ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-03 05:01:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_TILE] Fix incorrect computation of group mode PagedAttention (#1688)

Browse Source 
* Allow getting batch size from splitkv tile partitioner

* Fix wrong paged-kvcache impl for group mode

* Fix wrong example code for page-kvcache

* Undo changes in fmha_fwd.cpp

* Always use 2D block table

* Add is_gappy kernel argument for paged-kvcache

The is_gappy argument is used for differentiating seqstart_k_ptr usage
in flash-attention & xformers

* Remove out-of-date comments

* Remove no-longer used method

* Fix wrong # page-block calculation

* Fix wrong comment

---------

Co-authored-by: Qianfeng <qianfeng.zhang@amd.com>
This commit is contained in:
Po Yen Chen
2024-11-26 20:37:54 +08:00
committed by GitHub
parent b6bcd76d88
commit cf2d635ea2
5 changed files with 105 additions and 64 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
example/ck_tile/01_fmha/fmha_fwd.cpp
View File

@@ -1046,6 +1046,7 @@ bool run(const ck_tile::ArgParser& arg_parser)
(0 < page_block_size ? block_table_buf.GetDeviceBuffer() : nullptr);
args.batch_stride_block_table = batch_stride_block_table;
args.page_block_size = page_block_size;
args.is_gappy = false; // use 'false' for flash-attention integration
args.cache_batch_idx =
(use_cache_batch_idx ? cache_batch_idx_buf.GetDeviceBuffer() : nullptr);