feat(diffusion): add request-gated VAE fast paths

Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>
[Bugfix] rainfusion support all shape video (#6000 )
2026-08-11 09:31:27 +08:00 · 2026-08-11 05:10:50 +08:00 · 2026-08-11 05:08:31 +08:00 · 2026-08-10 22:20:37 +08:00 · 2026-08-10 21:55:01 +08:00 · 2026-08-10 19:47:03 +08:00
61 changed files with 2751 additions and 1384 deletions
@@ -1,6 +1,11 @@
 # CODEOWNERS is used to automatically assign PR reviewers
 # Format: path pattern @username or @team
 # More info: https://docs.github.com/en/repositories/managing-your-repositorys-settings-and-features/customizing-your-repository/about-code-owners
+#
+# The vllm_omni/** source-tree rules below are derived from the `owners` and
+# `primary_code_paths` front matter in docs/design/module/*.md. When updating
+# ownership, update the design doc first, then regenerate the matching rule here.
+# Last-match-wins: more specific rules must come AFTER broader ones.

 # Default owners (fallback reviewers for all files)
 * @Gaohan123 @hsliuustc0106 @ywang96
@@ -21,25 +26,10 @@
 # CI
 /.buildkite/ @yenuo26 @congw729 @NickCao

-# Engine and model_executor
-/vllm_omni/core/ @tzhouam @yinpeiqi @fake0fan @Sy0307
-/vllm_omni/engine/ @tzhouam @yinpeiqi @fake0fan @Sy0307
-/vllm_omni/entrypoints/ @tzhouam @yinpeiqi @fake0fan @alex-jw-brooks @NickCao
-/vllm_omni/model_executor/ @tzhouam @gcanlin
-/vllm_omni/worker/ @tzhouam @gcanlin
+# ComfyUI / external apps
+/apps/ @fhfuih @alex-jw-brooks

-# Diffusion
-/vllm_omni/diffusion/ @Isotr0py @princepride @SamitHuang @wtomin @ZJY0516 @RuixiangMa @david6666666 @xuechendi
-/vllm_omni/lora/ @princepride @SamitHuang @wtomin @ZJY0516 @RuixiangMa @yuanheng-zhao
-
-# Diffusion CPU offloading
-/vllm_omni/diffusion/offloader/ @Isotr0py @princepride @SamitHuang @wtomin @ZJY0516 @RuixiangMa @yuanheng-zhao @xuechendi
-
-# Omni and TTS support
-/vllm_omni/transformers_utils/ @tzhouam @yuanheng-zhao
-/vllm_omni/model_executor/models/ @linyueqian @ZeldaHuang @princepride @yuanheng-zhao @amy-why-3459 @Bounty-hunter @Sy0307 @gcanlin @NickCao
-
-# Benchmarks
+# Benchmarks (out-of-tree)
 /benchmarks/accuracy/ @david6666666
 /benchmarks/build_dataset/ @linyueqian @yuanheng-zhao
 /benchmarks/diffusion/ @Isotr0py @princepride @SamitHuang @wtomin @ZJY0516 @RuixiangMa @xuechendi
@@ -47,20 +37,112 @@
 /benchmarks/glm_image/ @princepride @JaredforReal
 /benchmarks/tts/ @linyueqian @ZeldaHuang @princepride @Sy0307 @gcanlin

-# Usability and Observability
-/vllm_omni/deploy/ @tzhouam @lishunyang12
-/vllm_omni/metrics/ @lishunyang12
-/vllm_omni/profiler/ @gcanlin
+# =====================================================================
+# Module design ownership (docs/design/module/*.md)
+# Each section header cites the source design doc.
+# =====================================================================

-# ComfyUI
-/apps/ @fhfuih @alex-jw-brooks
+# Autoregressive runtime — docs/design/module/ar_runtime.md
+/vllm_omni/core/ @tzhouam @yinpeiqi @fake0fan @Sy0307 @Gaohan123
+/vllm_omni/worker/ @tzhouam @yinpeiqi @fake0fan @Sy0307 @Gaohan123

-# Large Scale Serving
+# Engine orchestration & stage runtime — docs/design/module/engine_orchestration.md, stage_runtime.md
+/vllm_omni/engine/ @tzhouam @fake0fan
+# File-level overrides: input/output modality contracts own these engine files
+# — docs/design/module/input_output_modality_contracts.md
+/vllm_omni/engine/messages.py @Sy0307 @amy-why-3459 @Gaohan123 @alex-jw-brooks
+/vllm_omni/engine/serialization.py @Sy0307 @amy-why-3459 @Gaohan123 @alex-jw-brooks
+/vllm_omni/engine/mm_outputs.py @Sy0307 @amy-why-3459 @Gaohan123 @alex-jw-brooks
+/vllm_omni/engine/output_modality.py @Sy0307 @amy-why-3459 @Gaohan123 @alex-jw-brooks
+/vllm_omni/engine/output_processor.py @Sy0307 @amy-why-3459 @Gaohan123 @alex-jw-brooks
+
+# Input/output modality contracts — docs/design/module/input_output_modality_contracts.md
+/vllm_omni/inputs/ @Sy0307 @amy-why-3459 @Gaohan123 @alex-jw-brooks
+/vllm_omni/outputs/ @Sy0307 @amy-why-3459 @Gaohan123 @alex-jw-brooks
+/vllm_omni/request.py @Sy0307 @amy-why-3459 @Gaohan123 @alex-jw-brooks
+/vllm_omni/data_entry_keys.py @Sy0307 @amy-why-3459 @Gaohan123 @alex-jw-brooks
+
+# Entrypoints and serving boundaries — docs/design/module/entrypoints.md
+/vllm_omni/entrypoints/ @alex-jw-brooks @linyueqian @NickCao
+# File-level overrides within entrypoints/
+/vllm_omni/entrypoints/openai/errors.py @alex-jw-brooks @NickCao
+/vllm_omni/entrypoints/cli/benchmark/ @alex-jw-brooks @Bounty-hunter
+
+# Error contracts — docs/design/module/error_contracts.md
+/vllm_omni/errors.py @alex-jw-brooks @NickCao
+
+# OmniConnector (inter-stage transport) — docs/design/module/omni_connector.md
 /vllm_omni/distributed/ @princepride @yuanheng-zhao @xuechendi @natureofnature @fake0fan

-# Quantization and configuration
+# Model integration — docs/design/module/model_integration.md
+/vllm_omni/model_executor/ @tzhouam @gcanlin
+/vllm_omni/model_executor/models/ @tzhouam @gcanlin @Sy0307 @amy-why-3459 @linyueqian
+/vllm_omni/model_extras/ @tzhouam @gcanlin @princepride
+/vllm_omni/plugins/ @gcanlin @tjtanaa @xuechendi
+# Config override within model_executor — docs/design/module/vllm_omni_config.md
+/vllm_omni/model_executor/stage_configs/ @lishunyang12 @alex-jw-brooks
+
+# Execution platforms (hardware backends) — docs/design/module/execution_platforms.md
+/vllm_omni/platforms/ @gcanlin @tjtanaa @xuechendi
+/vllm_omni/attention/ @gcanlin @tjtanaa @xuechendi @david6666666 @lishunyang12
+
+# Configuration & deploy — docs/design/module/vllm_omni_config.md
 /vllm_omni/config/ @lishunyang12 @alex-jw-brooks
+/vllm_omni/deploy/ @lishunyang12 @alex-jw-brooks
+
+# Quantization — docs/design/module/quantization.md
 /vllm_omni/quantization/ @david6666666 @Isotr0py @lishunyang12

-# Hardware plugins and accelerators
-/vllm_omni/platforms/ @gcanlin @tjtanaa @xuechendi
+# Observability — docs/design/module/observability.md
+/vllm_omni/metrics/ @lishunyang12 @vraiti
+
+# Profiling — docs/design/module/profiling.md
+/vllm_omni/profiler/ @gcanlin
+
+# Benchmarking (in-tree) — docs/design/module/benchmarking.md
+/vllm_omni/benchmarks/ @alex-jw-brooks @Bounty-hunter
+
+# LoRA (no dedicated design doc; ownership intentionally narrowed)
+/vllm_omni/lora/ @SamitHuang @wtomin
+
+# Transformers utils (related path only in design docs; preserves prior ownership)
+/vllm_omni/transformers_utils/ @tzhouam @yuanheng-zhao
+
+# =====================================================================
+# Diffusion module family — docs/design/module/diffusion/
+# Umbrella rule first; sub-module rules after it win via last-match.
+# =====================================================================
+
+# Diffusion umbrella (top-level files: data.py, forward_context.py, etc.)
+# — docs/design/module/diffusion/index.md
+/vllm_omni/diffusion/ @Isotr0py @princepride @SamitHuang @wtomin @ZJY0516 @RuixiangMa @david6666666 @xuechendi
+
+# Cache management — docs/design/module/cache_management.md
+/vllm_omni/diffusion/cache/ @Isotr0py @princepride @SamitHuang
+
+# Diffusion quantization — docs/design/module/quantization.md
+/vllm_omni/diffusion/quantization/ @david6666666 @Isotr0py @lishunyang12
+
+# Diffusion profiling — docs/design/module/profiling.md
+/vllm_omni/diffusion/profiler/ @gcanlin
+
+# Diffusion runtime (scheduler, executor, worker)
+# — docs/design/module/diffusion/diffusion_runtime.md, continuous_batching.md
+/vllm_omni/diffusion/sched/ @Isotr0py @princepride @SamitHuang @fhfuih
+/vllm_omni/diffusion/executor/ @Isotr0py @princepride @SamitHuang @fhfuih
+/vllm_omni/diffusion/worker/ @Isotr0py @princepride @SamitHuang @fhfuih
+
+# Diffusion model integration (pipelines, registration, checkpoint loading)
+# — docs/design/module/diffusion/diffusion_model_integration.md
+/vllm_omni/diffusion/models/ @fhfuih @Bounty-hunter @wtomin
+/vllm_omni/diffusion/model_loader/ @fhfuih @Bounty-hunter @wtomin @RuixiangMa
+/vllm_omni/diffusion/lora/ @fhfuih @Bounty-hunter @wtomin @RuixiangMa
+
+# Diffusion parallelism (rank topology, TP/SP/PP/VAE patch)
+# — docs/design/module/diffusion/parallelism.md
+/vllm_omni/diffusion/distributed/ @princepride @yuanheng-zhao @xuechendi @Bounty-hunter
+/vllm_omni/diffusion/attention/parallel/ @princepride @yuanheng-zhao @xuechendi @Bounty-hunter
+
+# Diffusion offloader (CPU offloading, DLO)
+# — docs/design/module/diffusion/offloader.md
+/vllm_omni/diffusion/offloader/ @yuanheng-zhao @david6666666 @lishunyang12
@@ -66,7 +66,10 @@ repos:
      entry: python tools/pre_commit/check_tts_adapter.py
      language: python
      pass_filenames: false
-      files: ^vllm_omni/entrypoints/openai/(serving_speech\.py|tts_adapters/__init__\.py)$
+      files: ^(vllm_omni/entrypoints/openai/(serving_speech\.py|tts_adapters/__init__\.py)|tools/pre_commit/check_tts_adapter\.py)$
+      # The budget-is-now-slack reminder is printed on a PASSING run, and
+      # pre-commit hides output from passing hooks unless the hook is verbose.
+      verbose: true

    # Keep `suggestion` last
    - id: suggestion
@@ -16,31 +16,21 @@ nav:
  - Examples:
    - examples/README.md
    - Offline Inference:
-      - GLM-Image Multistage End-to-End Inference: user_guide/examples/offline_inference/glm_image.md
-      - HunyuanImage-3.0 Image-to-Text Inference: user_guide/examples/offline_inference/hunyuan_image3.md
      - Image-To-Image: user_guide/examples/offline_inference/image_to_image.md
      - Image-To-Video: user_guide/examples/offline_inference/image_to_video.md
-      - InternVLA-A1: user_guide/examples/offline_inference/internvla_a1.md
-      - MiMo-Audio: user_guide/examples/offline_inference/mimo_audio.md
-      - Qwen2.5-Omni: user_guide/examples/offline_inference/qwen2_5_omni.md
-      - Qwen3-Omni: user_guide/examples/offline_inference/qwen3_omni.md
+      - Speech-To-Video: user_guide/examples/offline_inference/speech_to_video.md
      - Text-To-Audio: user_guide/examples/offline_inference/text_to_audio.md
      - Text-To-Image: user_guide/examples/offline_inference/text_to_image.md
      - Text-To-Speech: user_guide/examples/offline_inference/text_to_speech.md
      - Text-To-Video: user_guide/examples/offline_inference/text_to_video.md
+      - X-To-Text: user_guide/examples/offline_inference/x_to_text.md
      - X-To-Video-Audio: user_guide/examples/offline_inference/x_to_video_audio.md
    - Online Serving:
-      - BAGEL-7B-MoT: user_guide/examples/online_serving/bagel.md
-      - vLLM-Omni Helm Chart: user_guide/examples/online_serving/chart-helm.md
-      - Diffusers Backend Adapter: user_guide/examples/online_serving/diffusers_pipeline_adapter.md
-      - GLM-Image Online Serving: user_guide/examples/online_serving/glm_image.md
      - Image-To-Image: user_guide/examples/online_serving/image_to_image.md
      - Image-To-Video: user_guide/examples/online_serving/image_to_video.md
-      - MiMo-Audio: user_guide/examples/online_serving/mimo_audio.md
-      - Qwen2.5-Omni: user_guide/examples/online_serving/qwen2_5_omni.md
-      - Qwen3-Omni: user_guide/examples/online_serving/qwen3_omni.md
-      - Text-To-Speech: user_guide/examples/online_serving/text_to_speech.md
+      - Speech-To-Video: user_guide/examples/online_serving/speech_to_video.md
      - Text-To-Image: user_guide/examples/online_serving/text_to_image.md
+      - Text-To-Speech: user_guide/examples/online_serving/text_to_speech.md
      - Text-To-Video: user_guide/examples/online_serving/text_to_video.md
  - General:
    - usage/*
@@ -7,6 +7,7 @@ owners:
  - "@yinpeiqi"
  - "@fake0fan"
  - "@Sy0307"
+  - "@Gaohan123"
 primary_code_paths:
  - vllm_omni/core/**
  - vllm_omni/worker/**
@@ -4,6 +4,7 @@ kind: module
 status: draft
 owners:
  - "@alex-jw-brooks"
+  - "@Bounty-hunter"
 primary_code_paths:
  - vllm_omni/benchmarks/**
  - vllm_omni/entrypoints/cli/benchmark/**
@@ -5,6 +5,7 @@ status: draft
 owners:
  - "@Isotr0py"
  - "@princepride"
+  - "@SamitHuang"
 primary_code_paths:
  - vllm_omni/diffusion/cache/**
 related_code_paths:
@@ -3,9 +3,10 @@ title: Diffusion Model Integration
 kind: module
 status: draft
 owners:
-  - "@Isotr0py"
+  - "@fhfuih"
+  - "@Bounty-hunter"
  - "@wtomin"
-  - "@ZJY0516"
+  - "@RuixiangMa"
 primary_code_paths:
  - vllm_omni/diffusion/models/**
  - vllm_omni/diffusion/model_loader/**
@@ -6,6 +6,7 @@ owners:
  - "@Isotr0py"
  - "@princepride"
  - "@SamitHuang"
+  - "@fhfuih"
 primary_code_paths:
  - vllm_omni/diffusion/sched/**
  - vllm_omni/diffusion/executor/**
@@ -4,7 +4,8 @@ kind: module
 status: draft
 owners:
  - "@yuanheng-zhao"
-  - "@Isotr0py"
+  - "@david6666666"
+  - "@lishunyang12"
 primary_code_paths:
  - vllm_omni/diffusion/offloader/**
 related_code_paths:
@@ -6,6 +6,7 @@ owners:
  - "@princepride"
  - "@yuanheng-zhao"
  - "@xuechendi"
+  - "@Bounty-hunter"
 primary_code_paths:
  - vllm_omni/diffusion/distributed/**
  - vllm_omni/diffusion/attention/parallel/**
@@ -6,6 +6,7 @@ architecture_state: current-plus-in-flight
 owners:
  - "@alex-jw-brooks"
  - "@linyueqian"
+  - "@NickCao"
 document_stewards:
  - "@hsliuustc0106"
  - "@Gaohan123"
@@ -8,8 +8,15 @@ owners:
  - "@xuechendi"
 primary_code_paths:
  - vllm_omni/platforms/**
-related_code_paths:
  - vllm_omni/attention/**
+primary_code_path_owners:
+  # Additional owners scoped to attention kernels (quantized/custom attention)
+  - paths:
+      - vllm_omni/attention/**
+    owners:
+      - "@david6666666"
+      - "@lishunyang12"
+related_code_paths: []
 depends_on:
  - ar_runtime.md
  - diffusion/index.md
@@ -6,6 +6,8 @@ architecture_state: current
 owners:
  - "@Sy0307"
  - "@amy-why-3459"
+  - "@Gaohan123"
+  - "@alex-jw-brooks"
 document_stewards:
  - "@hsliuustc0106"
  - "@Gaohan123"
@@ -9,6 +9,14 @@ primary_code_paths:
  - vllm_omni/model_executor/**
  - vllm_omni/model_extras/**
  - vllm_omni/plugins/**
+primary_code_path_owners:
+  # Additional owners scoped to a sub-path of model_executor/**
+  - paths:
+      - vllm_omni/model_executor/models/**
+    owners:
+      - "@Sy0307"
+      - "@amy-why-3459"
+      - "@linyueqian"
 related_code_paths:
  - vllm_omni/transformers_utils/**
  - vllm_omni/tokenizers/**
@@ -4,6 +4,7 @@ kind: module
 status: draft
 owners:
  - "@lishunyang12"
+  - "@vraiti"
 primary_code_paths:
  - vllm_omni/metrics/**
 related_code_paths:
@@ -24,6 +24,23 @@ SERVING_MODE_TITLES = {
    "online_serving": "Online serving",
 }

+# Keep the Examples navigation focused on reusable modality/task entry points.
+# Model-specific scripts remain available under ``examples/`` and are surfaced
+# from these shared pages instead of getting one page per model.
+GENERAL_EXAMPLE_SLUGS = frozenset(
+    {
+        "image_to_image",
+        "image_to_video",
+        "speech_to_video",
+        "text_to_audio",
+        "text_to_image",
+        "text_to_speech",
+        "text_to_video",
+        "x_to_text",
+        "x_to_video_audio",
+    }
+)
+

 def load_model_display_names() -> dict[str, str]:
    try:
@@ -46,6 +63,13 @@ def load_model_display_names() -> dict[str, str]:
 MODEL_DISPLAY_NAMES = load_model_display_names()


+def is_general_example(example: "Example") -> bool:
+    """Return whether an example should get a generated serving page."""
+    if example.category not in SERVING_MODE_TITLES:
+        return True
+    return example.path.stem in GENERAL_EXAMPLE_SLUGS
+
+
 def fix_case(text: str) -> str:
    subs = {
        "api": "API",
@@ -340,22 +364,21 @@ def update_nav_file(examples: list[Example]):
            examples_by_category[category] = []
        examples_by_category[category].append(example)

-    # Build the new Examples section - start with preserved items
-    examples_section = preserved_items.copy()
-
-    # Add examples grouped by category, sorted by category name
-    for category in sorted(examples_by_category.keys()):
-        category_examples = sorted(examples_by_category[category], key=lambda e: e.path.stem)
-        category_items = []
-        for example in category_examples:
+    def category_items(category: str) -> list[dict[str, str]]:
+        items = []
+        for example in sorted(examples_by_category[category], key=lambda e: e.path.stem):
            doc_path = EXAMPLE_DOC_DIR / example.category / f"{example.path.stem}.md"
            rel_path = doc_path.relative_to(ROOT_DIR / "docs")
-            category_items.append({example.nav_title: rel_path.as_posix()})
+            items.append({example.nav_title: rel_path.as_posix()})
+        return items

-        if category_items:
-            # Format category name (e.g., "offline_inference" -> "Offline Inference")
+    # Build the new Examples section - start with preserved items.
+    examples_section = preserved_items.copy()
+    for category in sorted(examples_by_category.keys()):
+        items = category_items(category)
+        if items:
            category_title = fix_case(category.replace("_", " ").title())
-            examples_section.append({category_title: category_items})
+            examples_section.append({category_title: items})

    # Update the nav structure
    nav_list[user_guide_idx]["User Guide"][examples_idx]["Examples"] = examples_section
@@ -390,6 +413,8 @@ def on_startup(command: Literal["build", "gh-deploy", "serve"], dirty: bool):
        for path in category.glob("*/*.md"):
            examples.append(Example(path.parent, category.stem))

+    examples = [example for example in examples if is_general_example(example)]
+
    # Generate the example documentation
    for example in sorted(examples, key=lambda e: e.path.stem):
        example_name = f"{example.path.stem}.md"
@@ -17,84 +17,93 @@ th {
 }
 </style>

-| Architecture | Models | Example HF Models | NVIDIA GPU | AMD GPU | Ascend NPU | Intel GPU |
-|--------------|--------|-------------------|------------|---------|-----|-----------|
-| `Qwen3OmniMoeForConditionalGeneration` | Qwen3-Omni | `Qwen/Qwen3-Omni-30B-A3B-Instruct` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `Qwen2_5OmniForConditionalGeneration` | Qwen2.5-Omni | `Qwen/Qwen2.5-Omni-7B`, `Qwen/Qwen2.5-Omni-3B` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `MingFlashOmniForConditionalGeneration` + `MingImagePipeline` | Ming-flash-omni-2.0 (omni-speech + imagegen<sup>1</sup>) | `Jonathan1909/Ming-flash-omni-2.0` | ✅︎ |   |   |   |
-| `BagelForConditionalGeneration` | BAGEL (DiT-only) | `ByteDance-Seed/BAGEL-7B-MoT` | ✅︎ | ✅︎ | | ✅︎ |
-| `InternVLAA1Pipeline` | InternVLA-A1 | `InternRobotics/InternVLA-A1-3B` | ✅︎ | ✅︎ | | |
-| `Gr00tN1d7Pipeline` | GR00T N1.7 | `nvidia/GR00T-N1.7-3B` | ✅︎ | | | |
-| `HunyuanImage3ForCausalMM` | HunyuanImage3.0 (DiT-only) | `tencent/HunyuanImage-3.0`, `tencent/HunyuanImage-3.0-Instruct` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `QwenImagePipeline` | Qwen-Image | `Qwen/Qwen-Image` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `QwenImagePipeline` | Qwen-Image-2512 | `Qwen/Qwen-Image-2512` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `QwenImageEditPipeline` | Qwen-Image-Edit | `Qwen/Qwen-Image-Edit` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `QwenImageEditPlusPipeline` | Qwen-Image-Edit-2509 | `Qwen/Qwen-Image-Edit-2509` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `QwenImageLayeredPipeline` | Qwen-Image-Layered | `Qwen/Qwen-Image-Layered` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `QwenImageEditPlusPipeline` | Qwen-Image-Edit-2511 | `Qwen/Qwen-Image-Edit-2511` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `GlmImagePipeline` | GLM-Image | `zai-org/GLM-Image` | ✅︎ | ✅︎ | | |
-| `ZImagePipeline` | Z-Image | `Tongyi-MAI/Z-Image-Turbo` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `Krea2Pipeline` | Krea 2 (Raw + Turbo) | `krea/Krea-2-Raw`, `krea/Krea-2-Turbo` | ✅︎ | | | |
-| `WanPipeline` | Wan2.1-T2V, Wan2.2-T2V, Wan2.2-TI2V | `Wan-AI/Wan2.1-T2V-1.3B-Diffusers`, `Wan-AI/Wan2.1-T2V-14B-Diffusers`, `Wan-AI/Wan2.2-T2V-A14B-Diffusers`, `Wan-AI/Wan2.2-TI2V-5B-Diffusers` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `WanImageToVideoPipeline` | Wan2.2-I2V | `Wan-AI/Wan2.2-I2V-A14B-Diffusers` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `Cosmos3OmniDiffusersPipeline` | Cosmos3 T2I, T2V, I2V, V2V, T2V with sound, action policy | `nvidia/Cosmos3-Nano`, `nvidia/Cosmos3-Super` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `Wan22S2VPipeline` | Wan2.2-S2V | `Wan-AI/Wan2.2-S2V-14B` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `Wan22VACEPipeline` | Wan2.1-VACE | `Wan-AI/Wan2.1-VACE-1.3B-diffusers`, `Wan-AI/Wan2.1-VACE-14B-diffusers` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `Wan22VACEPipeline` | Wan2.2-VACE | `Pyros13/Wan2.2-VACE-Fun-A14B-Diffusers` | ✅︎ |   |   |   |
-| `LTX2Pipeline` | LTX-2 / LTX-2.3 one-stage T2V and I2V | `Lightricks/LTX-2`, `diffusers/LTX-2.3-Diffusers` | ✅︎ | ✅︎ | | |
-| `LTX2DistilledPipeline` | LTX-2 distilled two-stage T2V and I2V | `rootonchair/LTX-2-19b-distilled` | ✅︎ | ✅︎ | | |
-| `LingBotVideoPipeline` | LingBot-Video dense and MoE T2I, T2V, TI2V | `robbyant/lingbot-video-dense-1.3b`, `robbyant/lingbot-video-moe-30b-a3b` | ✅︎ | | | |
-| `MiniMaxH3Pipeline` | MiniMax H3 T2VA, first/last-keyframe FL2VA, and mixed-reference Ref2VA | `MiniMaxAI/MiniMax-H3` | ✅︎ | | | |
-| `DreamZeroPipeline` | DreamZero-DROID | `GEAR-Dreams/DreamZero-DROID` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `HeliosPipeline`, `HeliosPyramidPipeline` | Helios | `BestWishYsh/Helios-Base`, `BestWishYsh/Helios-Mid`, `BestWishYsh/Helios-Distilled` | ✅︎ | ✅︎ | ✅︎ | |
-| `MagiHumanPipeline` | MagiHuman | `SII-GAIR/daVinci-MagiHuman-Base-1080p` | ✅︎ | ✅︎ | | |
-| `OvisImagePipeline` | Ovis-Image | `OvisAI/Ovis-Image` | ✅︎ | ✅︎ | | ✅︎ |
-| `LongcatImagePipeline` | LongCat-Image | `meituan-longcat/LongCat-Image` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `LongCatImageEditPipeline` | LongCat-Image-Edit | `meituan-longcat/LongCat-Image-Edit` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `BooguImagePipeline` | Boogu-Image | `Boogu/Boogu-Image-0.1-Base`, `Boogu/Boogu-Image-0.1-Edit` | ✅︎ | | | |
-| `StableDiffusionXLPipeline` | Stable-Diffusion-XL | `stabilityai/stable-diffusion-xl-base-1.0` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `StableDiffusion3Pipeline` | Stable-Diffusion-3 | `stabilityai/stable-diffusion-3.5-medium` | ✅︎ | ✅︎ | | ✅︎ |
-| `CosyVoice3Model` | CosyVoice3 | `FunAudioLLM/Fun-CosyVoice3-0.5B-2512` | ✅︎ | ✅︎ | | ✅︎ |
-| `OmniVoiceModel` | OmniVoice | `k2-fsa/OmniVoice` | ✅︎ | | | |
-| `VoxCPM2TalkerForConditionalGeneration` | VoxCPM2 | `openbmb/VoxCPM2` | ✅︎ | | | |
-| `MammothModa2ForConditionalGeneration` | MammothModa2-Preview | `bytedance-research/MammothModa2-Preview` | ✅︎ | ✅︎ | | |
-| `MammothModa2ForConditionalGeneration` | MammothModa2-Dev (AR-only image understanding) | `bytedance-research/MammothModa2-Dev` | ✅︎ | | | |
-| `Flux2KleinPipeline` | FLUX.2-klein | `black-forest-labs/FLUX.2-klein-4B`, `black-forest-labs/FLUX.2-klein-9B` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `FluxKontextPipeline` | FLUX.1-Kontext-dev | `black-forest-labs/FLUX.1-Kontext-dev` | ✅︎ | ✅︎ | | |
-| `FluxPipeline` | FLUX.1-dev | `black-forest-labs/FLUX.1-dev` | ✅︎ | ✅︎ | | ✅︎ |
-| `FluxPipeline` | FLUX.1-schnell | `black-forest-labs/FLUX.1-schnell` | ✅︎ | ✅︎ | | ✅︎ |
-| `OmniGen2Pipeline` | OmniGen2 | `OmniGen2/OmniGen2` | ✅︎ | ✅︎ | | ✅︎ |
-| `StableAudioPipeline` | Stable-Audio-Open | `stabilityai/stable-audio-open-1.0` | ✅︎ | ✅︎ | | ✅︎ |
-| `SoulXSingerPipeline` | SoulX-Singer (SVS) | `Soul-AILab/SoulX-Singer` | ✅︎ | | | |
-| `SoulXSingerSVCPipeline` | SoulX-Singer-SVC | `Soul-AILab/SoulX-Singer` (`model-svc.pt`) | ✅︎ | | | |
-| `AudioXPipeline` | AudioX | `zhangj1an/AudioX` | ✅︎ | ✅︎ | | |
-| `Qwen3TTSForConditionalGeneration` | Qwen3-TTS-12Hz-1.7B-CustomVoice | `Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `Qwen3TTSForConditionalGeneration` | Qwen3-TTS-12Hz-1.7B-VoiceDesign | `Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `Qwen3TTSForConditionalGeneration` | Qwen3-TTS-12Hz-1.7B-Base | `Qwen/Qwen3-TTS-12Hz-1.7B-Base` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-| `MingTTSForConditionalGeneration` | Ming-omni-tts dense 0.5B | `inclusionAI/Ming-omni-tts-0.5B` | ✅︎ | | | |
-| `GLMTTSForConditionalGeneration` | GLM-TTS | `zai-org/GLM-TTS` | ✅︎ | | | |
-| `MossTTSNanoForCausalLM` | MOSS-TTS-Nano | `OpenMOSS-Team/MOSS-TTS-Nano` | ✅︎ | | | |
-| `MossTTSDelayModel` | MOSS-TTS, MOSS-TTSD, MOSS-SoundEffect, MOSS-VoiceGenerator | `OpenMOSS-Team/MOSS-VoiceGenerator` | ✅︎ | | | |
-| `MossTTSRealtime` | MOSS-TTS-Realtime | `OpenMOSS-Team/MOSS-TTS-Realtime` | ✅︎ | | | |
-| `HiggsAudioV2ForConditionalGeneration` | Higgs-Audio v2 | `bosonai/higgs-audio-v2-generation-3B-base` | ✅︎ | | | |
-| `HiggsMultimodalQwen3ForConditionalGeneration` | Higgs-Audio v3 (TTS) | `bosonai/higgs-audio-v3-tts-4b` | ✅︎ | | | |
-| `IndexTTS2TalkerForConditionalGeneration` | IndexTTS-2 | `IndexTeam/IndexTTS-2` | ✅︎ | | | |
-| `NextStep11Pipeline` | NextStep-1.1 | `stepfun-ai/NextStep-1.1` | ✅︎ | ✅︎ | | ✅︎ |
-| `MiMoAudioModel` | MiMo-Audio-7B-Instruct | `XiaomiMiMo/MiMo-Audio-7B-Instruct` | ✅︎ | ✅︎ | | |
-| `MiMoV2ASRForCausalLM` | MiMo-V2.5-ASR | `XiaomiMiMo/MiMo-V2.5-ASR` | ✅︎ | ✅︎ | | |
-| `Flux2Pipeline` | FLUX.2-dev | `black-forest-labs/FLUX.2-dev` | ✅︎ | ✅︎ | | |
-| `FishSpeechSlowARForConditionalGeneration` | Fish Speech S2 Pro | `fishaudio/s2-pro` | ✅︎ | ✅︎ | | ✅︎ |
-| `DreamIDOmniPipeline` | DreamID-Omni | `XuGuo699/DreamID-Omni` | ✅︎ | ✅︎ | | |
-| `SenseNovaU1Pipeline` | SenseNova-U1 (DiT-only) | `SenseNova/SenseNova-U1-8B-MoT` | ✅︎ | | | |
-| `LancePipeline` | Lance | `bytedance-research/Lance` | ✅︎ | | | |
-| `HunyuanVideo15Pipeline` | HunyuanVideo-1.5-T2V | `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-480p_t2v`, `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-720p_t2v` | ✅︎ | ✅︎ | | |
-| `HunyuanVideo15ImageToVideoPipeline` | HunyuanVideo-1.5-I2V | `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-480p_i2v`, `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-720p_i2v` | ✅︎ | ✅︎ | | |
-| `VoxtralTTSForConditionalGeneration` | Voxtral TTS | `mistralai/Voxtral-4B-TTS-2603` | ✅︎ | ✅︎ | | |
-| `CovoAudioForConditionalGeneration` | Covo-Audio-Chat | `tencent/Covo-Audio-Chat` | ✅︎ | | | |
-|`DyninOmniForConditionalGeneration` | Dynin-Omni | `snu-aidas/Dynin-Omni` | ✅︎ | | | |
-| `MiniCPMO45OmniForConditionalGeneration` | MiniCPM-o 4.5 | `openbmb/MiniCPM-o-4_5` | ✅︎ | | ✅︎ | |
-| `ErnieImagePipeline` | ERNIE-Image | `baidu/ERNIE-Image`, `baidu/ERNIE-Image-Turbo` | ✅︎ | ✅︎ | ✅︎ | ✅︎ |
-|`HiDreamImagePipeline` | HiDream-I1-Full | `HiDream-ai/HiDream-I1-Full` | ✅︎ | ✅︎ | | |
+| Architecture | Models | Example HF Models | NVIDIA GPU | AMD GPU | Ascend NPU | Intel GPU | Recipe |
+|--------------|--------|-------------------|------------|---------|-----|-----------|--------|
+| `Qwen3OmniMoeForConditionalGeneration` | Qwen3-Omni | `Qwen/Qwen3-Omni-30B-A3B-Instruct` | ✅︎ | | | | [Repository](https://github.com/vllm-project/vllm-omni/blob/main/recipes/Qwen/Qwen3-Omni.md) |
+| `Qwen2_5OmniForConditionalGeneration` | Qwen2.5-Omni | `Qwen/Qwen2.5-Omni-7B`, `Qwen/Qwen2.5-Omni-3B` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `MingFlashOmniForConditionalGeneration` + `MingImagePipeline` | Ming-flash-omni-2.0 (omni-speech + imagegen<sup>1</sup>) | `Jonathan1909/Ming-flash-omni-2.0` | ✅︎ |   |   |   | — |
+| `BagelForConditionalGeneration` | BAGEL (DiT-only) | `ByteDance-Seed/BAGEL-7B-MoT` | ✅︎ | ✅︎ | | ✅︎ | — |
+| `InternVLAA1Pipeline` | InternVLA-A1 | `InternRobotics/InternVLA-A1-3B` | ✅︎ | ✅︎ | | | — |
+| `Gr00tN1d7Pipeline` | GR00T N1.7 | `nvidia/GR00T-N1.7-3B` | ✅︎ | | | | — |
+| `HunyuanImage3ForCausalMM` | HunyuanImage3.0 (DiT-only) | `tencent/HunyuanImage-3.0`, `tencent/HunyuanImage-3.0-Instruct` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `QwenImagePipeline` | Qwen-Image | `Qwen/Qwen-Image` | ✅︎ | | | | [Published](https://recipes.vllm.ai/Qwen/Qwen-Image) |
+| `QwenImagePipeline` | Qwen-Image-2512 | `Qwen/Qwen-Image-2512` | ✅︎ | | | | [Repository](https://github.com/vllm-project/vllm-omni/blob/main/recipes/Qwen/Qwen-Image-2512.md) |
+| `QwenImageEditPipeline` | Qwen-Image-Edit | `Qwen/Qwen-Image-Edit` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `QwenImageEditPlusPipeline` | Qwen-Image-Edit-2509 | `Qwen/Qwen-Image-Edit-2509` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `QwenImageLayeredPipeline` | Qwen-Image-Layered | `Qwen/Qwen-Image-Layered` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `QwenImageEditPlusPipeline` | Qwen-Image-Edit-2511 | `Qwen/Qwen-Image-Edit-2511` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `GlmImagePipeline` | GLM-Image | `zai-org/GLM-Image` | ✅︎ | ✅︎ | | | — |
+| `ZImagePipeline` | Z-Image | `Tongyi-MAI/Z-Image-Turbo` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `Krea2Pipeline` | Krea 2 (Raw + Turbo) | `krea/Krea-2-Raw`, `krea/Krea-2-Turbo` | ✅︎ | | | | — |
+| `WanPipeline` | Wan2.1-T2V, Wan2.2-T2V, Wan2.2-TI2V | `Wan-AI/Wan2.1-T2V-1.3B-Diffusers`, `Wan-AI/Wan2.1-T2V-14B-Diffusers`, `Wan-AI/Wan2.2-T2V-A14B-Diffusers`, `Wan-AI/Wan2.2-TI2V-5B-Diffusers` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `WanImageToVideoPipeline` | Wan2.2-I2V | `Wan-AI/Wan2.2-I2V-A14B-Diffusers` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `Cosmos3OmniDiffusersPipeline` | Cosmos3 T2I, T2V, I2V, V2V, T2V with sound, action policy | `nvidia/Cosmos3-Nano`, `nvidia/Cosmos3-Super` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `Wan22S2VPipeline` | Wan2.2-S2V | `Wan-AI/Wan2.2-S2V-14B` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `Wan22VACEPipeline` | Wan2.1-VACE | `Wan-AI/Wan2.1-VACE-1.3B-diffusers`, `Wan-AI/Wan2.1-VACE-14B-diffusers` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `Wan22VACEPipeline` | Wan2.2-VACE | `Pyros13/Wan2.2-VACE-Fun-A14B-Diffusers` | ✅︎ |   |   |   | — |
+| `LTX2Pipeline` | LTX-2 / LTX-2.3 one-stage T2V and I2V | `Lightricks/LTX-2`, `diffusers/LTX-2.3-Diffusers` | ✅︎ | ✅︎ | | | — |
+| `LTX2DistilledPipeline` | LTX-2 distilled two-stage T2V and I2V | `rootonchair/LTX-2-19b-distilled` | ✅︎ | ✅︎ | | | — |
+| `LingBotVideoPipeline` | LingBot-Video dense and MoE T2I, T2V, TI2V | `robbyant/lingbot-video-dense-1.3b`, `robbyant/lingbot-video-moe-30b-a3b` | ✅︎ | | | | — |
+| `MiniMaxH3Pipeline` | MiniMax H3 T2VA, first/last-keyframe FL2VA, and mixed-reference Ref2VA | `MiniMaxAI/MiniMax-H3` | ✅︎ | ✅︎ | | | [Published](https://recipes.vllm.ai/MiniMaxAI/MiniMax-H3) |
+| `DreamZeroPipeline` | DreamZero-DROID | `GEAR-Dreams/DreamZero-DROID` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `HeliosPipeline`, `HeliosPyramidPipeline` | Helios | `BestWishYsh/Helios-Base`, `BestWishYsh/Helios-Mid`, `BestWishYsh/Helios-Distilled` | ✅︎ | ✅︎ | ✅︎ | | — |
+| `MagiHumanPipeline` | MagiHuman | `SII-GAIR/daVinci-MagiHuman-Base-1080p` | ✅︎ | ✅︎ | | | — |
+| `OvisImagePipeline` | Ovis-Image | `OvisAI/Ovis-Image` | ✅︎ | ✅︎ | | ✅︎ | — |
+| `LongcatImagePipeline` | LongCat-Image | `meituan-longcat/LongCat-Image` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `LongCatImageEditPipeline` | LongCat-Image-Edit | `meituan-longcat/LongCat-Image-Edit` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `BooguImagePipeline` | Boogu-Image | `Boogu/Boogu-Image-0.1-Base`, `Boogu/Boogu-Image-0.1-Edit` | ✅︎ | | | | — |
+| `StableDiffusionXLPipeline` | Stable-Diffusion-XL | `stabilityai/stable-diffusion-xl-base-1.0` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `StableDiffusion3Pipeline` | Stable-Diffusion-3 | `stabilityai/stable-diffusion-3.5-medium` | ✅︎ | ✅︎ | | ✅︎ | — |
+| `CosyVoice3Model` | CosyVoice3 | `FunAudioLLM/Fun-CosyVoice3-0.5B-2512` | ✅︎ | ✅︎ | | ✅︎ | — |
+| `OmniVoiceModel` | OmniVoice | `k2-fsa/OmniVoice` | ✅︎ | | | | — |
+| `VoxCPM2TalkerForConditionalGeneration` | VoxCPM2 | `openbmb/VoxCPM2` | ✅︎ | | | | — |
+| `MammothModa2ForConditionalGeneration` | MammothModa2-Preview | `bytedance-research/MammothModa2-Preview` | ✅︎ | ✅︎ | | | — |
+| `MammothModa2ForConditionalGeneration` | MammothModa2-Dev (AR-only image understanding) | `bytedance-research/MammothModa2-Dev` | ✅︎ | | | | — |
+| `Flux2KleinPipeline` | FLUX.2-klein | `black-forest-labs/FLUX.2-klein-4B`, `black-forest-labs/FLUX.2-klein-9B` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `FluxKontextPipeline` | FLUX.1-Kontext-dev | `black-forest-labs/FLUX.1-Kontext-dev` | ✅︎ | ✅︎ | | | — |
+| `FluxPipeline` | FLUX.1-dev | `black-forest-labs/FLUX.1-dev` | ✅︎ | ✅︎ | | ✅︎ | — |
+| `FluxPipeline` | FLUX.1-schnell | `black-forest-labs/FLUX.1-schnell` | ✅︎ | ✅︎ | | ✅︎ | — |
+| `OmniGen2Pipeline` | OmniGen2 | `OmniGen2/OmniGen2` | ✅︎ | ✅︎ | | ✅︎ | — |
+| `StableAudioPipeline` | Stable-Audio-Open | `stabilityai/stable-audio-open-1.0` | ✅︎ | ✅︎ | | ✅︎ | — |
+| `SoulXSingerPipeline` | SoulX-Singer (SVS) | `Soul-AILab/SoulX-Singer` | ✅︎ | | | | — |
+| `SoulXSingerSVCPipeline` | SoulX-Singer-SVC | `Soul-AILab/SoulX-Singer` (`model-svc.pt`) | ✅︎ | | | | — |
+| `AudioXPipeline` | AudioX | `zhangj1an/AudioX` | ✅︎ | ✅︎ | | | — |
+| `Qwen3TTSForConditionalGeneration` | Qwen3-TTS-12Hz-1.7B-CustomVoice | `Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice` | ✅︎ | | | | [Published](https://recipes.vllm.ai/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice) |
+| `Qwen3TTSForConditionalGeneration` | Qwen3-TTS-12Hz-1.7B-VoiceDesign | `Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `Qwen3TTSForConditionalGeneration` | Qwen3-TTS-12Hz-1.7B-Base | `Qwen/Qwen3-TTS-12Hz-1.7B-Base` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+| `MingTTSForConditionalGeneration` | Ming-omni-tts dense 0.5B | `inclusionAI/Ming-omni-tts-0.5B` | ✅︎ | | | | — |
+| `GLMTTSForConditionalGeneration` | GLM-TTS | `zai-org/GLM-TTS` | ✅︎ | | | | — |
+| `MossTTSNanoForCausalLM` | MOSS-TTS-Nano | `OpenMOSS-Team/MOSS-TTS-Nano` | ✅︎ | | | | — |
+| `MossTTSDelayModel` | MOSS-TTS, MOSS-TTSD, MOSS-SoundEffect, MOSS-VoiceGenerator | `OpenMOSS-Team/MOSS-VoiceGenerator` | ✅︎ | | | | — |
+| `MossTTSRealtime` | MOSS-TTS-Realtime | `OpenMOSS-Team/MOSS-TTS-Realtime` | ✅︎ | | | | — |
+| `HiggsAudioV2ForConditionalGeneration` | Higgs-Audio v2 | `bosonai/higgs-audio-v2-generation-3B-base` | ✅︎ | | | | — |
+| `HiggsMultimodalQwen3ForConditionalGeneration` | Higgs-Audio v3 (TTS) | `bosonai/higgs-audio-v3-tts-4b` | ✅︎ | | | | [Published](https://recipes.vllm.ai/bosonai/higgs-audio-v3-tts-4b) |
+| `IndexTTS2TalkerForConditionalGeneration` | IndexTTS-2 | `IndexTeam/IndexTTS-2` | ✅︎ | | | | — |
+| `NextStep11Pipeline` | NextStep-1.1 | `stepfun-ai/NextStep-1.1` | ✅︎ | ✅︎ | | ✅︎ | — |
+| `MiMoAudioModel` | MiMo-Audio-7B-Instruct | `XiaomiMiMo/MiMo-Audio-7B-Instruct` | ✅︎ | ✅︎ | | | — |
+| `MiMoV2ASRForCausalLM` | MiMo-V2.5-ASR | `XiaomiMiMo/MiMo-V2.5-ASR` | ✅︎ | ✅︎ | | | — |
+| `Flux2Pipeline` | FLUX.2-dev | `black-forest-labs/FLUX.2-dev` | ✅︎ | ✅︎ | | | — |
+| `FishSpeechSlowARForConditionalGeneration` | Fish Speech S2 Pro | `fishaudio/s2-pro` | ✅︎ | ✅︎ | | ✅︎ | — |
+| `DreamIDOmniPipeline` | DreamID-Omni | `XuGuo699/DreamID-Omni` | ✅︎ | ✅︎ | | | — |
+| `SenseNovaU1Pipeline` | SenseNova-U1 (DiT-only) | `SenseNova/SenseNova-U1-8B-MoT` | ✅︎ | | | | — |
+| `LancePipeline` | Lance | `bytedance-research/Lance` | ✅︎ | | | | — |
+| `HunyuanVideo15Pipeline` | HunyuanVideo-1.5-T2V | `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-480p_t2v`, `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-720p_t2v` | ✅︎ | ✅︎ | | | — |
+| `HunyuanVideo15ImageToVideoPipeline` | HunyuanVideo-1.5-I2V | `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-480p_i2v`, `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-720p_i2v` | ✅︎ | ✅︎ | | | — |
+| `VoxtralTTSForConditionalGeneration` | Voxtral TTS | `mistralai/Voxtral-4B-TTS-2603` | ✅︎ | ✅︎ | | | — |
+| `CovoAudioForConditionalGeneration` | Covo-Audio-Chat | `tencent/Covo-Audio-Chat` | ✅︎ | | | | — |
+|`DyninOmniForConditionalGeneration` | Dynin-Omni | `snu-aidas/Dynin-Omni` | ✅︎ | | | | — |
+| `MiniCPMO45OmniForConditionalGeneration` | MiniCPM-o 4.5 | `openbmb/MiniCPM-o-4_5` | ✅︎ | | ✅︎ | | — |
+| `ErnieImagePipeline` | ERNIE-Image | `baidu/ERNIE-Image`, `baidu/ERNIE-Image-Turbo` | ✅︎ | ✅︎ | ✅︎ | ✅︎ | — |
+|`HiDreamImagePipeline` | HiDream-I1-Full | `HiDream-ai/HiDream-I1-Full` | ✅︎ | ✅︎ | | | — |
+
+The `Recipe` column links the corresponding deployment recipe. `Published` is
+preferred when a corresponding `recipes.vllm.ai` page exists; otherwise the
+link points to the repository recipe. `—` means that no recipe link has been
+added for that model row yet.
+
+For rows with an audited recipe link, the hardware checkmarks reflect hardware
+explicitly documented in that recipe. Other rows retain the implementation
+support metadata until their recipe is audited.

 ✅︎ indicates the model is supported on that backend. Empty cells mean not listed as supported on that backend.

@@ -283,23 +283,20 @@ blocks — use Ulysses SP (`ring_degree=1`).

 ### Which geometries run sparse

-Sparsity is only applied when the video segment is a **multiple of 128 rows**, where the row count
-is `latent_t × (height / 32) × (width / 32)`. Otherwise `rf_v2`'s block mask and the kernel's own
-tiling disagree on the block count, and the block straddling the seam mixes video and prefix rows
-that selection may then drop. Any resolution still runs — an unaligned geometry falls back to dense
-attention and logs `RAINFUSION_ATTN staying dense` with the row count it computed — but it gets no
-speedup, so pick an aligned geometry when you want one.
+MindIE-SD rf_v2 handles arbitrary video grids by spatially rearranging video first and, when
+necessary, promoting a real-video suffix to its always-kept prefix segment. It never pads: rf_v2
+does not consume an attention mask for padded keys. The remaining video rows stay sparse, while the
+promoted rows remain visible to every query and key.

-Alignment is necessary, not sufficient, for good quality. `rf_v2` groups video positions into 8x8
-spatial tiles (two tiles fill one 128-row block), which is what makes a selected block a compact
-patch of the frame. When the latent `h` or `w` is not a multiple of 8, the leftover rows or columns
-are peeled off and appended as a flat run instead of being tiled, so they get pooled with spatially
-distant positions and selection can no longer rank them meaningfully. This is silent, and invisible
-to a `sparsity=0` check, because a fully populated mask does not care how blocks are grouped.
+For example, MiniMax-H3 at 1344x768 has grid `(62, 24, 42)`: its 62496 video rows leave a 32-row
+128-block residue, and its two non-8-wide latent columns form a 2928-row rearranged suffix.
+MindIE-SD promotes 2976 rows, leaving 59520 sparse rows (`465 × 128`). This keeps the block mask
+and kernel tiling aligned, at the cost of retaining more blocks.

-"Multiple of 8" on the latent grid means **width and height that are multiples of 256**. Aligned
-resolutions off that grid still run sparse; they just lose more fidelity at the same `sparsity`,
-which you can buy back with `start_step`.
+8x8-aligned spatial grids remain preferable for performance. A latent `h` or `w` not divisible by 8
+creates an always-kept residual suffix and lowers realized sparsity. "Multiple of 8" on the latent
+grid means **width and height that are multiples of 256**. This path requires a MindIE-SD release
+that implements protected video tails for rf_v2.

 ## End-to-End Benchmark (BF16, sm_120 RTX Pro 6000 Blackwell)

@@ -99,7 +99,7 @@ a native vLLM-Omni pipeline for those cases.
 ### Attention Backends

 The diffusers backend converts
-[vLLM-Omni standard of attention backend setting](../../../docs/user_guide/diffusion/attention_backends.md)
+[vLLM-Omni standard of attention backend setting](../../diffusion/attention_backends.md)
 to [diffusers standard](https://huggingface.co/docs/diffusers/optimization/attention_backends#available-backends).

 Specifically for `FLASH_ATTN`, it will first attempt to use FlashAttention-3 and then FlashAttention-2.
@@ -1,247 +0,0 @@
-#!/usr/bin/env python3
-# SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-"""Verify a ModelOpt FP8 diffusers checkpoint exported by
-quantize_wan2_2_modelopt_fp8.py / quantize_hunyuanvideo_15_modelopt_fp8.py.
-
-Three checks:
-  A. transformer/config.json has a sane quantization_config block.
-  B. transformer/*.safetensors contains FP8 (float8_e4m3fn) quantized tensors.
-  C. transformer disk size is materially smaller than a BF16 baseline.
-
-Example:
-    python examples/quantization/check_modelopt_fp8_export.py \\
-        --output ./hv15-480p-modelopt-fp8
-
-    # Optional: compare disk size against a local or HF BF16 baseline.
-    python examples/quantization/check_modelopt_fp8_export.py \\
-        --output ./hv15-480p-modelopt-fp8 \\
-        --baseline hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-480p_t2v
-"""
-
-from __future__ import annotations
-
-import argparse
-import json
-import sys
-from collections import Counter
-from pathlib import Path
-
-SUPPORTED_ALGOS = ("FP8",)
-
-
-def _check_config(transformer_dir: Path) -> tuple[int, str | None]:
-    """Returns (status, quant_algo). status: 0 pass, 1 fail, 2 warn."""
-    cfg_path = transformer_dir / "config.json"
-    if not cfg_path.exists():
-        print(f"[FAIL] {cfg_path} missing.")
-        return 1, None
-
-    with cfg_path.open(encoding="utf-8") as f:
-        cfg = json.load(f)
-
-    qc = cfg.get("quantization_config")
-    if not isinstance(qc, dict):
-        print(f"[FAIL] No `quantization_config` block in {cfg_path}.")
-        return 1, None
-
-    print(f"[A] quantization_config from {cfg_path}:")
-    print(json.dumps(qc, indent=2))
-
-    quant_algo = qc.get("quant_algo")
-    issues = []
-    if qc.get("quant_method") != "modelopt":
-        issues.append(f"quant_method={qc.get('quant_method')!r} (expected 'modelopt')")
-    if quant_algo not in SUPPORTED_ALGOS:
-        issues.append(
-            f"quant_algo={quant_algo!r} (expected one of {SUPPORTED_ALGOS} — vllm-omni adapter may not auto-detect)"
-        )
-
-    if issues:
-        print("[A] WARN — config looks incomplete:")
-        for issue in issues:
-            print(f"    - {issue}")
-        return 2, quant_algo
-    print(f"[A] PASS — config looks correct (quant_algo={quant_algo}).")
-    return 0, quant_algo
-
-
-def _read_safetensors_header(path: Path) -> dict:
-    """Read the JSON header of a safetensors file. Bypass-safe — doesn't materialize tensors.
-
-    Returns {tensor_name: {'dtype': 'F8_E4M3', 'shape': [...], 'data_offsets': [...]}}.
-    Header dtype strings: F8_E4M3, F8_E5M2, BF16, F16, F32, F64, I8, I16, I32, I64, BOOL, U8, ...
-    """
-    import struct
-
-    with open(path, "rb") as f:
-        header_len = struct.unpack("<Q", f.read(8))[0]
-        header = json.loads(f.read(header_len))
-    header.pop("__metadata__", None)
-    return header
-
-
-def _classify_weight_scale_granularity(weight_scale_shapes: list[list[int]]) -> str:
-    """Infer per-tensor vs per-channel vs per-block from sample weight_scale shapes.
-
-    ModelOpt block-wise produces shapes like `[16, 1, 16, 1]` (broadcasting dims of 1
-    interleaved with block-count dims). We count "meaningful" dims — ones with size > 1 —
-    and classify: 0 meaningful dims = per-tensor (scalar), 1 = per-channel, 2+ = per-block.
-    """
-    if not weight_scale_shapes:
-        return "no weight_scale tensors found"
-
-    def meaningful_dims(shape: list[int]) -> int:
-        return sum(1 for d in shape if d > 1)
-
-    per_tensor = sum(1 for s in weight_scale_shapes if meaningful_dims(s) == 0)
-    per_channel = sum(1 for s in weight_scale_shapes if meaningful_dims(s) == 1)
-    per_block = sum(1 for s in weight_scale_shapes if meaningful_dims(s) >= 2)
-    total = len(weight_scale_shapes)
-    if per_tensor == total:
-        return "per-tensor (all scalar scales)"
-    if per_channel == total:
-        return "per-channel (1 meaningful dim)"
-    if per_block == total:
-        return "per-block (2+ meaningful dims — e.g. [M//bm, 1, N//bn, 1] for tiles)"
-    return f"mixed: per-tensor={per_tensor}, per-channel={per_channel}, per-block={per_block} of {total}"
-
-
-def _check_safetensors(transformer_dir: Path) -> int:
-    """Returns 0 on pass, 1 on fail. Reads on-disk dtype from the safetensors header."""
-    files = sorted(transformer_dir.glob("*.safetensors"))
-    if not files:
-        print(f"[FAIL] No *.safetensors in {transformer_dir}.")
-        return 1
-
-    header_dtype_counts: Counter[str] = Counter()
-    sample_quant_weight_keys: list[str] = []
-    sample_scale_keys: list[str] = []
-    weight_scale_shapes: list[list[int]] = []
-    sample_weight_scale_entries: list[tuple[str, list[int]]] = []
-
-    for f in files:
-        try:
-            header = _read_safetensors_header(f)
-        except Exception as exc:
-            print(f"[B] WARN — could not parse header of {f}: {exc}")
-            continue
-        for k, info in header.items():
-            dtype = info.get("dtype", "?")
-            header_dtype_counts[dtype] += 1
-            # FP8 stores weights as F8_E4M3 directly.
-            if dtype.startswith("F8") and k.endswith(".weight") and len(sample_quant_weight_keys) < 5:
-                sample_quant_weight_keys.append(k)
-            if k.endswith(("_scale", ".weight_scale", ".input_scale", "_scale_inv")) and len(sample_scale_keys) < 5:
-                sample_scale_keys.append(k)
-            if k.endswith(".weight_scale"):
-                weight_scale_shapes.append(info.get("shape", []))
-                if len(sample_weight_scale_entries) < 5:
-                    sample_weight_scale_entries.append((k, info.get("shape", [])))
-
-    print(f"\n[B] On-disk dtype counts across {len(files)} safetensors file(s) (from header, not get_tensor):")
-    for dtype, count in sorted(header_dtype_counts.items(), key=lambda kv: -kv[1]):
-        marker = "  <-- FP8" if dtype.startswith("F8") else ""
-        print(f"    {dtype:10s} {count:>6d}{marker}")
-
-    quant_count = sum(c for d, c in header_dtype_counts.items() if d.startswith("F8"))
-    if quant_count == 0:
-        print("[B] FAIL — no FP8 tensors on disk. Calibration likely did not actually quantize the weights.")
-        return 1
-
-    print(f"[B] PASS — {quant_count} FP8 tensors stored on disk.")
-    if sample_quant_weight_keys:
-        print(f"    sample quantized weight tensors: {sample_quant_weight_keys[:3]}")
-    if sample_scale_keys:
-        print(f"    sample scale tensors:            {sample_scale_keys[:3]}")
-    print("    (Note: torch's get_tensor() may return these as bf16 views on some versions —")
-    print("     irrelevant; vLLM's loader uses native FP8 ops.)")
-
-    # Weight-scale granularity — per-tensor (scalar) vs per-channel (1-D) vs per-block (N-D).
-    print(f"\n    weight_scale granularity: {_classify_weight_scale_granularity(weight_scale_shapes)}")
-    for key, shape in sample_weight_scale_entries[:3]:
-        print(f"      {key}: shape {shape}")
-    return 0
-
-
-def _disk_size_gib(p: Path) -> float:
-    return sum(f.stat().st_size for f in p.rglob("*") if f.is_file()) / (1024**3)
-
-
-def _check_size_vs_baseline(transformer_dir: Path, baseline: str | None) -> int:
-    """Returns 0 always (informational only)."""
-    quant_size = _disk_size_gib(transformer_dir)
-    print(f"\n[C] FP8 transformer disk size: {quant_size:.2f} GiB")
-
-    if baseline is None:
-        print("[C] SKIP — pass --baseline <path or HF id> to compare against BF16.")
-        return 0
-
-    baseline_path = Path(baseline)
-    if not baseline_path.exists():
-        # Try HF download.
-        try:
-            from huggingface_hub import snapshot_download
-        except ImportError:
-            print("[C] SKIP — huggingface_hub not installed and baseline not a local path.")
-            return 0
-        print(f"    Downloading baseline transformer from HF: {baseline}")
-        baseline_path = Path(snapshot_download(baseline, allow_patterns=["transformer/*"]))
-
-    bf16_dir = baseline_path / "transformer" if (baseline_path / "transformer").exists() else baseline_path
-    bf16_size = _disk_size_gib(bf16_dir)
-    if bf16_size == 0:
-        print(f"[C] WARN — baseline transformer dir empty: {bf16_dir}")
-        return 0
-
-    # Expected reduction for FP8: ~50%.
-    min_reduction = 30
-    reduction = (1 - quant_size / bf16_size) * 100
-    print(f"[C] BF16 baseline transformer disk size: {bf16_size:.2f} GiB ({bf16_dir})")
-    print(f"[C] Disk reduction: {reduction:.1f}%  (FP8 transformer is {quant_size / bf16_size:.0%} of BF16)")
-    if reduction < min_reduction:
-        print(
-            f"[C] WARN — FP8 should typically reduce disk by ~40-50%; <{min_reduction}% suggests partial quantization."
-        )
-    return 0
-
-
-def main() -> None:
-    p = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
-    p.add_argument("--output", required=True, help="Path to the exported ModelOpt FP8 checkpoint root.")
-    p.add_argument(
-        "--baseline",
-        default=None,
-        help="Optional BF16 baseline (local diffusers dir or HF id) for disk-size comparison.",
-    )
-    args = p.parse_args()
-
-    out_root = Path(args.output).expanduser().resolve()
-    transformer_dir = out_root / "transformer"
-    if not transformer_dir.exists():
-        print(f"[FAIL] {transformer_dir} does not exist.")
-        sys.exit(1)
-
-    print(f"Checking: {out_root}\n")
-
-    fail = 0
-    config_status, _quant_algo = _check_config(transformer_dir)
-    fail |= config_status
-    fail |= _check_safetensors(transformer_dir)
-    _check_size_vs_baseline(transformer_dir, args.baseline)
-
-    print()
-    if fail == 0:
-        print("=" * 60)
-        print("ALL CHECKS PASSED — checkpoint looks ready for vllm-omni serving.")
-    elif fail == 1:
-        print("=" * 60)
-        print("FAILURES detected — calibration may need to be re-run.")
-        sys.exit(1)
-    else:
-        print("=" * 60)
-        print("WARNINGS only — checkpoint may serve but with caveats. See [A] above.")
-
-
-if __name__ == "__main__":
-    main()
@@ -1,448 +0,0 @@
-#!/usr/bin/env python3
-# SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-"""Quantize HunyuanVideo-1.5 (480p T2V) to a ModelOpt FP8 Hugging Face checkpoint.
-
-Calibrates the DiT transformer using a small video prompt set and exports a
-diffusers-style directory whose transformer carries ModelOpt FP8 metadata.
-The exported checkpoint is consumable by vllm-omni's ModelOpt FP8 adapter
-(see vllm_omni/diffusion/model_loader/checkpoint_adapters/modelopt.py).
-
-Layers kept full precision match the #2728 / #2795 pattern: modulation,
-AdaLayerNorm, entry/exit projections, embeddings, the token refiner path,
-and final proj_out. MHA quantizers are off by default; HV-1.5 self-attention
-empirically degrades under FP8 (see #2920 ablation).
-
-Example:
-    python examples/quantization/quantize_hunyuanvideo_15_modelopt_fp8.py \\
-        --model hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-480p_t2v \\
-        --output ./hv15-480p-modelopt-fp8 \\
-        --overwrite
-"""
-
-from __future__ import annotations
-
-import argparse
-import copy
-import json
-import re
-import shutil
-import sys
-from pathlib import Path
-from typing import Any
-
-import torch
-from diffusers import DiffusionPipeline
-
-DEFAULT_PROMPTS = [
-    "A dog running across a field of golden wheat.",
-    "An astronaut riding a horse across the surface of Mars, red dust swirling, cinematic wide shot.",
-    "A hummingbird hovering in front of a vibrant red flower, slow motion, macro shot.",
-    "A crackling campfire at night under a starry sky, sparks rising into the dark.",
-    "An underwater shot of a coral reef with tropical fish swimming by, sun rays piercing the water.",
-    "A close-up of a blooming rose covered in morning dew, soft natural light.",
-    "A peaceful mountain village at dawn, mist rolling over the rooftops, cinematic establishing shot.",
-    "A skateboarder doing a kickflip in an urban plaza, slow motion, golden hour lighting.",
-]
-
-
-def _build_parser() -> argparse.ArgumentParser:
-    p = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
-    p.add_argument("--model", required=True, help="Input HV-1.5 diffusers directory or HF id.")
-    p.add_argument("--output", required=True, help="Output directory for the ModelOpt FP8 checkpoint.")
-    p.add_argument("--dtype", choices=("bfloat16", "float16"), default="bfloat16")
-    p.add_argument("--height", type=int, default=480)
-    p.add_argument("--width", type=int, default=832)
-    p.add_argument(
-        "--num-frames",
-        type=int,
-        default=33,
-        help="Frames per calibration sample. 33 matches the typical short benchmark.",
-    )
-    p.add_argument("--guidance-scale", type=float, default=6.0)
-    p.add_argument(
-        "--calib-steps",
-        type=int,
-        default=10,
-        help="Denoising steps per calibration prompt (10 is enough for amax statistics).",
-    )
-    p.add_argument("--calib-size", type=int, default=8, help="How many prompts to use for calibration.")
-    p.add_argument("--seed", type=int, default=42)
-    p.add_argument(
-        "--prompt",
-        action="append",
-        default=[],
-        help="Custom calibration prompt. Repeat to provide multiple.",
-    )
-    p.add_argument(
-        "--quantize-mha",
-        action="store_true",
-        help="Enable FP8 attention K/V/softmax quantizers. Off by default — empirically degrades HV-1.5 video output.",
-    )
-    p.add_argument(
-        "--weight-block-size",
-        type=str,
-        default=None,
-        help="Per-block weight quantization as 'M,N' (e.g. '128,128' for 128x128 tiles). "
-        "Default: per-tensor (one scale per linear). Block-wise typically gives tighter quality at "
-        "negligible memory cost. Static FP8 is exempt from upstream vLLM's online block-wise gate.",
-    )
-    p.add_argument("--overwrite", action="store_true", help="Replace an existing output directory.")
-    return p
-
-
-def _parse_block_size(spec: str | None) -> list[int] | None:
-    if spec is None:
-        return None
-    parts = [int(x) for x in spec.split(",") if x.strip()]
-    if len(parts) != 2:
-        raise SystemExit(f"--weight-block-size must be 'M,N' (2 ints), got {spec!r}")
-    return parts
-
-
-def _require_modelopt() -> Any:
-    try:
-        import modelopt.torch.quantization as mtq
-    except ModuleNotFoundError as exc:
-        raise SystemExit(
-            "NVIDIA ModelOpt is not installed. Install with:\n"
-            "  pip install 'nvidia-modelopt[all]'\n"
-            f"Original error: {exc}"
-        ) from exc
-    return mtq
-
-
-def _ensure_paths(args: argparse.Namespace) -> tuple[str, Path]:
-    model_path = args.model
-    output_dir = Path(args.output).expanduser().resolve()
-    if output_dir.exists():
-        if not args.overwrite:
-            raise SystemExit(f"Output directory already exists: {output_dir}\nPass --overwrite to replace it.")
-        shutil.rmtree(output_dir)
-    return model_path, output_dir
-
-
-def _select_dtype(name: str) -> torch.dtype:
-    return {"bfloat16": torch.bfloat16, "float16": torch.float16}[name]
-
-
-def _build_prompts(args: argparse.Namespace) -> list[str]:
-    prompts = args.prompt or DEFAULT_PROMPTS
-    if args.calib_size <= 0:
-        raise SystemExit("--calib-size must be positive.")
-    if len(prompts) < args.calib_size:
-        repeats = (args.calib_size + len(prompts) - 1) // len(prompts)
-        prompts = (prompts * repeats)[: args.calib_size]
-    return prompts[: args.calib_size]
-
-
-# Layers to KEEP at full precision (mirror of the #2920 wiring + #2728/#2795 skip pattern).
-# - x_embedder, image_embedder, context_embedder*, time_embed*, cond_type_embed: entry/embedding
-# - norm_out, norm1*.linear, norm1_context*.linear, norm2*, norm2_context*: AdaLayerNorm modulation
-# - proj_out: final output projection
-# - token_refiner*: text-encoder refinement uses diffusers raw nn.Linear
-def _filter_func_hv15(name: str) -> bool:
-    pattern = re.compile(
-        r"(proj_out.*|"
-        r".*(x_embedder|image_embedder|context_embedder|context_embedder_2|"
-        r"time_embed|cond_type_embed|"
-        r"norm_out|norm1\.linear|norm1_context\.linear|norm2|norm2_context|"
-        r"token_refiner).*)"
-    )
-    return pattern.match(name) is not None
-
-
-def _mha_filter_func(name: str) -> bool:
-    pattern = re.compile(
-        r".*(q_bmm_quantizer|k_bmm_quantizer|v_bmm_quantizer|softmax_quantizer|bmm2_output_quantizer).*"
-    )
-    return pattern.match(name) is not None
-
-
-def _disable_known_problematic_quantizers(mtq: Any, backbone: torch.nn.Module, *, quantize_mha: bool) -> None:
-    if not hasattr(mtq, "disable_quantizer"):
-        return
-    mtq.disable_quantizer(backbone, _filter_func_hv15)
-    if not quantize_mha:
-        mtq.disable_quantizer(backbone, _mha_filter_func)
-
-
-def _load_pipeline(model_path: str, dtype: torch.dtype) -> DiffusionPipeline:
-    pipe = DiffusionPipeline.from_pretrained(model_path, torch_dtype=dtype)
-    if hasattr(pipe, "set_progress_bar_config"):
-        pipe.set_progress_bar_config(disable=True)
-    pipe.to("cuda")
-    return pipe
-
-
-def _build_forward_loop(pipe: DiffusionPipeline, args: argparse.Namespace, prompts: list[str]):
-    generator = torch.Generator(device="cuda")
-
-    # Try to set guidance on the pipeline's guider object up front (modern
-    # diffusers HV-1.5 uses a Guider abstraction, not a per-call kwarg). Falls
-    # back silently — calibration uses whatever default the pipeline ships with.
-    guider = getattr(pipe, "guider", None)
-    if guider is not None and hasattr(guider, "guidance_scale"):
-        try:
-            guider.guidance_scale = args.guidance_scale
-        except Exception:
-            pass
-
-    base_kwargs = dict(
-        height=args.height,
-        width=args.width,
-        num_frames=args.num_frames,
-        num_inference_steps=args.calib_steps,
-        output_type="latent",
-    )
-
-    def forward_loop(*_unused_args, **_unused_kwargs) -> None:
-        with torch.inference_mode():
-            for idx, prompt in enumerate(prompts):
-                generator.manual_seed(args.seed + idx)
-                # Try with guidance_scale first; fall back without on TypeError
-                # for pipelines (like HV-1.5) that take CFG via guider config.
-                try:
-                    pipe(prompt=prompt, generator=generator, guidance_scale=args.guidance_scale, **base_kwargs)
-                except TypeError as exc:
-                    if "guidance_scale" not in str(exc):
-                        raise
-                    pipe(prompt=prompt, generator=generator, **base_kwargs)
-
-    return forward_loop
-
-
-def _summarize_export(output_dir: Path) -> None:
-    cfg_path = output_dir / "transformer" / "config.json"
-    if not cfg_path.exists():
-        print(f"[warn] {cfg_path} missing.", file=sys.stderr)
-        return
-    with cfg_path.open(encoding="utf-8") as f:
-        cfg = json.load(f)
-    qc = cfg.get("quantization_config")
-    if not isinstance(qc, dict):
-        print("[warn] No quantization_config in transformer/config.json.", file=sys.stderr)
-        return
-    print("Export summary:")
-    print(f"  quant_method: {qc.get('quant_method')}")
-    print(f"  quant_algo:   {qc.get('quant_algo')}")
-    producer = qc.get("producer")
-    if isinstance(producer, dict):
-        print(f"  producer:     {producer.get('name')} {producer.get('version')}")
-    print(f"  config path:  {cfg_path}")
-
-
-def _force_export_quantized_weights(backbone: torch.nn.Module, dtype: torch.dtype) -> int:
-    """Convert in-memory weights of quantized modules to actual FP8 storage.
-
-    `export_hf_checkpoint` skips this step for unknown model types (HV-1.5 isn't
-    in ModelOpt's recognized-model registry), so we must call the per-weight
-    export helper ourselves. Same workaround as the HunyuanImage-3 calibration
-    helper.
-    """
-    from modelopt.torch.export.quant_utils import (
-        QUANTIZATION_NONE,
-        get_quantization_format,
-        quantizer_attr_names,
-        weight_attr_names,
-    )
-    from modelopt.torch.export.unified_export_hf import _export_quantized_weight
-
-    exported = 0
-    for name, module in backbone.named_modules():
-        try:
-            quantization_format = get_quantization_format(module)
-        except Exception as exc:
-            print(f"[warn] Could not inspect quantization format for {name}: {exc}", file=sys.stderr)
-            continue
-        if quantization_format == QUANTIZATION_NONE:
-            continue
-        for weight_name in weight_attr_names(module):
-            quantizer_attrs = quantizer_attr_names(weight_name)
-            weight_quantizer = getattr(module, quantizer_attrs.weight_quantizer, None)
-            if weight_quantizer is None or not getattr(weight_quantizer, "is_enabled", False):
-                continue
-            _export_quantized_weight(module, dtype, weight_name)
-            exported += 1
-    return exported
-
-
-def _hv15_quant_config_block(weight_block_size: list[int] | None = None) -> dict:
-    """Mirror ModelOpt FP8 metadata expected by vllm-omni's adapter (#2913).
-
-    Same shape as the HunyuanImage-3 author's _hunyuan_quant_config(). When
-    `weight_block_size` is given, advertise block-wise weight quantization in
-    the saved metadata (so consumers know to expect multi-element scale tensors).
-    """
-    weights_cfg: dict = {"dynamic": False, "num_bits": 8, "type": "float"}
-    if weight_block_size is not None:
-        weights_cfg["strategy"] = "block"
-        weights_cfg["block_structure"] = f"{weight_block_size[0]}x{weight_block_size[1]}"
-    return {
-        "config_groups": {
-            "group_0": {
-                "input_activations": {"dynamic": False, "num_bits": 8, "type": "float"},
-                "weights": weights_cfg,
-                "targets": ["Linear"],
-            }
-        },
-        "ignore": [
-            "context_embedder*",
-            "context_embedder_2*",
-            "cond_type_embed*",
-            "image_embedder*",
-            "norm1.linear*",
-            "norm1_context.linear*",
-            "norm2*",
-            "norm2_context*",
-            "norm_out*",
-            "proj_out*",
-            "time_embed*",
-            "token_refiner*",
-            "x_embedder*",
-        ],
-        "producer": {"name": "modelopt"},
-        "quant_algo": "FP8",
-        "quant_method": "modelopt",
-    }
-
-
-def _patch_quant_config(output_dir: Path, weight_block_size: list[int] | None = None) -> None:
-    """Inject quant_algo: FP8 + config_groups into transformer/config.json so
-    vllm-omni's adapter (#2913) recognises the checkpoint as ModelOpt FP8."""
-    cfg_path = output_dir / "transformer" / "config.json"
-    with cfg_path.open(encoding="utf-8") as f:
-        cfg = json.load(f)
-
-    new_qc = _hv15_quant_config_block(weight_block_size=weight_block_size)
-    existing = cfg.get("quantization_config")
-    if isinstance(existing, dict):
-        producer = existing.get("producer")
-        if isinstance(producer, dict):
-            new_qc["producer"] = producer
-
-    cfg["quantization_config"] = new_qc
-    with cfg_path.open("w", encoding="utf-8") as f:
-        json.dump(cfg, f, indent=2)
-
-
-def _save_pipeline_with_fp8_transformer(
-    pipe: DiffusionPipeline,
-    model_path: str,
-    output_dir: Path,
-    max_shard_size: str = "5GB",
-) -> None:
-    """Save the pipeline with the (now FP8) transformer.
-
-    Copies the source directory verbatim except for `transformer/`, then
-    saves the transformer with quantizers hidden so the state dict contains
-    only the FP8 weights + scale tensors.
-    """
-    from modelopt.torch.export.diffusers_utils import hide_quantizers_from_state_dict
-
-    src = Path(model_path)
-    if not src.exists():
-        from huggingface_hub import snapshot_download
-
-        src = Path(snapshot_download(model_path))
-
-    if output_dir.exists():
-        shutil.rmtree(output_dir)
-    shutil.copytree(src, output_dir, ignore=shutil.ignore_patterns("transformer"))
-
-    transformer_out = output_dir / "transformer"
-    # `hide_quantizers_from_state_dict` walks named_modules(); pass the actual
-    # nn.Module (transformer), not the diffusers Pipeline wrapper.
-    with hide_quantizers_from_state_dict(pipe.transformer):
-        pipe.transformer.save_pretrained(
-            str(transformer_out),
-            safe_serialization=True,
-            max_shard_size=max_shard_size,
-        )
-
-
-def main() -> None:
-    args = _build_parser().parse_args()
-    if not torch.cuda.is_available():
-        raise SystemExit("CUDA is required for ModelOpt FP8 quantization.")
-
-    mtq = _require_modelopt()
-    model_path, output_dir = _ensure_paths(args)
-    dtype = _select_dtype(args.dtype)
-    prompts = _build_prompts(args)
-
-    print("Quantization plan:")
-    weight_block_size = _parse_block_size(args.weight_block_size)
-
-    print(f"  input:           {args.model}")
-    print(f"  output:          {output_dir}")
-    print(f"  dtype:           {dtype}")
-    print(f"  height/width:    {args.height}x{args.width}")
-    print(f"  num_frames:      {args.num_frames}")
-    print(f"  calib_size:      {len(prompts)}")
-    print(f"  calib_steps:     {args.calib_steps}")
-    print(f"  quantize_mha:    {args.quantize_mha}")
-    print(
-        f"  weight strategy: {'block-wise ' + str(weight_block_size) if weight_block_size else 'per-tensor (default)'}"
-    )
-
-    pipe = _load_pipeline(model_path, dtype)
-    backbone = pipe.transformer
-
-    quant_config = copy.deepcopy(mtq.FP8_DEFAULT_CFG)
-    if weight_block_size is not None:
-        # Switch from per-tensor (default) to block-wise weight quantization.
-        # ModelOpt's wildcard "*weight_quantizer" matches every linear's weight quantizer.
-        quant_config["quant_cfg"]["*weight_quantizer"] = {
-            "num_bits": (4, 3),  # E4M3 (FP8 weights, same as default)
-            "block_sizes": {-1: weight_block_size[1], -2: weight_block_size[0]},
-        }
-        print(
-            f"  -> overriding weight quantizer with block_sizes={weight_block_size} "
-            f"({weight_block_size[0]}x{weight_block_size[1]} tiles)"
-        )
-
-    forward_loop = _build_forward_loop(pipe, args, prompts)
-    quantized = mtq.quantize(backbone, quant_config, forward_loop)
-    if quantized is not None:
-        pipe.transformer = quantized
-        backbone = quantized
-
-    _disable_known_problematic_quantizers(mtq, backbone, quantize_mha=args.quantize_mha)
-
-    print("\nForcing FP8 weight serialization (HV-1.5 isn't in ModelOpt's recognized-model registry,")
-    print("so we have to call the per-weight export helper ourselves)...")
-    exported = _force_export_quantized_weights(backbone, dtype)
-    print(f"  -> {exported} weights converted to FP8 in memory")
-    if exported == 0:
-        raise SystemExit(
-            "No quantized weights were exported. Calibration may have skipped every layer "
-            "(check the disable_quantizer regex) or `mtq.quantize` did not actually wrap any "
-            "weight quantizers."
-        )
-
-    print("\nSaving pipeline with FP8 transformer...")
-    _save_pipeline_with_fp8_transformer(pipe, model_path, output_dir)
-    _patch_quant_config(output_dir, weight_block_size=weight_block_size)
-    print(f"Saved to: {output_dir}")
-    _summarize_export(output_dir)
-
-    print("\nNext: validate the checkpoint with vllm-omni:")
-    print(
-        "  python examples/offline_inference/text_to_video/text_to_video.py \\\n"
-        f"    --model {output_dir} \\\n"
-        "    --quantization fp8 \\\n"
-        "    --prompt 'A dog running across a field of golden wheat.' \\\n"
-        f"    --height {args.height} --width {args.width} --num-frames {args.num_frames} \\\n"
-        "    --num-inference-steps 30 --guidance-scale 6.0 --seed 42 \\\n"
-        "    --output outputs/hv15_modelopt_fp8.mp4 \\\n"
-        "    --enforce-eager"
-    )
-    print(
-        "\n  (--quantization fp8 is auto-upgraded to ModelOpt FP8 at runtime because the "
-        "checkpoint's config.json has modelopt metadata.)"
-    )
-
-
-if __name__ == "__main__":
-    main()
@@ -1,471 +0,0 @@
-#!/usr/bin/env python3
-# SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-"""Quantize Wan2.2 (TI2V-5B, 704x1280 T2V) to a ModelOpt FP8 Hugging Face checkpoint.
-
-Calibrates the DiT transformer using a small video prompt set and exports a
-diffusers-style directory whose transformer carries ModelOpt FP8 metadata.
-The exported checkpoint is consumable by vllm-omni's ModelOpt FP8 adapter
-(see vllm_omni/diffusion/model_loader/checkpoint_adapters/modelopt.py).
-
-Layers kept full precision match the #2728 / #2795 pattern: condition embedder
-(time/text/image), patch embedding, modulation (scale_shift_table), final
-norm + proj_out, and sequence-parallel helpers. All attention + FFN linears
-are quantized — static calibration handles the numerics that online FP8
-couldn't (see #2920 ablation).
-
-Default target is `Wan-AI/Wan2.2-TI2V-5B-Diffusers`, the dense 5B variant that
-fits 80GB BF16. The A14B MoE variants need 2+ GPUs and are out of scope here.
-
-Example:
-    python examples/quantization/quantize_wan2_2_modelopt_fp8.py \\
-        --model Wan-AI/Wan2.2-TI2V-5B-Diffusers \\
-        --output ./wan22-ti2v-modelopt-fp8 \\
-        --overwrite
-"""
-
-from __future__ import annotations
-
-import argparse
-import copy
-import json
-import re
-import shutil
-import sys
-from pathlib import Path
-from typing import Any
-
-import torch
-from diffusers import DiffusionPipeline
-
-DEFAULT_PROMPTS = [
-    "A dog running across a field of golden wheat.",
-    "An astronaut riding a horse across the surface of Mars, red dust swirling, cinematic wide shot.",
-    "A hummingbird hovering in front of a vibrant red flower, slow motion, macro shot.",
-    "A crackling campfire at night under a starry sky, sparks rising into the dark.",
-    "An underwater shot of a coral reef with tropical fish swimming by, sun rays piercing the water.",
-    "A close-up of a blooming rose covered in morning dew, soft natural light.",
-    "A peaceful mountain village at dawn, mist rolling over the rooftops, cinematic establishing shot.",
-    "A skateboarder doing a kickflip in an urban plaza, slow motion, golden hour lighting.",
-]
-
-
-def _build_parser() -> argparse.ArgumentParser:
-    p = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
-    p.add_argument("--model", required=True, help="Input Wan2.2 diffusers directory or HF id.")
-    p.add_argument("--output", required=True, help="Output directory for the ModelOpt FP8 checkpoint.")
-    p.add_argument("--dtype", choices=("bfloat16", "float16"), default="bfloat16")
-    p.add_argument("--height", type=int, default=704, help="Calibration video height (Wan2.2 TI2V-5B native: 704).")
-    p.add_argument("--width", type=int, default=1280, help="Calibration video width (Wan2.2 TI2V-5B native: 1280).")
-    p.add_argument(
-        "--num-frames",
-        type=int,
-        default=49,
-        help="Frames per calibration sample. 49 matches the typical short benchmark; "
-        "use 17 to reduce memory pressure during calibration.",
-    )
-    p.add_argument("--guidance-scale", type=float, default=5.0)
-    p.add_argument(
-        "--calib-steps",
-        type=int,
-        default=10,
-        help="Denoising steps per calibration prompt (10 is enough for amax statistics).",
-    )
-    p.add_argument("--calib-size", type=int, default=8, help="How many prompts to use for calibration.")
-    p.add_argument("--seed", type=int, default=42)
-    p.add_argument(
-        "--prompt",
-        action="append",
-        default=[],
-        help="Custom calibration prompt. Repeat to provide multiple.",
-    )
-    p.add_argument(
-        "--quantize-mha",
-        action="store_true",
-        help="Enable FP8 attention K/V/softmax quantizers. Off by default — Wan2.2's long attention "
-        "sequences amplified FP8 drift in the online ablation (see #2920).",
-    )
-    p.add_argument(
-        "--weight-block-size",
-        type=str,
-        default=None,
-        help="Per-block weight quantization as 'M,N' (e.g. '128,128'). Default per-tensor. "
-        "Note: vllm-omni's ModelOpt adapter may not yet dispatch block-wise scales — check #2924 "
-        "for the HV-1.5 investigation status before relying on this.",
-    )
-    p.add_argument("--overwrite", action="store_true", help="Replace an existing output directory.")
-    return p
-
-
-def _parse_block_size(spec: str | None) -> list[int] | None:
-    if spec is None:
-        return None
-    parts = [int(x) for x in spec.split(",") if x.strip()]
-    if len(parts) != 2:
-        raise SystemExit(f"--weight-block-size must be 'M,N' (2 ints), got {spec!r}")
-    return parts
-
-
-def _require_modelopt() -> Any:
-    try:
-        import modelopt.torch.quantization as mtq
-    except ModuleNotFoundError as exc:
-        raise SystemExit(
-            "NVIDIA ModelOpt is not installed. Install with:\n"
-            "  pip install 'nvidia-modelopt[all]'\n"
-            f"Original error: {exc}"
-        ) from exc
-    return mtq
-
-
-def _ensure_paths(args: argparse.Namespace) -> tuple[str, Path]:
-    model_path = args.model
-    output_dir = Path(args.output).expanduser().resolve()
-    if output_dir.exists():
-        if not args.overwrite:
-            raise SystemExit(f"Output directory already exists: {output_dir}\nPass --overwrite to replace it.")
-        shutil.rmtree(output_dir)
-    return model_path, output_dir
-
-
-def _select_dtype(name: str) -> torch.dtype:
-    return {"bfloat16": torch.bfloat16, "float16": torch.float16}[name]
-
-
-def _build_prompts(args: argparse.Namespace) -> list[str]:
-    prompts = args.prompt or DEFAULT_PROMPTS
-    if args.calib_size <= 0:
-        raise SystemExit("--calib-size must be positive.")
-    if len(prompts) < args.calib_size:
-        repeats = (args.calib_size + len(prompts) - 1) // len(prompts)
-        prompts = (prompts * repeats)[: args.calib_size]
-    return prompts[: args.calib_size]
-
-
-# Layers to KEEP at full precision. Wan2.2's module naming:
-# - condition_embedder: time_embedder, time_proj, text_embedder, image_embedder (I2V)
-# - patch_embedding: Conv3dLayer (already not Linear, belt-and-suspenders skip)
-# - scale_shift_table: nn.Parameter modulation (not Linear, but pattern guard)
-# - norm_out: AdaLayerNorm final
-# - proj_out: final nn.Linear
-# - timestep_proj_prepare / output_scale_shift_prepare: SP helpers
-def _filter_func_wan22(name: str) -> bool:
-    pattern = re.compile(
-        r"(proj_out.*|"
-        r".*(condition_embedder|patch_embedding|"
-        r"norm_out|scale_shift_table|"
-        r"timestep_proj_prepare|output_scale_shift_prepare).*)"
-    )
-    return pattern.match(name) is not None
-
-
-def _mha_filter_func(name: str) -> bool:
-    pattern = re.compile(
-        r".*(q_bmm_quantizer|k_bmm_quantizer|v_bmm_quantizer|softmax_quantizer|bmm2_output_quantizer).*"
-    )
-    return pattern.match(name) is not None
-
-
-def _disable_known_problematic_quantizers(mtq: Any, backbone: torch.nn.Module, *, quantize_mha: bool) -> None:
-    if not hasattr(mtq, "disable_quantizer"):
-        return
-    mtq.disable_quantizer(backbone, _filter_func_wan22)
-    if not quantize_mha:
-        mtq.disable_quantizer(backbone, _mha_filter_func)
-
-
-def _load_pipeline(model_path: str, dtype: torch.dtype) -> DiffusionPipeline:
-    pipe = DiffusionPipeline.from_pretrained(model_path, torch_dtype=dtype)
-    if hasattr(pipe, "set_progress_bar_config"):
-        pipe.set_progress_bar_config(disable=True)
-    pipe.to("cuda")
-    return pipe
-
-
-def _build_forward_loop(pipe: DiffusionPipeline, args: argparse.Namespace, prompts: list[str]):
-    generator = torch.Generator(device="cuda")
-
-    # Try setting guidance on the pipeline's guider if present (newer diffusers APIs).
-    guider = getattr(pipe, "guider", None)
-    if guider is not None and hasattr(guider, "guidance_scale"):
-        try:
-            guider.guidance_scale = args.guidance_scale
-        except Exception:
-            pass
-
-    base_kwargs = dict(
-        height=args.height,
-        width=args.width,
-        num_frames=args.num_frames,
-        num_inference_steps=args.calib_steps,
-        output_type="latent",
-    )
-
-    def forward_loop(*_unused_args, **_unused_kwargs) -> None:
-        with torch.inference_mode():
-            for idx, prompt in enumerate(prompts):
-                generator.manual_seed(args.seed + idx)
-                # Try with guidance_scale first; fall back without on TypeError
-                # for pipelines that take CFG via guider config only.
-                try:
-                    pipe(prompt=prompt, generator=generator, guidance_scale=args.guidance_scale, **base_kwargs)
-                except TypeError as exc:
-                    if "guidance_scale" not in str(exc):
-                        raise
-                    pipe(prompt=prompt, generator=generator, **base_kwargs)
-
-    return forward_loop
-
-
-def _summarize_export(output_dir: Path) -> None:
-    cfg_path = output_dir / "transformer" / "config.json"
-    if not cfg_path.exists():
-        print(f"[warn] {cfg_path} missing.", file=sys.stderr)
-        return
-    with cfg_path.open(encoding="utf-8") as f:
-        cfg = json.load(f)
-    qc = cfg.get("quantization_config")
-    if not isinstance(qc, dict):
-        print("[warn] No quantization_config in transformer/config.json.", file=sys.stderr)
-        return
-    print("Export summary:")
-    print(f"  quant_method: {qc.get('quant_method')}")
-    print(f"  quant_algo:   {qc.get('quant_algo')}")
-    producer = qc.get("producer")
-    if isinstance(producer, dict):
-        print(f"  producer:     {producer.get('name')} {producer.get('version')}")
-    print(f"  config path:  {cfg_path}")
-
-
-def _force_export_quantized_weights(backbone: torch.nn.Module, dtype: torch.dtype) -> int:
-    """Convert in-memory weights of quantized modules to actual FP8 storage.
-
-    `export_hf_checkpoint` skips this step for unknown model types (Wan2.2 isn't
-    in ModelOpt's recognized-model registry), so we must call the per-weight
-    export helper ourselves. Same workaround as the HunyuanVideo-1.5 / HunyuanImage-3
-    calibration helpers.
-    """
-    from modelopt.torch.export.quant_utils import (
-        QUANTIZATION_NONE,
-        get_quantization_format,
-        quantizer_attr_names,
-        weight_attr_names,
-    )
-    from modelopt.torch.export.unified_export_hf import _export_quantized_weight
-
-    exported = 0
-    for name, module in backbone.named_modules():
-        try:
-            quantization_format = get_quantization_format(module)
-        except Exception as exc:
-            print(f"[warn] Could not inspect quantization format for {name}: {exc}", file=sys.stderr)
-            continue
-        if quantization_format == QUANTIZATION_NONE:
-            continue
-        for weight_name in weight_attr_names(module):
-            quantizer_attrs = quantizer_attr_names(weight_name)
-            weight_quantizer = getattr(module, quantizer_attrs.weight_quantizer, None)
-            if weight_quantizer is None or not getattr(weight_quantizer, "is_enabled", False):
-                continue
-            _export_quantized_weight(module, dtype, weight_name)
-            exported += 1
-    return exported
-
-
-def _wan22_quant_config_block(weight_block_size: list[int] | None = None) -> dict:
-    """Mirror ModelOpt FP8 metadata expected by vllm-omni's adapter (#2913)."""
-    weights_cfg: dict = {"dynamic": False, "num_bits": 8, "type": "float"}
-    if weight_block_size is not None:
-        weights_cfg["strategy"] = "block"
-        weights_cfg["block_structure"] = f"{weight_block_size[0]}x{weight_block_size[1]}"
-    return {
-        "config_groups": {
-            "group_0": {
-                "input_activations": {"dynamic": False, "num_bits": 8, "type": "float"},
-                "weights": weights_cfg,
-                "targets": ["Linear"],
-            }
-        },
-        "ignore": [
-            "condition_embedder*",
-            "norm_out*",
-            "output_scale_shift_prepare*",
-            "patch_embedding*",
-            "proj_out*",
-            "scale_shift_table*",
-            "timestep_proj_prepare*",
-        ],
-        "producer": {"name": "modelopt"},
-        "quant_algo": "FP8",
-        "quant_method": "modelopt",
-    }
-
-
-def _patch_quant_config(output_dir: Path, weight_block_size: list[int] | None = None) -> None:
-    """Inject quant_algo: FP8 + config_groups into each transformer's config.json
-    so vllm-omni's adapter (#2913) recognises the checkpoint as ModelOpt FP8.
-
-    Patches both ``transformer`` and ``transformer_2`` (A14B) when present."""
-    for sub in ("transformer", "transformer_2"):
-        cfg_path = output_dir / sub / "config.json"
-        if not cfg_path.exists():
-            continue
-        with cfg_path.open(encoding="utf-8") as f:
-            cfg = json.load(f)
-
-        new_qc = _wan22_quant_config_block(weight_block_size=weight_block_size)
-        existing = cfg.get("quantization_config")
-        if isinstance(existing, dict):
-            producer = existing.get("producer")
-            if isinstance(producer, dict):
-                new_qc["producer"] = producer
-
-        cfg["quantization_config"] = new_qc
-        with cfg_path.open("w", encoding="utf-8") as f:
-            json.dump(cfg, f, indent=2)
-
-
-def _list_transformers(pipe: DiffusionPipeline) -> list[tuple[str, torch.nn.Module]]:
-    """Return [(attr_name, module)] for every DiT on the pipeline.
-
-    Wan2.2 A14B has both ``transformer`` (low-noise) and ``transformer_2``
-    (high-noise). TI2V-5B has only ``transformer``.
-    """
-    out = []
-    for attr in ("transformer", "transformer_2"):
-        mod = getattr(pipe, attr, None)
-        if mod is not None:
-            out.append((attr, mod))
-    return out
-
-
-def _save_pipeline_with_fp8_transformer(
-    pipe: DiffusionPipeline,
-    model_path: str,
-    output_dir: Path,
-    max_shard_size: str = "5GB",
-) -> None:
-    """Copy source dir verbatim minus transformer dirs, then save each quantized DiT.
-
-    Handles A14B's dual transformer (``transformer`` + ``transformer_2``) as well
-    as the single-DiT TI2V-5B.
-    """
-    from modelopt.torch.export.diffusers_utils import hide_quantizers_from_state_dict
-
-    src = Path(model_path)
-    if not src.exists():
-        from huggingface_hub import snapshot_download
-
-        src = Path(snapshot_download(model_path))
-
-    if output_dir.exists():
-        shutil.rmtree(output_dir)
-    shutil.copytree(src, output_dir, ignore=shutil.ignore_patterns("transformer", "transformer_2"))
-
-    for attr, backbone in _list_transformers(pipe):
-        transformer_out = output_dir / attr
-        # Pass the nn.Module (transformer), not the Pipeline wrapper.
-        with hide_quantizers_from_state_dict(backbone):
-            backbone.save_pretrained(
-                str(transformer_out),
-                safe_serialization=True,
-                max_shard_size=max_shard_size,
-            )
-
-
-def main() -> None:
-    args = _build_parser().parse_args()
-    if not torch.cuda.is_available():
-        raise SystemExit("CUDA is required for ModelOpt FP8 quantization.")
-
-    mtq = _require_modelopt()
-    model_path, output_dir = _ensure_paths(args)
-    dtype = _select_dtype(args.dtype)
-    prompts = _build_prompts(args)
-    weight_block_size = _parse_block_size(args.weight_block_size)
-
-    print("Quantization plan:")
-    print(f"  input:           {args.model}")
-    print(f"  output:          {output_dir}")
-    print(f"  dtype:           {dtype}")
-    print(f"  height/width:    {args.height}x{args.width}")
-    print(f"  num_frames:      {args.num_frames}")
-    print(f"  calib_size:      {len(prompts)}")
-    print(f"  calib_steps:     {args.calib_steps}")
-    print(f"  quantize_mha:    {args.quantize_mha}")
-    print(
-        f"  weight strategy: {'block-wise ' + str(weight_block_size) if weight_block_size else 'per-tensor (default)'}"
-    )
-
-    pipe = _load_pipeline(model_path, dtype)
-    transformers = _list_transformers(pipe)
-    if not transformers:
-        raise SystemExit("Pipeline has no transformer or transformer_2 attribute.")
-    print(f"  found {len(transformers)} transformer(s): {', '.join(a for a, _ in transformers)}")
-
-    quant_config = copy.deepcopy(mtq.FP8_DEFAULT_CFG)
-    if weight_block_size is not None:
-        quant_config["quant_cfg"]["*weight_quantizer"] = {
-            "num_bits": (4, 3),
-            "block_sizes": {-1: weight_block_size[1], -2: weight_block_size[0]},
-        }
-        print(
-            f"  -> overriding weight quantizer with block_sizes={weight_block_size} "
-            f"({weight_block_size[0]}x{weight_block_size[1]} tiles)"
-        )
-
-    forward_loop = _build_forward_loop(pipe, args, prompts)
-
-    # Quantize each DiT. The forward_loop runs the full pipeline, so quantizers
-    # on every attached transformer (A14B: transformer + transformer_2) get
-    # calibrated during the same passes.
-    for attr, backbone in transformers:
-        print(f"\nQuantizing {attr}...")
-        quantized = mtq.quantize(backbone, quant_config, forward_loop)
-        if quantized is not None:
-            setattr(pipe, attr, quantized)
-        _disable_known_problematic_quantizers(mtq, getattr(pipe, attr), quantize_mha=args.quantize_mha)
-
-    # Export weights to FP8 only AFTER all transformers are calibrated. Exporting
-    # inside the loop would convert transformer's weights to real FP8, and the
-    # next transformer's forward_loop (full pipeline) would then fake-quantize an
-    # already-FP8 weight -> "fake_e4m3fy not implemented for Float8_e4m3fn".
-    print("\nForcing FP8 weight serialization (Wan2.2 isn't in ModelOpt's recognized-model registry,")
-    print("so we have to call the per-weight export helper ourselves)...")
-    total_exported = 0
-    for attr, backbone in _list_transformers(pipe):
-        exported = _force_export_quantized_weights(backbone, dtype)
-        print(f"  -> {exported} weights converted to FP8 in {attr}")
-        total_exported += exported
-
-    if total_exported == 0:
-        raise SystemExit(
-            "No quantized weights were exported. Calibration may have skipped every layer "
-            "(check the disable_quantizer regex) or `mtq.quantize` did not actually wrap any "
-            "weight quantizers."
-        )
-
-    print("\nSaving pipeline with FP8 transformer(s)...")
-    _save_pipeline_with_fp8_transformer(pipe, model_path, output_dir)
-    _patch_quant_config(output_dir, weight_block_size=weight_block_size)
-    print(f"Saved to: {output_dir}")
-    _summarize_export(output_dir)
-
-    print("\nNext: validate the checkpoint with vllm-omni:")
-    print(
-        "  python examples/offline_inference/text_to_video/text_to_video.py \\\n"
-        f"    --model {output_dir} \\\n"
-        "    --quantization fp8 \\\n"
-        "    --prompt 'A dog running across a field of golden wheat.' \\\n"
-        f"    --height {args.height} --width {args.width} --num-frames {args.num_frames} \\\n"
-        "    --num-inference-steps 30 --guidance-scale 5.0 --seed 42 \\\n"
-        "    --output outputs/wan22_modelopt_fp8.mp4"
-    )
-    print(
-        "\n  (--quantization fp8 is auto-upgraded to ModelOpt FP8 at runtime because the "
-        "checkpoint's config.json has modelopt metadata.)"
-    )
-
-
-if __name__ == "__main__":
-    main()
@@ -79,7 +79,8 @@ sharing the tokenizer, processor, text encoder, and VAEs from `FL2VA`.
 ### Multi-NPU: 768P combined-service configuration

 Use Ulysses sequence parallelism degree 8, text-encoder tensor parallelism
-degree 8, native tiled VAE patch parallelism degree 8, and layerwise offload:
+degree 8, native tiled VAE patch parallelism degree 8, and distributed
+layerwise offload:

 ```bash
 export ASCEND_RT_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
@@ -98,7 +99,7 @@ vllm serve "${MODEL}" \
  --usp 8 \
  --ring 1 \
  --text-encoder-tp-size 8 \
-  --enable-layerwise-offload \
+  --enable-distributed-layerwise-offload \
  --vae-parallel-mode tile \
  --vae-use-tiling \
  --vae-patch-parallel-size 8 \
@@ -112,6 +113,11 @@ H3 is CFG-distilled, so `--cfg-parallel-size` must remain 1.
 The same endpoint accepts `task=t2va`, `task=fl2va`, and `task=ref2va`; no
 partition restart is required. Layerwise offload applies to both DiTs.

+On Atlas 800I A3 (64 GB HBM per device) the combined service does not fit at
+768P without offloading or sharding: use distributed layerwise offload (as
+above) or HSDP — see
+[§ Memory and attention optimizations](#memory-and-attention-optimizations-a3).
+
 ### Optional optimizations

 Two independent optimizations may be enabled on top of the configuration
@@ -134,6 +140,63 @@ Keep `--ring 1` when using RainFusion: the `rf_v2` kernel ranks key blocks
 over the whole sequence, so ring parallelism would split away the keys it
 needs. Scale with `--usp` instead.

+## Memory and attention optimizations (A3)
+
+### Fitting 768P into 64 GB HBM: distributed layerwise offload or HSDP
+
+Each Atlas 800I A3 NPU has 64 GB of HBM, which is not enough for the
+combined MiniMax-H3 service at 768P. Enable **one** of the following at
+server startup:
+
+**Distributed layerwise offload** — DiT layers are offloaded with parameters
+gathered across the parallel group instead of replicated per rank:
+
+```bash
+  --enable-distributed-layerwise-offload
+```
+
+Measured peak NPU memory is about 45 GB per device for Ref2VA with a 13.88 s
+input video generating a 15 s 768P (1344x768) output video. Host (CPU) memory
+usage is high with this option.
+
+**HSDP** — hybrid sharded data parallelism for the DiT parameters. The
+multi-stream memory-reuse knob is mandatory with this flag:
+
+```bash
+export MULTI_STREAM_MEMORY_REUSE=2
+```
+
+```bash
+  --use-hsdp
+```
+
+Host memory usage is small, but large shapes may still OOM; HBM usage
+optimizations for this path are ongoing.
+
+### FLASH_ATTN backend with MindIE-SD
+
+Keep `--diffusion-attention-backend FLASH_ATTN` and install MindIE-SD (see
+[§ Environment](#environment)). On NPU this backend carries most of the
+memory-reduction and performance work: a mask-free packed varlen path that
+never materializes the quadratic `full_qk` padding mask, and K/V prefix
+slicing, both driven by the packed `cu_seqlens` metadata emitted by the H3
+transformer.
+
+### Optional: LaserAttention fused kernel
+
+For an additional attention speedup, select the Ascend LaserAttention fused
+kernel before starting the server:
+
+```bash
+export MINDIE_SD_FA_TYPE="ascend_laser_attention"
+```
+
+This requires the FLASH_ATTN backend and MindIE-SD. H3 automatically applies
+exact power-of-two input pre-scaling (`laser_input_scale=256`) so the
+kernel's fp16 workspace cannot overflow on outlier activations. Measured
+speedup numbers will be added here.
+
+
 ## HTTP API examples

 The request format is identical to the GPU recipe; see
@@ -154,8 +217,8 @@ configuration above:

 | Workload | Configuration |
 |----------|---------------|
-| T2VA, 209 frames, 1344x768 | TE TP8, layerwise offload, Ulysses 8, VPP8 tile, regional compile |
-| Ref2VA (prompt + video), 124 frames, 1344x768 | TE TP8, layerwise offload, Ulysses 8, VPP8 tile, regional compile |
+| T2VA, 209 frames, 1344x768 | TE TP8, distributed layerwise offload, Ulysses 8, VPP8 tile, regional compile |
+| Ref2VA (prompt + video), 124 frames, 1344x768 | TE TP8, distributed layerwise offload, Ulysses 8, VPP8 tile, regional compile |

 These measurements describe the validated shapes rather than a general
 throughput guarantee.
@@ -4,6 +4,17 @@ This recipe uses online FP8 weight quantization, tiled VAE decode, and a single
 resident partition. GB10 is a unified-memory platform, so unlike the discrete-GPU
 recipes it uses **no offload of any kind** — see the capacity note below.

+Validated on:
+
+- Host: DGX Spark (GB10), aarch64
+- GPUs: 1 (unified memory)
+- Driver: <FILL_DRIVER_VERSION>
+- vLLM: 0.26.0
+- vLLM-Omni: `main` at `e1aa6eae75c460cd1893bc320546e81e66973831`
+- Workloads: T2VA and Ref2VA, 960x576, `duration=8.0`, 50 steps, `flow_shift=12`,
+  `seed=1101`
+
+
 ## Capacity requirements

 | Resource | DGX Spark (GB10) |
@@ -11,7 +22,8 @@ recipes it uses **no offload of any kind** — see the capacity note below.
 | Unified memory | 128 GB LPDDR5X (~121 GiB usable) |
 | GPUs | 1 |
 | Checkpoint storage | 135 GiB per partition |
-| Measured peak allocator high-water (FP8) | 97.7 GiB |
+| Measured peak allocator high-water, T2VA (FP8) | 97.7 GiB |
+| Measured peak allocator high-water, Ref2VA (FP8) | 102.8 GiB |

 `FL2VA` and `Ref2VA` are separate 135 GiB checkpoint partitions. Start one server
 at a time.
@@ -29,9 +41,21 @@ math relative to the RTX and datacenter recipes:
  fit in 121 GiB. Online FP8 quantizes the DiT only (62 GiB to ~31 GiB); the
  Qwen3-VL text encoder and both VAEs stay BF16.

-With ~97.7 GiB of the pool held by the allocator at peak, there is roughly 23 GiB
-of headroom for the OS, page cache, and any other process on the box. Treat this
-platform as single-tenant while a request is in flight.
+With ~97.7 GiB of the pool held by the allocator at peak, T2VA leaves roughly
+23 GiB of headroom for the OS, page cache, and any other process on the box.
+Ref2VA is tighter on both counts. The allocator high-water rises to 102.8 GiB,
+and the allocator figure understates true pool usage.
+
+Sampling `/proc/meminfo` at 1 Hz across a 50-step Ref2VA request on a host with
+`MemTotal: 127600524 kB` (121.7 GiB), `MemAvailable` started at 8.1 GiB, fell to
+a trough of **7.4 GiB**, and recovered to 8.9 GiB. At the trough the pool held
+roughly **114.3 GiB** — 11.4 GiB more than the 102.8 GiB the allocator reports,
+which is the CUDA context, non-PyTorch allocations, and page cache the header
+does not count.
+
+Treat this platform as strictly single-tenant for Ref2VA. There is no room for a
+second process, and adding reference images, reference videos, or a larger output
+shape will run into the OOM killer.

 ## One GB10: 960x576, 8 seconds

@@ -94,6 +118,52 @@ ffprobe -v error -show_entries \
 Both requests return `200 OK` with an MP4 body: H.264 video at the requested
 geometry plus a 32 kHz stereo AAC track.

+## One GB10: Ref2VA, 960x576, 8 seconds
+
+`FL2VA` and `Ref2VA` are separate 135 GiB partitions and only one fits at a time.
+Stop the FL2VA server before starting this one.
+
+```bash
+CUDA_VISIBLE_DEVICES=0 \
+FLASHINFER_DISABLE_VERSION_CHECK=1 \
+VLLM_WORKER_MULTIPROC_METHOD=spawn \
+VLLM_OMNI_VIDEO_SYNC_TIMEOUT=14400 \
+vllm serve /path/to/MiniMax-H3/Ref2VA \
+  --omni --trust-remote-code --host 0.0.0.0 --port 8000 \
+  --init-timeout 3600 \
+  --num-gpus 1 --tensor-parallel-size 1 --text-encoder-tp-size 1 \
+  --usp 1 --ring 1 --vae-patch-parallel-size 1 \
+  --vae-parallel-mode tile --vae-use-tiling \
+  --quantization fp8 --enforce-eager \
+  --diffusion-attention-backend CUDNN_ATTN
+```
+
+The 7200 s timeout used for T2VA is too short here: a 50-step Ref2VA request
+takes over an hour on this platform. Use 14400 s.
+
+Unlike `t2va`, `ref2va` defaults to a 16:9 output ratio, so `aspect_ratio` can be
+omitted when `width` and `height` are supplied.
+
+```bash
+curl --fail-with-body -sS -X POST http://127.0.0.1:8000/v1/videos/sync \
+  -D ref2va-headers.txt \
+  -w 'client_e2e_s=%{time_total}\n' \
+  --max-time 14400 \
+  -F 'input_reference=@/path/to/reference.jpg;type=image/jpeg' \
+  -F 'prompt=A hand rests on a wooden table in warm evening light, then slowly opens and turns toward the camera, with quiet room ambience.' \
+  -F 'width=960' \
+  -F 'height=576' \
+  -F 'fps=24' \
+  -F 'num_inference_steps=50' \
+  -F 'flow_shift=12' \
+  -F 'seed=1101' \
+  -F 'extra_params={"task":"ref2va","duration":8.0,"audio_flow_shift":3.0}' \
+  -o h3-gb10-ref2va.mp4
+```
+
+The response is `200 OK` with an MP4 body: H.264 at the requested geometry plus a
+32 kHz stereo AAC track. `ffprobe` reports 8.032 s for `duration=8.0`.
+
 ## Measured latency

 Single GB10, FP8, eager, cuDNN attention, tiled VAE, one request at a time:
@@ -102,32 +172,69 @@ Single GB10, FP8, eager, cuDNN attention, tiled VAE, one request at a time:
 | --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
 | T2VA, 4 s, smoke shape | 10 | — | — | 8.68 s | — | 0.49 s | 95.9 s | — |
 | T2VA, 8 s, 960x576 | 50 | 0.25 s | 2088 s | 42.61 s | 70 s | 4.84 s | 2169.4 s (36 min 9 s) | 97.7 GiB |
+| Ref2VA, 8 s, 960x576, one image (3 runs) | 10 | 8.43-10.94 s | 665.5-748.8 s | 66.55-74.88 s | 67.9-69.8 s | 3.63 s | 770.3-861.1 s | 102.8 GiB |
+| Ref2VA, 8 s, 960x576, one image | 50 | 9.66 s | 4039.6 s | 80.79 s | 69.4 s | 3.63 s | 4157.0 s (69 min 17 s) | 102.8 GiB |

 How these were measured:

 - Stage times come from `--enable-diffusion-pipeline-profiler`, read from the
-  `X-Stage-Durations` response header of `/v1/videos/sync`.
+  `X-Stage-Durations` response header of `/v1/videos/sync`. The keys are
+  `MiniMaxH3Pipeline.encode_prompt` (text encode),
+  `MiniMaxH3Pipeline.diffuse` (denoise), and `MiniMaxH3Pipeline.decode`
+  (VAE decode).
+- MP4 mux is not a header field. It is the
+  `Video response encoding (MP4 bytes)` line in the server log.
 - Peak memory is the `X-Peak-Memory-MB` header, which reports the CUDA allocator
  high-water mark (`torch.cuda.max_memory_reserved`) for the request. It excludes
  the CUDA context and non-PyTorch allocations, so true pool usage is somewhat
  higher. `nvidia-smi --query-gpu=memory.used` is not usable on GB10 — unified
-  memory has no discrete VRAM to report and the field returns `[N/A]`.
- Text encode and VAE decode were measured on a separate 10-step run at the same
-  960x576 / 8 s geometry. Both are independent of step count, so they transfer to
-  the 50-step row unchanged.
+  memory has no discrete VRAM to report and the field returns `[N/A]`. See the
+  capacity section for a `/proc/meminfo` cross-check.
+- For T2VA, text encode and VAE decode were measured on a separate 10-step run at
+  the same 960x576 / 8 s geometry. Both are independent of step count, so they
+  transfer to the 50-step row unchanged. The Ref2VA MP4 mux figure transfers the
+  same way, from the 10-step run at identical output geometry and duration.
+- Ref2VA peak memory is byte-identical between the 10-step and 50-step runs
+  (`X-Peak-Memory-MB: 105318.000` in both), confirming that peak is set by output
+  geometry and reference input rather than step count.
 - Stage times do not sum exactly to end-to-end. Queueing, result transfer, and
  engine bookkeeping sit outside the profiled stages.

 Denoise dominates at 96% of end-to-end, so step count and geometry are the only
 knobs that materially change the wall clock on this platform.

-Repeated identical 10-step requests varied by up to 23% in denoise time
+Ref2VA costs 1.92x the wall clock of T2VA at the same geometry and step count:
+4157.0 s versus 2169.4 s. Text encoding shows the most dramatic ratio — 9.66 s
+versus 0.25 s, 39x — but at 0.2% of end-to-end it is not what makes Ref2VA slow.
+The cost is in denoise. A single reference image expands the Qwen3-VL
+presentation to 6883 tokens (logged as `MiniMax H3 ref2va Qwen presentation:
+6883 tokens, 1 reference images`), and those tokens sit in the attention context
+of every denoise step. Per-step time rises from 42.61 s to 80.79 s, and that
+1.90x multiplies across all 50 steps. VAE decode is unchanged at ~69 s, as
+expected for an identical output shape.
+
+Repeated identical 10-step T2VA requests varied by up to 23% in denoise time
 (397 s to 490 s) on a machine that had already been under sustained load. Treat
 these as single-run measurements on one machine, not a throughput guarantee, and
 expect the slower end of that range for back-to-back jobs.

+Ref2VA per-step time is load-dependent in the same way. On an idle machine the
+first 10-step request ran at 66.55 s per step. The 50-step request that followed
+occupied the box for 69 consecutive minutes and averaged 80.79 s per step. Two
+further 10-step requests taken immediately afterwards settled at 74.88 s and
+74.31 s — 12.5% slower than the idle run, and within 0.8% of each other, so this
+is a stable thermal state rather than measurement noise.
+
+Plan for 81 s per step when sizing a full 50-step Ref2VA job; 66 s per step is
+only reachable on a cold machine and does not hold for the length of one request.
+
+Ref2VA text encode converged downward across the four runs (10.94 s, 9.66 s,
+8.52 s, 8.43 s) as first-touch costs fell away; treat ~8.5 s as steady state.
+VAE decode was stable at 67.9-69.8 s throughout.
+
 Set `VLLM_OMNI_VIDEO_SYNC_TIMEOUT` well above the expected request time — the
-default 1800 s is shorter than a 50-step run on this platform.
+default 1800 s is shorter than a 50-step run on this platform, and 7200 s is
+shorter than a 50-step Ref2VA run.

 ## Known limitations

@@ -139,11 +246,17 @@ default 1800 s is shorter than a 50-step run on this platform.
 - `t2va` rejects requests without an explicit `aspect_ratio`. When `width` and
  `height` are both supplied they take precedence and `aspect_ratio` only has to
  name one of the supported ratios.
- Sustained load degrades throughput. See the variance note above.
+- Sustained load degrades throughput. See the variance notes above.
 - Only one partition is resident. For Ref2VA, stop the FL2VA server and restart
-  the same command with `/path/to/MiniMax-H3/Ref2VA`. Ref2VA reference video
-  count and prompt length increase activation memory; begin with one request at
-  a time and keep the short edge at 576.
+  with `/path/to/MiniMax-H3/Ref2VA`. Ref2VA with one reference image is measured
+  above; reference videos and additional images push the token count higher and
+  were not measured. With `MemAvailable` bottoming out at 7.4 GiB during a
+  single-image request, there is very little room to grow — re-measure before
+  adding references or raising the output shape.
+- Ref2VA requires a vLLM-Omni build newer than `v0.26.0`. See the version note
+  under **Validated on**: FP8 weight loading and image-only Ref2VA are both
+  broken on `release/v0.26.0`.suggestion is  `v0.26.1`.
 - Online FP8 is incompatible with layerwise offload — the offload path produces a
  weight stride the Cutlass FP8 kernel rejects. This is not a practical
  restriction here since offload is unusable on GB10 anyway.
+```
@@ -19,6 +19,11 @@ add files for this repo. Use one Markdown file per model family by default.
 > Individual pages resolve at `https://recipes.vllm.ai/<provider>/<model-id>`,
 > e.g. [`bosonai/higgs-audio-v3-tts-4b`](https://recipes.vllm.ai/bosonai/higgs-audio-v3-tts-4b).

+The `Recipe` column in the Supported Models documentation links to the
+published recipe when available and otherwise to the corresponding recipe in
+this directory. The recipe files remain the source of truth for commands and
+validation details.
+
 Example layout:

 ```text
@@ -9,6 +9,10 @@ This script tests two main scenarios:
 2. Case 2: Comparing FlashAttention and SDPA backends for batch_size=2 with padding
 """

+import sys
+from types import SimpleNamespace
+from unittest.mock import Mock
+
 import pytest
 import torch

@@ -371,6 +375,264 @@ def test_packed_varlen_metadata_must_be_complete(monkeypatch):
        impl.forward_cuda(query, query, query, metadata)


+# ---------------------------------------------------------------------------
+# NPU mask-free packed attention (PR #5891).
+#
+# These tests cover the backend-side NPU branches that ``forward_fa_npu`` adds:
+#   - boundary resolution in ``_resolve_packed_seq_npu`` (pure Python, no device
+#     sync, no MindIE-SD import);
+#   - env dispatch in ``forward_fa_npu`` (which op a packed forward takes);
+#   - the laser input pre-scaling math in ``_forward_prefix_kv_slice_npu``.
+#
+# They run on CPU without a real NPU by injecting a fake ``mindiesd`` module
+# into ``sys.modules`` (same pattern as ``test_paged_attention`` /
+# ``benchmarks/conftest``). No production code is exercised for real kernels.
+
+
+def _npu_impl(causal: bool = False) -> FlashAttentionImpl:
+    return FlashAttentionImpl(num_heads=2, head_size=4, softmax_scale=0.5, causal=causal)
+
+
+def _fake_mindiesd(monkeypatch, *, attention_forward=None, attention_forward_varlen=None):
+    """Install a stub ``mindiesd`` so local ``from mindiesd import ...`` works.
+
+    Restored automatically by monkeypatch. Defaults are no-op callables so the
+    top-level import in ``forward_fa_npu`` succeeds even when the test mocks the
+    instance methods that would otherwise call them.
+    """
+    fake = SimpleNamespace(
+        attention_forward=attention_forward or Mock(return_value=torch.zeros(1)),
+        attention_forward_varlen=attention_forward_varlen or Mock(return_value=torch.zeros(1)),
+    )
+    monkeypatch.setitem(sys.modules, "mindiesd", fake)
+    return fake
+
+
+# --- Test group A: boundary resolution (_resolve_packed_seq_npu) -------------
+
+
+def test_resolve_packed_seq_accepts_real_plus_pad_contract():
+    impl = _npu_impl()
+    q = torch.randn(1, 8, 2, 4)
+    cu = torch.tensor([0, 5, 8], dtype=torch.int32)
+    extra = {"cu_seqlens_q": cu, "cu_seqlens_k": cu, "max_seqlen_q": 5, "max_seqlen_k": 5}
+
+    assert impl._resolve_packed_seq_npu(q, q, extra) == ([5, 8], [5, 8])
+
+
+def test_resolve_packed_seq_single_document_no_padding():
+    impl = _npu_impl()
+    q = torch.randn(1, 8, 2, 4)
+    cu = torch.tensor([0, 8], dtype=torch.int32)
+    extra = {"cu_seqlens_q": cu, "cu_seqlens_k": cu, "max_seqlen_q": 8, "max_seqlen_k": 8}
+
+    assert impl._resolve_packed_seq_npu(q, q, extra) == ([8], [8])
+
+
+@pytest.mark.parametrize(
+    "case",
+    [
+        "causal",
+        "missing_key",
+        "batch_ne_1",
+        "cu_too_long",
+        "cu_qk_mismatch",
+        "max_seqlen_not_int",
+        "used_gt_total",
+        "used_zero",
+        "pad_longer_than_real",
+    ],
+)
+def test_resolve_packed_seq_rejects_malformed_metadata(case):
+    """Any metadata outside the exact [real, pad] contract returns None so the
+    caller falls back to the masked path."""
+    cu3 = torch.tensor([0, 5, 8], dtype=torch.int32)
+    cu2 = torch.tensor([0, 8], dtype=torch.int32)
+    cu4 = torch.tensor([0, 3, 5, 8], dtype=torch.int32)
+    cu_short = torch.tensor([0, 3, 8], dtype=torch.int32)
+    base = {"cu_seqlens_q": cu3, "cu_seqlens_k": cu3, "max_seqlen_q": 5, "max_seqlen_k": 5}
+    q = torch.randn(1, 8, 2, 4)
+
+    impl = _npu_impl(causal=(case == "causal"))
+    if case == "missing_key":
+        extra = {"cu_seqlens_q": cu3, "max_seqlen_q": 5}  # 3 of 4 keys absent
+    elif case == "batch_ne_1":
+        q = torch.randn(2, 8, 2, 4)
+        extra = base
+    elif case == "cu_too_long":
+        extra = {**base, "cu_seqlens_q": cu4, "cu_seqlens_k": cu4}
+    elif case == "cu_qk_mismatch":
+        extra = {**base, "cu_seqlens_q": cu3, "cu_seqlens_k": cu2}
+    elif case == "max_seqlen_not_int":
+        extra = {**base, "max_seqlen_q": 5.0}
+    elif case == "used_gt_total":
+        extra = {**base, "max_seqlen_q": 9}  # 9 > total length 8
+    elif case == "used_zero":
+        extra = {**base, "max_seqlen_q": 0, "max_seqlen_k": 0}
+    elif case == "pad_longer_than_real":
+        # real=3, pad=5: the real document must be the longer one.
+        extra = {
+            "cu_seqlens_q": cu_short,
+            "cu_seqlens_k": cu_short,
+            "max_seqlen_q": 3,
+            "max_seqlen_k": 3,
+        }
+    else:
+        extra = base
+
+    assert impl._resolve_packed_seq_npu(q, q, extra) is None
+
+
+# --- Test group B: env dispatch in forward_fa_npu (current behavior) --------
+
+
+def _npu_impl_with_mocked_paths() -> tuple[FlashAttentionImpl, torch.Tensor]:
+    """FlashAttentionImpl whose two packed paths are Mocks returning a sentinel.
+
+    Lets a test assert which branch ``forward_fa_npu`` dispatched to without
+    running any real kernel.
+    """
+    impl = _npu_impl()
+    sentinel = torch.tensor([1.0])
+    impl._forward_varlen_packed_npu = Mock(return_value=sentinel)
+    impl._forward_prefix_kv_slice_npu = Mock(return_value=sentinel)
+    return impl, sentinel
+
+
+def test_npu_env_unset_uses_varlen(monkeypatch):
+    monkeypatch.delenv("MINDIE_SD_FA_TYPE", raising=False)
+    _fake_mindiesd(monkeypatch)
+    impl, _ = _npu_impl_with_mocked_paths()
+    metadata = AttentionMetadata(extra={"npu_attn_varlen": True})
+
+    impl.forward_fa_npu(torch.randn(1, 8, 2, 4), *[torch.randn(1, 8, 2, 4)] * 2, metadata)
+
+    impl._forward_varlen_packed_npu.assert_called_once()
+    impl._forward_prefix_kv_slice_npu.assert_not_called()
+
+
+def test_npu_env_laser_uses_slice(monkeypatch):
+    monkeypatch.setenv("MINDIE_SD_FA_TYPE", "ascend_laser_attention")
+    _fake_mindiesd(monkeypatch)
+    impl, _ = _npu_impl_with_mocked_paths()
+    metadata = AttentionMetadata(extra={"npu_attn_varlen": True})
+
+    impl.forward_fa_npu(torch.randn(1, 8, 2, 4), *[torch.randn(1, 8, 2, 4)] * 2, metadata)
+
+    impl._forward_prefix_kv_slice_npu.assert_called_once()
+    impl._forward_varlen_packed_npu.assert_not_called()
+
+
+@pytest.mark.parametrize("fa_type", ["prompt_flash_attn", "fused_attn_score", "ascend_laser_attenton", "garbage"])
+def test_npu_env_non_laser_value_falls_back_to_varlen(monkeypatch, fa_type):
+    """Locks in the CURRENT dispatch behavior (PR #5891).
+
+    Any value other than ``ascend_laser_attention`` — including the valid
+    ``prompt_flash_attn``/``fused_attn_score`` and typos — is silently routed to
+    the TND varlen op, which cannot honor ``MINDIE_SD_FA_TYPE``. This is review
+    comment #1 on the PR and is intentionally captured here as a regression
+    net: update this test when the dispatch starts honoring/rejecting these.
+    """
+    monkeypatch.setenv("MINDIE_SD_FA_TYPE", fa_type)
+    _fake_mindiesd(monkeypatch)
+    impl, _ = _npu_impl_with_mocked_paths()
+    metadata = AttentionMetadata(extra={"npu_attn_varlen": True})
+
+    impl.forward_fa_npu(torch.randn(1, 8, 2, 4), *[torch.randn(1, 8, 2, 4)] * 2, metadata)
+
+    impl._forward_varlen_packed_npu.assert_called_once()
+    impl._forward_prefix_kv_slice_npu.assert_not_called()
+
+
+def test_npu_varlen_opt_in_unset_takes_mask_path(monkeypatch):
+    """Models that do not set ``npu_attn_varlen`` (e.g. Wan/Cosmos) keep using
+    the existing masked attention_forward path; the new branches are skipped."""
+    fake_forward = Mock(return_value=torch.zeros(1, 8, 2, 4))
+    _fake_mindiesd(monkeypatch, attention_forward=fake_forward)
+    impl, _ = _npu_impl_with_mocked_paths()
+    metadata = AttentionMetadata(
+        attn_mask=torch.tensor([[True, True, True, True, True, False, False, False]]),
+        extra={},  # no npu_attn_varlen opt-in
+    )
+    q = torch.randn(1, 8, 2, 4)
+
+    out = impl.forward_fa_npu(q, q, q, metadata)
+
+    impl._forward_varlen_packed_npu.assert_not_called()
+    impl._forward_prefix_kv_slice_npu.assert_not_called()
+    fake_forward.assert_called_once()
+    assert out is fake_forward.return_value
+
+
+# --- Test group C: laser input pre-scaling in _forward_prefix_kv_slice_npu --
+
+
+def test_prefix_kv_slice_applies_laser_input_scaling(monkeypatch):
+    monkeypatch.setenv("MINDIE_SD_FA_TYPE", "ascend_laser_attention")
+    captured: dict = {}
+
+    def fake_attention_forward(query, key, value, **kwargs):
+        captured["query"] = query
+        captured["key"] = key
+        captured["value"] = value
+        captured.update(kwargs)
+        return torch.ones_like(query)  # deterministic stand-in kernel output
+
+    _fake_mindiesd(monkeypatch, attention_forward=fake_attention_forward)
+    impl = _npu_impl()  # softmax_scale == 0.5
+    q = torch.randn(1, 8, 2, 4)
+    cu = torch.tensor([0, 5, 8], dtype=torch.int32)
+    extra = {
+        "cu_seqlens_q": cu,
+        "cu_seqlens_k": cu,
+        "max_seqlen_q": 5,
+        "max_seqlen_k": 5,
+        "laser_input_scale": 256.0,
+    }
+
+    out = impl._forward_prefix_kv_slice_npu(q, q, q, extra)
+
+    # K/V sliced to the valid prefix (used_k == 5); q keeps full length.
+    assert captured["key"].shape == (1, 5, 2, 4)
+    assert captured["value"].shape == (1, 5, 2, 4)
+    assert captured["query"].shape == (1, 8, 2, 4)
+    assert captured["attn_mask"] is None
+    # Pre-divide q/k/v by the factor and compensate the kernel scale by its
+    # square (exact for power-of-two); output scaled back by the factor.
+    torch.testing.assert_close(captured["query"], q / 256.0)
+    torch.testing.assert_close(captured["key"], q[:, :5] / 256.0)
+    assert captured["scale"] == pytest.approx(0.5 * 256.0**2)
+    torch.testing.assert_close(out, torch.ones(1, 8, 2, 4) * 256.0)
+
+
+def test_prefix_kv_slice_no_scaling_without_factor(monkeypatch):
+    monkeypatch.setenv("MINDIE_SD_FA_TYPE", "ascend_laser_attention")
+    captured: dict = {}
+
+    def fake_attention_forward(query, key, value, **kwargs):
+        captured["query"] = query
+        captured.update(kwargs)
+        return torch.ones_like(query)
+
+    _fake_mindiesd(monkeypatch, attention_forward=fake_attention_forward)
+    impl = _npu_impl()
+    q = torch.randn(1, 8, 2, 4)
+    cu = torch.tensor([0, 5, 8], dtype=torch.int32)
+    extra = {
+        "cu_seqlens_q": cu,
+        "cu_seqlens_k": cu,
+        "max_seqlen_q": 5,
+        "max_seqlen_k": 5,
+        # no laser_input_scale → no pre-scaling
+    }
+
+    out = impl._forward_prefix_kv_slice_npu(q, q, q, extra)
+
+    torch.testing.assert_close(captured["query"], q)
+    assert captured["scale"] == pytest.approx(0.5)
+    torch.testing.assert_close(out, torch.ones(1, 8, 2, 4))
+
+
 if __name__ == "__main__":
    print("Running FlashAttention Padding Tests...")
    print("=" * 60)
@@ -5,10 +5,9 @@

 The grids here are the ones MiniMax-H3 FL2VA actually produces for an 8.7s clip:
 1280x768 gives a 62x24x40 latent grid whose 59520 video rows land on the 128-row
-kernel block, and 1344x768 gives 62x24x42 whose 62496 rows straddle it. Only the
-first can run sparse: rf_v2 tiles the sequence as a whole but pools its block
-mask from the video and prefix segments separately, so the two only agree on the
-block count when the video segment ends on a boundary.
+kernel block, and 1344x768 gives 62x24x42 whose 62496 rows do not. Both are
+handed to MindIE-SD: its rf_v2 preprocessing promotes an irregular video tail
+to the always-kept prefix segment before generating the block mask.
 """

 import dataclasses
@@ -57,10 +56,13 @@ def test_block_aligned_video_segment_runs_sparse():
    assert plan.latent_shape == list(ALIGNED_GRID)


-def test_misaligned_video_segment_stays_dense():
-    # Realigning the sequence would mean padding it, and rf_v2 ignores attn_mask:
-    # pad keys would take a share of every real query's softmax denominator.
-    assert make_impl()._resolve_plan(make_metadata(MISALIGNED_GRID)) is None
+def test_misaligned_video_segment_runs_sparse_with_updated_mindiesd():
+    plan = make_impl()._resolve_plan(make_metadata(MISALIGNED_GRID))
+
+    assert plan is not None
+    assert plan.prefix_len == PREFIX_ROWS
+    assert plan.used_len == PREFIX_ROWS + 62496
+    assert plan.latent_shape == list(MISALIGNED_GRID)


@pytest.mark.parametrize("prefix_len", [1, 127, _BLOCK_SIZE, 710, 900])
@@ -155,3 +157,28 @@ def test_fully_populated_mask_reproduces_dense_attention():
    ).transpose(1, 2)
    error = (out.float() - reference.float()).abs().mean() / reference.float().abs().mean()
    assert error < 2e-3, f"mean relative error {error:.4%} against dense attention"
+
+
+@pytest.mark.skipif(not current_omni_platform.is_npu(), reason="rf_v2 runs on Ascend NPU only.")
+def test_irregular_video_tail_reproduces_dense_attention_with_updated_mindiesd():
+    """vLLM forwards the full irregular grid; MindIE-SD owns its tail handling."""
+    torch.manual_seed(1)
+    prefix_len, grid = 384, (4, 16, 10)
+    video_len = grid[0] * grid[1] * grid[2]
+    used = prefix_len + video_len
+    heads, head_dim = 4, 128
+    q, k, v = (torch.randn(1, used, heads, head_dim, dtype=torch.bfloat16, device="npu") for _ in range(3))
+
+    impl = make_impl()
+    impl.rainfusion = dataclasses.replace(impl.rainfusion, sparsity=0.0)
+    plan = RainFusionPlan(prefix_len=prefix_len, used_len=used, latent_shape=list(grid))
+    out = impl._forward_sparse_npu(q, k, v, plan)
+
+    reference = torch.nn.functional.scaled_dot_product_attention(
+        q.transpose(1, 2),
+        k.transpose(1, 2),
+        v.transpose(1, 2),
+        scale=impl.softmax_scale,
+    ).transpose(1, 2)
+    error = (out.float() - reference.float()).abs().mean() / reference.float().abs().mean()
+    assert error < 2e-3, f"mean relative error {error:.4%} against dense attention"
@@ -488,6 +488,14 @@ class _MiniMaxH3FakeAttention(nn.Module):
    class FakeBackend:
        supports_prefix_kv_slicing = False

+        # Mirror AttentionBackend.supports_packed_mask_free: H3's packed
+        # attention path calls attn_backend.supports_packed_mask_free() to decide
+        # whether to build the padding mask. Return False so the fake exercises
+        # the masked path (the abstract backend default).
+        @classmethod
+        def supports_packed_mask_free(cls) -> bool:
+            return False
+
    def __init__(self, **kwargs):
        del kwargs
        super().__init__()
@@ -492,6 +492,10 @@ def test_cudnn_packed_attention_uses_python_length_without_padding_mask():
    class FakeBackend:
        supports_prefix_kv_slicing = True

+        @classmethod
+        def supports_packed_mask_free(cls) -> bool:
+            return False
+
    class FakeAttention(torch.nn.Module):
        attn_backend = FakeBackend

@@ -526,6 +530,48 @@ def test_cudnn_packed_attention_uses_python_length_without_padding_mask():
    assert attention.attention.metadata.extra["valid_kv_length"] == 5


+def test_packed_attention_skips_mask_for_packed_mask_free_backend():
+    from vllm_omni.diffusion.models.minimax_h3.minimax_h3_transformer import (
+        MiniMaxH3Attention,
+    )
+
+    class FakeBackend:
+        supports_prefix_kv_slicing = False
+
+        @classmethod
+        def supports_packed_mask_free(cls) -> bool:
+            return True
+
+    class FakeAttention(torch.nn.Module):
+        attn_backend = FakeBackend
+
+        def __init__(self):
+            super().__init__()
+            self.metadata = None
+
+        def forward(self, query, key, value, metadata):
+            self.metadata = metadata
+            return query
+
+    attention = object.__new__(MiniMaxH3Attention)
+    torch.nn.Module.__init__(attention)
+    attention.attention = FakeAttention()
+    q = torch.randn(8, 2, 4)
+
+    attention._run_packed_attention(
+        q,
+        q,
+        q,
+        cu_seqlens=torch.tensor([0, 5, 8], dtype=torch.int32),
+        max_seqlen=5,
+        packed_total=8,
+    )
+
+    assert attention.attention.metadata.attn_mask is None
+    assert attention.attention.metadata.extra["valid_kv_length"] == 5
+    assert attention.attention.metadata.extra["npu_attn_varlen"] is True
+
+
 def test_packed_attention_keeps_padding_mask_for_other_backends():
    from vllm_omni.diffusion.models.minimax_h3.minimax_h3_transformer import (
        MiniMaxH3Attention,
@@ -534,6 +580,10 @@ def test_packed_attention_keeps_padding_mask_for_other_backends():
    class FakeBackend:
        supports_prefix_kv_slicing = False

+        @classmethod
+        def supports_packed_mask_free(cls) -> bool:
+            return False
+
    class FakeAttention(torch.nn.Module):
        attn_backend = FakeBackend

@@ -69,6 +69,7 @@ def test_h3_fused_rope_matches_reference_and_preserves_unrotated_dims():

    attention = object.__new__(MiniMaxH3Attention)
    nn.Module.__init__(attention)
+    attention.rot_dim = 96
    attention.rope = RotaryEmbedding(is_neox_style=True, half_head_dim=False)
    attention.rope._forward_method = attention.rope.forward_native

@@ -113,6 +113,14 @@ class TestCreateDefaultDiffusion:
        stages = StageConfigFactory.create_default_diffusion({"model": "x"})
        assert stages[0]["engine_args"]["cache_backend"] == "none"

+    def test_vae_decode_precision_roundtrip(self):
+        od = _roundtrip_diffusion_config(model="x", vae_decode_precision="bf16")
+        assert od.vae_decode_precision == "bf16"
+
+    def test_invalid_vae_decode_precision_is_rejected(self):
+        with pytest.raises(ValueError, match="vae_decode_precision"):
+            OmniDiffusionConfig(model="x", vae_decode_precision="int8")
+
    def test_single_gpu_default_devices(self):
        stages = StageConfigFactory.create_default_diffusion({"model": "x"})
        assert stages[0]["runtime"]["devices"] == "0"
@@ -512,6 +512,35 @@ def test_execute_model_accepts_bare_diffusion_output_from_single_request_pipelin
    assert runner.pipeline.last_req.num_reqs == 1


+@pytest.mark.core_model
+@pytest.mark.cpu
+@pytest.mark.parametrize(
+    ("quality", "expected_enabled"),
+    [(None, False), ("lossless", False), ("high", True)],
+)
+def test_execute_model_scopes_vae_fast_path_to_request_quality(monkeypatch, quality, expected_enabled):
+    runner = _make_runner(cache_backend=None, cache_backend_name="none")
+    req = _make_request()
+    req.sampling_params.quality = quality
+    gate_calls = []
+
+    @contextmanager
+    def _record_gate(pipeline, enabled):
+        gate_calls.append((pipeline, enabled, "enter"))
+        yield
+        gate_calls.append((pipeline, enabled, "exit"))
+
+    monkeypatch.setattr(model_runner_module, "set_forward_context", _noop_forward_context)
+    monkeypatch.setattr(model_runner_module, "use_pipeline_vae_fast_path", _record_gate)
+
+    DiffusionModelRunner.execute_model(runner, req)
+
+    assert gate_calls == [
+        (runner.pipeline, expected_enabled, "enter"),
+        (runner.pipeline, expected_enabled, "exit"),
+    ]
+
+
 class _BatchPipeline:
    """Pipeline returning a configurable list of outputs from forward()."""

@@ -579,6 +608,18 @@ def test_execute_model_batch_rejects_output_count_mismatch(monkeypatch):
        DiffusionModelRunner.execute_model_batch(runner, sched, runner.od_config)


+@pytest.mark.core_model
+@pytest.mark.cpu
+def test_execute_model_batch_rejects_mixed_vae_quality() -> None:
+    runner = _make_batch_runner(_BatchPipeline(outputs=[DiffusionOutput(output="a"), DiffusionOutput(output="b")]))
+    sched = _make_scheduler_output(num_reqs=2)
+    sched.scheduled_new_reqs[0].req.sampling_params.quality = "lossless"
+    sched.scheduled_new_reqs[1].req.sampling_params.quality = "high"
+
+    with pytest.raises(RuntimeError, match="cannot mix request quality levels"):
+        DiffusionModelRunner.execute_model_batch(runner, sched, runner.od_config)
+
+
@pytest.mark.core_model
@pytest.mark.cpu
 def test_execute_model_batch_routes_one_output_per_request(monkeypatch):
@@ -728,6 +769,19 @@ def test_execute_model_runs_forward_after_kv_receive(monkeypatch):
@pytest.mark.core_model
@pytest.mark.cpu
 def test_load_model_clears_cache_backend_for_unsupported_pipeline(monkeypatch):
+    memory_pool_active = False
+    optimization_pool_states = []
+
+    @contextmanager
+    def _memory_pool_context(tag):
+        nonlocal memory_pool_active
+        assert tag == "weights"
+        memory_pool_active = True
+        try:
+            yield
+        finally:
+            memory_pool_active = False
+
    class _DummyLoader:
        def __init__(self, load_config, od_config=None):
            del load_config, od_config
@@ -775,16 +829,22 @@ def test_load_model_clears_cache_backend_for_unsupported_pipeline(monkeypatch):
    monkeypatch.setattr(model_runner_module, "LoadConfig", lambda: object())
    monkeypatch.setattr(model_runner_module, "DiffusersPipelineLoader", _DummyLoader)
    monkeypatch.setattr(model_runner_module, "DeviceMemoryProfiler", _DummyMemoryProfiler)
+    monkeypatch.setattr(
+        model_runner_module,
+        "optimize_pipeline_vaes",
+        lambda pipeline, od_config: optimization_pool_states.append(memory_pool_active),
+    )
    monkeypatch.setattr(model_runner_module, "get_offload_backend", lambda od_config, device: None)
    monkeypatch.setattr(
        model_runner_module, "get_cache_backend", lambda cache_backend, cache_config: dummy_cache_backend
    )

-    DiffusionModelRunner.load_model(runner)
+    DiffusionModelRunner.load_model(runner, memory_pool_context_fn=_memory_pool_context)

    assert runner.cache_backend is None
    assert runner.od_config.cache_backend is None
    assert dummy_cache_backend.enabled is False
+    assert optimization_pool_states == [True]


@pytest.mark.core_model
@@ -102,6 +102,7 @@ class TestDiffusionWorkerSleep:

    def test_sleep_level_1(self, mocker: MockerFixture, mock_gpu_worker):
        """Test sleep mode level 1 (offload weights only)."""
+        clear_vae_caches = mocker.patch("vllm_omni.diffusion.vae_optimizations.clear_pipeline_vae_fast_path_caches")
        mock_allocator_class = patch_cumem_allocator(mocker)
        mock_platform = mocker.patch("vllm_omni.diffusion.worker.diffusion_worker.current_omni_platform")
        mock_platform.get_free_memory.side_effect = [10 * 1024**3, 12 * 1024**3]
@@ -128,6 +129,7 @@ class TestDiffusionWorkerSleep:

        # Verify sleep was called with correct tags
        mock_allocator.sleep.assert_called_once_with(offload_tags=("weights",))
+        clear_vae_caches.assert_called_once_with(mock_gpu_worker.model_runner.pipeline)
        assert bool(result) is True
        # Verify buffers were NOT saved (level 1 doesn't save buffers)
        assert len(mock_gpu_worker._sleep_saved_buffers) == 0
@@ -0,0 +1,300 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Correctness contracts for request-gated VAE decoder fast paths."""
+
+from __future__ import annotations
+
+import pytest
+import torch
+import torch.nn.functional as F
+from torch import nn
+
+from tests.helpers.mark import hardware_test
+from vllm_omni.diffusion.vae_optimizations.gate import (
+    VaeFastPathGate,
+    register_vae_fast_path_gate,
+    use_pipeline_vae_fast_path,
+    use_vae_fast_path,
+)
+
+pytestmark = [pytest.mark.diffusion]
+
+
+class _PipelineWithVae(nn.Module):
+    def __init__(self, vae: nn.Module) -> None:
+        super().__init__()
+        self.vae = vae
+
+
+class _DistributedExecutorConfig:
+    def __init__(self, parallel_mode: str) -> None:
+        self.parallel_mode = parallel_mode
+
+
+@pytest.mark.core_model
+@pytest.mark.cpu
+def test_vae_fast_path_gate_restores_nested_state() -> None:
+    vae = nn.Identity()
+    gate = VaeFastPathGate()
+    register_vae_fast_path_gate(vae, gate)
+
+    with use_vae_fast_path(vae, True):
+        assert gate.enabled
+        with use_vae_fast_path(vae, False):
+            assert not gate.enabled
+        assert gate.enabled
+
+    assert not gate.enabled
+
+
+@pytest.mark.core_model
+@pytest.mark.cpu
+def test_pipeline_vae_fast_path_controls_discovered_vae() -> None:
+    vae = nn.Identity()
+    pipeline = _PipelineWithVae(vae)
+    gate = VaeFastPathGate()
+    register_vae_fast_path_gate(vae, gate)
+
+    with use_pipeline_vae_fast_path(pipeline, True):
+        assert gate.enabled
+
+    assert not gate.enabled
+
+
+@pytest.mark.core_model
+@pytest.mark.cpu
+def test_image_group_norm_fast_path_uses_compile_safe_fallback(monkeypatch) -> None:
+    from vllm_omni.diffusion.vae_optimizations import image
+
+    gate = VaeFastPathGate()
+    gate.enabled = True
+    norm = image.FusedGroupNormSiLU(nn.GroupNorm(2, 4), gate)
+    inputs = torch.randn(1, 4, 3, 3)
+
+    monkeypatch.setattr(torch.compiler, "is_compiling", lambda: True)
+    monkeypatch.setattr(
+        image,
+        "group_norm_silu_4d",
+        lambda *args, **kwargs: pytest.fail("eager Triton kernel entered during compilation"),
+    )
+
+    actual = norm(inputs)
+    expected = F.silu(F.group_norm(inputs, 2, norm.weight, norm.bias, norm.eps))
+    torch.testing.assert_close(actual, expected)
+
+
+def _small_image_vae() -> nn.Module:
+    from diffusers.models.autoencoders import AutoencoderKL
+
+    return AutoencoderKL(
+        in_channels=3,
+        out_channels=3,
+        down_block_types=("DownEncoderBlock2D", "DownEncoderBlock2D"),
+        up_block_types=("UpDecoderBlock2D", "UpDecoderBlock2D"),
+        block_out_channels=(32, 32),
+        layers_per_block=1,
+        latent_channels=4,
+        norm_num_groups=8,
+        sample_size=16,
+    )
+
+
+def _small_wan_vae(vae_cls: type[nn.Module] | None = None) -> nn.Module:
+    if vae_cls is None:
+        from diffusers.models.autoencoders import AutoencoderKLWan
+
+        vae_cls = AutoencoderKLWan
+
+    return vae_cls(
+        base_dim=8,
+        decoder_base_dim=8,
+        z_dim=4,
+        dim_mult=[1, 2],
+        num_res_blocks=1,
+        attn_scales=[],
+        temperal_downsample=[False],
+        latents_mean=[0.0] * 4,
+        latents_std=[1.0] * 4,
+        scale_factor_temporal=1,
+        scale_factor_spatial=2,
+    )
+
+
+@pytest.mark.core_model
+@hardware_test(res={"cuda": "L4"}, num_cards=1)
+@torch.no_grad()
+def test_image_vae_install_preserves_lossless_path_and_state_dict() -> None:
+    from vllm_omni.diffusion.vae_optimizations import clear_pipeline_vae_fast_path_caches
+    from vllm_omni.diffusion.vae_optimizations.image import (
+        FusedGroupNormSiLU,
+        FusedUpsample2xConv2d,
+        maybe_optimize_image_vae,
+    )
+
+    torch.manual_seed(0)
+    vae = _small_image_vae().to(device="cuda", dtype=torch.bfloat16).eval()
+    latents = torch.randn(1, 4, 8, 8, device="cuda", dtype=torch.bfloat16)
+    parameter_names = {name for name, _ in vae.named_parameters()}
+    state_dict = {name: value.clone() for name, value in vae.state_dict().items()}
+    reference = vae.decode(latents).sample
+
+    maybe_optimize_image_vae(vae)
+    maybe_optimize_image_vae(vae)
+
+    assert {name for name, _ in vae.named_parameters()} == parameter_names
+    vae.load_state_dict(state_dict, strict=True)
+    assert any(isinstance(module, FusedGroupNormSiLU) for module in vae.modules())
+    assert any(isinstance(module, FusedUpsample2xConv2d) for module in vae.modules())
+    assert torch.equal(vae.decode(latents).sample, reference)
+
+    with use_vae_fast_path(vae, True):
+        fast = vae.decode(latents).sample
+
+    torch.testing.assert_close(fast.float(), reference.float(), atol=0.1, rtol=0)
+    assert any(
+        buffer is not None
+        for module in vae.modules()
+        for name, buffer in module._buffers.items()
+        if name in {"_fused_weight", "_vllm_folded_value_weight", "_vllm_folded_value_bias"}
+    )
+    clear_pipeline_vae_fast_path_caches(_PipelineWithVae(vae))
+    assert all(
+        buffer is None
+        for module in vae.modules()
+        for name, buffer in module._buffers.items()
+        if name in {"_fused_weight", "_vllm_folded_value_weight", "_vllm_folded_value_bias"}
+    )
+    with use_vae_fast_path(vae, True):
+        torch.testing.assert_close(vae.decode(latents).sample.float(), reference.float(), atol=0.1, rtol=0)
+    assert torch.equal(vae.decode(latents).sample, reference)
+    assert vae.decoder.conv_in.weight.is_contiguous()
+
+
+@pytest.mark.core_model
+@hardware_test(res={"cuda": "L4"}, num_cards=1)
+@torch.no_grad()
+@pytest.mark.parametrize("dtype", [torch.float32, torch.bfloat16])
+def test_channels_last_group_norm_silu_matches_reference(dtype: torch.dtype) -> None:
+    from vllm_omni.diffusion.vae_optimizations.triton_group_norm_silu import group_norm_silu_4d
+
+    torch.manual_seed(0)
+    x = torch.randn(1, 64, 16, 16, device="cuda", dtype=dtype).contiguous(memory_format=torch.channels_last)
+    weight = torch.randn(64, device="cuda", dtype=dtype)
+    bias = torch.randn(64, device="cuda", dtype=dtype)
+
+    actual = group_norm_silu_4d(x, weight, bias, num_groups=8, eps=1e-5)
+    expected = F.silu(F.group_norm(x, 8, weight, bias))
+
+    assert actual is not None
+    atol, rtol = (1e-5, 1e-5) if dtype == torch.float32 else (7e-2, 2e-2)
+    torch.testing.assert_close(actual, expected, atol=atol, rtol=rtol)
+
+
+@pytest.mark.core_model
+@hardware_test(res={"cuda": "L4"}, num_cards=1)
+@torch.no_grad()
+@pytest.mark.parametrize(
+    "activation_dtype,affine_dtype,atol,rtol",
+    [
+        (torch.float32, torch.float32, 1e-5, 1e-5),
+        (torch.bfloat16, torch.float32, 1.5e-1, 3e-2),
+    ],
+)
+def test_wan_rmsnorm_silu_matches_reference(
+    activation_dtype: torch.dtype,
+    affine_dtype: torch.dtype,
+    atol: float,
+    rtol: float,
+) -> None:
+    from vllm_omni.diffusion.vae_optimizations.triton_wan_rmsnorm_silu import wan_rmsnorm_silu
+
+    torch.manual_seed(0)
+    x = torch.randn(1, 96, 3, 10, 14, device="cuda", dtype=activation_dtype).contiguous(
+        memory_format=torch.channels_last_3d
+    )
+    gamma = torch.randn(96, 1, 1, 1, device="cuda", dtype=affine_dtype)
+    bias = torch.randn_like(gamma)
+    normalized = F.normalize(x.float() if activation_dtype != torch.float32 else x, dim=1).to(activation_dtype)
+    expected = F.silu(normalized * 96**0.5 * gamma + bias)
+
+    actual = wan_rmsnorm_silu(x, gamma, bias)
+
+    assert actual is not None
+    assert actual.stride() == x.stride()
+    torch.testing.assert_close(actual, expected, atol=atol, rtol=rtol)
+
+
+@pytest.mark.core_model
+@hardware_test(res={"cuda": "L4"}, num_cards=1)
+@torch.no_grad()
+def test_wan_vae_install_preserves_lossless_path_and_state_dict() -> None:
+    from vllm_omni.diffusion.vae_optimizations.wan import (
+        FusedWanRMSNormSiLU,
+        maybe_optimize_wan_vae,
+    )
+
+    torch.manual_seed(0)
+    vae = _small_wan_vae().cuda().eval()
+    latents = torch.randn(1, 4, 1, 8, 8, device="cuda")
+    parameter_names = {name for name, _ in vae.named_parameters()}
+    state_dict = {name: value.clone() for name, value in vae.state_dict().items()}
+    reference = vae.decode(latents).sample
+
+    maybe_optimize_wan_vae(vae)
+    maybe_optimize_wan_vae(vae)
+
+    assert {name for name, _ in vae.named_parameters()} == parameter_names
+    vae.load_state_dict(state_dict, strict=True)
+    assert any(isinstance(module, FusedWanRMSNormSiLU) for module in vae.modules())
+    assert torch.equal(vae.decode(latents).sample, reference)
+
+    with use_vae_fast_path(vae, True):
+        fast = vae.decode(latents).sample
+
+    torch.testing.assert_close(fast, reference, atol=5e-4, rtol=1e-4)
+    assert torch.equal(vae.decode(latents).sample, reference)
+
+
+@pytest.mark.core_model
+@pytest.mark.cpu
+@pytest.mark.parametrize("parallel_mode", ["auto", "spatial_shard_height", "spatial_shard_width"])
+def test_wan_fast_path_skips_spatial_shard_mode(monkeypatch: pytest.MonkeyPatch, parallel_mode: str) -> None:
+    import vllm_omni.diffusion.vae_optimizations.wan as wan_optimizations
+
+    monkeypatch.setattr(wan_optimizations, "_HAS_TRITON", True)
+    vae = _small_wan_vae()
+    vae.distributed_executor = _DistributedExecutorConfig(parallel_mode)
+
+    wan_optimizations.maybe_optimize_wan_vae(vae)
+
+    assert not any(isinstance(module, wan_optimizations.FusedWanRMSNormSiLU) for module in vae.modules())
+
+
+@pytest.mark.core_model
+@hardware_test(res={"cuda": "L4"}, num_cards=1)
+@torch.no_grad()
+@pytest.mark.parametrize(
+    ("precision", "decode_dtype"),
+    [("fp32", torch.float32), ("fp16", torch.float16), ("bf16", torch.bfloat16)],
+)
+def test_wan_decode_precision_keeps_encode_fp32(precision: str, decode_dtype: torch.dtype) -> None:
+    from vllm_omni.diffusion.distributed.autoencoders.autoencoder_kl_wan import DistributedAutoencoderKLWan
+    from vllm_omni.diffusion.vae_optimizations.precision import configure_wan_decode_precision
+
+    torch.manual_seed(0)
+    vae = _small_wan_vae(DistributedAutoencoderKLWan).to(device="cuda", dtype=torch.bfloat16).eval()
+    configure_wan_decode_precision(vae, precision)
+
+    assert next(vae.encoder.parameters()).dtype == torch.float32
+    assert next(vae.quant_conv.parameters()).dtype == torch.float32
+    assert next(vae.decoder.parameters()).dtype == decode_dtype
+    assert next(vae.post_quant_conv.parameters()).dtype == decode_dtype
+
+    _, grid_spec = vae.tile_split(torch.randn(1, 4, 1, 8, 8, device="cuda", dtype=decode_dtype))
+    assert grid_spec.output_dtype == decode_dtype
+
+    encoded = vae.encode(torch.randn(1, 3, 1, 16, 16, device="cuda", dtype=torch.bfloat16))
+    decoded = vae.decode(torch.randn(1, 4, 1, 8, 8, device="cuda", dtype=torch.float32))
+
+    assert encoded.latent_dist.parameters.dtype == torch.float32
+    assert decoded.sample.dtype == decode_dtype
@@ -3,6 +3,7 @@

 import importlib.util
 from pathlib import Path
+from types import SimpleNamespace

 import pytest

@@ -43,3 +44,11 @@ def test_model_display_names_require_string_mapping(tmp_path, monkeypatch):

    with pytest.raises(ValueError, match="MODEL_DISPLAY_NAMES_FILE"):
        generate_examples.load_model_display_names()
+
+
+def test_only_general_serving_examples_get_generated_pages():
+    general_example = SimpleNamespace(category="offline_inference", path=Path("text_to_image"))
+    model_example = SimpleNamespace(category="offline_inference", path=Path("qwen3_omni"))
+
+    assert generate_examples.is_general_example(general_example)
+    assert not generate_examples.is_general_example(model_example)
@@ -76,6 +76,17 @@ def test_serve_parser_accepts_four_way_cfg_parallelism() -> None:
    assert args.cfg_parallel_size == 4


+def test_serve_parser_accepts_wan_decode_precision() -> None:
+    parser = TrackingArgumentParser()
+    subparsers = parser.add_subparsers(dest="subcommand")
+    OmniServeCommand().subparser_init(subparsers)
+
+    args = parser.parse_args(["serve", "fake-model", "--omni", "--vae-decode-precision", "bf16"])
+
+    assert args.vae_decode_precision == "bf16"
+    assert args.get_explicit_kwargs_dict()["vae_decode_precision"] == "bf16"
+
+
 def _make_headless_args(**kwargs) -> TrackingNamespace:
    defaults = {
        "model": "fake-model",
@@ -123,18 +123,29 @@ def _legacy_detectors(tree: ast.AST) -> list[tuple[int, str]]:
    return found


-def _check(label: str, path: Path, actual: int, budget: int, sample: list[str], budget_name: str) -> list[str]:
+def _check(
+    label: str,
+    path: Path,
+    actual: int,
+    budget: int,
+    sample: list[str],
+    budget_name: str,
+    notices: list[str],
+) -> list[str]:
    if actual > budget:
        listing = "\n".join(f"      {line}" for line in sample)
        return [
            f"{path}: {actual} {label} exceeds the budget of {budget}.\n{listing}\n\n{_GUIDANCE}\n",
        ]
    if actual < budget:
-        return [
-            f"{path}: {actual} {label}, down from a budget of {budget}. "
-            f"Lower {budget_name} to {actual} in {Path(__file__).relative_to(REPO_ROOT)} "
-            f"so the ratchet holds the new ground.",
-        ]
+        # Progress is never a build failure. Removing branches is the goal, so a
+        # PR that does it must not hand-edit this file to stay green -- nor break
+        # `main` for everyone else when it merges from a stale base.
+        notices.append(
+            f"{path}: {actual} {label}, below the budget of {budget}. "
+            f"Lower {budget_name} to {actual} in "
+            f"{Path(__file__).relative_to(REPO_ROOT)} to hold the new ground."
+        )
    return []


@@ -152,6 +163,7 @@ def main(argv: list[str] | None = None) -> int:
            print(f"check_tts_adapter: expected file not found: {path}", file=sys.stderr)
            return 1

+    notices: list[str] = []
    branches = _model_type_branches(ast.parse(serving_path.read_text()))
    detectors = _legacy_detectors(ast.parse(adapters_path.read_text()))

@@ -162,6 +174,7 @@ def main(argv: list[str] | None = None) -> int:
        MAX_MODEL_TYPE_BRANCHES,
        [f"line {lineno}: {src}" for lineno, src in branches],
        "MAX_MODEL_TYPE_BRANCHES",
+        notices,
    )
    errors += _check(
        "unmigrated model types in LEGACY_TTS_DETECTORS",
@@ -170,8 +183,11 @@ def main(argv: list[str] | None = None) -> int:
        MAX_LEGACY_DETECTORS,
        [f"line {lineno}: {name}" for lineno, name in detectors],
        "MAX_LEGACY_DETECTORS",
+        notices,
    )

+    for notice in notices:
+        print(f"check_tts_adapter: {notice}", file=sys.stderr)
    for error in errors:
        print(f"check_tts_adapter: {error}", file=sys.stderr)
    return 1 if errors else 0
@@ -573,6 +573,7 @@ class _DiffusionConfigProjection:
    fa_deterministic: bool = False
    vae_use_slicing: bool = False
    vae_use_tiling: bool = False
+    vae_decode_precision: Literal["fp32", "fp16", "bf16"] | None = None
    mask_strategy_file_path: str | None = None
    skip_time_steps: int = 15
    VSA_sparsity: float = 0.0
@@ -401,6 +401,7 @@ class StageDeployConfig:
    step_execution: bool | None = None
    vae_use_slicing: bool | None = None
    vae_use_tiling: bool | None = None
+    vae_decode_precision: str | None = None
    boundary_ratio: float | None = None
    flow_shift: float | None = None
    diffusion_kv_cache_dtype: str | None = None
@@ -20,6 +20,17 @@ class AttentionBackend(ABC):
    # avoid a slower masked-attention plan when tail padding is not semantic.
    supports_prefix_kv_slicing: bool = False

+    @classmethod
+    def supports_packed_mask_free(cls) -> bool:
+        """Whether packed attention never reads attn_mask on this platform.
+
+        When True, models that pack a [real, pad] two-document layout and
+        carry cu_seqlens/max_seqlen in ``AttentionMetadata.extra`` may skip
+        constructing the padding mask entirely. Backends whose mask-free
+        behavior is platform-dependent must check current_omni_platform.
+        """
+        return False
+
    # ``OmniPlatformEnum`` values this backend runs on; None means unrestricted.
    # Platform resolution rejects an explicit selection outside this set, so a
    # hardware-specific backend fails before the model is built.
@@ -121,6 +132,17 @@ class AttentionMetadata:
    #     the packed cu_seqlens tensors.
    #   "valid_kv_length": int — contiguous valid K/V prefix length for a
    #     backend that advertises supports_prefix_kv_slicing.
+    #   "npu_attn_varlen": bool — model opt-in for the NPU packed varlen path
+    #     (TND npu_fusion_attention driven by cu_seqlens, mask never read).
+    #     Requires the [real, pad] two-document packing contract; see
+    #     FlashAttentionImpl._forward_varlen_packed_npu.
+    #   "laser_input_scale": float — model opt-in input pre-scale for the NPU
+    #     ascend_laser_attention path. The kernel stores unscaled QK^T in an
+    #     fp16 workspace, so outlier activations overflow 65504 into NaN rows;
+    #     with this set (>1), q/k/v are divided by the factor before the op,
+    #     the kernel scale_value is multiplied by its square, and the output
+    #     is scaled back (exact for power-of-two factors). Absent means no
+    #     pre-scaling. See FlashAttentionImpl._forward_prefix_kv_slice_npu.

    # Piecewise attention metadata (mixed causal/full masks).
    # full_attn_spans: per-sample [start, end) spans in global coordinates using full attention.
@@ -1,6 +1,7 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project

+import os
 from functools import partial

 import torch
@@ -16,6 +17,7 @@ from vllm_omni.diffusion.attention.backends.utils.piecewise_attn import (
    piecewise_attn,
 )
 from vllm_omni.diffusion.config import get_current_diffusion_config_or_none
+from vllm_omni.platforms import current_omni_platform

 logger = init_logger(__name__)

@@ -24,6 +26,15 @@ class FlashAttentionBackend(AttentionBackend):
    accept_output_buffer: bool = True
    supports_piecewise_spans: bool = True

+    @classmethod
+    def supports_packed_mask_free(cls) -> bool:
+        # CUDA: forward_cuda dispatches the packed cu_seqlens varlen path
+        # before ever reading attn_mask. NPU: forward_fa_npu honors the
+        # npu_attn_varlen opt-in and rebuilds the padding mask itself if the
+        # packed contract fails. XPU reads attn_mask, so models must keep
+        # constructing it there.
+        return current_omni_platform.is_cuda() or current_omni_platform.is_npu()
+
    @classmethod
    def supports_attention_mask(cls) -> bool:
        return True
@@ -415,7 +426,36 @@ class FlashAttentionImpl(AttentionImpl):
                "For installation details, see https://gitcode.com/Ascend/MindIE-SD"
                "Otherwise, use SDPA backend by setting DIFFUSION_ATTENTION_BACKEND=TORCH_SDPA"
            )
+        # Opt-in mask-free paths (mirror the CUDA cu_seqlens behavior): the
+        # model marks extra["npu_attn_varlen"] and carries packed metadata, so
+        # the padding document is excluded without reading or materializing
+        # the quadratic full_qk mask.
+        #   - default: packed varlen via mindiesd attention_forward_varlen;
+        #   - when MINDIE_SD_FA_TYPE=ascend_laser_attention: prefix K/V
+        #     slicing via mindiesd attention_forward (so the env-selected op
+        #     type is honored; the TND varlen op cannot be).
+        extra = attn_metadata.extra if attn_metadata else {}
+        if extra.get("npu_attn_varlen", False):
+            if os.environ.get("MINDIE_SD_FA_TYPE") == "ascend_laser_attention":
+                out = self._forward_prefix_kv_slice_npu(query, key, value, extra)
+            else:
+                out = self._forward_varlen_packed_npu(query, key, value, extra)
+            if out is not None:
+                return out
        attention_mask = attn_metadata.attn_mask if attn_metadata else None
+        if attention_mask is None and extra.get("npu_attn_varlen", False):
+            # Models skip mask construction when this opt-in is set
+            # (FlashAttentionBackend.supports_packed_mask_free). The packed
+            # paths above declined (contract mismatch), so rebuild the padding
+            # mask here to keep the masked fallback correct.
+            used = extra.get("valid_kv_length")
+            if not isinstance(used, int) or not 0 < used <= query.shape[1]:
+                raise ValueError(
+                    "npu_attn_varlen packed metadata is unusable and no attn_mask "
+                    f"was constructed (valid_kv_length={used!r}, seq_len={query.shape[1]}); "
+                    "refusing to run unmasked attention over padding rows."
+                )
+            attention_mask = torch.arange(query.shape[1], device=query.device)[None] < used

        # NPU aclnnFlashAttentionScore requires mask shape to be one of:
        # [B, N, Sq, Skv], [B, 1, Sq, Skv], [1, 1, Sq, Skv], or [Sq, Skv]
@@ -433,3 +473,161 @@ class FlashAttentionImpl(AttentionImpl):
            op_type="fused_attn_score",
            layout=layout,
        )
+
+    def _resolve_packed_seq_npu(
+        self,
+        query: torch.Tensor,
+        key: torch.Tensor,
+        extra: dict,
+    ) -> tuple[list[int], list[int]] | None:
+        """Resolve packed document boundaries (cumulative end offsets per doc).
+
+        Returns (seq_q, seq_k), e.g. ([used_q, total_q], [used_k, total_k]),
+        or None unless the metadata matches the exact contract MiniMax-H3
+        produces:
+          - q/k/v are [1, T, N, D] (BSND, single packed batch);
+          - cu_seqlens describe a [real, pad] two-document packing, with the
+            padding document as a strict suffix;
+          - max_seqlen_q/k are Python ints equal to the real document length
+            (so boundaries are derived without any device sync).
+        """
+        if self.causal:
+            return None
+        packed_keys = ("cu_seqlens_q", "cu_seqlens_k", "max_seqlen_q", "max_seqlen_k")
+        if any(k not in extra for k in packed_keys):
+            return None
+        cu_q, cu_k = extra["cu_seqlens_q"], extra["cu_seqlens_k"]
+        used_q, used_k = extra["max_seqlen_q"], extra["max_seqlen_k"]
+        if query.shape[0] != 1 or key.shape[0] != 1:
+            return None
+        if not isinstance(used_q, int) or not isinstance(used_k, int):
+            return None
+        total_q, total_k = query.shape[1], key.shape[1]
+        # .shape is host-side metadata: counting documents never syncs.
+        if cu_q.shape[0] != cu_k.shape[0] or cu_q.shape[0] > 3:
+            return None
+        if not (0 < used_q <= total_q) or not (0 < used_k <= total_k):
+            return None
+        if used_q == total_q and used_k == total_k:
+            # No padding document: one full-length document.
+            return [total_q], [total_k]
+        if cu_q.shape[0] == 3 and used_q >= total_q - used_q and used_k >= total_k - used_k:
+            # [real, pad] packing: the real document must be the longer one
+            # (consistent with the max_seqlen naming).
+            return [used_q, total_q], [used_k, total_k]
+        return None
+
+    def _forward_varlen_packed_npu(
+        self,
+        query: torch.Tensor,
+        key: torch.Tensor,
+        value: torch.Tensor,
+        extra: dict,
+    ) -> torch.Tensor | None:
+        """Packed varlen attention on NPU via mindiesd attention_forward_varlen.
+
+        Returns None (caller falls back to the mask path) when the packed
+        contract does not hold; see _resolve_packed_seq_npu.
+        """
+        resolved = self._resolve_packed_seq_npu(query, key, extra)
+        if resolved is None:
+            return None
+        seq_q, seq_k = resolved
+
+        try:
+            from mindiesd import attention_forward_varlen
+        except ImportError:
+            raise ImportError(
+                "FlashAttentionBackend NPU implementation requires MindIE-SD. "
+                "Please install MindIE-SD to enable NPU attention support. "
+                "For installation details, see https://gitcode.com/Ascend/MindIE-SD"
+                "Otherwise, use SDPA backend by setting DIFFUSION_ATTENTION_BACKEND=TORCH_SDPA"
+            )
+        q = query.squeeze(0)  # [T, N, D] == TND
+        k = key.squeeze(0)
+        v = value.squeeze(0)
+        # attention_forward_varlen wants cu_seqlens as host lists of cumulative
+        # offsets and forwards cu[1:] as actual_seq_qlen/actual_seq_kvlen.
+        out = attention_forward_varlen(
+            q,
+            k,
+            v,
+            [0, *seq_q],
+            [0, *seq_k],
+            softmax_scale=self.softmax_scale,
+        )
+        return out.unsqueeze(0)
+
+    def _forward_prefix_kv_slice_npu(
+        self,
+        query: torch.Tensor,
+        key: torch.Tensor,
+        value: torch.Tensor,
+        extra: dict,
+    ) -> torch.Tensor | None:
+        """Mask-free attention by slicing K/V to the valid prefix (zero-copy
+        views), then mindiesd attention_forward without a mask.
+
+        Same numerical contract as the varlen path: the padding document is a
+        strict suffix, so dropping it from K/V is identical to masking it out.
+        Query keeps full length; outputs on padding rows are never consumed
+        downstream. Returns None when the packed contract does not hold.
+
+        Used when MINDIE_SD_FA_TYPE=ascend_laser_attention: the TND varlen op
+        cannot honor the env-selected op type, while attention_forward can.
+
+        The laser kernel stores unscaled S=QK^T in an fp16 GM workspace, so
+        bf16 activations with large outliers overflow 65504 into ±inf and the
+        affected rows turn NaN. Models may opt into exact power-of-two input
+        pre-scaling via extra["laser_input_scale"] (see abstract.py): q/k/v
+        are divided by the factor, the kernel scale_value is multiplied by its
+        square, and the output is scaled back. Absent or invalid factor means
+        no pre-scaling.
+        """
+        resolved = self._resolve_packed_seq_npu(query, key, extra)
+        if resolved is None:
+            return None
+        _, seq_k = resolved
+        used_k = seq_k[0]  # real document length (first cumulative end)
+
+        try:
+            from mindiesd import attention_forward
+        except ImportError:
+            raise ImportError(
+                "FlashAttentionBackend NPU implementation requires MindIE-SD. "
+                "Please install MindIE-SD to enable NPU attention support. "
+                "For installation details, see https://gitcode.com/Ascend/MindIE-SD"
+                "Otherwise, use SDPA backend by setting DIFFUSION_ATTENTION_BACKEND=TORCH_SDPA"
+            )
+        # mindiesd always takes BSND input regardless of `layout`; the arg
+        # selects the op-internal layout and laser only supports BNSD, so do
+        # NOT forward the model's qkv_layout ("BSND" for MiniMax-H3) here.
+        layout = "BNSD"
+        key = key[:, :used_k]
+        value = value[:, :used_k]
+
+        input_scale = extra.get("laser_input_scale")
+        preserve_input_range = isinstance(input_scale, (int, float)) and input_scale > 1
+        if preserve_input_range:
+            query = query / input_scale
+            key = key / input_scale
+            value = value / input_scale
+        scale = self.softmax_scale
+        if preserve_input_range:
+            scale *= input_scale**2
+
+        def _postprocess(out: torch.Tensor) -> torch.Tensor:
+            return out * input_scale if preserve_input_range else out
+
+        return _postprocess(
+            attention_forward(
+                query,
+                key,
+                value,
+                attn_mask=None,
+                opt_mode="manual",
+                op_type="fused_attn_score",  # MINDIE_SD_FA_TYPE env overrides this
+                layout=layout,
+                scale=scale,
+            )
+        )
@@ -128,9 +128,9 @@ class RainFusionAttentionImpl(AttentionImpl):
    whose extent the model publishes as ``AttentionMetadata.video_layout``. Every
    other case — warmup denoise steps, exempt layers, a layer that does not declare
    ``qkv_layout="BSND"``, sequences without a published video segment, video
-    segments too short to pay for block selection or not aligned to the kernel's
-    block size — delegates to FlashAttention, so a model can select this backend
-    unconditionally.
+    segments too short to pay for block selection — delegates to FlashAttention,
+    so a model can select this backend unconditionally. MindIE-SD handles an
+    irregular video tail internally, retaining it outside the sparse blocks.
    """

    def __init__(
@@ -292,25 +292,6 @@ class RainFusionAttentionImpl(AttentionImpl):
                _MIN_VIDEO_BLOCKS,
            )
            return None
-        if video_len % _BLOCK_SIZE != 0:
-            # rf_v2 tiles the sequence into ceil((video + prefix) / 128) blocks but
-            # builds its block mask from the two segments pooled apart. The counts
-            # only agree when the video segment ends on a block boundary; otherwise
-            # a prefix block is forced on out of range, and the block straddling the
-            # seam holds real prefix rows that selection may drop. Padding the
-            # sequence to realign it is not an option because rf_v2 ignores
-            # attn_mask, so pad keys would take a share of every softmax denominator.
-            logger.warning_once(
-                "RAINFUSION_ATTN staying dense: latent grid %s gives %d video rows, which is not a "
-                "multiple of the %d-row kernel block. Choose a geometry whose latent_t * h * w is a "
-                "multiple of %d to enable sparsity.",
-                tuple(latent_shape),
-                video_len,
-                _BLOCK_SIZE,
-                _BLOCK_SIZE,
-            )
-            return None
-
        logger.info_once(
            "RAINFUSION_ATTN active: sparsity=%.2f, start_step=%d, exempt_layers=%d, "
            "latent_grid=%s, prefix_rows=%d, video_rows=%d. Realized sparsity is lower than nominal "
@@ -711,6 +711,9 @@ class OmniDiffusionConfig:
    # VAE memory optimization parameters
    vae_use_slicing: bool = False
    vae_use_tiling: bool = False
+    # Optional Wan decode-only precision. When set, the VAE encoder remains
+    # FP32 while decoder-side modules use the selected precision.
+    vae_decode_precision: str | None = None

    # STA (Sliding Tile Attention) parameters
    mask_strategy_file_path: str | None = None
@@ -923,6 +926,11 @@ class OmniDiffusionConfig:
        )

    def __post_init__(self):
+        if self.vae_decode_precision not in (None, "fp32", "fp16", "bf16"):
+            raise ValueError(
+                "vae_decode_precision must be one of {'fp32', 'fp16', 'bf16'} or None, "
+                f"got {self.vae_decode_precision!r}"
+            )
        if self.diffusion_compile_granularity not in {"regional", "full"}:
            raise ValueError(
                "diffusion_compile_granularity must be 'regional' or 'full', "
@@ -106,7 +106,7 @@ class DistributedAutoencoderKLWan(OmniAutoencoderKLWan, DistributedVaeMixin):
            split_dims=(3, 4),
            grid_shape=(tiletask_list[-1].grid_coord[0] + 1, tiletask_list[-1].grid_coord[1] + 1),
            tile_spec=tile_spec,
-            output_dtype=self.dtype,
+            output_dtype=getattr(self, "_vllm_decode_dtype", self.dtype),
        )
        return tiletask_list, grid_spec

@@ -68,6 +68,7 @@ class RMSNorm(CustomOp):
    def __init__(self, hidden_size: int, eps: float = 1e-6, dtype: torch.dtype = torch.float32) -> None:
        super().__init__()
        self.weight = nn.Parameter(torch.ones(hidden_size, dtype=dtype))
+        self.hidden_size = hidden_size
        self.variance_epsilon = eps

    def _forward_fused(self, x: torch.Tensor) -> torch.Tensor:
@@ -107,6 +108,19 @@ class RMSNorm(CustomOp):
        except Exception:
            return self.forward_native(x)

+    def forward_musa(
+        self,
+        x: torch.Tensor,
+    ) -> torch.Tensor:
+        # Preserve the aten::rms_norm graph so dynamic Inductor can fuse the
+        # H3 Q/K norm with its inline RoPE path on MUSA.
+        return F.rms_norm(
+            x,
+            (self.hidden_size,),
+            self.weight,
+            self.variance_epsilon,
+        )
+
    def forward_npu(
        self,
        x: torch.Tensor,
@@ -193,7 +193,16 @@ class RotaryEmbedding(CustomOp):
        cos: torch.Tensor,
        sin: torch.Tensor,
    ) -> torch.Tensor:
-        return self.forward_native(x, cos, sin)
+        # Keep the full-dimension NeoX form inline for MUSA compilation.
+        # Other layouts use the shared native implementation.
+        if self.half_head_dim or self.interleaved:
+            return self.forward_native(x, cos, sin)
+        if cos.dim() == 3:
+            cos, sin = cos[0], sin[0]
+        cos = cos.unsqueeze(-2)
+        sin = sin.unsqueeze(-2)
+        x1, x2 = x.chunk(2, dim=-1)
+        return x * cos + torch.cat((-x2, x1), dim=-1) * sin

    def forward_native(
        self,
@@ -107,6 +107,15 @@ MINIMAX_H3_FP32_BUFFER_NAMES = frozenset({"rope.inv_freq"})
 # lookup and masked out afterwards).
 MINIMAX_H3_ADALN_MODALITY_NUM = 3

+# Opt-in fp16-range protection for the NPU ascend_laser_attention kernel
+# (consumed only via the "laser_input_scale" extra key; other backends and
+# platforms ignore it). The kernel stores unscaled QK^T in an fp16 GM
+# workspace, and H3's outlier activations (per-element amax in the hundreds)
+# push dot products past fp16 max 65504, turning whole 128-row blocks NaN.
+# 256 is a power of two, so pre-dividing q/k/v and the compensating
+# kernel-scale/output multiplies are exact in floating point.
+MINIMAX_H3_LASER_INPUT_SCALE = 256.0
+

 def _required_kwarg(kwargs: dict[str, Any], key: str) -> Any:
    if key not in kwargs or kwargs[key] is None:
@@ -333,6 +342,7 @@ class MiniMaxH3Attention(nn.Module):
        )
        self.num_heads = self.qkv_proj.num_heads
        self.num_kv_heads = self.qkv_proj.num_kv_heads
+        self.rot_dim = 6 * arch.rope_inv_freq_len
        self.q_norm = _norm(arch.attention_head_dim, eps=arch.qk_norm_eps)
        self.k_norm = _norm(arch.attention_head_dim, eps=arch.qk_norm_eps)
        self.rope = RotaryEmbedding(is_neox_style=True, half_head_dim=False)
@@ -365,7 +375,7 @@ class MiniMaxH3Attention(nn.Module):
        x: [T, heads, head_dim]; freqs: [T, rot_dim]. In the unfused path, cos/sin
        are cast to the activation dtype before the elementwise math.
        """
-        rot_dim = freqs.shape[-1]
+        rot_dim = self.rot_dim
        x_rot, x_pass = x[..., :rot_dim], x[..., rot_dim:]
        cos = torch.cos(freqs).to(x.dtype)  # [T, rot_dim]
        sin = torch.sin(freqs).to(x.dtype)
@@ -401,11 +411,16 @@ class MiniMaxH3Attention(nn.Module):
        used = min(max_seqlen, packed_total)
        attn_mask = None
        # Ring attention can dispatch to a different implementation from the
-        # configured backend, so this no-mask fast path is local-only.
-        prefix_slice = (
-            not getattr(self.attention, "use_ring", False) and self.attention.attn_backend.supports_prefix_kv_slicing
+        # configured backend, so the no-mask fast paths are local-only.
+        # supports_prefix_kv_slicing: backend slices K/V itself (cuDNN).
+        # supports_packed_mask_free: backend consumes the packed metadata
+        # without ever reading attn_mask (CUDA packed varlen, NPU
+        # npu_attn_varlen opt-in with its own fallback rebuild).
+        no_mask = not getattr(self.attention, "use_ring", False) and (
+            self.attention.attn_backend.supports_prefix_kv_slicing
+            or self.attention.attn_backend.supports_packed_mask_free()
        )
-        if used < packed_total and not prefix_slice:
+        if used < packed_total and not no_mask:
            attn_mask = torch.arange(packed_total, device=q.device)[None] < used
        metadata = AttentionMetadata(
            attn_mask=attn_mask,
@@ -415,6 +430,15 @@ class MiniMaxH3Attention(nn.Module):
                "max_seqlen_q": max_seqlen,
                "max_seqlen_k": max_seqlen,
                "valid_kv_length": used,
+                # Opt the NPU flash backend into the packed varlen path so the
+                # quadratic full_qk mask is never materialized. Ring attention
+                # is excluded: it keeps the aligned padding rows for its
+                # fixed-size P2P buffers and still needs the mask.
+                "npu_attn_varlen": not getattr(self.attention, "use_ring", False),
+                # fp16-range protection for the ascend_laser_attention kernel
+                # (see MINIMAX_H3_LASER_INPUT_SCALE). Ignored by every other
+                # backend/path.
+                "laser_input_scale": MINIMAX_H3_LASER_INPUT_SCALE,
            },
            video_layout=video_layout,
        )
@@ -0,0 +1,19 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Optional VAE decoder optimizations."""
+
+from vllm_omni.diffusion.vae_optimizations.gate import (
+    use_pipeline_vae_fast_path,
+    use_vae_fast_path,
+)
+from vllm_omni.diffusion.vae_optimizations.optimize import (
+    clear_pipeline_vae_fast_path_caches,
+    optimize_pipeline_vaes,
+)
+
+__all__ = [
+    "clear_pipeline_vae_fast_path_caches",
+    "optimize_pipeline_vaes",
+    "use_pipeline_vae_fast_path",
+    "use_vae_fast_path",
+]
@@ -0,0 +1,67 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Decode-scoped gates for optional, non-bit-exact VAE fast paths."""
+
+from __future__ import annotations
+
+from collections.abc import Iterator
+from contextlib import ExitStack, contextmanager
+from weakref import WeakKeyDictionary
+
+from torch import nn
+
+from vllm_omni.diffusion.offloader.module_collector import ModuleDiscovery
+
+
+class VaeFastPathGate:
+    """Mutable flag shared by wrappers installed on one VAE instance."""
+
+    __slots__ = ("enabled",)
+
+    def __init__(self) -> None:
+        self.enabled = False
+
+
+_VAE_FAST_PATH_GATES: WeakKeyDictionary[nn.Module, VaeFastPathGate] = WeakKeyDictionary()
+
+
+def register_vae_fast_path_gate(vae: nn.Module, gate: VaeFastPathGate) -> None:
+    """Associate an installed fast path with its owning VAE."""
+
+    _VAE_FAST_PATH_GATES[vae] = gate
+
+
+@contextmanager
+def use_vae_fast_path(vae: nn.Module, enabled: bool) -> Iterator[None]:
+    """Enable an installed fast path for one decode and always restore it."""
+
+    gate = _VAE_FAST_PATH_GATES.get(vae)
+    if gate is None:
+        yield
+        return
+
+    previous_enabled = gate.enabled
+    gate.enabled = enabled
+    try:
+        yield
+    finally:
+        gate.enabled = previous_enabled
+
+
+@contextmanager
+def use_pipeline_vae_fast_path(pipeline: nn.Module, enabled: bool) -> Iterator[None]:
+    """Apply one request's quality choice to every declared VAE component."""
+
+    vaes = ModuleDiscovery.discover(pipeline).vaes
+    with ExitStack() as stack:
+        for vae in vaes:
+            stack.enter_context(use_vae_fast_path(vae, enabled))
+        yield
+
+
+__all__ = [
+    "VaeFastPathGate",
+    "register_vae_fast_path_gate",
+    "use_pipeline_vae_fast_path",
+    "use_vae_fast_path",
+]
@@ -0,0 +1,354 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Quality-gated CUDA fast paths for Diffusers-style 2D KL VAE decoders."""
+
+from __future__ import annotations
+
+from types import MethodType
+from typing import Any
+
+import torch
+import torch.nn.functional as F
+from torch import nn
+from vllm.logger import init_logger
+
+from vllm_omni.diffusion.vae_optimizations.gate import (
+    VaeFastPathGate,
+    register_vae_fast_path_gate,
+)
+
+logger = init_logger(__name__)
+
+try:
+    from vllm_omni.diffusion.vae_optimizations.triton_group_norm_silu import (
+        group_norm_silu_4d,
+        group_norm_silu_rows,
+    )
+
+    _HAS_TRITON = True
+except ImportError:  # pragma: no cover - exercised on builds without Triton
+    _HAS_TRITON = False
+
+
+class FusedGroupNormSiLU(nn.Module):
+    """Quality-gated GroupNorm + SiLU with an exact eager fallback."""
+
+    def __init__(self, norm: nn.GroupNorm, gate: VaeFastPathGate) -> None:
+        super().__init__()
+        self.num_groups = norm.num_groups
+        self.num_channels = norm.num_channels
+        self.eps = norm.eps
+        self.affine = norm.affine
+        # Register parameters directly so state-dict FQNs do not change.
+        self.weight = norm.weight
+        self.bias = norm.bias
+        self._vllm_gate = gate
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if self._vllm_gate.enabled and x.dim() == 4 and not torch.compiler.is_compiling():
+            output = group_norm_silu_4d(
+                x,
+                self.weight,
+                self.bias,
+                self.num_groups,
+                self.eps,
+                apply_silu=True,
+            )
+            if output is not None:
+                return output
+        return F.silu(F.group_norm(x, self.num_groups, self.weight, self.bias, self.eps))
+
+
+def _install_norm_silu(decoder: nn.Module, resnet_cls: type[nn.Module], gate: VaeFastPathGate) -> int:
+    def is_fusable(norm: object) -> bool:
+        return type(norm) is nn.GroupNorm and norm.affine and norm.weight is not None and norm.bias is not None
+
+    count = 0
+    for module in decoder.modules():
+        if (
+            type(module) is resnet_cls
+            and module.time_emb_proj is None
+            and module.time_embedding_norm in ("default", "group")
+            and module.upsample is None
+            and module.downsample is None
+            and isinstance(module.nonlinearity, nn.SiLU)
+            and is_fusable(module.norm1)
+            and is_fusable(module.norm2)
+        ):
+            module.norm1 = FusedGroupNormSiLU(module.norm1, gate)
+            module.norm2 = FusedGroupNormSiLU(module.norm2, gate)
+            module.nonlinearity = nn.Identity()
+            count += 2
+
+    if (
+        type(getattr(decoder, "conv_norm_out", None)) is nn.GroupNorm
+        and isinstance(getattr(decoder, "conv_act", None), nn.SiLU)
+        and is_fusable(decoder.conv_norm_out)
+    ):
+        decoder.conv_norm_out = FusedGroupNormSiLU(decoder.conv_norm_out, gate)
+        decoder.conv_act = nn.Identity()
+        count += 1
+    return count
+
+
+_UPSAMPLE_TAP_MAP = {0: (2,), 1: (1, 2), 2: (0, 1), 3: (0,)}
+
+
+def _fold_upsample2x_conv2d_weight(conv: nn.Conv2d) -> torch.Tensor:
+    weight = conv.weight.detach().float()
+    output_channels, input_channels = weight.shape[:2]
+    folded = weight.new_zeros(input_channels, output_channels, 4, 4)
+    for row in range(4):
+        for column in range(4):
+            value = weight.new_zeros(output_channels, input_channels)
+            for source_row in _UPSAMPLE_TAP_MAP[row]:
+                for source_column in _UPSAMPLE_TAP_MAP[column]:
+                    value += weight[:, :, source_row, source_column]
+            folded[:, :, row, column] = value.t()
+    folded = folded.to(conv.weight.dtype)
+    if conv.weight.is_contiguous(memory_format=torch.channels_last):
+        folded = folded.contiguous(memory_format=torch.channels_last)
+    return folded.to(conv.weight.device)
+
+
+class FusedUpsample2xConv2d(nn.Module):
+    """Quality-gated nearest-2x + Conv2d replacement using ConvTranspose2d."""
+
+    def __init__(self, upsample: nn.Module, gate: VaeFastPathGate) -> None:
+        super().__init__()
+        # Keep only ``conv`` registered so checkpoint parameter names stay
+        # ``...upsamplers.N.conv.*``. The original module is an eager fallback.
+        object.__setattr__(self, "_original", upsample)
+        self.conv = upsample.conv
+        self.channels = upsample.channels
+        self._vllm_gate = gate
+        self.register_buffer("_fused_weight", None, persistent=False)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        output_size: int | None = None,
+        *args: Any,
+        **kwargs: Any,
+    ) -> torch.Tensor:
+        if not self._vllm_gate.enabled or output_size is not None or hidden_states.shape[1] != self.channels:
+            return self._original(hidden_states, output_size=output_size, *args, **kwargs)
+
+        weight = self._fused_weight
+        if weight is None or weight.device != self.conv.weight.device or weight.dtype != self.conv.weight.dtype:
+            weight = _fold_upsample2x_conv2d_weight(self.conv)
+            self._fused_weight = weight
+        return F.conv_transpose2d(hidden_states, weight, self.conv.bias, stride=2, padding=1)
+
+
+def _install_fused_upsample(decoder: nn.Module, upsample_cls: type[nn.Module], gate: VaeFastPathGate) -> int:
+    count = 0
+    for block in decoder.up_blocks:
+        upsamplers = getattr(block, "upsamplers", None)
+        if not upsamplers:
+            continue
+        for index, upsample in enumerate(upsamplers):
+            if type(upsample) is not upsample_cls:
+                continue
+            conv = getattr(upsample, "conv", None)
+            if (
+                upsample.use_conv
+                and not upsample.use_conv_transpose
+                and upsample.interpolate
+                and upsample.norm is None
+                and upsample.name == "conv"
+                and type(conv) is nn.Conv2d
+                and conv.kernel_size == (3, 3)
+                and conv.stride == (1, 1)
+                and conv.padding == (1, 1)
+                and conv.dilation == (1, 1)
+                and conv.groups == 1
+                and conv.padding_mode == "zeros"
+            ):
+                upsamplers[index] = FusedUpsample2xConv2d(upsample, gate)
+                count += 1
+    return count
+
+
+def _fold_attention_value_projection(module: nn.Module) -> tuple[torch.Tensor, torch.Tensor]:
+    value_weight = module.to_v.weight.detach().float()
+    value_bias = module.to_v.bias.detach().float()
+    output_weight = module.to_out[0].weight.detach().float()
+    output_bias = module.to_out[0].bias.detach().float()
+    dtype = module.to_v.weight.dtype
+    return (
+        (output_weight @ value_weight).to(dtype).contiguous(),
+        (output_weight @ value_bias + output_bias).to(dtype),
+    )
+
+
+def _attention_fast_forward(
+    self: nn.Module,
+    hidden_states: torch.Tensor,
+    encoder_hidden_states: torch.Tensor | None = None,
+    attention_mask: torch.Tensor | None = None,
+    temb: torch.Tensor | None = None,
+    **cross_attention_kwargs: Any,
+) -> torch.Tensor:
+    if (
+        not self._vllm_gate.enabled
+        or encoder_hidden_states is not None
+        or attention_mask is not None
+        or temb is not None
+        or hidden_states.ndim != 4
+    ):
+        return type(self).forward(
+            self,
+            hidden_states,
+            encoder_hidden_states=encoder_hidden_states,
+            attention_mask=attention_mask,
+            temb=temb,
+            **cross_attention_kwargs,
+        )
+
+    residual = hidden_states
+    batch_size, channels, height, width = hidden_states.shape
+    hidden_states = hidden_states.permute(0, 2, 3, 1).reshape(batch_size, height * width, channels)
+
+    if self.group_norm is not None:
+        norm = self.group_norm
+        normalized = None
+        if not torch.compiler.is_compiling():
+            normalized = group_norm_silu_rows(
+                hidden_states,
+                norm.weight,
+                norm.bias,
+                norm.num_groups,
+                norm.eps,
+                apply_silu=False,
+            )
+        hidden_states = normalized if normalized is not None else norm(hidden_states.transpose(1, 2)).transpose(1, 2)
+
+    query = self.to_q(hidden_states)
+    key = self.to_k(hidden_states)
+    folded_weight = self._vllm_folded_value_weight
+    folded_bias = self._vllm_folded_value_bias
+    if (
+        folded_weight is None
+        or folded_bias is None
+        or folded_weight.device != hidden_states.device
+        or folded_weight.dtype != self.to_v.weight.dtype
+    ):
+        folded_weight, folded_bias = _fold_attention_value_projection(self)
+        self._vllm_folded_value_weight = folded_weight
+        self._vllm_folded_value_bias = folded_bias
+    value = F.linear(hidden_states, folded_weight, folded_bias)
+
+    output = F.scaled_dot_product_attention(query.unsqueeze(1), key.unsqueeze(1), value.unsqueeze(1))
+    output = output.squeeze(1).to(query.dtype)
+    output = self.to_out[1](output)
+    output = output.reshape(batch_size, height, width, channels).permute(0, 3, 1, 2)
+    if self.residual_connection:
+        output = output + residual
+    return output / self.rescale_output_factor
+
+
+def _attention_fast_compatible(module: nn.Module, attention_cls: type, processor_cls: type) -> bool:
+    return (
+        type(module) is attention_cls
+        and isinstance(module.processor, processor_cls)
+        and module.heads == 1
+        and module.scale_qk
+        and module.spatial_norm is None
+        and not module.norm_cross
+        and getattr(module, "norm_q", None) is None
+        and getattr(module, "norm_k", None) is None
+        and getattr(module, "add_k_proj", None) is None
+        and type(module.to_q) is nn.Linear
+        and type(module.to_k) is nn.Linear
+        and type(module.to_v) is nn.Linear
+        and type(module.to_out[0]) is nn.Linear
+        and isinstance(module.to_out[1], nn.Dropout)
+        and module.to_v.bias is not None
+        and module.to_out[0].bias is not None
+    )
+
+
+def _decoder_layout_forward(self: nn.Module, *args: Any, **kwargs: Any) -> torch.Tensor:
+    use_channels_last = self._vllm_gate.enabled
+    if use_channels_last != self._vllm_channels_last:
+        memory_format = torch.channels_last if use_channels_last else torch.contiguous_format
+        self.to(memory_format=memory_format)
+        self._vllm_channels_last = use_channels_last
+        logger.debug(
+            "%s decoder switched to %s layout",
+            self._vllm_label,
+            "channels_last" if use_channels_last else "contiguous",
+        )
+    return type(self).forward(self, *args, **kwargs)
+
+
+def maybe_optimize_image_vae(vae: nn.Module) -> nn.Module:
+    """Install image VAE decoder fast paths when the structure is fully supported."""
+
+    if not _HAS_TRITON or getattr(vae, "_vllm_vae_fast_path_installed", False):
+        return vae
+
+    from diffusers.models.attention_processor import Attention, AttnProcessor2_0
+    from diffusers.models.autoencoders import AutoencoderKL
+    from diffusers.models.autoencoders.vae import Decoder
+    from diffusers.models.resnet import ResnetBlock2D
+    from diffusers.models.upsampling import Upsample2D
+
+    supported_types: tuple[type[nn.Module], ...] = (AutoencoderKL,)
+    try:
+        from diffusers.models.autoencoders.autoencoder_kl_flux2 import AutoencoderKLFlux2
+
+        supported_types += (AutoencoderKLFlux2,)
+    except ImportError:  # pragma: no cover - older supported Diffusers builds
+        pass
+
+    if not isinstance(vae, supported_types):
+        return vae
+    decoder = getattr(vae, "decoder", None)
+    if type(decoder) is not Decoder:
+        return vae
+
+    attention_modules = [
+        module for module in decoder.modules() if _attention_fast_compatible(module, Attention, AttnProcessor2_0)
+    ]
+    attention_count = sum(isinstance(module, Attention) for module in decoder.modules())
+    if len(attention_modules) != attention_count:
+        logger.info(
+            "%s has %d/%d attention blocks without a layout-safe rewrite; skipping VAE fast paths",
+            type(vae).__name__,
+            attention_count - len(attention_modules),
+            attention_count,
+        )
+        return vae
+
+    gate = VaeFastPathGate()
+    decoder._vllm_gate = gate
+    decoder._vllm_label = type(vae).__name__
+    decoder._vllm_channels_last = False
+    decoder.forward = MethodType(_decoder_layout_forward, decoder)
+    upsample_count = _install_fused_upsample(decoder, Upsample2D, gate)
+    for module in attention_modules:
+        module._vllm_gate = gate
+        module.register_buffer("_vllm_folded_value_weight", None, persistent=False)
+        module.register_buffer("_vllm_folded_value_bias", None, persistent=False)
+        module.forward = MethodType(_attention_fast_forward, module)
+    norm_count = _install_norm_silu(decoder, ResnetBlock2D, gate)
+    register_vae_fast_path_gate(vae, gate)
+    vae._vllm_vae_fast_path_installed = True
+    logger.info(
+        "%s installed quality-gated VAE decoder fast paths "
+        "(%d upsamplers, %d attention blocks, %d GroupNorm+SiLU fusions)",
+        type(vae).__name__,
+        upsample_count,
+        len(attention_modules),
+        norm_count,
+    )
+    return vae
+
+
+__all__ = [
+    "FusedGroupNormSiLU",
+    "FusedUpsample2xConv2d",
+    "maybe_optimize_image_vae",
+]
@@ -0,0 +1,70 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Pipeline-level installation for optional VAE optimizations."""
+
+from __future__ import annotations
+
+from torch import nn
+from vllm.logger import init_logger
+
+from vllm_omni.diffusion.data import OmniDiffusionConfig
+from vllm_omni.diffusion.offloader.module_collector import ModuleDiscovery
+from vllm_omni.diffusion.vae_optimizations.image import maybe_optimize_image_vae
+from vllm_omni.diffusion.vae_optimizations.precision import configure_wan_decode_precision
+from vllm_omni.diffusion.vae_optimizations.wan import maybe_optimize_wan_vae
+from vllm_omni.platforms import current_omni_platform
+
+logger = init_logger(__name__)
+
+_FOLDED_CACHE_NAMES = (
+    "_fused_weight",
+    "_vllm_folded_value_weight",
+    "_vllm_folded_value_bias",
+)
+
+
+def optimize_pipeline_vaes(pipeline: nn.Module, od_config: OmniDiffusionConfig) -> None:
+    """Configure declared VAE components once after checkpoint loading."""
+
+    precision = getattr(od_config, "vae_decode_precision", None)
+    vaes = ModuleDiscovery.discover(pipeline).vaes
+    for vae in vaes:
+        # An explicit precision override is a serving contract, so failures on
+        # a recognized Wan VAE must abort startup instead of being hidden.
+        configure_wan_decode_precision(vae, precision)
+
+    if not current_omni_platform.is_cuda():
+        return
+
+    for vae in vaes:
+        for optimizer in (maybe_optimize_image_vae, maybe_optimize_wan_vae):
+            try:
+                optimizer(vae)
+            except Exception:
+                # A wrapper may already have installed its exact fallback
+                # before a later optional rewrite failed. With no registered
+                # enabled gate, those wrappers remain on the reference path.
+                logger.warning(
+                    "Failed to install optional VAE fast paths on %s; using the reference path",
+                    type(vae).__name__,
+                    exc_info=True,
+                )
+
+
+def clear_pipeline_vae_fast_path_caches(pipeline: nn.Module) -> None:
+    """Release lazily folded buffers before the worker enters sleep mode."""
+
+    cleared = 0
+    for vae in ModuleDiscovery.discover(pipeline).vaes:
+        if not getattr(vae, "_vllm_vae_fast_path_installed", False):
+            continue
+        for module in vae.modules():
+            for name in _FOLDED_CACHE_NAMES:
+                if name in module._buffers and module._buffers[name] is not None:
+                    setattr(module, name, None)
+                    cleared += 1
+    if cleared:
+        logger.debug("Released %d lazy VAE fast-path buffers before sleep", cleared)
+
+
+__all__ = ["clear_pipeline_vae_fast_path_caches", "optimize_pipeline_vaes"]
@@ -0,0 +1,91 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Decode-only precision configuration for structurally split Wan VAEs."""
+
+from __future__ import annotations
+
+from types import MethodType
+from typing import Any
+
+import torch
+from torch import nn
+from vllm.logger import init_logger
+
+logger = init_logger(__name__)
+
+PRECISION_TO_DTYPE = {
+    "fp32": torch.float32,
+    "fp16": torch.float16,
+    "bf16": torch.bfloat16,
+}
+
+
+def _decode_with_precision(
+    self: nn.Module,
+    latents: torch.Tensor,
+    *args: Any,
+    **kwargs: Any,
+) -> Any:
+    return type(self).decode(
+        self,
+        latents.to(dtype=self._vllm_decode_dtype),
+        *args,
+        **kwargs,
+    )
+
+
+def _encode_in_fp32(
+    self: nn.Module,
+    samples: torch.Tensor,
+    *args: Any,
+    **kwargs: Any,
+) -> Any:
+    return type(self).encode(
+        self,
+        samples.to(dtype=torch.float32),
+        *args,
+        **kwargs,
+    )
+
+
+def configure_wan_decode_precision(vae: nn.Module, precision: str | None) -> nn.Module:
+    """Keep Wan encode in FP32 while materializing only decode modules lower."""
+
+    if precision is None:
+        return vae
+
+    from diffusers.models.autoencoders import AutoencoderKLWan
+
+    if not isinstance(vae, AutoencoderKLWan):
+        return vae
+    try:
+        decode_dtype = PRECISION_TO_DTYPE[precision]
+    except KeyError as exc:
+        raise ValueError(
+            f"Unsupported vae_decode_precision={precision!r}; expected one of {sorted(PRECISION_TO_DTYPE)}"
+        ) from exc
+
+    encoder = getattr(vae, "encoder", None)
+    quant_conv = getattr(vae, "quant_conv", None)
+    decoder = getattr(vae, "decoder", None)
+    post_quant_conv = getattr(vae, "post_quant_conv", None)
+    if not all(isinstance(module, nn.Module) for module in (encoder, decoder, post_quant_conv)):
+        logger.info("Wan VAE lacks separate encoder/decoder modules; skipping decode-only precision")
+        return vae
+
+    encoder.to(dtype=torch.float32)
+    if isinstance(quant_conv, nn.Module):
+        quant_conv.to(dtype=torch.float32)
+    decoder.to(dtype=decode_dtype)
+    post_quant_conv.to(dtype=decode_dtype)
+    vae.encode = MethodType(_encode_in_fp32, vae)
+    vae._vllm_decode_dtype = decode_dtype
+    vae.decode = MethodType(_decode_with_precision, vae)
+    logger.info(
+        "Wan VAE configured with FP32 encode and %s decode",
+        precision,
+    )
+    return vae
+
+
+__all__ = ["PRECISION_TO_DTYPE", "configure_wan_decode_precision"]
@@ -0,0 +1,237 @@
+# ruff: noqa: N803
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Channels-last two-pass GroupNorm(+SiLU) Triton kernels.
+
+The kernels keep the channel dimension innermost throughout a 2D VAE
+decoder. Statistics are accumulated in FP32, then a separate apply pass
+performs the affine transform and optional SiLU epilogue. Unsupported inputs
+return ``None`` so callers can fail closed to the reference implementation.
+"""
+
+from __future__ import annotations
+
+import torch
+import triton
+import triton.language as tl
+
+_SUPPORTED_DTYPES = {torch.float16, torch.bfloat16, torch.float32}
+_MAX_CHANNELS = 2048
+
+
+@triton.jit
+def _gn_partial_rows_kernel(
+    x_ptr,
+    psum_ptr,
+    psq_ptr,
+    rows,
+    rows_per_prog,
+    C: tl.constexpr,
+    BLOCK_R: tl.constexpr,
+):
+    chunk = tl.program_id(0).to(tl.int64)
+    batch = tl.program_id(1).to(tl.int64)
+    num_chunks = tl.num_programs(0)
+    row0 = chunk * rows_per_prog
+    cols = tl.arange(0, C)
+    acc_sum = tl.zeros((C,), tl.float32)
+    acc_sq = tl.zeros((C,), tl.float32)
+    x_base = x_ptr + batch * rows * C
+    for row_offset in range(0, rows_per_prog, BLOCK_R):
+        row_indices = row0 + row_offset + tl.arange(0, BLOCK_R)
+        mask = row_indices < rows
+        x = tl.load(
+            x_base + row_indices[:, None] * C + cols[None, :],
+            mask=mask[:, None],
+            other=0.0,
+        ).to(tl.float32)
+        acc_sum += tl.sum(x, 0)
+        acc_sq += tl.sum(x * x, 0)
+    output_offsets = (batch * num_chunks + chunk) * C + cols
+    tl.store(psum_ptr + output_offsets, acc_sum)
+    tl.store(psq_ptr + output_offsets, acc_sq)
+
+
+@triton.jit
+def _gn_finalize_kernel(
+    psum_ptr,
+    psq_ptr,
+    weight_ptr,
+    bias_ptr,
+    scale_shift_ptr,
+    num_chunks,
+    group_numel,
+    eps,
+    channels,
+    CPG: tl.constexpr,
+    BLOCK_K: tl.constexpr,
+):
+    group = tl.program_id(0).to(tl.int64)
+    batch = tl.program_id(1).to(tl.int64)
+    cols = group * CPG + tl.arange(0, CPG)
+    acc_sum = tl.zeros((), tl.float32)
+    acc_sq = tl.zeros((), tl.float32)
+    for chunk0 in range(0, num_chunks, BLOCK_K):
+        chunks = chunk0 + tl.arange(0, BLOCK_K)
+        mask = chunks < num_chunks
+        offsets = (batch * num_chunks + chunks)[:, None] * channels + cols[None, :]
+        acc_sum += tl.sum(tl.load(psum_ptr + offsets, mask=mask[:, None], other=0.0))
+        acc_sq += tl.sum(tl.load(psq_ptr + offsets, mask=mask[:, None], other=0.0))
+    mean = acc_sum / group_numel
+    variance = acc_sq / group_numel - mean * mean
+    variance = tl.maximum(variance, 0.0)
+    reciprocal_std = tl.rsqrt(variance + eps)
+    weight = tl.load(weight_ptr + cols).to(tl.float32)
+    bias = tl.load(bias_ptr + cols).to(tl.float32)
+    scale = weight * reciprocal_std
+    shift = bias - mean * scale
+    tl.store(scale_shift_ptr + batch * 2 * channels + cols, scale)
+    tl.store(scale_shift_ptr + (batch * 2 + 1) * channels + cols, shift)
+
+
+@triton.jit
+def _gn_apply_rows_kernel(
+    x_ptr,
+    scale_shift_ptr,
+    y_ptr,
+    rows,
+    C: tl.constexpr,
+    BLOCK_R: tl.constexpr,
+    SILU: tl.constexpr,
+):
+    program = tl.program_id(0).to(tl.int64)
+    batch = tl.program_id(1).to(tl.int64)
+    cols = tl.arange(0, C)
+    scale = tl.load(scale_shift_ptr + batch * 2 * C + cols)
+    shift = tl.load(scale_shift_ptr + (batch * 2 + 1) * C + cols)
+    row_indices = program * BLOCK_R + tl.arange(0, BLOCK_R)
+    mask = row_indices < rows
+    offsets = batch * rows * C + row_indices[:, None] * C + cols[None, :]
+    x = tl.load(x_ptr + offsets, mask=mask[:, None], other=0.0).to(tl.float32)
+    y = x * scale[None, :] + shift[None, :]
+    if SILU:
+        y = y * tl.sigmoid(y)
+    tl.store(y_ptr + offsets, y.to(y_ptr.dtype.element_ty), mask=mask[:, None])
+
+
+def _group_norm_silu_rows(
+    x: torch.Tensor,
+    weight: torch.Tensor,
+    bias: torch.Tensor,
+    num_groups: int,
+    eps: float,
+    apply_silu: bool,
+) -> torch.Tensor:
+    batch_size, rows, channels = x.shape
+    channels_per_group = channels // num_groups
+    block_rows = max(1, 8192 // channels)
+    rows_per_program = block_rows * 32
+    num_chunks = triton.cdiv(rows, rows_per_program)
+    partial_sum = torch.empty((batch_size, num_chunks, channels), device=x.device, dtype=torch.float32)
+    partial_sq = torch.empty_like(partial_sum)
+    with torch.cuda.device(x.device):
+        _gn_partial_rows_kernel[(num_chunks, batch_size)](
+            x,
+            partial_sum,
+            partial_sq,
+            rows,
+            rows_per_program,
+            C=channels,
+            BLOCK_R=block_rows,
+            num_warps=4,
+        )
+        scale_shift = torch.empty((batch_size, 2, channels), device=x.device, dtype=torch.float32)
+        block_chunks = max(1, min(4096 // channels_per_group, triton.next_power_of_2(num_chunks)))
+        _gn_finalize_kernel[(num_groups, batch_size)](
+            partial_sum,
+            partial_sq,
+            weight,
+            bias,
+            scale_shift,
+            num_chunks,
+            rows * channels_per_group,
+            eps,
+            channels,
+            CPG=channels_per_group,
+            BLOCK_K=block_chunks,
+            num_warps=4,
+        )
+        output = torch.empty_like(x)
+        _gn_apply_rows_kernel[(triton.cdiv(rows, block_rows), batch_size)](
+            x,
+            scale_shift,
+            output,
+            rows,
+            C=channels,
+            BLOCK_R=block_rows,
+            SILU=apply_silu,
+            num_warps=4,
+        )
+    return output
+
+
+def _is_supported(
+    x: torch.Tensor,
+    weight: torch.Tensor | None,
+    bias: torch.Tensor | None,
+    num_groups: int,
+) -> bool:
+    if not (x.is_cuda and x.numel() > 0 and not torch.is_grad_enabled()):
+        return False
+    if x.requires_grad or x.dtype not in _SUPPORTED_DTYPES:
+        return False
+    if weight is None or bias is None:
+        return False
+    channels = x.shape[1] if x.dim() == 4 else x.shape[-1]
+    if weight.shape != (channels,) or bias.shape != (channels,):
+        return False
+    if not (
+        weight.device == bias.device == x.device
+        and weight.dtype == bias.dtype == x.dtype
+        and weight.is_contiguous()
+        and bias.is_contiguous()
+    ):
+        return False
+    if num_groups < 1 or channels % num_groups != 0:
+        return False
+    return triton.next_power_of_2(channels) == channels and channels <= _MAX_CHANNELS
+
+
+def group_norm_silu_4d(
+    x: torch.Tensor,
+    weight: torch.Tensor,
+    bias: torch.Tensor,
+    num_groups: int,
+    eps: float,
+    apply_silu: bool = True,
+) -> torch.Tensor | None:
+    """Run GroupNorm(+SiLU) on a channels-last ``(N, C, H, W)`` tensor."""
+
+    if x.dim() != 4 or not _is_supported(x, weight, bias, num_groups):
+        return None
+    batch_size, channels, height, width = x.shape
+    if not (channels > 1 and (height > 1 or width > 1) and x.is_contiguous(memory_format=torch.channels_last)):
+        return None
+    rows = x.permute(0, 2, 3, 1).reshape(batch_size, height * width, channels)
+    output = _group_norm_silu_rows(rows, weight, bias, num_groups, eps, apply_silu)
+    return output.reshape(batch_size, height, width, channels).permute(0, 3, 1, 2)
+
+
+def group_norm_silu_rows(
+    x: torch.Tensor,
+    weight: torch.Tensor,
+    bias: torch.Tensor,
+    num_groups: int,
+    eps: float,
+    apply_silu: bool = True,
+) -> torch.Tensor | None:
+    """Run GroupNorm(+SiLU) over contiguous ``(N, rows, C)`` input."""
+
+    if x.dim() != 3 or not x.is_contiguous():
+        return None
+    if not _is_supported(x, weight, bias, num_groups):
+        return None
+    return _group_norm_silu_rows(x, weight, bias, num_groups, eps, apply_silu)
+
+
+__all__ = ["group_norm_silu_4d", "group_norm_silu_rows"]
@@ -0,0 +1,160 @@
+# ruff: noqa: N803
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Channels-last-3D Wan VAE RMSNorm(+SiLU) Triton kernel."""
+
+from __future__ import annotations
+
+import torch
+import triton
+import triton.language as tl
+
+_SUPPORTED_DTYPES = {torch.float16, torch.bfloat16, torch.float32}
+_MAX_CHANNELS = 1024
+
+
+@triton.jit
+def _wan_rmsnorm_silu_kernel(
+    x_ptr,
+    gamma_ptr,
+    bias_ptr,
+    output_ptr,
+    channels: tl.constexpr,
+    time_size,
+    height_size,
+    width_size,
+    x_stride_batch,
+    x_stride_channel,
+    x_stride_time,
+    x_stride_height,
+    x_stride_width,
+    output_stride_batch,
+    output_stride_channel,
+    output_stride_time,
+    output_stride_height,
+    output_stride_width,
+    rms_scale,
+    eps,
+    has_bias: tl.constexpr,
+    BLOCK_C: tl.constexpr,
+):
+    row = tl.program_id(0).to(tl.int64)
+    channel_offsets = tl.arange(0, BLOCK_C)
+    channel_mask = channel_offsets < channels
+
+    width = row % width_size
+    remainder = row // width_size
+    height = remainder % height_size
+    remainder = remainder // height_size
+    time = remainder % time_size
+    batch = remainder // time_size
+
+    x_base = batch * x_stride_batch + time * x_stride_time + height * x_stride_height + width * x_stride_width
+    output_base = (
+        batch * output_stride_batch
+        + time * output_stride_time
+        + height * output_stride_height
+        + width * output_stride_width
+    )
+
+    x = tl.load(
+        x_ptr + x_base + channel_offsets * x_stride_channel,
+        mask=channel_mask,
+        other=0.0,
+    ).to(tl.float32)
+    norm = tl.sqrt(tl.sum(x * x, axis=0))
+    inverse_norm = 1.0 / tl.maximum(norm, eps)
+
+    # Match the eager dtype boundaries: normalize and scale in activation
+    # dtype, affine in the promoted dtype, then SiLU in FP32.
+    y = (x * inverse_norm).to(x_ptr.dtype.element_ty)
+    gamma = tl.load(gamma_ptr + channel_offsets, mask=channel_mask, other=1.0)
+    y = (y * rms_scale).to(x_ptr.dtype.element_ty)
+    y = (y.to(tl.float32) * gamma.to(tl.float32)).to(output_ptr.dtype.element_ty)
+    if has_bias:
+        bias = tl.load(bias_ptr + channel_offsets, mask=channel_mask, other=0.0)
+        y = (y.to(tl.float32) + bias.to(tl.float32)).to(output_ptr.dtype.element_ty)
+    y = y.to(tl.float32)
+    y = y * tl.sigmoid(y)
+    tl.store(
+        output_ptr + output_base + channel_offsets * output_stride_channel,
+        y,
+        mask=channel_mask,
+    )
+
+
+def _affine_supported(x: torch.Tensor, affine: torch.Tensor) -> bool:
+    return (
+        affine.is_cuda
+        and affine.device == x.device
+        and (affine.dtype == x.dtype or affine.dtype == torch.float32)
+        and affine.numel() == x.shape[1]
+    )
+
+
+def can_use_wan_rmsnorm_silu(
+    x: torch.Tensor,
+    gamma: torch.Tensor,
+    bias: torch.Tensor | None,
+) -> bool:
+    """Return whether ``wan_rmsnorm_silu`` supports this exact input."""
+
+    return (
+        x.is_cuda
+        and not torch.is_grad_enabled()
+        and not x.requires_grad
+        and x.dtype in _SUPPORTED_DTYPES
+        and x.ndim == 5
+        and x.numel() > 0
+        and 0 < x.shape[1] <= _MAX_CHANNELS
+        and x.is_contiguous(memory_format=torch.channels_last_3d)
+        and _affine_supported(x, gamma)
+        and (bias is None or _affine_supported(x, bias))
+    )
+
+
+def wan_rmsnorm_silu(
+    x: torch.Tensor,
+    gamma: torch.Tensor,
+    bias: torch.Tensor | None = None,
+    rms_scale: float | None = None,
+    eps: float = 1e-12,
+) -> torch.Tensor | None:
+    """Fuse ``SiLU(F.normalize(x, dim=1) * scale * gamma + bias)``."""
+
+    if not can_use_wan_rmsnorm_silu(x, gamma, bias):
+        return None
+
+    batch_size, channels, time_size, height_size, width_size = x.shape
+    output_dtype = torch.promote_types(x.dtype, gamma.dtype)
+    output = torch.empty_strided(x.shape, x.stride(), device=x.device, dtype=output_dtype)
+    block_channels = triton.next_power_of_2(channels)
+    num_warps = 1 if block_channels <= 64 else 4 if block_channels <= 512 else 8
+    gamma = gamma.reshape(channels).contiguous()
+    has_bias = bias is not None
+    bias = gamma if bias is None else bias.reshape(channels).contiguous()
+    if rms_scale is None:
+        rms_scale = channels**0.5
+
+    with torch.cuda.device(x.device):
+        _wan_rmsnorm_silu_kernel[(batch_size * time_size * height_size * width_size,)](
+            x,
+            gamma,
+            bias,
+            output,
+            channels,
+            time_size,
+            height_size,
+            width_size,
+            *x.stride(),
+            *output.stride(),
+            float(rms_scale),
+            eps,
+            has_bias,
+            BLOCK_C=block_channels,
+            num_warps=num_warps,
+        )
+    return output
+
+
+__all__ = ["can_use_wan_rmsnorm_silu", "wan_rmsnorm_silu"]
@@ -0,0 +1,162 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Quality-gated CUDA fast path for Diffusers Wan VAE decoders."""
+
+from __future__ import annotations
+
+from types import MethodType
+from typing import Any
+
+import torch
+import torch.nn.functional as F
+from torch import nn
+from vllm.logger import init_logger
+
+from vllm_omni.diffusion.vae_optimizations.gate import (
+    VaeFastPathGate,
+    register_vae_fast_path_gate,
+)
+
+logger = init_logger(__name__)
+
+try:
+    from vllm_omni.diffusion.vae_optimizations.triton_wan_rmsnorm_silu import (
+        wan_rmsnorm_silu,
+    )
+
+    _HAS_TRITON = True
+except ImportError:  # pragma: no cover - exercised on builds without Triton
+    _HAS_TRITON = False
+
+
+class FusedWanRMSNormSiLU(nn.Module):
+    """Quality-gated Wan RMSNorm + SiLU with an exact eager fallback."""
+
+    def __init__(self, norm: nn.Module, gate: VaeFastPathGate) -> None:
+        super().__init__()
+        self.gamma = norm.gamma
+        self.bias = norm.bias
+        self.scale = float(norm.scale)
+        self._vllm_gate = gate
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if self._vllm_gate.enabled and not torch.compiler.is_compiling():
+            bias = self.bias if isinstance(self.bias, torch.Tensor) else None
+            output = wan_rmsnorm_silu(x, self.gamma, bias, rms_scale=self.scale)
+            if output is not None:
+                return output
+
+        needs_fp32_normalize = x.dtype in (torch.float16, torch.bfloat16) or any(
+            token in str(x.dtype) for token in ("float4_", "float8_")
+        )
+        normalized = F.normalize(x.float() if needs_fp32_normalize else x, dim=1).to(x.dtype)
+        return F.silu(normalized * self.scale * self.gamma + self.bias)
+
+
+def _plain_silu(activation: object) -> bool:
+    return isinstance(activation, nn.SiLU) and not activation.inplace
+
+
+def _fusable_norm(norm: object, norm_cls: type) -> bool:
+    return (
+        type(norm) is norm_cls
+        and getattr(norm, "channel_first", False)
+        and isinstance(getattr(norm, "gamma", None), torch.Tensor)
+    )
+
+
+def _install_norm_silu(decoder: nn.Module, gate: VaeFastPathGate) -> int | None:
+    from diffusers.models.autoencoders.autoencoder_kl_wan import WanResidualBlock, WanRMS_norm
+
+    residual_blocks = [module for module in decoder.modules() if isinstance(module, WanResidualBlock)]
+    eligible = [
+        module
+        for module in residual_blocks
+        if type(module) is WanResidualBlock
+        and _plain_silu(module.nonlinearity)
+        and _fusable_norm(module.norm1, WanRMS_norm)
+        and _fusable_norm(module.norm2, WanRMS_norm)
+    ]
+    if len(eligible) != len(residual_blocks):
+        logger.info(
+            "Wan VAE has %d/%d non-standard residual blocks; skipping fast path",
+            len(residual_blocks) - len(eligible),
+            len(residual_blocks),
+        )
+        return None
+    if not (
+        _fusable_norm(getattr(decoder, "norm_out", None), WanRMS_norm)
+        and _plain_silu(getattr(decoder, "nonlinearity", None))
+    ):
+        logger.info("Wan VAE has a non-standard output head; skipping fast path")
+        return None
+
+    count = 0
+    for module in eligible:
+        module.norm1 = FusedWanRMSNormSiLU(module.norm1, gate)
+        module.norm2 = FusedWanRMSNormSiLU(module.norm2, gate)
+        module.nonlinearity = nn.Identity()
+        count += 2
+    decoder.norm_out = FusedWanRMSNormSiLU(decoder.norm_out, gate)
+    decoder.nonlinearity = nn.Identity()
+    return count + 1
+
+
+def _set_wan_decoder_memory_format(decoder: nn.Module, *, channels_last: bool) -> None:
+    for module in decoder.modules():
+        if isinstance(module, nn.Conv3d):
+            module.to(memory_format=torch.channels_last_3d if channels_last else torch.contiguous_format)
+        elif isinstance(module, (nn.Conv2d, nn.ConvTranspose2d)):
+            module.to(memory_format=torch.channels_last if channels_last else torch.contiguous_format)
+
+
+def _decoder_layout_forward(self: nn.Module, *args: Any, **kwargs: Any) -> torch.Tensor:
+    use_channels_last = self._vllm_gate.enabled
+    if use_channels_last != self._vllm_channels_last:
+        _set_wan_decoder_memory_format(self, channels_last=use_channels_last)
+        self._vllm_channels_last = use_channels_last
+        logger.debug(
+            "Wan VAE decoder switched to %s layout",
+            "channels_last_3d" if use_channels_last else "contiguous",
+        )
+    return type(self).forward(self, *args, **kwargs)
+
+
+def maybe_optimize_wan_vae(vae: nn.Module) -> nn.Module:
+    """Install the Wan decoder fast path when its full structure is supported."""
+
+    if not _HAS_TRITON or getattr(vae, "_vllm_vae_fast_path_installed", False):
+        return vae
+
+    from diffusers.models.autoencoders import AutoencoderKLWan
+    from diffusers.models.autoencoders.autoencoder_kl_wan import WanDecoder3d
+
+    if not isinstance(vae, AutoencoderKLWan):
+        return vae
+    decoder = getattr(vae, "decoder", None)
+    if type(decoder) is not WanDecoder3d:
+        return vae
+
+    distributed_executor = getattr(vae, "distributed_executor", None)
+    parallel_mode = getattr(distributed_executor, "parallel_mode", "tile")
+    if parallel_mode in ("auto", "spatial_shard_height", "spatial_shard_width"):
+        logger.info("Wan spatial-shard decode is configured; skipping quality-gated VAE fast path")
+        return vae
+
+    gate = VaeFastPathGate()
+    norm_count = _install_norm_silu(decoder, gate)
+    if norm_count is None:
+        return vae
+    decoder._vllm_gate = gate
+    decoder._vllm_channels_last = False
+    decoder.forward = MethodType(_decoder_layout_forward, decoder)
+    register_vae_fast_path_gate(vae, gate)
+    vae._vllm_vae_fast_path_installed = True
+    logger.info(
+        "Wan VAE installed quality-gated decoder fast path (%d RMSNorm+SiLU fusions)",
+        norm_count,
+    )
+    return vae
+
+
+__all__ = ["FusedWanRMSNormSiLU", "maybe_optimize_wan_vae"]
@@ -43,6 +43,10 @@ from vllm_omni.diffusion.offloader import get_offload_backend
 from vllm_omni.diffusion.registry import _NO_CACHE_ACCELERATION
 from vllm_omni.diffusion.request import OmniDiffusionRequest
 from vllm_omni.diffusion.sched.interface import DiffusionSchedulerOutput, KVPrefetchJob
+from vllm_omni.diffusion.vae_optimizations import (
+    optimize_pipeline_vaes,
+    use_pipeline_vae_fast_path,
+)
 from vllm_omni.diffusion.worker.input_batch import InputBatch, scatter_latents
 from vllm_omni.diffusion.worker.request_batch import DiffusionRequestBatch
 from vllm_omni.diffusion.worker.utils import (
@@ -243,6 +247,9 @@ class DiffusionModelRunner(OmniConnectorModelRunnerMixin):
                    custom_pipeline_name=custom_pipeline_name,
                    device=self.device,
                )
+                # Keep precision rematerialization in the tagged weights pool
+                # so sleep level 1 can offload every resulting parameter.
+                optimize_pipeline_vaes(self.pipeline, self.od_config)
        time_after_load = time.perf_counter()

        logger.info(
@@ -499,6 +506,11 @@ class DiffusionModelRunner(OmniConnectorModelRunnerMixin):
        if require_request_batch_support and not getattr(self.pipeline, "supports_request_batch", False):
            raise RuntimeError(f"{type(self.pipeline).__name__} does not support request-batch forward.")

+        quality = getattr(reqs[0].sampling_params, "quality", None)
+        if any(getattr(req.sampling_params, "quality", None) != quality for req in reqs[1:]):
+            raise RuntimeError("A diffusion worker batch cannot mix request quality levels.")
+        use_fast_vae = quality == "high"
+
        # Use no_grad() for HSDP compatibility, inference_mode() otherwise for
        # better perf. HSDP2's fully_shard pre-forward hooks need tensor version
        # counters, which inference tensors do not track.
@@ -523,7 +535,8 @@ class DiffusionModelRunner(OmniConnectorModelRunnerMixin):

            with set_forward_context(vllm_config=self.vllm_config, omni_diffusion_config=od_config):
                with record_function(record_name):
-                    raw_outputs = self.pipeline.forward(batch)
+                    with use_pipeline_vae_fast_path(self.pipeline, use_fast_vae):
+                        raw_outputs = self.pipeline.forward(batch)
                    outputs = _normalize_pipeline_outputs(
                        raw_outputs,
                        expected_count=len(reqs),
@@ -774,7 +787,11 @@ class DiffusionModelRunner(OmniConnectorModelRunnerMixin):

                            if should_decode:
                                clear_pipeline_stage_durations(self.pipeline)
-                                result = self.pipeline.post_decode(req)
+                                with use_pipeline_vae_fast_path(
+                                    self.pipeline,
+                                    getattr(req.sampling, "quality", None) == "high",
+                                ):
+                                    result = self.pipeline.post_decode(req)
                                if result is not None:
                                    self._attach_stepwise_metrics(
                                        req,
@@ -576,6 +576,14 @@ class DiffusionWorker:

        usage_before = allocator.get_current_usage()

+        if self.model_runner is not None and self.model_runner.pipeline is not None:
+            from vllm_omni.diffusion.vae_optimizations import clear_pipeline_vae_fast_path_caches
+
+            # Folded weights are allocated lazily on the request path rather
+            # than in the tagged weight pool. Drop them before sleep so they
+            # neither remain resident nor need a separate CPU backup.
+            clear_pipeline_vae_fast_path_caches(self.model_runner.pipeline)
+
        if level == 2 and self.model_runner is not None:
            if hasattr(self.model_runner, "graph_runners"):
                self.model_runner.graph_runners.clear()
@@ -491,6 +491,7 @@ class OrchestratorArgs:
    step_execution: bool = False
    vae_use_slicing: bool = False
    vae_use_tiling: bool = False
+    vae_decode_precision: str | None = None
    enable_multithread_weight_load: bool = True
    num_weight_load_threads: int = 4
    enable_cpu_offload: bool = False
@@ -1026,6 +1026,7 @@ class AsyncOmniEngine:
            "request_batch_max_wait_ms": kwargs.get("request_batch_max_wait_ms", 0.0),
            "vae_use_slicing": kwargs.get("vae_use_slicing", False),
            "vae_use_tiling": kwargs.get("vae_use_tiling", False),
+            "vae_decode_precision": kwargs.get("vae_decode_precision", None),
            "cache_backend": cache_backend,
            "cache_config": cache_config,
            "enable_cache_dit_summary": kwargs.get("enable_cache_dit_summary", False),
@@ -629,6 +629,15 @@ class OmniServeCommand(CLISubcommand):
            action="store_true",
            help="Enable VAE tiling for memory optimization (useful for mitigating OOM issues).",
        )
+        omni_config_group.add_argument(
+            "--vae-decode-precision",
+            choices=("fp32", "fp16", "bf16"),
+            default=None,
+            help=(
+                "Optional Wan decode-only VAE precision. The encoder remains FP32; "
+                "unset preserves the pipeline's existing precision behavior."
+            ),
+        )

        # Parallel weight loading (faster diffusion startup)
        omni_config_group.add_argument(
@@ -139,6 +139,18 @@ class NPUOmniPlatform(OmniPlatform, NPUPlatform):
                )
                backend_upper = "FLASH_ATTN"

+            if backend_upper == "FLASH_ATTN" and find_spec("mindiesd"):
+                # The NPU FLASH_ATTN backend imports mindiesd lazily at first
+                # forward, but CANN snapshots the custom-op registry at the
+                # first custom-op regInfo lookup in the process (e.g. a
+                # vllm-ascend custom op during model load/warmup). Import
+                # mindiesd here so its env.py prepends the mindiesd vendor
+                # dirs (aie_ascendc etc.) to ASCEND_CUSTOM_OPP_PATH before
+                # that snapshot; otherwise aclnnLaserAttention /
+                # FusedAttentionScore fail with EZ1001 "does not support
+                # opType" for the rest of the process.
+                import mindiesd  # noqa: F401
+
            backend = DiffusionAttentionBackendEnum[backend_upper]
            logger.debug("Using diffusion attention backend '%s'", backend_upper)
            return backend.get_path()
Author	SHA1	Message	Date
Hongsheng Liu	094abd70c1	feat(diffusion): add request-gated VAE fast paths Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>	2026-08-11 09:31:27 +08:00
fan2956andFan	c27623c29c	[Bugfix] rainfusion support all shape video (#6000 ) Signed-off-by: Fan <fan@FandeMacBook-Air.local> Co-authored-by: Fan <fan@FandeMacBook-Air.local>	2026-08-11 05:10:50 +08:00
Yueqian Lin	f8fa86fb1f	[BugFix] Don't fail the TTS ratchet when the branch count goes down (#6008 ) Signed-off-by: Yueqian Lin <linyueqian@outlook.com>	2026-08-11 05:08:31 +08:00
ooooooyeandHongsheng Liu	422d8fd8f9	[Bugfix][Diffusion] Restore supports_packed_mask_free on teacache FakeBackend (#5997 ) Signed-off-by: brandneway <gyuan4892@gmail.com> Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>	2026-08-10 22:20:37 +08:00
Hongsheng Liu	62e239bf1e	[Docs] Fix broken attention backend link (#5998 ) Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>	2026-08-10 21:55:01 +08:00
ooooooye	740f45abcd	[Perf][Diffusion] MiniMax-H3: opt into packed varlen attention on NPU to eliminate quadratic mask materialization (#5891 ) Signed-off-by: brandneway <gyuan4892@gmail.com>	2026-08-10 19:47:03 +08:00
Hongsheng Liu hsliu_ustc Copilot Autofix powered by AI	5e4b9f8c6e	[Misc] Align CODEOWNERS with module/feature design docs (#5958 ) Signed-off-by: hsliu_ustc <hsliu_ustc@noreply.gitcode.com> Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com> Co-authored-by: hsliu_ustc <hsliu_ustc@noreply.gitcode.com> Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>	2026-08-10 19:25:03 +08:00
Hongsheng Liu	ae811043e9	docs: keep shared task examples in navigation (#5987 ) Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>	2026-08-10 19:13:52 +08:00
R0CKSTAR	636d44e272	[Perf][MUSA] Restore MiniMax-H3 dynamic RoPE fusion (#5881 ) Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>	2026-08-10 17:24:46 +08:00
yiminghub2024	8e44e78038	Add Ref2VA measurements to the MiniMax-H3 DGX Spark (GB10) recipe (#5972 ) Signed-off-by: yiminghub2024 <482890@qq.com>	2026-08-10 15:51:15 +08:00
Hongsheng Liu	b98b7c8529	[Docs] Preserve generated quantization URLs and link recipes in supported models (#5969 ) Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>	2026-08-10 15:45:50 +08:00