biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Revert "[Kernel] Expand MoE weight loading + Add Fused Marlin MoE Kernel (#7527)" (#7764)

Browse Source
This commit is contained in:
Michael Goin
2024-08-21 23:42:14 -04:00
committed by GitHub
parent cde9183b40
commit aae74ef95c
15 changed files with 84 additions and 2374 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
csrc/moe/marlin_moe_ops.h
View File

@@ -1,12 +0,0 @@
#pragma once
#include <torch/all.h>
torch::Tensor marlin_gemm_moe(
const torch::Tensor& a, const torch::Tensor& b_q_weights,
const torch::Tensor& sorted_ids, const torch::Tensor& topk_weights,
const torch::Tensor& topk_ids, const torch::Tensor& b_scales,
const torch::Tensor& g_idx, const torch::Tensor& perm,
torch::Tensor& workspace, int64_t size_m, int64_t size_n, int64_t size_k,
bool is_k_full, int64_t num_experts, int64_t topk, int64_t moe_block_size,
bool replicate_input, bool apply_weights);