Commit Graph

266 Commits

Author SHA1 Message Date
Hubert Lu
51e2eaa458 [AMD] Support fast_topk kernels in sgl-kernel (#15172) 2025-12-19 22:19:09 -08:00
66RING
46be74b4b4 [diffusion] kernel: timestep embedding kernel implementation (#12995)
Co-authored-by: 戚余航 <qiyuhang@bytedance.com>
Co-authored-by: Qi Yuhang <45795032+HydraQYH@users.noreply.github.com>
2025-12-19 20:59:50 +08:00
sunxxuns
f2d64e6782 [amd] Add deterministic all-reduce kernel for AMD (ROCm) (#15340)
Co-authored-by: Thomas Wang <1am9trash@gmail.com>
2025-12-18 23:36:03 -08:00
Bruce-x-1997
793c96c3d2 [perf]optimize w4afp8 kernel on deepseek-v3-0324 (#12921)
Signed-off-by: bruce.xu <bruce.x@gmicloud.ai>
2025-12-18 18:13:22 +08:00
Kevin_Xiong
4792d1f452 [sgl-kernel][1/2] Fused qk_norm_rope for GLM4.6 (#15141) 2025-12-18 17:07:04 +08:00
Xiaoyu Zhang
56d12b4aea Fix warp illegal instruction in kimi k2 thinking PCG (#15306) 2025-12-18 16:58:23 +08:00
blzheng
e27635a02d [CPU] Add 4D input support for ROPE in sgl-kernel (#9337) 2025-12-16 17:27:39 +08:00
blzheng
d16ff357db [CPU] Add Gemma3RMSNorm kernel in sgl-kernel and add ut (#9324) 2025-12-15 00:24:02 -08:00
zyl_keep_moving
a9ce1623cd [kernel][moe] add moe topk fast (#13969)
Co-authored-by: Xiaoyu Zhang <35585791+BBuf@users.noreply.github.com>
2025-12-14 22:26:40 +08:00
Zaili Wang
d6bd2d1126 [CPU] layernorm & fused add-layernorm kernels (#14074) 2025-12-11 16:58:23 -08:00
Yibo Cai
7871593cc8 [cpu] Implement all gather/reduce for arm64 cpu (#12527) 2025-12-08 19:03:04 +08:00
Qiaolin Yu
aeff0d386b Fix amd rope definition (#14556) 2025-12-07 23:47:03 -08:00
blzheng
d257bf87b9 [CPU] add mamba fla kernels for Qwen3-next (#12324) 2025-12-06 14:16:23 +08:00
Qiaolin Yu
cb8df87fc1 [1/2] Add rope kernel in sgl-kernel (#14334) 2025-12-04 16:45:44 +08:00
jianan-gu
70d2587324 [CPU] Optimize small oc GEMM for Qwen3-next on CPU (#12446)
Co-authored-by: Zheng, Beilei <beilei.zheng@intel.com>
2025-12-04 00:38:47 -08:00
Ma Mingfei
f90b400431 [CPU] add support for mamba causal conv1d for qwen3-next (#12309) 2025-12-04 13:41:42 +08:00
Ma Mingfei
46f21a5956 use faster covnersion from float8_e4m3fn to bfloat16 (#12316) 2025-12-04 12:34:05 +08:00
Qi Yuhang
16ff892c18 [sgl-kernel][Feat][B200][1/N] Support MXFP8 Grouped GEMM in Blackwell (#13731)
Co-authored-by: Yineng Zhang <me@zhyncs.com>
2025-12-04 10:09:09 +08:00
blzheng
974c562a25 [CPU] add fused_qkvzba_split_reshape_cat kernel for Qwen3-next (#12330) 2025-12-03 23:46:08 +08:00
Xuan Liao
c233e9d7a9 [CPU] Support chunk_gated_delta_rule kernel for Qwen3-Next (#12441) 2025-12-03 17:03:48 +08:00
Xiaoyu Zhang
c5947ecd85 Opt moe align block size kernel (#14133) 2025-12-02 19:13:55 +08:00
Xiaoyu Zhang
3de09aadbc Add new moe wna16 marlin gemm (#14122) 2025-12-01 23:07:53 +08:00
Zaili Wang
0b0b2607ca [CPU] Apply uv as package manager (#14106) 2025-11-28 10:36:53 -08:00
Yuan Luo
e12c78aab6 [sgl-kernel][1/2] Fused qk_norm_rope for Qwen3-MoE (#14036)
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
2025-11-28 12:25:15 +08:00
Xiaoyu Zhang
ecefc7904f [sgl-kernel Code Clean] Remove useless lightning_attention kernel (#13819) 2025-11-24 18:26:25 +08:00
YanbingJiang
acde21d8d5 Add fused_rmsnorm_gated_cpu kernel for CPU to support Qwen3-Next (#11577) 2025-11-21 01:33:31 +08:00
Roger Young
e72cf13693 Support moe topk sigmoid kernel (#13049)
Co-authored-by: xuebi <xuebi@minimaxi.com>
2025-11-20 00:24:37 +08:00
Xiaoyu Zhang
820e13c9c1 [opt kimi k2 3/n] opt kimi_k2 moe_fused_gate kernel (#13374) 2025-11-18 15:36:31 +08:00
iLeGend
20e59f9510 Add FP32 dtype support for RoPE - Part1 (#13181) 2025-11-15 11:37:18 -08:00
Xiaoyu Zhang
1d3d42bda0 [opt kimi k2 1 / n] Add kimi k2 moe fused gate (#13287) 2025-11-15 17:14:19 +08:00
wangyxbh
5027739f2c [CPU] Use covt_e4m3_bf16 to optim BF16 to FP8 convert (#12191) 2025-11-14 17:36:51 +08:00
Binyao Jiang
9db3add319 Update GDN causal conv1d cuda kernel - prepare for new changes (#13188) 2025-11-13 14:09:47 -08:00
Fan Yin
2966367a31 [sgl-kernel] support custom fp8 flashmla kernel (#13087) 2025-11-13 12:45:21 -08:00
haoyangli-amd
ea10a9d165 [bug][rocm]fix qr when variable inp (#11609)
Signed-off-by: Haoyang Li <lihaoyang0109@gmail.com>
2025-11-11 01:43:48 -08:00
Xiaoyu Zhang
05559a4a90 Support hidden_dim % 4 == 0 in per_token_quant_fp8 (#12883) 2025-11-10 17:13:14 +08:00
huangtingwei
a5e5088dfb Fix errors of page head kernels in sgl-kernel for ROCm (#12604) 2025-11-10 15:15:50 +08:00
hlu1
b8ddc296f4 [sgl-kernel][Deepseek V3.2] Add row_starts to topk kernel (#12582)
Signed-off-by: Hao Lu <14827759+hlu1@users.noreply.github.com>
2025-11-07 18:33:27 -08:00
jianan-gu
6e6009fb6b [CPU] Fix TP padding case with weight block size (#8243) 2025-11-07 03:24:48 +08:00
Lianmin Zheng
20315697f4 move all get_stream in sgl_kernel to c++ to reduce the launch overhead (#12521) 2025-11-02 13:15:05 -08:00
bingps
15ed27d7d4 [Fix] concat_mla_absorb_q_kernel fails for long inputs (#12453) 2025-11-02 11:52:06 -08:00
Xiaoyu Zhang
95191ebdca Migrate weak_ref_tensor to sgl-kernel (#12505) 2025-11-02 10:55:39 +08:00
Lianmin Zheng
c0652d907b Clean up sgl kernel (#12413)
Co-authored-by: Byron Hsu <byronhsu1230@gmail.com>
2025-10-31 01:13:34 -07:00
AichenF
ed1044ac1b support cutlass fp4 kernel in sm120 (#11737) 2025-10-29 12:25:16 -07:00
Qi Yuhang
ce112c07fe [sgl-kernel][4/N]Support Expert Specialization Grouped GEMM (#12080) 2025-10-27 01:20:01 -07:00
Joonchen Liau
f7dc2f334b [sgl-kernel] feat: Support sm120 cutlass fp8 gemm kernel (#9403) 2025-10-26 23:45:45 -07:00
huangtingwei
3e6281d0aa [HiCache]Page head layout IO kernel (#11615) 2025-10-26 15:53:50 +08:00
Jinwu
13bf565d60 [2/N]Support DeepSeek-R1 w4a8 low latency deepep (#8464)
Co-authored-by: Hank Han <hanhan7630@outlook.com>
Co-authored-by: Shangchuan Huang <2510421000@qq.com>
2025-10-24 17:41:16 -07:00
blzheng
13fb8b5489 [CPU] Optimize FP16 decode_attention_cpu (#10652) 2025-10-22 21:39:51 -07:00
Fan Yin
23afdfd1c2 [sgl-kernel] support flashmla libtorch (#11717) 2025-10-21 21:17:50 -07:00
Serge Panev
2b1da821b5 [NVIDIA] Add new SMs support for Spark & Thor (#11287)
Signed-off-by: Serge Panev <spanev@nvidia.com>
2025-10-22 02:02:24 +08:00