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

pre-commit autoupdate (#17380)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-04-29 14:46:55 +01:00
committed by GitHub
parent 00ee37efa2
commit 40896bdf3f
9 changed files with 37 additions and 37 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
csrc/quantization/gptq_marlin/gptq_marlin.cu
View File

@@ -173,8 +173,8 @@ dequant<half, vllm::kU4B8.id()>(int q) {
const int HI = 0x00f000f0;
const int EX = 0x64006400;
// Guarantee that the `(a & b) | c` operations are LOP3s.
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
// We want signed int4 outputs, hence we fuse the `-8` symmetric zero point
// directly into `SUB` and `ADD`.
const int SUB = 0x64086408;
@@ -197,9 +197,9 @@ dequant<nv_bfloat16, vllm::kU4B8.id()>(int q) {
// Guarantee that the `(a & b) | c` operations are LOP3s.
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
q >>= 4;
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
typename ScalarType<nv_bfloat16>::FragB frag_b;
static constexpr uint32_t MUL = 0x3F803F80;
@@ -221,8 +221,8 @@ dequant<half, vllm::kU4.id()>(int q) {
const int HI = 0x00f000f0;
const int EX = 0x64006400;
// Guarantee that the `(a & b) | c` operations are LOP3s.
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
const int SUB = 0x64006400;
const int MUL = 0x2c002c00;
@@ -244,9 +244,9 @@ dequant<nv_bfloat16, vllm::kU4.id()>(int q) {
// Guarantee that the `(a & b) | c` operations are LOP3s.
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
q >>= 4;
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
typename ScalarType<nv_bfloat16>::FragB frag_b;
static constexpr uint32_t MUL = 0x3F803F80;