apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)

Browse Source 
* init

* init

* fix format

* add

* add files

* add ut

* fix some

* add ut

* add more

* add

* fix pre-commit

* fix pre-commit

* fix cover

* skip long seq

* add

* add

* fix

* remove not need

* fix set attr

* fix comments

* fix comments

* fix failed tests

---------

Co-authored-by: gongweibao <gognweibao@baidu.com>
This commit is contained in:
gongweibao
2026-03-16 21:32:43 +08:00
committed by GitHub
parent d113397b09
commit a6351dea0b
61 changed files with 1595 additions and 171 deletions
Download Patch File Download Diff File Expand all files Collapse all files
custom_ops/gpu_ops/mla_attn/kernel_traits.cuh
+4 -3
View File
@@ -90,13 +90,14 @@ struct AttentionKernelTraits {
static constexpr bool USE_TMA_LOAD_KV = USE_TMA_LOAD_KV_;
static constexpr int GROUP_SIZE = GROUP_SIZE_;
static constexpr int BLOCK_SHAPE_Q = BLOCK_SHAPE_Q_;
static_assert(BLOCK_SHAPE_Q % 64 == 0);
static_assert(BLOCK_SHAPE_Q % 64 == 0,
"BLOCK_SHAPE_Q must be a multiple of 64");
static constexpr int BLOCK_SHAPE_KV = BLOCK_SHAPE_KV_;
static constexpr int HEAD_DIM_QK = HEAD_DIM_QK_;
static constexpr int HEAD_DIM_VO = HEAD_DIM_VO_;
static constexpr int NUM_PER_STAGE = BLOCK_SHAPE_KV * HEAD_DIM_QK;
static_assert(HEAD_DIM_QK % 32 == 0);
static_assert(HEAD_DIM_VO % 32 == 0);
static_assert(HEAD_DIM_QK % 32 == 0, "HEAD_DIM_QK must be a multiple of 32");
static_assert(HEAD_DIM_VO % 32 == 0, "HEAD_DIM_VO must be a multiple of 32");
static constexpr int NUM_WARPS = 12;
static constexpr int NUM_THREADS = 384;