mirror of
https://github.com/kvcache-ai/ktransformers.git
synced 2025-09-10 23:34:35 +00:00
24 lines
No EOL
923 B
C
24 lines
No EOL
923 B
C
/**
|
|
* @Description :
|
|
* @Author : Azure
|
|
* @Date : 2024-07-22 09:27:55
|
|
* @Version : 1.0.0
|
|
* @LastEditors : Azure
|
|
* @LastEditTime : 2024-07-26 08:35:00
|
|
* @Copyright (c) 2024 by KVCache.AI, All Rights Reserved.
|
|
**/
|
|
#pragma once
|
|
|
|
#include <torch/library.h>
|
|
#include <torch/extension.h>
|
|
#include <torch/torch.h>
|
|
|
|
torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
|
|
torch::Tensor& b_scales, torch::Tensor& g_idx,
|
|
torch::Tensor& perm, torch::Tensor& workspace,
|
|
int64_t num_bits, int64_t size_m, int64_t size_n,
|
|
int64_t size_k, bool is_k_full);
|
|
|
|
// torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm,
|
|
// int64_t size_k, int64_t size_n,
|
|
// int64_t num_bits);
|