Commit Graph
27 Commits
Author SHA1 Message Date
Trevor Morris a91e90d9a3 [2/2] Fuse routed scaling factor into select_experts (#8690) 2025-08-20 15:10:16 -07:00
Trevor Morris eff4eb3fdd Add fp4 quantize before all-gather for Flashinfer cutlass MoE DP (max throughput) (#7667) 2025-08-15 22:08:11 -07:00
Trevor Morris 13c48dcf88 [1/2][resubmit again] sgl-kernel: Fuse routed scaling factor into moe_fused_gate (#9088) 2025-08-12 20:12:38 -07:00
Trevor Morris 52e1f52f32 [bugfix] Fix missing args in bench one batch (#8877) 2025-08-09 01:34:03 -07:00
Trevor MorrisandYineng Zhang a60f88b5a4 Add unit test for flashinfer fp4 moe (#8330)
Co-authored-by: Yineng Zhang <me@zhyncs.com>
2025-08-08 17:55:37 -07:00
Trevor Morris 591c232f7c [1/2][resubmit] sgl-kernel: Fuse routed scaling factor into moe_fused_gate (select_experts) (#8770) 2025-08-08 17:55:06 -07:00
Trevor Morris c0e84297c2 Use reduce scatter for DP (#8539) 2025-08-06 16:21:26 -07:00
Trevor Morris 9bd4872a34 [bugfix] Fix typo in modelopt quant: 'FusedMoE' object has no attribute 'local_num_experts' (#8768) 2025-08-04 11:08:08 -07:00
Trevor Morris 32f2815451 Do layernorm before allgather for DP attention (#8631) 2025-08-03 00:53:08 -07:00
Trevor Morris 89caf7a3c6 [bugfix] Apply routed scaling factor to cutlass_fused_experts_fp8 (#8688) 2025-08-01 19:00:24 -07:00
Trevor Morris f642524fd9 [1/2] sgl-kernel: Fuse routed scaling factor into select_experts (#8364) 2025-08-01 18:14:24 -07:00
Trevor Morris 6a7528e623 [bugfix] Fix page size for create_flashmla_kv_indices_triton() for cutlass mla (#8685) 2025-08-01 14:28:04 -07:00
Trevor Morris b7170cc820 [bugfix] Fix flashinfer cutlass EP moe after MoE refactor (#8630) 2025-07-31 13:57:08 -07:00
Trevor Morris 58c468f404 Fix FP4 MoE accuracy from missing routed_scaling_factor (#8333) 2025-07-25 16:40:23 -07:00
Trevor Morris bb9b608c86 [PD][NIXL] Set is_sorted=False to fix NIXL_ERR_NOT_FOUND (#7330) 2025-06-26 10:39:39 -07:00
Trevor Morris 5f527834a8 [PD] NIXL: Register kv args in advance and cleanup finished requests (#6717) 2025-06-24 11:26:09 -07:00
5962e70d8d FlashInfer NVFP4 MoE with EP & 2-stream shared expert (#7327)
Co-authored-by: JieXin Liang <Alcanderian@users.noreply.github.com>
Co-authored-by: alcanderian <alcanderian@gmail.com>
2025-06-22 13:38:47 -07:00
Trevor Morris e806f708c9 [PD] Make bootstrap code common between NIXL and Mooncake (#6473) 2025-05-27 12:47:38 -07:00
Trevor Morris 7adf245ba2 [Metrics] Add KV events publishing (#6098) 2025-05-19 14:19:54 -07:00
Trevor Morris 0ab3f437ab Cutlass MLA: Disable split kv due to https://github.com/NVIDIA/cutlass/issues/2274 (#6101) 2025-05-08 18:44:30 -07:00
Trevor Morris 8d463fe351 Cutlass MLA decode - fix dtype error (#5868) 2025-04-28 21:12:58 -07:00
Trevor Morris 84810da4ae Add Cutlass MLA attention backend (#5390) 2025-04-27 20:58:53 -07:00
Trevor Morris 4dce1cc608 [PD] Add NIXL transfer backend (#5477) 2025-04-22 01:36:12 +08:00
Trevor Morris e8f62b20ca BLackwell cutlass mla: Add check for bad page size/block num combinations (#5431) 2025-04-15 14:07:42 -07:00
Trevor Morris f65b8d5c89 Blackwell Cutlass MLA kernel (#5142) 2025-04-11 22:16:51 -07:00
Trevor Morris 11d760d56a FP4 weight loading and inference (2/2) (#3972) 2025-04-08 17:26:21 -07:00
Trevor Morris e9f8e42318 Support FP4 gemm (1/2) (#3899) 2025-03-24 19:50:23 -07:00