@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / fedf64332e

Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 9, 05:00 UTC

[Bugfix]: Fix display errors in TORCH_CHECK messages (#32942)

Signed-off-by: linhaifeng <1371675203@qq.com> Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>

linhaifeng committed Feb 1, 2026 at 01:48 UTC fedf64332e9940c8119f584a74b9c08bab6852b7

6 files changed +8 -8

csrc/cpu/sgl-kernels/gemm.cpp

+1 -1

@@ -265,7 +265,7 @@ void tinygemm_kernel(

265 // mb_size = 4

266 case 0x42: LAUNCH_TINYGEMM_KERNEL_NN(4, 32); break;

267 case 0x44: LAUNCH_TINYGEMM_KERNEL_NN(4, 64); break;

268 - default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", "nb_size");

268 + default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", nb_size);

269 }

270 }

271 }

csrc/cpu/sgl-kernels/gemm_fp8.cpp

+1 -1

@@ -324,7 +324,7 @@ void tinygemm_kernel(

324 case 0x22: LAUNCH_TINYGEMM_KERNEL_NN(2, 32); break;

325 case 0x32: LAUNCH_TINYGEMM_KERNEL_NN(3, 32); break;

326 case 0x42: LAUNCH_TINYGEMM_KERNEL_NN(4, 32); break;

327 - default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", "nb_size");

327 + default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", nb_size);

328 }

329 }

330 }

csrc/cpu/sgl-kernels/gemm_int8.cpp

+1 -1

@@ -180,7 +180,7 @@ void tinygemm_kernel(

180 // mb_size = 4

181 case 0x42: LAUNCH_TINYGEMM_KERNEL_NN(4, 32); break;

182 case 0x44: LAUNCH_TINYGEMM_KERNEL_NN(4, 64); break;

183 - default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", "nb_size");

183 + default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", nb_size);

184 }

185 }

186 }

csrc/cpu/sgl-kernels/moe.cpp

+2 -2

@@ -398,7 +398,7 @@ void tinygemm_kernel(

398 case 0x32: LAUNCH_TINYGEMM_KERNEL_NN(3, 32); break;

399 // mb_size = 4

400 case 0x42: LAUNCH_TINYGEMM_KERNEL_NN(4, 32); break;

401 - default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", "nb_size");

401 + default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", nb_size);

402 }

403 }

404 }

@@ -511,7 +511,7 @@ void tinygemm_kernel(

511 case 0x32: LAUNCH_TINYGEMM_KERNEL_NN2(3, 32); break;

512 // mb_size = 4

513 case 0x42: LAUNCH_TINYGEMM_KERNEL_NN2(4, 32); break;

514 - default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", "nb_size");

514 + default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", nb_size);

515 }

516 }

517 }

csrc/cpu/sgl-kernels/moe_int8.cpp

+2 -2

@@ -271,7 +271,7 @@ void tinygemm_kernel(

271 case 0x22: LAUNCH_TINYGEMM_KERNEL_VNNI(2, 32); break;

272 case 0x32: LAUNCH_TINYGEMM_KERNEL_VNNI(3, 32); break;

273 case 0x42: LAUNCH_TINYGEMM_KERNEL_VNNI(4, 32); break;

274 - default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", "nb_size");

274 + default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", nb_size);

275 }

276 }

277 }

@@ -401,7 +401,7 @@ void tinygemm_kernel(

401 case 0x22: LAUNCH_TINYGEMM_KERNEL_VNNI2(2, 32); break;

402 case 0x32: LAUNCH_TINYGEMM_KERNEL_VNNI2(3, 32); break;

403 case 0x42: LAUNCH_TINYGEMM_KERNEL_VNNI2(4, 32); break;

404 - default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", "nb_size");

404 + default: TORCH_CHECK(false, "Unexpected block size, ", mb_size, "x", nb_size);

405 }

406 }

407 }

csrc/moe/marlin_moe_wna16/ops.cu

+1 -1

@@ -770,7 +770,7 @@ torch::Tensor moe_wna16_marlin_gemm(

770 b_bias = b_bias_or_none.value();

771 TORCH_CHECK(b_bias.device().is_cuda(), "b_bias is not on GPU");

772 TORCH_CHECK(b_bias.is_contiguous(), "b_bias is not contiguous");

773 - TORCH_CHECK(b_bias.size(1) == size_n, "b_bias.size(0) != size_n");

773 + TORCH_CHECK(b_bias.size(1) == size_n, "b_bias.size(1) != size_n");

774 TORCH_CHECK(b_bias.stride(1) == 1, "b_bias.stride(1) != 1");

775 } else {

776 b_bias = torch::empty({0}, options);