ThomasX 3a5928ef53 feat: add page-aligned index validation for CP HiCache direct transfers ...

Add validate_page_aligned_token_indices utility and apply it across
HiCache write/load paths, NSA indexer transfers, and CUDA direct copy
kernels to reject malformed (partial, misaligned, non-contiguous) page
groups before they reach native transfer code. Also validate supported
CP HiCache backend/layout combinations at server startup.

2026-05-10 02:02:54 +08:00

..

spatial

[Feature] CUDA Green Context Support (#7649)

2025-07-15 02:49:16 +08:00

speculative

[speculative decoding] rename lookahead to ngram (#11010)

2025-09-28 21:06:59 -07:00

conftest.py

Add sgl-kernel CI test for Blackwell (B200) (#13301)

2025-11-20 19:02:42 -08:00

test_activation.py

…

test_amd_deterministic_custom_allreduce.py

[AMD CI] Add 2-GPU sgl-kernel Tests (#17555)

2026-01-22 21:48:52 -08:00

test_amd_nccl_allreduce_determinism.py

[AMD CI] Add 2-GPU sgl-kernel Tests (#17555)

2026-01-22 21:48:52 -08:00

test_apply_token_bitmask_inplace.py

…

test_awq_dequant.py

Fix AWQ Dequant and Weight Loading of deepseek v2 (#6842)

2025-06-17 13:45:10 -07:00

test_bmm_fp8.py

…

test_causal_conv1d.py

Add mamba kernel (#10234)

2025-09-09 12:58:43 -07:00

test_copy.py

Super tiny add UT for copy_to_gpu_no_ce (#12270)

2025-11-04 09:40:51 +08:00

test_custom_allreduce.py

…

test_cutlass_mla.py

Disable kernel cutlass_mla_decode on SM103 (#10058)

2025-09-06 01:35:18 -07:00

test_cutlass_w4a8_moe_mm.py

[kernel slimming] Clean many useless sgl-kernel deprecated kernels (#20277)

2026-03-14 16:45:54 +08:00

test_dsv3_fused_a_gemm.py

Add dsv3 fused a gemm to sgl-kernel (#7630)

2025-06-29 02:52:24 -07:00

test_dsv3_router_gemm.py

[Kimi K2] dsv3_router_gemm supports NUM_EXPERTS == 384 (#8013)

2025-08-01 22:01:24 +08:00

test_es_fp8_blockwise_moe.py

[sgl-kernel] fix b200 kernel ci (#13907)

2025-11-30 10:15:37 -08:00

test_es_mxfp8_blockscaled_moe.py

[sgl-kernel][Feat][B200][1/N] Support MXFP8 Grouped GEMM in Blackwell (#13731)

2025-12-04 10:09:09 +08:00

test_flash_attention.py

Reduce overhead for fa by not calling heavy CUDA property check (#7375)

2025-08-20 16:26:28 +08:00

test_flash_attn_sparse.py

[Tiny] Rename test_sparse_flash_attn.py to fix CI (#16895)

2026-01-11 18:18:29 +08:00

test_flashmla.py

[sgl-kernel] fix b200 kernel ci (#13907)

2025-11-30 10:15:37 -08:00

test_fp8_blockwise_gemm.py

…

test_fp8_blockwise_moe.py

Support running FP4 Deepseek on SM120. (#11708)

2025-10-27 17:37:49 -07:00

test_fp8_gemm.py

…

test_fused_qk_norm_rope.py

[sgl-kernel][1/2] Fused qk_norm_rope for GLM4.6 (#15141)

2025-12-18 17:07:04 +08:00

test_gguf.py

[sgl-kernel] Optimize gguf test (#11667)

2025-10-15 15:45:53 -07:00

test_gptq_kernel.py

[2/n]decouple quantization implementation from vLLM dependency (#8112)

2025-08-14 03:19:03 -07:00

test_hadamard.py

[kernel slimming] Move fast_hadamard_transform to jit_kernel (#18475)

2026-02-14 23:06:21 +08:00

test_int8_gemm.py

Skip some tests on Blackwell (#9777)

2025-08-28 20:00:32 -07:00

test_kimi_k2_moe_fused_gate.py

[opt kimi k2 1 / n] Add kimi k2 moe fused gate (#13287)

2025-11-15 17:14:19 +08:00

test_kvcacheio.py

feat: add page-aligned index validation for CP HiCache direct transfers

2026-05-10 02:02:54 +08:00

test_merge_state_v2.py

Add main for merge state tests (#6492)

2025-05-21 21:56:25 -07:00

test_merge_state.py

Add main for merge state tests (#6492)

2025-05-21 21:56:25 -07:00

test_moe_align.py

Opt moe align block size kernel (#14133)

2025-12-02 19:13:55 +08:00

test_moe_fused_gate.py

[CI] Fix test_moe_fused_gate error (#17844)

2026-01-28 12:03:17 +08:00

test_moe_topk_sigmoid.py

Support moe topk sigmoid kernel (#13049)

2025-11-20 00:24:37 +08:00

test_moe_topk_softmax.py

[kernel][moe] add moe topk fast (#13969)

2025-12-14 22:26:40 +08:00

test_mscclpp.py

support 1 shot allreduce in 1-node and 2-node using mscclpp (#6277)

2025-06-04 22:11:24 -07:00

test_norm.py

Restruct sgl-kernel benchmark (#10861)

2025-09-25 07:45:25 +08:00

test_per_token_group_quant_8bit.py

update pre-commit config (#18860)

2026-02-16 00:18:31 +08:00

test_per_token_quant_fp8.py

Support hidden_dim % 4 == 0 in per_token_quant_fp8 (#12883)

2025-11-10 17:13:14 +08:00

test_qserve_w4a8_per_chn_gemm.py

[1/2] Support Qserve (#6457)

2025-05-21 19:48:59 -07:00

test_qserve_w4a8_per_group_gemm.py

[1/2] Support Qserve (#6457)

2025-05-21 19:48:59 -07:00

test_sampling.py

use flashinfer.sampling (#18696)

2026-02-26 10:02:38 +08:00

test_topk.py

[sgl-kernel][Deepseek V3.2] Add row_starts to topk kernel (#12582)

2025-11-07 18:33:27 -08:00

test_torch_defaults_reset.py

Restore torch defaults between sgl-kernel tests (#11131)

2025-11-03 23:51:23 -08:00

utils.py

Skip some tests on Blackwell (#9777)

2025-08-28 20:00:32 -07:00