Commit Graph
62 Commits
Author SHA1 Message Date
Ke Bao b6267de5ae Upgrade python version in lint ci (#16221) 2025-12-31 18:22:53 +08:00
Ke Bao b973202526 Split tp model worker init (#16165) 2025-12-30 23:36:03 +08:00
Ke Bao b3817fa93b Split model_worker init function (#16160) 2025-12-30 21:39:11 +08:00
Ke Bao 059428bd8a Tiny remove additional args in init_memory_pool (#16158) 2025-12-30 21:38:06 +08:00
Ke Bao 3de23274ee Clean up swa handling in fa3 backend (#15877) 2025-12-30 08:51:40 +08:00
Ke Bao faecd37ed4 Add Mimo-v2-flash model to ci test (#15887) 2025-12-27 14:18:08 +08:00
Ke Bao c28c536c91 Fix swa available memory check (#15867) 2025-12-26 13:19:01 +08:00
Ke Bao 7b7e357f61 Separate swa and local attention chunk cache eviction (#15820) 2025-12-26 09:34:22 +08:00
Ke Bao 2f66b0671b Fix chunk_kda_fwd missing argument (#15851) 2025-12-26 09:32:35 +08:00
Ke BaoandLiangsheng Yin cbf9f13493 Adjust server args for Mimo-v2-flash model (#15803)
Co-authored-by: Liangsheng Yin <lsyincs@gmail.com>
2025-12-25 16:04:39 +08:00
Ke Bao 8fe3e37468 Support piecewise cuda graph for dsv3 fp4 (#15531) 2025-12-21 14:50:32 +08:00
Ke Bao b399e3ac4f Support piecewise cuda graph for fused marlin moe (#15100) 2025-12-16 20:05:32 +08:00
Ke Bao 5b5571a8da Apply back moe_sum_reduce for fused_marlin_moe (#14829) 2025-12-11 00:39:41 +08:00
Ke Bao 0b3f002daf Update release-whl-kernel.yml (#13921) 2025-11-25 19:05:38 +08:00
Ke Bao 0d41ddfbd0 Temporarily disable test_vision_openai_server_a CI (#13331) 2025-11-15 19:46:49 +08:00
Ke Bao 8e9f05ece1 Update marlin moe kernel interface (#13322) 2025-11-15 17:10:39 +08:00
Ke Bao 2a96e302cb Revert moe sum reduce for marlin moe (#13314) 2025-11-15 15:57:41 +08:00
Ke Bao 4edb240112 Fuse routed_scaling_factor to fused_marlin_moe (#12998) 2025-11-13 00:18:58 +08:00
Ke Bao a7cc02e36e Fix run suite sanity check (#13133) 2025-11-12 15:29:10 +08:00
Ke Bao e38994dd71 Update rope dtype config (#13037) 2025-11-11 22:47:29 +08:00
Ke Bao 58b12ccb46 Support piecewise cuda graph for deepseek v3 (#12996) 2025-11-10 23:18:03 +08:00
Ke Bao db24d34603 Support piecewise cuda graph for MLA (#11812) 2025-11-10 09:13:48 +08:00
Ke Bao bb00e24f87 Adjust server launch time in ci (#12917) 2025-11-09 20:41:29 +08:00
Ke Bao b5e0417392 Add kimi k2 thinking to ci (#12907) 2025-11-09 16:10:32 +08:00
Ke Bao b142831a26 Fix empty server args in marlin moe test (#12904) 2025-11-09 13:30:47 +08:00
Ke Bao 5236290399 Update CODEOWNERS (#12897) 2025-11-09 09:37:11 +08:00
Ke Bao 44f594d832 Apply moe_reduce_sum kernel for fused_marlin_moe (#12888) 2025-11-09 01:31:05 +08:00
Ke Bao 0fe9c1f70b Fix piecewise cuda graph ci test (#12836) 2025-11-08 00:25:00 +08:00
Ke Bao cd135bfe30 Update dsv3 quantization auto setting for sm100 (#12778) 2025-11-06 22:39:13 +08:00
Ke Bao 7cee07a067 Fix skip layer in get_quant_method (#12632) 2025-11-04 23:27:46 +08:00
Ke Bao 69193f7122 Filter tokenizer warning for kimi models (#12485) 2025-11-01 16:27:31 +08:00
Ke Baoandyizhang2077 a4bf5c6ad2 Support Kimi Linear (#12469)
Co-authored-by: yizhang2077 <1109276519@qq.com>
2025-10-31 14:03:35 -07:00
Ke Bao f7f9e41b36 Fix run benchmark (#12473) 2025-11-01 02:39:48 +08:00
Ke Bao 0095e01874 Fix lint in deepseek-ocr (#12470) 2025-11-01 00:08:19 +08:00
Ke BaoandHanming Lu 24bc3fb0f9 EAGLE cache fix for SWARadixCache (#11231)
Co-authored-by: Hanming Lu <69857889+hanming-lu@users.noreply.github.com>
2025-10-07 18:21:37 +08:00
Ke Bao 31b49c0b51 EAGLE cache fix for HiCache (#11215) 2025-10-04 16:53:53 -07:00
Ke Bao 91847e382a Fix eagle radix cache (#10846) 2025-09-30 22:59:20 +08:00
Ke Bao 424591d53d Fix spec filter batch when target extend (#10991) 2025-09-30 14:44:02 +08:00
Ke Bao 60d7beda6b Add split tile size for Triton attention (#10425) 2025-09-14 17:35:49 -07:00
Ke Bao 603b3446dc Fix FA3 swa spec verify topk>1 (#9658) 2025-08-26 15:03:14 -07:00
Ke Bao e0ce171d79 Fix triton backend eagle illegal memory access (#9344) 2025-08-19 20:16:26 -07:00
Ke Bao be1a3cd9b4 Fix swa eagle verify accuracy for Triton backend (#9279) 2025-08-17 12:52:02 -07:00
Ke Bao 0ff6d1fce1 Support FA3 backend for gpt-oss (#9028) 2025-08-13 10:41:50 -07:00
Ke Bao 94f44b88d1 Update fa3 interface and add unit test (#9150) 2025-08-13 20:05:02 +08:00
Ke Bao 0475448ee3 Optimize triton swa kernel by skipping computation (#8860) 2025-08-06 21:37:50 +08:00
Ke Bao 399e7ec8b3 Refine naming (#8868) 2025-08-06 21:37:02 +08:00
Ke Bao 4fc5f2f977 Add unit test for triton swa kernel (#8853) 2025-08-06 16:10:38 +08:00
Ke Bao 8128e08d36 Turn off hybrid cache by default (#8839) 2025-08-06 09:53:45 +08:00
Ke Bao 0242bb9c74 Fix triton kernels topk with keyword arguments (#8732) 2025-08-03 10:45:15 -07:00
Ke Bao e252192679 Fix deepgemm masked grouped gemm jit compile (#8679) 2025-08-01 15:37:59 -07:00
Ke Bao 33f0de337d chore: bump v0.4.10.post1 (#8652) 2025-08-01 12:07:30 +08:00
Ke Bao 8fbcfd0723 Update step3v default config (#8626) 2025-08-01 00:49:26 +08:00
Ke Bao 3c307dc057 Fix hf3fs_fuse import error (#8623) 2025-07-31 22:42:31 +08:00
Ke Bao 5973675bc3 Fix moe align kernel test (#8531) 2025-07-29 11:03:02 -07:00
Ke Bao b582159246 Update PR template (#8465) 2025-07-28 22:12:36 +08:00
Ke Bao 8af145b7dc Fix test_moe_fused_gate_combined sgl-kernel ci test (#8374) 2025-07-26 09:30:12 +08:00
Ke Bao e885bfdc6a Fix sgl-kernel ci test (#8284) 2025-07-23 14:01:47 +08:00
Ke BaoandMick e2d66f60c8 Skip llama4 vision module loading when multimodal disabled (#8272)
Co-authored-by: Mick <mickjagger19@icloud.com>
2025-07-23 12:41:25 +08:00
Ke Bao 6936be3221 Remve router gemm output dtype conversion (#8204) 2025-07-21 15:37:00 +08:00
Ke Bao c9e8613c97 Apply fused sorted token ids padding (#8193) 2025-07-21 11:19:48 +08:00
Ke Bao 465968b2e3 Fix dtype error in CI (#8197) 2025-07-21 00:27:55 +08:00
Ke Bao a3398d8478 Optimize moe align block size kernel (#7794) 2025-07-07 09:20:30 +08:00