Commit Graph
3191 Commits
Author SHA1 Message Date
akshatvishu 3084b5d107 [BugFix][TTS] Restore MOSS v1 codec streaming (#6420)
Signed-off-by: akshatvishu <akshatnayak197@gmail.com>
2026-09-18 20:06:57 +08:00
Tianyu Guo ab4cf24b3e Optimize Qwen-Image 2.1 attention and decode graph reuse (#7790)
Signed-off-by: Tianyu Guo <guoty9@mail2.sysu.edu.cn>
2026-09-18 19:37:19 +08:00
Tianyu Guo f930348e79 Preserve Qwen-Image 2.1 condition latent order (#7766)
Signed-off-by: Tianyu Guo <guoty9@mail2.sysu.edu.cn>
2026-09-18 18:07:45 +08:00
chickeytonandClaude Fable 5.1 6df62d5157 [CI/Build][MiniCPM-o] Fix the Nightly tests (#7758)
Signed-off-by: chickeyton <ngton2014@gmail.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
2026-09-18 08:58:33 +00:00
holykie aa2fadac42 [Perf][Diffusion] Activate NPU/GPU matmul for LoRA delta computation when weights are on CPU (#7590)
Signed-off-by: holykie <116775684+holykie@users.noreply.github.com>
2026-09-18 16:51:23 +08:00
Zeng Chuang eb2075981d [Test][Ascend] Share HunyuanVideo-1.5 E2E across GPU and NPU (#7547)
Signed-off-by: zengchuang <zengchuang3@huawei.com>
2026-09-18 08:49:08 +00:00
ooooooye 823af92347 [Bugfix][Ascend NPU][Diffusion] Load over-wide unquantized fallback weights straight into host memory under DLOFix/mxfp online offload after quant (#7009)
Signed-off-by: brandneway <gyuan4892@gmail.com>
2026-09-18 07:36:48 +00:00
WeiQing Chen 5d3e6dc1b4 [Bugfix][MiniMax-H3] Use Hopper-safe modulation precision (#7693)
Signed-off-by: david6666666 <530634352@qq.com>
2026-09-18 15:16:45 +08:00
Tianyu Guo b0a309e5e7 Isolate Qwen-Image 2.1 prefill and decode scheduling (#7770)
Signed-off-by: Tianyu Guo <guoty9@mail2.sysu.edu.cn>
2026-09-18 14:11:22 +08:00
Tianyu GuoandIsotr0py a26013145c Align Qwen-Image 2.1 preprocessing, CFG defaults and prefix cache (#7769)
Signed-off-by: Tianyu Guo <guoty9@mail2.sysu.edu.cn>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
2026-09-18 13:53:15 +08:00
Tianyu Guo 936c0cc7fe Align Qwen-Image-2.1 VAE normalization with Diffusers (#7768)
Signed-off-by: Tianyu Guo <guoty9@mail2.sysu.edu.cn>
2026-09-18 13:51:06 +08:00
Tianyu Guo 93d7a06e7f Use pre-norm text features for Qwen-Image 2.1 (#7765)
Signed-off-by: Tianyu Guo <guoty9@mail2.sysu.edu.cn>
2026-09-18 13:50:36 +08:00
Tianyu Guo eb54a3186b Separate Qwen-Image 2.1 batches by condition image layout (#7764)
Signed-off-by: Tianyu Guo <guoty9@mail2.sysu.edu.cn>
2026-09-18 13:47:35 +08:00
wtz2333 9fbc36ab2d [AR-Diffusion] Bound KV metadata for long rollouts (#7498)
Signed-off-by: wtz2333 <2955110911@qq.com>
2026-09-18 12:06:02 +08:00
1c7476ec19 [CI/Build][ROCm] Stabilize shared AMD test lanes (#7706)
Signed-off-by: andyluo7 <andy.luo@amd.com>
Signed-off-by: haic0 <haic0@users.noreply.github.com>
Co-authored-by: haic0 <haic0@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-09-18 11:13:07 +08:00
Isotr0py 3affc7766d Enable VAE tiling for Qwen-Image-2.1 (#5)
* init

Signed-off-by: Isotr0py <mozf@inferact.ai>

* fix

Signed-off-by: Isotr0py <mozf@inferact.ai>

* test(diffusion): add Qwen-Image-2.1 unit tests

Signed-off-by: Isotr0py <mozf@inferact.ai>

* fix(diffusion): Qwen-Image-2.1 VAE spatial ratio, tiling e2e verification

- AutoencoderKLQwenImage21: derive spatial_compression_ratio from the
  architecture (2 ** len(temperal_downsample) = 16x) instead of trusting
  the checkpoint's scale_factor_spatial=8, which corrupted tiled
  decode/encode geometry (output cropped to half size) and tiling gating.
- pipeline: _read_vae_scale_factor prefers the architecture-derived value
  for the same reason.
- models/qwen_image_21/__init__: lazy exports to break the circular import
  distributed VAE -> package __init__ -> pipeline -> distributed VAE.
- Add VAE unit tests: ratio derivation, tiled vs untiled decode/encode
  consistency, distributed tile_split/exec/merge == tiled_decode.
- e2e matrix (GB200, seed 42, 20 steps) all green; see e2e_out/qwen21_vae/.
- examples: qwen_image_21_edit.py gains --vae-use-tiling.
- docs: mark VAE patch-parallel verified for Qwen-Image-2.1; recipe notes
  the default-tile banding limitation at 2048px+.

* clean tests

Signed-off-by: Isotr0py <mozf@inferact.ai>

* fix cudnn determination

Signed-off-by: Isotr0py <mozf@inferact.ai>

---------

Signed-off-by: Isotr0py <mozf@inferact.ai>
2026-09-18 10:02:50 +08:00
Isotr0py 4fc3cf8c60 Enable CUDA Graph for DiT decode layers (#4)
* init

Signed-off-by: Isotr0py <mozf@inferact.ai>

* fix

Signed-off-by: Isotr0py <mozf@inferact.ai>

* test(diffusion): add Qwen-Image-2.1 unit tests

Signed-off-by: Isotr0py <mozf@inferact.ai>

* feat(qwen_image_21): opt-in CUDA graph capture for KV-cache decode steps

Decode steps (1..N) recompute only the target image's tokens at a fixed
shape, but the per-request dict KV cache made addresses unstable, so the
step could not be captured. Prefill now copies prefix K/V into persistent
fixed-address buffers (keyed by branch/batch/prefix_len/target_tokens/
dtype/device/backend) and re-points the caller's cache dicts at them;
decode replays a graph captured per key over the full 32-block forward.
Eager fallback (logged) for non-CUDA, SP/ring/TP>1, HSDP, compiled or
offload-hooked blocks, KV-cache quantization, dynamic LoRA wrappers,
padded text masks, unknown keys, and capture failure.

Off by default; opt in via OmniDiffusionConfig.enable_cuda_graph_decode
or the QwenImage21Transformer2DModel constructor kwarg.

E2E (GB200, T2I 1024x1024, 50 steps, CFG 4.0, CUDNN_ATTN):
171.2 -> 142.1 ms/step, 8.70s -> 7.24s total (1.20x), 44.9 dB PSNR
vs eager output.

* clean

Signed-off-by: Isotr0py <mozf@inferact.ai>

---------

Signed-off-by: Isotr0py <mozf@inferact.ai>
2026-09-18 10:02:39 +08:00
Zhenqiang Xuandzqxu 242c3cddeb [Hardware][Ascend] Integrate MindIE dense and sparse quantized attention for Wan2.2 T2V (#7561)
Signed-off-by: zqxu <xuzhenqiang3@huawei.com>
Co-authored-by: zqxu <xuzhenqiang3@huawei.com>
2026-09-18 01:28:27 +00:00
Anjie Hou f8197e120f [5/N] Schedule chunk transfers as their own stage (#7406)
Signed-off-by: specture724 <specture724@gmail.com>
2026-09-18 09:27:56 +08:00
Alex Brooks 6bda549ac2 [Bugfix] Fix Parallel State Initialization For MoE + Diffusion (#7676)
Signed-off-by: Alex Brooks <albrooks@redhat.com>
2026-09-17 17:15:37 -06:00
Joshna-Medisetty 77d8de70da [XPU][HunyuanImage3] Fix HunyuanImage-3.0 text-to-image on XPU (#7674)
Signed-off-by: Joshna Medisetty <joshna.medisetty@intel.com>
2026-09-17 19:23:27 +00:00
Nick CaoandCodex b81f2c5d4b [Misc] treewide: remove shm_threshold_bytes from deploy config (#7522)
Signed-off-by: Nick Cao <ncao@redhat.com>
Co-authored-by: Codex <noreply@openai.com>
2026-09-17 19:22:14 +00:00
dongbo910220 a6570b6814 [Diffusion] Fuse ERNIE-Image Q/K RoPE for eager execution (#7502)
Signed-off-by: dongbo910220 <1275604947@qq.com>
2026-09-17 19:06:45 +00:00
BeatSeat 9cdc57d2e7 [Bugfix][MAGI-2] Fix layerwise offload OOM by pointing offload block attrs to 'block' (#7523) (#7540)
Signed-off-by: BeatSeat <wendavid552@gmail.com>
2026-09-17 19:02:08 +00:00
dongbo910220 97912de8d8 [Kernel] Fuse LongCat paired Q/K RoPE (#7500)
Signed-off-by: dongbo910220 <1275604947@qq.com>
2026-09-18 02:27:57 +08:00
2ab5d1701a [4/N] Unify worker-side MP4 encoding and wire Wan into it (#7048)
Signed-off-by: specture724 <specture724@gmail.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
2026-09-17 23:33:56 +08:00
amy-why-3459 882bf8bc95 [Frontend] Add a shared realtime web UI for MiniCPM-o and Qwen3-Omni (#7222) (#7585)
Signed-off-by: amy-why-3459 <wuhaiyan17@huawei.com>
2026-09-17 23:32:31 +08:00
Canlin Guo eb1abbadf8 [Feature][TTS] MOSS-TTS: place the reference-audio encoder on the code2wav stage's GPU (#7724)
Signed-off-by: Canlin Guo <canlinguosdu@gmail.com>
2026-09-17 05:51:26 -07:00
howard.shan c61b30255b [Qwen3-Omni] Remove redundant text embedding in DeepStack path (#7573)
Signed-off-by: howard-shan <shanwenhao668@gmail.com>
Signed-off-by: howard.shan <shanwenhao668@gmail.com>
2026-09-17 05:47:10 -07:00
Tiago de Souza Fernandes 5b22f5562f fix(minicpmo45): deadline-align native duplex silence continuation (#7059)
Signed-off-by: Tiagosf00 <tiagotsf2000@gmail.com>
2026-09-17 05:26:34 -07:00
Chenchao Xu d0755b5716 [Model] Add π0.5 (Pi0.5) VLA model support (#6950)
Signed-off-by: chenchaox <chenchaox@nvidia.com>
2026-09-17 05:18:51 -07:00
ooooooye 873e9bff7c [Doc][NPU] Refresh MiniMax-H3 Ascend recipes for 0.28-era configurations (#7720)
Signed-off-by: brandneway <gyuan4892@gmail.com>
2026-09-17 11:05:15 +00:00
Clodagh Walsh bf74544a35 [Voxtral] Follow up to raise BadRequest for ref_audio (#7649)
Signed-off-by: Clodagh Walsh <clodaghwalsh17@gmail.com>
2026-09-17 03:56:58 -07:00
Tianyao WuandClaude f5e4f5fa42 [Bugfix][MiniCPM-o] Size the duplex HD slice reservation from the frame (#7654)
Signed-off-by: Tianyao Wu <rayroy31@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>
2026-09-17 03:35:46 -07:00
psv666 e3be42e052 [Core] Give the OpenAI Realtime wire codec a reusable home (RFC #6592 P0a) (#7640)
Signed-off-by: psv666 <2693925048@qq.com>
2026-09-17 02:33:40 -07:00
MrDongslsandMr.Dong 825ab20b07 [Model][PersonaPlex] add per-row RingKV offsets and active rows (#7670)
Signed-off-by: Mr.Dong <1042542469@qq.com>
Co-authored-by: Mr.Dong <1042542469@qq.com>
2026-09-17 01:50:40 -07:00
WeiQing Chen cb439f3e84 [Diffusion][Model] Add direct FastH3 8-Step V2 checkpoint loading (#7610)
Signed-off-by: david6666666 <530634352@qq.com>
2026-09-17 08:41:52 +00:00
Alex Brooks 8c3d3407d1 [CI] Fix Structured Structured Diffusion Config Wan Test (#7707)
Signed-off-by: Alex Brooks <albrooks@redhat.com>
2026-09-17 00:01:01 -07:00
LOGO127 224152566a [Model][PersonaPlex] Add paced multi-session validation (#7428)
Signed-off-by: luozijian <luozijian0924@gmail.com>
2026-09-16 23:04:37 -07:00
Sy03 4200768f59 [Frontend] Scale Omni serving across API processes (#6923)
Signed-off-by: Sy03 <1370724210@qq.com>
2026-09-16 22:29:31 -07:00
a769d9357b [Bugfix][NPU] Fix talker_mtp ACL graph capture missing force_uniform_decode (#7008)
Signed-off-by: Wallbreazzz <wallbreazzz@users.noreply.github.com>
Signed-off-by: Wallbreazzz <110282866+Wallbreazzz@users.noreply.github.com>
Co-authored-by: Canlin Guo <canlinguosdu@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
2026-09-16 21:50:38 -07:00
maithilijoshi20 f1a6e7ce2a [Core] Retain structured configs through runtime startup (#6849)
Signed-off-by: maithilijoshi20 <97733343+maithilijoshi20@users.noreply.github.com>
2026-09-16 20:47:38 -07:00
HAAZZZEEEE becfefd440 [Hardware][Ascend] Add Wan2.2 MXFP4 UOS, Smooth and W4A8 fallback (#7210)
Signed-off-by: HAAZZZEEEE <wang-taicheng@qq.com>
2026-09-17 03:11:40 +00:00
Linze Shi 4fab34c8f4 [Bugfix] Update ComfyUI serving-mode mock after helper rename (#7694)
Signed-off-by: Linze-Shi <linzeshi0@gmail.com>
2026-09-17 02:45:31 +00:00
WeiQing Chen 72cc012cfd [CI][bugfix] Fix MiniMax H3 FP8 quality test remote-code loading (#7617) 2026-09-16 18:52:23 -07:00
Bojiang Li d81777f297 [Doc] Scope Spark FP8 guidance to the validated revision (#7536)
Signed-off-by: bojiang-li <327132355+bojiang-li@users.noreply.github.com>
2026-09-17 08:30:47 +08:00
Linze Shi c8b3a17bd3 [Frontend] Allow explicit turn-based serving (#7675)
Signed-off-by: Linze-Shi <linzeshi0@gmail.com>
2026-09-16 17:29:57 -07:00
Aman c32aaeb2da [Bugfix] Fix Qwen3-TTS word timestamps with async chunking (#7544)
Signed-off-by: bezdarnost <amanurumbekov@gmail.com>
2026-09-16 15:07:52 -07:00
Weiming Liao 63032c91c6 [Frontend] Add ComfyUI Generate Music node with initial MiniMax Music 3 support (#7516)
Signed-off-by: Weiming Liao <liaowm5@gmail.com>
2026-09-16 13:43:43 -07:00
Yueqian Lin 5aadcbb0ab [Bugfix] Fix Breeze-TTS-2 repeated silence and audio EOS handling 2026-09-16 18:40:12 +00:00