Skip to content

Commit 7eca34a

Browse files
hmellorlk-chen
hmellor
authored and
lk-chen
committed
pre-commit autoupdate (vllm-project#17380)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
1 parent d9b8d17 commit 7eca34a

File tree

9 files changed

+37
-37
lines changed

9 files changed

+37
-37
lines changed

‎.pre-commit-config.yaml

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -12,29 +12,29 @@ repos:
1212
- id: yapf
1313
args: [--in-place, --verbose]
1414
- repo: https://github.com/astral-sh/ruff-pre-commit
15-
rev: v0.9.3
15+
rev: v0.11.7
1616
hooks:
1717
- id: ruff
1818
args: [--output-format, github, --fix]
1919
- repo: https://github.com/codespell-project/codespell
20-
rev: v2.4.0
20+
rev: v2.4.1
2121
hooks:
2222
- id: codespell
2323
additional_dependencies: ['tomli']
2424
args: ['--toml', 'pyproject.toml']
2525
- repo: https://github.com/PyCQA/isort
26-
rev: 0a0b7a830386ba6a31c2ec8316849ae4d1b8240d # 6.0.0
26+
rev: 6.0.1
2727
hooks:
2828
- id: isort
2929
- repo: https://github.com/pre-commit/mirrors-clang-format
30-
rev: v19.1.7
30+
rev: v20.1.3
3131
hooks:
3232
- id: clang-format
3333
exclude: 'csrc/(moe/topk_softmax_kernels.cu|quantization/gguf/(ggml-common.h|dequantize.cuh|vecdotq.cuh|mmq.cuh|mmvq.cuh))|vllm/third_party/.*'
3434
types_or: [c++, cuda]
3535
args: [--style=file, --verbose]
3636
- repo: https://github.com/jackdewinter/pymarkdown
37-
rev: v0.9.27
37+
rev: v0.9.29
3838
hooks:
3939
- id: pymarkdown
4040
args: [fix]
@@ -43,7 +43,7 @@ repos:
4343
hooks:
4444
- id: actionlint
4545
- repo: https://github.com/astral-sh/uv-pre-commit
46-
rev: 0.6.2
46+
rev: 0.6.17
4747
hooks:
4848
- id: pip-compile
4949
args: [requirements/test.in, -o, requirements/test.txt]

‎csrc/moe/marlin_kernels/marlin_moe_kernel.h

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -138,8 +138,8 @@ __device__ inline FragB dequant<vllm::kU4B8.id()>(int q) {
138138
const int HI = 0x00f000f0;
139139
const int EX = 0x64006400;
140140
// Guarantee that the `(a & b) | c` operations are LOP3s.
141-
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
142-
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
141+
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
142+
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
143143
// We want signed int4 outputs, hence we fuse the `-8` symmetric zero point
144144
// directly into `SUB` and `ADD`.
145145
const int SUB = 0x64086408;
@@ -182,8 +182,8 @@ __device__ inline FragB dequant<vllm::kU4.id()>(int q) {
182182
const int HI = 0x00f000f0;
183183
const int EX = 0x64006400;
184184
// Guarantee that the `(a & b) | c` operations are LOP3s.
185-
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
186-
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
185+
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
186+
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
187187

188188
const int SUB = 0x64006400;
189189
const int MUL = 0x2c002c00;

‎csrc/moe/marlin_moe_wna16/marlin_template.h

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -209,8 +209,8 @@ __device__ inline typename ScalarType<half>::FragB dequant<half, 4>(
209209
const int HI = 0x00f000f0;
210210
const int EX = 0x64006400;
211211
// Guarantee that the `(a & b) | c` operations are LOP3s.
212-
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
213-
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
212+
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
213+
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
214214
// We want signed int4 outputs, hence we fuse the `-8` symmetric zero point
215215
// directly into `SUB` and `ADD`.
216216
const int SUB = 0x64086408;
@@ -233,9 +233,9 @@ dequant<nv_bfloat16, 4>(int q,
233233

234234
// Guarantee that the `(a & b) | c` operations are LOP3s.
235235

236-
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
236+
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
237237
q >>= 4;
238-
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
238+
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
239239

240240
static constexpr uint32_t MUL = 0x3F803F80;
241241
static constexpr uint32_t ADD = 0xC308C308;

‎csrc/moe/moe_wna16_utils.h

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -108,11 +108,11 @@ __device__ inline void dequant<half2, 4>(int q, half2* res) {
108108
const int MUL = 0x2c002c00;
109109
const int ADD = 0xd400d400;
110110

111-
int lo0 = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
112-
int hi0 = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
111+
int lo0 = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
112+
int hi0 = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
113113
q >>= 8;
114-
int lo1 = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
115-
int hi1 = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
114+
int lo1 = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
115+
int hi1 = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
116116

117117
res[0] = __hsub2(*reinterpret_cast<half2*>(&lo0),
118118
*reinterpret_cast<const half2*>(&SUB));
@@ -149,13 +149,13 @@ __device__ inline void dequant<nv_bfloat162, 4>(int q, nv_bfloat162* res) {
149149
static constexpr uint32_t MASK = 0x000f000f;
150150
static constexpr uint32_t EX = 0x43004300;
151151

152-
int lo0 = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
152+
int lo0 = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
153153
q >>= 4;
154-
int hi0 = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
154+
int hi0 = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
155155
q >>= 4;
156-
int lo1 = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
156+
int lo1 = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
157157
q >>= 4;
158-
int hi1 = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
158+
int hi1 = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
159159

160160
static constexpr uint32_t MUL = 0x3F803F80;
161161
static constexpr uint32_t ADD = 0xC300C300;

‎csrc/quantization/gptq_allspark/allspark_qgemm_w8a16.cu

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -347,7 +347,7 @@ struct ComputeTile_W8A16_PerC_MtilexNtilex32_multistage_SM8x_SplitK {
347347
for (int n_idx = 0; n_idx < WARP_NITER; ++n_idx) {
348348
hmma16816_f32<FType>(
349349
C_frag[m_idx][n_idx], A_frag[reg_buf_idx][m_idx],
350-
reinterpret_cast<uint32_t(&)[2]>(BF_frag[reg_buf_idx][n_idx]));
350+
reinterpret_cast<uint32_t (&)[2]>(BF_frag[reg_buf_idx][n_idx]));
351351
}
352352
}
353353
}

‎csrc/quantization/gptq_marlin/gptq_marlin.cu

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -173,8 +173,8 @@ dequant<half, vllm::kU4B8.id()>(int q) {
173173
const int HI = 0x00f000f0;
174174
const int EX = 0x64006400;
175175
// Guarantee that the `(a & b) | c` operations are LOP3s.
176-
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
177-
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
176+
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
177+
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
178178
// We want signed int4 outputs, hence we fuse the `-8` symmetric zero point
179179
// directly into `SUB` and `ADD`.
180180
const int SUB = 0x64086408;
@@ -197,9 +197,9 @@ dequant<nv_bfloat16, vllm::kU4B8.id()>(int q) {
197197

198198
// Guarantee that the `(a & b) | c` operations are LOP3s.
199199

200-
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
200+
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
201201
q >>= 4;
202-
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
202+
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
203203

204204
typename ScalarType<nv_bfloat16>::FragB frag_b;
205205
static constexpr uint32_t MUL = 0x3F803F80;
@@ -221,8 +221,8 @@ dequant<half, vllm::kU4.id()>(int q) {
221221
const int HI = 0x00f000f0;
222222
const int EX = 0x64006400;
223223
// Guarantee that the `(a & b) | c` operations are LOP3s.
224-
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
225-
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
224+
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
225+
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
226226

227227
const int SUB = 0x64006400;
228228
const int MUL = 0x2c002c00;
@@ -244,9 +244,9 @@ dequant<nv_bfloat16, vllm::kU4.id()>(int q) {
244244

245245
// Guarantee that the `(a & b) | c` operations are LOP3s.
246246

247-
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
247+
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
248248
q >>= 4;
249-
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, MASK, EX);
249+
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
250250

251251
typename ScalarType<nv_bfloat16>::FragB frag_b;
252252
static constexpr uint32_t MUL = 0x3F803F80;

‎csrc/quantization/marlin/dense/marlin_cuda_kernel.cu

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -96,8 +96,8 @@ __device__ inline FragB dequant(int q) {
9696
const int HI = 0x00f000f0;
9797
const int EX = 0x64006400;
9898
// Guarantee that the `(a & b) | c` operations are LOP3s.
99-
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
100-
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
99+
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
100+
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
101101
// We want signed int4 outputs, hence we fuse the `-8` symmetric zero point
102102
// directly into `SUB` and `ADD`.
103103
const int SUB = 0x64086408;

‎csrc/quantization/marlin/qqq/marlin_qqq_gemm_kernel.cu

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -141,8 +141,8 @@ __device__ inline FragB dequant_per_group(int q, FragS_GROUP& frag_s, int i) {
141141
static constexpr uint32_t HI = 0x00f000f0;
142142
static constexpr uint32_t EX = 0x64006400;
143143
// Guarantee that the `(a & b) | c` operations are LOP3s.
144-
uint32_t t0 = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
145-
uint32_t t1 = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
144+
uint32_t t0 = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
145+
uint32_t t1 = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
146146
// We want signed int4 outputs, hence we fuse the `-8` symmetric zero point
147147
// directly into `SUB` and `ADD`.
148148
static constexpr uint32_t SUB = 0x64086408;

‎csrc/quantization/marlin/sparse/common/mma.h

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -127,8 +127,8 @@ __device__ inline FragB dequant_4bit(int q) {
127127
const int HI = 0x00f000f0;
128128
const int EX = 0x64006400;
129129
// Guarantee that the `(a & b) | c` operations are LOP3s.
130-
int lo = lop3 < (0xf0 & 0xcc) | 0xaa > (q, LO, EX);
131-
int hi = lop3 < (0xf0 & 0xcc) | 0xaa > (q, HI, EX);
130+
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
131+
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
132132
// We want signed int4 outputs, hence we fuse the `-8` symmetric zero point
133133
// directly into `SUB` and `ADD`.
134134
const int SUB = 0x64086408;

0 commit comments

Comments
 (0)