tzhouam 034d579d77 Update vLLM precompiled wheel version and refactor various components for improved functionality and error handling. Changes include modifying the get_name method in DiffusionQuantizationConfig to use class-level attributes, enhancing metadata handling in StageEngineCoreClient, and adjusting the AsyncOmni class constructor for better parameter management. Additionally, improved error handling in omni_snapshot_download and optimized initial chunk size computation in talker2code2wav_async_chunk for dynamic load adaptation. ...

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>

2026-03-19 05:14:34 +00:00

..

attention

[XPU] Enable FLASH_ATTN on XPU (#1332)

2026-02-26 20:28:09 +08:00

cache

[Feature] Add cache-dit support for HunyuanImage3 (#1848)

2026-03-18 10:18:27 +00:00

distributed

[Feat] Enable expert parallel for diffusion MoE layers (#1323)

2026-03-11 10:45:36 +08:00

executor

[Bugfix] Race condition in MultiprocExecutor when concurent access to Scheduler (#1448)

2026-02-26 21:03:38 +08:00

hooks

[Bugfix] fix layer-wise offload incompatible with cache-dit (#1786)

2026-03-13 15:28:13 +08:00

layers

[Feat] support SP for FLUX.2-klein (#1250)

2026-03-15 12:18:12 +08:00

lora

[BugFix] Fix Max Rank Handling in LoRA (#1397)

2026-03-14 02:42:54 +00:00

model_loader

[Bugfix] Fix cpu offload and quantization compatibility (#1473)

2026-03-14 18:55:30 +00:00

models

Fix OmniGen2 transformer config loading for HF models (#1934)

2026-03-19 10:21:54 +08:00

offloader

[BugFix] Enable CPU offloading and Cache-DiT together on Diffusion Model (#1723)

2026-03-09 22:25:41 +08:00

profiler

[Core] Implement Diffusion Profiler Support (#709)

2026-01-20 11:35:48 +08:00

quantization

Update vLLM precompiled wheel version and refactor various components for improved functionality and error handling. Changes include modifying the get_name method in DiffusionQuantizationConfig to use class-level attributes, enhancing metadata handling in StageEngineCoreClient, and adjusting the AsyncOmni class constructor for better parameter management. Additionally, improved error handling in omni_snapshot_download and optimized initial chunk size computation in talker2code2wav_async_chunk for dynamic load adaptation.

2026-03-19 05:14:34 +00:00

utils

Dev/rebase v0170 (#1639)

2026-03-09 17:29:38 +08:00

worker

[Bugfix] Fix cpu offload and quantization compatibility (#1473)

2026-03-14 18:55:30 +00:00

__init__.py

[Diffusion] Qwen image support (#82)

2025-11-28 14:22:11 +08:00

compile.py

[core] add torch compile for diffusion (#684)

2026-01-09 04:05:44 +13:00

data.py

fix: propagate parallel_config through create_default_diffusion (#1878)

2026-03-14 17:07:13 +08:00

diffusion_engine.py

[Bug]: fix CUDA OOM during diffusion post-processing (#1670)

2026-03-18 14:59:58 +00:00

envs.py

[Frontend] Rewrite video API for async job lifecycle (#1665)

2026-03-13 17:11:06 +08:00

forward_context.py

[Bugfix]: SP attention not enabling when _sp_plan hooks are not applied (#1704)

2026-03-11 10:49:42 +08:00

ipc.py

[Perf] Reduce IPC overhead for single-stage diffusion serving for Wan2.2 (#1715)

2026-03-10 00:03:51 +13:00

registry.py

[Diffusion] [Model] Dreamid omni (#1855)

2026-03-13 21:53:06 +08:00

request.py

[Frontend][Model] Support batch request with refined OmniDiffusionReq… (#797)

2026-01-28 16:07:42 +08:00

scheduler.py

[Perf] Reduce IPC overhead for single-stage diffusion serving for Wan2.2 (#1715)

2026-03-10 00:03:51 +13:00

stage_diffusion_client.py

[Entrypoint][Refactor] vLLM-Omni Entrypoint Refactoring (#1908)

2026-03-18 17:33:53 +08:00