[diffusion] Keep LongLive2 components resident on large GPUs - #35993
Merged
Merged
Conversation
Contributor
|
Preview deployment for your docs. Learn more about Mintlify Previews.
💡 Tip: Enable Workflows to automatically generate PRs for you. |
github-actions
Bot
added
documentation
BBuf
marked this pull request as ready for review
August 23, 2026 13:08
BBuf
requested review from
AgainstEntropy,
HaiShaw,
JustinTong0323,
mickqian,
ping1jing2,
sogalin,
wisclmy0611,
yichiche and
zijiexia
as code owners
August 23, 2026 13:09
Collaborator
Author
|
/tag-and-rerun-ci extra |
BBuf
force-pushed
the
bbuf/longlive2-high-memory-residency
branch
from
August 23, 2026 14:11
fdc114c to
9d3ecd7
Compare
longxin9715
added a commit
to longxin9715/sglang
that referenced
this pull request
Aug 24, 2026
…n-transport1 * 'main' of https://github.com/sgl-project/sglang: (326 commits) [diffusion] feat: cache LoRA-merged weights in files the page cache can hold (sgl-project#36062) [diffusion] Speed up LingBot high-quality VAE decode (sgl-project#36024) [diffusion] Honor XDG cache for model overlays (sgl-project#36019) Support streaming session on NPU (sgl-project#32597) fix(xpu): read enable_deterministic_inference from the config bag (sgl-project#36149) xeon ci fail fast strategy change (sgl-project#36146) [diffusion] Fix Hunyuan QKV pack indexing at production video shapes (sgl-project#36009) [diffusion] Refresh quality and BCG benchmark skills (sgl-project#36016) [MoE] Gather the cutlass MoE activation and its scales in one launch (sgl-project#34915) [diffusion] feat: add plain component weight overrides (sgl-project#36086) [diffusion] feat: support loading mixed w4a8 text encoders (sgl-project#36037) [diffusion] Default Hunyuan VAE to tiled decode (sgl-project#36012) fix(xpu): enable compressed-tensors FP8 W8A8 on XPU (RedHatAI FP8-dynamic models) (sgl-project#33057) chore: move cuda_vmm_utils.py under srt/utils/ (sgl-project#36053) [Intel XPU] Add xpu pass for biased_topk and hash_topk (sgl-project#33323) [CPU] Fix NUMA/core binding for DP ranks (sgl-project#32856) [Fix] Harden FlashAttention CUDA graph metadata bounds (sgl-project#35454) [XPU] Use a fused GDN kernel from sgl-kernel for Qwen3.5 (sgl-project#33354) [diffusion] Fuse LongCat-Image QKNorm and interleaved RoPE (sgl-project#35995) [diffusion] Keep LongLive2 components resident on large GPUs (sgl-project#35993) ... # Conflicts: # python/sglang/srt/multimodal/processors/base_processor.py # python/sglang/srt/server_args.py
saturn-acc
pushed a commit
to saturn-acc/sglang
that referenced
this pull request
Aug 31, 2026
jakki-amd
pushed a commit
to jakki-amd/sglang
that referenced
this pull request
Sep 9, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation
LongLive2 inherits the Wan memory defaults, which layerwise-offload the text encoder and VAE even when a large GPU has enough memory to keep the full pipeline resident. On H200 this adds roughly 280 ms of host/device traffic to every request.
Changes
H200 benchmark
Base:
3c69a4c744525be0146629fc5933e3ddab609fb2Model:
Rabinovich/LongLive-2.0-5B-DiffusersShape: 832x480, 61 frames, 4 steps, seed 42
Prompt:
A curious raccoonMethod: fresh process per sample, request warmup, ABBA order, two measured samples per variant
The high-quality profile reduced GPU memcpy time from about 617 ms to 6.7 ms. Denoise time is unchanged; the gain comes from removing repeated encoder/VAE transfers.
Output check
Main and PR videos are byte-identical in every repeated run:
0f57a6ddd031773cea922d0b36bec3f436b1096a2432eee81a1cce53c67524ea344b36d477d62891c28bb1777c54743955accaa41f10336b31a2c119b6a7eebfTests
pytest -q python/sglang/multimodal_gen/test/unit/test_server_args.py -k 'pipeline_configs_declare_auto_tune_hints or longlive_residency_scales_with_available_memory'pytest -q python/sglang/multimodal_gen/test/unit/test_longlive2_pipeline_config.pypre-commit run --files docs/cookbook/diffusion/LongLive/LongLive-2.0.mdx python/sglang/multimodal_gen/configs/pipeline_configs/longlive2.py python/sglang/multimodal_gen/test/unit/test_server_args.pyThe combined unit files report 180 passed plus 48 subtests and one unrelated LTX2 device-mode failure. The same failure reproduces on the base commit on H200.
CI States
Latest PR Test (Base): ✅ Run #32644662988
Latest PR Test (Extra): ✅ Run #32651539133
Latest PR Test (AMD ROCm 7.2): ⏳ Run #32644662882