mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-03 05:01:25 +00:00
2920604786
* hack for cap logits * fix bug * Re-format files * Allow specifying logits_soft_cap through APIs * Support turn on/off logits_soft_cap in async pipeline * Do not generate non-verified kernels * Align receipt used in Aiter * Sync logits soft-capping across pipelines * Re-enable some hdim pipelines * fix perf * Add attention variant for logits_soft_cap * Add newline at end-of-file * Fix performance * Add comment to explain logits_soft_cap pre-processing * Unify code * Unify floating-point literal style * Use class data member to slience the compilation error * [CK_TILE] Update attention customizaton interface: add LogitsMask() (#2133) * Send 'mask' along with variant params to the LogitsMask() * Send block indices to the variant * Add indices parameters in variant interface * Fix fmha bwd codegen error * Allow switch logits_soft_cap impl * Eliminate register spills * Fix compilation errors * Fix wrong LSE * Fix LSE for splitkv kernel * Sync splitkv pipeline changes * Add batch_prefill kernel/pipeline * Fix codegen error * Undo changes in CMakeLists.txt * Merge pipeline filtering check * Use different code path if kHasLogitsSoftCap=false * Remove [[maybe_unused]] attribute * Use pre-existing compile-time flag to instantiate templates * Sync pipeline changes * Update CHANGELOG.md --------- Co-authored-by: Bernard <bernaliu@amd.com> Co-authored-by: coderfeli <coderfeli@163.com>
54 lines
3.6 KiB
C++
54 lines
3.6 KiB
C++
// SPDX-License-Identifier: MIT
|
|
// Copyright (c) 2018-2025, Advanced Micro Devices, Inc. All rights reserved.
|
|
|
|
#pragma once
|
|
|
|
#include "ck_tile/ops/fmha/block/block_attention_bias_enum.hpp"
|
|
#include "ck_tile/ops/fmha/block/block_dropout.hpp"
|
|
#include "ck_tile/ops/fmha/block/block_masking.hpp"
|
|
#include "ck_tile/ops/fmha/block/block_position_encoding.hpp"
|
|
#include "ck_tile/ops/fmha/block/block_rotary_embedding.hpp"
|
|
#include "ck_tile/ops/fmha/block/page_block_navigator.hpp"
|
|
#include "ck_tile/ops/fmha/block/variants.hpp"
|
|
#include "ck_tile/ops/fmha/kernel/fmha_batch_prefill_kernel.hpp"
|
|
#include "ck_tile/ops/fmha/kernel/fmha_bwd_kernel.hpp"
|
|
#include "ck_tile/ops/fmha/kernel/fmha_fwd_appendkv_kernel.hpp"
|
|
#include "ck_tile/ops/fmha/kernel/fmha_fwd_appendkv_tile_partitioner.hpp"
|
|
#include "ck_tile/ops/fmha/kernel/fmha_fwd_kernel.hpp"
|
|
#include "ck_tile/ops/fmha/kernel/fmha_fwd_splitkv_combine_kernel.hpp"
|
|
#include "ck_tile/ops/fmha/kernel/fmha_fwd_splitkv_kernel.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_batch_prefill_pipeline_qr_ks_vs_async.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_batch_prefill_pipeline_qr_ks_vs_async_default_policy.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_bwd_convert_dq.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_bwd_dot_do_o.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_bwd_dq_dk_dv_pipeline_kr_ktr_vr.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_bwd_dq_dk_dv_pipeline_kr_ktr_vr_iglp.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_bwd_pipeline_default_policy.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_bwd_pipeline_enum.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_bwd_pipeline_problem.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_fwd_appendkv_pipeline.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_fwd_appendkv_pipeline_default_policy.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_combine_pipeline.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_combine_pipeline_default_policy.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_pipeline_nwarp_sshuffle_qr_ks_vs.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_pipeline_nwarp_sshuffle_qr_ks_vs_default_policy.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_pipeline_qr_ks_vs.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_pipeline_qr_ks_vs_default_policy.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_enum.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_problem.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_async.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_async_default_policy.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_default_policy.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_fp8.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_whole_k_prefetch.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_whole_k_prefetch_default_policy.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qs_ks_vs.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qs_ks_vs_default_policy.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qx_ks_vs_custom_policy.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/tile_fmha_shape.hpp"
|
|
#include "ck_tile/ops/fmha/pipeline/tile_fmha_traits.hpp"
|
|
#include "ck_tile/ops/common/generic_2d_block_shape.hpp"
|
|
#include "ck_tile/ops/common/tensor_layout.hpp"
|
|
#include "ck_tile/ops/common/utils.hpp"
|