composable_kernel

mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-14 18:17:44 +00:00

Files

AMD-dteng 12103d0f17 enable bias feature that add bias before adding residual (for rtpllm project) (#1741 )

* 1. enable bias feature that add bias before adding residual; 2. change block size from 128->64 when m<64 in fp16

* delete comment

* 1.remove fmha change 2.change buffer name from bias to xbias

* Now bias can be used independently from fadd

* change kbias to kxbias

---------

Co-authored-by: feli <felix.li@amd.com>

[ROCm/composable_kernel commit: d5c8a334ca]

2025-01-08 17:51:06 +08:00

01_fmha

[CK_TILE] fmha fwd splitkv optimization for decode (seqlen_q=1) (#1789 )

2025-01-07 18:49:24 +08:00

02_layernorm2d

enable bias feature that add bias before adding residual (for rtpllm project) (#1741 )

2025-01-08 17:51:06 +08:00

03_gemm

[CK TILE] GEMM and Batched GEMM SplitK support (#1724 )

2024-12-28 14:40:17 +01:00

04_img2col

[CK_TILE] Update example README files & fix script compatibility issue (#1548 )