cuda/cublas/index.html#cublasltmatmuldescattributes-t).
So you need update the implementation to copy the GELU input to EPILOGUE_AUX_POINTER
cublasLtEpilogue_t e; | ||
e = CUBLASLT_EPILOGUE_DEFAULT; | ||
e = CUBLASLT_EPILOGUE_RELU; | ||
e = CUBLASLT_EPILOGUE_GELU; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please also add E2E tests for those 2 epilogues in: https://github.com/oneapi-src/SYCLomatic-test/blob/SYCLomatic/features/feature_case/cublasLt/matmul.cu
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
OK
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please update the PR.
Co-authored-by: Jiang, Zhiwei <[email protected]>
Refer to #2460 . Closing this. |
cc @zhiweij1 @zhimingwang36