[Public release 26/04] Introducing Mega MoE, FP4 Indexer and other features/fixes (#304)

* Merge with private repo

* Update README

* Update README

* Update README

* Add PyTorch requirements

* Fix sync scopes for MQA logits (#256)

* Update README
This commit is contained in:
Chenggang Zhao
2026-04-17 09:45:14 +08:00
committed by GitHub
parent d30fc36c8f
commit 7f2a703ed5
109 changed files with 12101 additions and 3219 deletions
+6 -1
View File
@@ -3,12 +3,14 @@
#include <deep_gemm/impls/sm90_fp8_gemm_1d1d.cuh>
#include <deep_gemm/impls/sm90_fp8_gemm_1d2d.cuh>
#include <deep_gemm/impls/sm100_bf16_gemm.cuh>
#include <deep_gemm/impls/sm100_fp8_gemm_1d1d.cuh>
#include <deep_gemm/impls/sm100_fp8_fp4_gemm_1d1d.cuh>
// Attention kernels
#include <deep_gemm/impls/sm90_fp8_mqa_logits.cuh>
#include <deep_gemm/impls/sm90_fp8_paged_mqa_logits.cuh>
#include <deep_gemm/impls/sm100_fp4_mqa_logits.cuh>
#include <deep_gemm/impls/sm100_fp8_mqa_logits.cuh>
#include <deep_gemm/impls/sm100_fp4_paged_mqa_logits.cuh>
#include <deep_gemm/impls/sm100_fp8_paged_mqa_logits.cuh>
// Einsum kernels
@@ -23,6 +25,9 @@
#include <deep_gemm/impls/smxx_layout.cuh>
#include <deep_gemm/impls/smxx_clean_logits.cuh>
// Mega kernels
#include <deep_gemm/impls/sm100_fp8_fp4_mega_moe.cuh>
using namespace deep_gemm;
int main() {