Illia Silin f46a6ffad8 Fix the fp8 gemm for large tensors on MI300. (#1011) ...

* Fix the fp8 conversion

* Try clipping value before conversion

* Fix return

* Simplify with a const

* reduce the gemm input tensor values to reduce round-off error

* replace if-else with lambda

* fix syntax

---------

Co-authored-by: Rostyslav Geyyer <rosty.geyyer@amd.com>

2023-10-27 21:10:47 -07:00

..

CMakeLists.txt

Refactoring cmake files to build data types separately. (#932)

2023-09-20 22:15:56 -07:00

common.hpp

update copyright headers (#726)

2023-05-31 18:46:57 -05:00

groupnorm_sigmoid_mul_fp16.cpp

Layernorm and groupnorm support to save mean and inverse std in forward (#929)

2023-10-19 07:36:29 +08:00

groupnorm_splitk_fp16.cpp

Layernorm and groupnorm support to save mean and inverse std in forward (#929)

2023-10-19 07:36:29 +08:00

groupnorm_swish_fp16.cpp

Layernorm and groupnorm support to save mean and inverse std in forward (#929)

2023-10-19 07:36:29 +08:00

run_groupnorm_example.inc

Fix the fp8 gemm for large tensors on MI300. (#1011)

2023-10-27 21:10:47 -07:00