minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity
13,225 Commits 3 Branches 0 Tags
bdfd5da53e67bf6a7977995b4e4ba5da304f059a
Commit Graph
100 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Mick 24bcb37efb [diffusion] fix: fix diffusion LoRA consistency cases (#26327) 2026-05-28 06:29:35 +08:00
Mick f70e604101 [diffusion] fix: fix diffusion serve warmup defaults (#26247) 2026-05-27 15:43:47 +08:00
Mick 63a89bf1a9 [kernel] reuse wna16 marlin moe workspace (#26112) 2026-05-26 13:19:13 +08:00
Mick 8f2a4e70f8 [SRT] minor: reuse req input id array for unpadded ids (#26232) 2026-05-26 08:58:32 +08:00
Mick c05756da7a [SRT] fix flashInfer allreduce fusion not used on blackwell (#26197) 2026-05-25 18:07:15 +08:00
Mick e1463bb2c2 [VLM] try to reuse precomputed padded input ids in scheduler instead of padding (#26097) 2026-05-25 10:27:04 +08:00
Mick 64e2b54a8f [VLM] feat: accept grid_thws from preprocessed metadata for kimi (#26149) 2026-05-25 09:10:25 +08:00
Mick 72c1582d4e [VLM] fix: fix only the grids from last split mm item is collected for qwen-vl (#26094) 2026-05-25 09:09:46 +08:00
Mick 5c3775823e [diffusion] chore: use model-aware vae channels_last_3d policy (#26214) 2026-05-25 00:25:34 +08:00
Mick b6f71d5850 [VLM] avoid extra cuda-ipc staging for preprocessed input (#26096) 2026-05-24 19:48:06 +08:00
Mick 6447596501 [VLM] feat: replace small H2D calls with a single one for qwen-vl (#26167) 2026-05-24 18:27:17 +08:00
Mick 4c2b32bfbf [VLM] accept precomputed multimodal metadata (#26101) 2026-05-24 15:43:21 +08:00
Mick 826a4de062 [srt] store req input ids as arrays (#26165) 2026-05-24 15:09:36 +08:00
Mick d6d9f12444 [VLM] adopt simplified get_rope_index for image-only requests (#26100) 2026-05-24 11:51:24 +08:00
Mick 774b29dade [VLM] feat: early-return in mm processor if the input is preprocessed (#26117) 2026-05-23 17:02:44 +08:00
Mick 19b60a4f9e [VLM] reuse pretokenized ids from preprocessed input for qwen-vl (#26116) 2026-05-23 16:01:04 +08:00
Mick c8cea6d4aa [diffusion] feat: auto-select vae channels_last_3d (#26121) 2026-05-23 10:20:30 +08:00
Mick b801a27ce6 [diffusion] CI: disable torch compile in nightly comparison (#26120) 2026-05-22 23:36:47 +08:00
Mick f5ed2687ec [diffusion] CI: guard gt publishing from corrupt output (#26044) 2026-05-22 19:07:20 +08:00
Mick 16b3edc84f [diffusion] feat: enable warmup for sglang serve by default (#25988) 2026-05-22 08:54:48 +08:00
Mick f6d98a17ba [diffusion] optimize: reuse cached dynamic lora weights (#25893) 2026-05-22 08:54:01 +08:00
Mick ca9dc17be4 [diffusion] chore: adjust layer wise-offload strategy (#25930) 2026-05-21 23:48:58 +08:00
Mick 1ac3e33622 [diffusion] optimize: enable inference mode in pipeline executor (#25891) 2026-05-21 13:20:24 +08:00
Mick a7b3ced334 [diffusion] fix: fix LTX2 resident defaults and stage profiling (#25596) 2026-05-19 10:41:28 +08:00
Mick 110bbdcad7 [diffusion] fix: use dynamic LoRA for LTX2 original stage-two (#25591) 2026-05-18 23:05:29 +08:00
Mick 89e501c5a8 [diffusion] CI: tighten selected perf baselines (#25510) 2026-05-17 23:35:59 +08:00
Mick eccfd6dea7 [diffusion] feat: configure encoder as layerwise-offload by default (#25517) 2026-05-17 20:47:48 +08:00
Mick c1d9e37a52 [diffusion] feat: add memory-aware component load order (#25457) 2026-05-17 13:22:55 +08:00
Mick 7f37ffae9d [diffusion] CI: fix nightly CI (#25241) 2026-05-16 16:55:30 +08:00
Mick 416fdbbb3d [diffusion] feat: generalize layerwise offload residency mixin to all components (#24593) 2026-05-16 11:44:46 +08:00
Mick ff70aeac30 [diffusion] feat: add performance mode server args (#24491) 2026-05-14 00:57:46 +08:00
Mick 6d30b571b2 [diffusion] fix: fix single-step flow-match timesteps (#24708) 2026-05-11 18:09:24 +08:00
Mick 6e5b4de01a [diffusion] fix: further align ltx2.3 accuracy with tp (#24660) 2026-05-11 13:42:08 +08:00
Mick 73b8eda103 [diffusion] fix: fix FA3 varlen out argument handling (#24688) 2026-05-08 19:01:49 +08:00
Mick 17888fa92a [diffusion] doc: update ltx2 multi-gpu deployment guide (#24682) 2026-05-08 18:38:05 +08:00
Mick 2afb450501 [diffusion] optimize: optimize frame returns path (#24616) 2026-05-08 12:10:09 +08:00
Mick b0225a69dc [diffusion] optimize: precompute LTX2 guidance perturbation states (#24494) 2026-05-08 01:18:42 +08:00
Mick 2e642ea187 [diffusion] chore: align LTX-2 with official (#24313) 2026-05-07 08:46:28 +08:00
Mick fbebfdec9a [diffusion] fix: fix diffusion FSDP sharding (#24431) 2026-05-06 14:55:51 +08:00
Mick d23ef408f7 [diffusion] fix: fix RowParallel LoRA merged forwarding (#24410) 2026-05-06 00:30:16 +08:00
Mick cc54d8e8d0 [diffusion] chore: clean CUDA cache only at explicit release points (#24397) 2026-05-05 22:30:43 +08:00
Mick 177babcc38 [diffusion] optimize: fuse LTX2 split rotary embedding (#24411) 2026-05-05 16:07:40 +08:00
Mick 04926e1d9f [diffusion] feat: cache encoder results for default negative prompt (#24304) 2026-05-05 11:56:01 +08:00
Mick e483e60b72 [diffusion] CI: pin diffusion consistency GT revision (#24400) 2026-05-05 11:53:22 +08:00
Mick 2f7d99b7f7 [diffusion] cli: support component attention backend overrides (#24320) 2026-05-05 08:39:27 +08:00
Mick 1be3163011 [diffusion] fix: use direct all-to-all for USP collectives (#24366) 2026-05-05 00:08:48 +08:00
Mick c611a3fb78 [diffusion] chore: disable VAE cpu offload by default (#24315) 2026-05-04 08:24:51 +08:00
Mick 5925572c95 [diffusion] CI: switch CI data references to sgl-project/ci-data (#24299) 2026-05-03 23:05:12 +08:00
Mick 2bfc5d3bb1 [diffusion] optimize LTX2.3 HQ denoising split passes (#24298) 2026-05-03 16:37:46 +08:00
Mick 8b12367b92 [diffusion] CI: add diffusion GT generation (official implementation) to the workflow (#24270) 2026-05-03 15:10:35 +08:00
Mick b7d4647568 [diffusion] CI: change ground truth repo (#24219) 2026-05-01 21:25:40 -07:00
Mick 193b977572 [diffusion] chore: clean scheduler (#24229) 2026-05-02 09:30:06 +08:00
Mick 6fa499cd6d [diffusion] CI: remove parametrized-only from diffusion PR test (#24202) 2026-05-01 14:15:22 +08:00
Mick 9d84268705 [diffusion] refactor: introduce component residency manager (#23771) 2026-05-01 11:10:41 +08:00
Mick 0b1fbdba15 [diffusion] CI: change ground truth upload path and improve publish script (#24120) 2026-04-30 12:26:10 +08:00
Mick eeb7b5c433 [diffusion] fix: align encoder of flux klein with official (#24008) 2026-04-29 22:42:16 +08:00
Mick e5c2a9b6cf [diffusion] fix: improve LTX2.3 reference accuracy controls (#24022) 2026-04-29 21:39:27 +08:00
Mick 71d2227a78 [diffusion] CI: update ground truth with official output (#23714) 2026-04-29 13:51:31 +08:00
Mick 144038fbae [diffusion] chore: change default seed to 42 (#23836) 2026-04-28 20:39:23 +08:00
Mick a392ae8879 [diffusion] feat: accelerate multiple-outputs generation (#23759) 2026-04-27 01:47:33 +08:00
Mick d49a0377de [diffusion] refactor: make timestep scheduler request-local (#23716) 2026-04-26 15:59:53 +08:00
Mick 03849496ad jit_kernel: tolerate FA3 kernels without out arg (#23717) 2026-04-25 23:42:33 +08:00
Mick 05696527ea [diffusion] feat: support LoRA for LTX2.3 (#23649) 2026-04-25 01:52:41 +08:00
Mick b3b03369a5 [diffusion] fix: unify LTX-2.3 HQ codepath gates for all LTX-2.3 variants (#23624) 2026-04-24 17:44:08 +08:00
Mick cd1fa7506a [diffusion] model: support LTX2.3 high quality pipeline (#23366) 2026-04-24 14:18:20 +08:00
Mick c0166355ae [diffusion] CI: minor refactor CI (#23576) 2026-04-24 08:48:31 +08:00
Mick 4d3c7e781a [diffusion] CI: do not retry consistency failures (#23517) 2026-04-23 12:39:37 +08:00
Mick 4323fce82a fix: dot-boundary match in is_layer_skipped for FP8 modules_to_not_convert (#23467) 2026-04-22 22:16:22 +08:00
Mick 9a0fd2ff0c [diffusion] optimize: default to in-memory loading for URL/base64 image inputs (#23118) 2026-04-20 23:29:02 +08:00
Mick 0be6ab04dd [diffusion] refactor: LTX2.3 code cleanup (#23207) 2026-04-20 19:02:05 +08:00
Mick 0d94c3366a [diffusion] feat: introduce ltx-2-two-stage device manager (#22869) 2026-04-18 11:04:33 +08:00
Mick 5de89ea942 [diffusion] CI: fix auto-partition (#23076) 2026-04-17 22:37:24 +08:00
Mick 0b2058853d [diffusion] doc: update doc (#23052) 2026-04-17 16:23:46 +08:00
Mick 3d2d57c6cc [diffusion] refactor: extract LTX2 image encoding from denoising stage (#22976) 2026-04-17 08:35:15 +08:00
Mick 29f56cb230 CI: fix lint (#22991) 2026-04-17 02:09:04 +08:00
Mick 80718492dd [diffusion] CI: reset thresholds (#22854) 2026-04-15 21:11:00 +08:00
Mick e95c2e73bd [diffusion] CI: refactor diffusion ci and reduce redundancy (#22810) 2026-04-15 10:12:29 +08:00
MickandClaude Opus 4.6 d2f479e544 [diffusion] chore: auto-enable best parallel setting if unspecified (#22763) 
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-04-15 00:02:05 +08:00
Mick c5e95080d2 [diffusion] model: support Ltx 2.3 two stage ti2v (#22667) 2026-04-14 22:10:08 +08:00
Mick d524f110ac [diffusion] refactor: streamline denoising stages (#22633) 2026-04-13 13:34:37 +08:00
Mick bf022e177c Revert "[Diffusion] Add FLUX.1-dev ModelOpt NVFP4 support (#22574)" (#22649) 2026-04-13 11:17:32 +08:00
Mick 495ef8ec64 [diffusion] model: support LTX2.3 two stage (#22182) 2026-04-12 22:15:57 +08:00
Mick 0b4f5c9fcb [diffusion] CI: improve readability and fix bug of early-return (#22507) 2026-04-11 10:08:44 +08:00
Mick 7c6b5c095c [diffusion] fix: fix flux2 i2i accuracy (#22423) 2026-04-10 16:16:51 +08:00
Mick 9709192ce9 [diffusion] feat: support FLUX.2-small-decoder (#22414) 2026-04-09 15:53:14 +08:00
Mick 355fcbcc17 [diffusion] fix: fix cache dit refresh none mask (#22374) 2026-04-09 11:58:24 +08:00
Mick 4ac6fa0d87 [diffusion] fix: fix loading multiple ckpts with different precision for a same module (#22360) 2026-04-09 02:44:19 +08:00
Mick eca62ab8f4 UX: clean loggings (#22174) 2026-04-08 09:46:38 +08:00
Mick e7bc23cdab [diffusion] CI: fix consistency check (#22251) 2026-04-07 23:43:18 +08:00
Mick 82c41a2d9e [diffusion] model: support LTX2.3 (#22111) 2026-04-06 12:26:30 +08:00
MickandClaude Opus 4.6 efee62efa6 [diffusion] CI: improve diffusion comparison benchmark setting for realistic perf and auto-discover ut (#22086) 
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
 2026-04-04 23:20:37 +08:00
Mickgemini-code-assist[bot]Claude Opus 4.6
 		151f727163 [diffusion] fix: fix gated repo failing the generate cmd (#22040) 
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
 2026-04-04 00:43:11 +08:00
Mick 030fb1c4b1 refactor: replace mm_inputs dict with MultimodalProcessorOutput (#21738) 2026-04-03 23:26:37 +08:00
Mick 838f815e9f [diffusion] CI: temporarily disable accuracy ci (#22031) 2026-04-03 17:39:29 +08:00
Mick 2278a321ca [diffusion] chore: fix stage profiler for multi-stage denoising (#21955) 2026-04-03 01:16:38 +08:00
Mick 7bba319f1e [diffusion] fix: respect --prompt-path (#21756) 2026-04-01 16:47:59 +08:00
Mick f9debd6514 [diffusion] CI: improve ci reliability (#21763) 2026-04-01 10:06:57 +08:00
MickandClaude Opus 4.6 7790645b82 [diffusion] UX: replace deprecated ORJSONResponse with orjson_response (#21755) 
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
 2026-03-31 21:41:33 +08:00
Mick 0a1fb42869 [diffusion] CI: relax pr-test threshold (#21682) 2026-03-30 20:23:46 +08:00
Mick b76730701b [diffusion] feat: enhance overlay mechanism (#21648) 2026-03-30 19:45:34 +08:00