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

[2/n] Migrate per_token_group_quant to torch stable ABI (#36058)

Browse Source 
Signed-off-by: Mikayla Gawarecki <mikaylagawarecki@gmail.com>
This commit is contained in:
mikaylagawarecki
2026-03-25 13:15:13 -04:00
committed by GitHub
parent 1ac2ef2e53
commit bf4cc9ed2d
22 changed files with 207 additions and 133 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
csrc/quantization/w8a8/int8/per_token_group_quant.cu
View File

@@ -1,12 +0,0 @@
#include <ATen/cuda/CUDAContext.h>
#include <torch/all.h>
#include "quantization/w8a8/per_token_group_quant_8bit.h"
void per_token_group_quant_int8(const torch::Tensor& input,
torch::Tensor& output_q,
torch::Tensor& output_s, int64_t group_size,
double eps, double int8_min, double int8_max) {
per_token_group_quant_8bit(input, output_q, output_s, group_size, eps,
int8_min, int8_max);
}

2
csrc/quantization/w8a8/int8/scaled_quant.cu
View File

@@ -5,7 +5,7 @@
#include <cmath>
#include "dispatch_utils.h"
#include "quantization/vectorization_utils.cuh"
#include "libtorch_stable/quantization/vectorization_utils.cuh"
#include "cub_helpers.h"
static inline __device__ int8_t float_to_int8_rn(float x) {