[Docs] Redirect the former PD guide URL

Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>

.buildkite/cuda/test-nightly.yml

@@ -345,7 +345,7 @@ steps:
         commands:
           - export DIFFUSION_BENCHMARK_DIR=tests/dfx/perf/results
           - export DIFFUSION_ATTENTION_BACKEND=FLASH_ATTN
-          - export CACHE_DIT_VERSION=1.3.0
+          - export CACHE_DIT_VERSION=1.5.0
           - |
             set +e
             pytest -s -v tests/dfx/perf/scripts/run_diffusion_benchmark.py --test-config-file tests/dfx/perf/tests/test_qwen_image_vllm_omni.json
@@ -361,7 +361,7 @@ steps:
         commands:
           - export DIFFUSION_BENCHMARK_DIR=tests/dfx/perf/results
           - export DIFFUSION_ATTENTION_BACKEND=FLASH_ATTN
-          - export CACHE_DIT_VERSION=1.3.0
+          - export CACHE_DIT_VERSION=1.5.0
           - |
             set +e
             pytest -s -v tests/dfx/perf/scripts/run_diffusion_benchmark.py --test-config-file tests/dfx/perf/tests/test_qwen_image_edit_vllm_omni.json
@@ -377,7 +377,7 @@ steps:
         commands:
           - export DIFFUSION_BENCHMARK_DIR=tests/dfx/perf/results
           - export DIFFUSION_ATTENTION_BACKEND=FLASH_ATTN
-          - export CACHE_DIT_VERSION=1.3.0
+          - export CACHE_DIT_VERSION=1.5.0
           - |
             set +e
             pytest -s -v tests/dfx/perf/scripts/run_diffusion_benchmark.py --test-config-file tests/dfx/perf/tests/test_qwen_image_edit_2511_vllm_omni.json
@@ -393,7 +393,7 @@ steps:
         commands:
           - export DIFFUSION_BENCHMARK_DIR=tests/dfx/perf/results
           - export DIFFUSION_ATTENTION_BACKEND=FLASH_ATTN
-          - export CACHE_DIT_VERSION=1.3.0
+          - export CACHE_DIT_VERSION=1.5.0
           - |
             set +e
             pytest -s -v tests/dfx/perf/scripts/run_diffusion_benchmark.py --test-config-file tests/dfx/perf/tests/test_qwen_image_layered_vllm_omni.json
@@ -409,7 +409,7 @@ steps:
         commands:
           - export DIFFUSION_BENCHMARK_DIR=tests/dfx/perf/results
           - export DIFFUSION_ATTENTION_BACKEND=FLASH_ATTN
-          - export CACHE_DIT_VERSION=1.3.0
+          - export CACHE_DIT_VERSION=1.5.0
           - |
             set +e
             pytest -s -v tests/dfx/perf/scripts/run_diffusion_benchmark.py --test-config-file tests/dfx/perf/tests/test_bagel_vllm_omni.json

.buildkite/cuda/test-ready.yml

@@ -227,6 +227,18 @@ steps:
             "
         mirror_hardwares: h100_1
 
+      - label: "Omni · MiniCPM-o 4.5 Perf Test"
+        timeout_in_minutes: 60
+        source_file_dependencies:
+          - tests/dfx/perf/tests/test_minicpmo_4_5_ready.json
+          - vllm_omni/deploy/minicpmo_4_5.yaml
+          - vllm_omni/model_executor/models/minicpmo_4_5/
+          - vllm_omni/model_executor/stage_input_processors/minicpmo_4_5_omni.py
+        commands:
+          - export BENCHMARK_DIR=tests/dfx/perf/results
+          - pytest -s -v tests/dfx/perf/scripts/run_benchmark.py --test-config-file tests/dfx/perf/tests/test_minicpmo_4_5_ready.json -m 'core_model and cuda'
+        mirror_hardwares: h100_1
+
       - label: "TTS · VoxCPM2 Test"
         source_file_dependencies:
           - tests/e2e/online_serving/test_voxcpm2_tts.py

.buildkite/intel/pipeline-intel.yml

@@ -10,7 +10,7 @@ steps:
           DOCKER_BUILDKIT: "1"
           # Buildkite will automatically replace this with the actual commit hash
           VLLM_IMAGE_TAG: "${BUILDKITE_COMMIT}"
-          VLLM_VERSION: "v0.26.0"
+          VLLM_VERSION: "v0.27.0"
         priority: 100
         timeout_in_minutes: 60
         soft_fail: false

.buildkite/intel/scripts/run-xpu-test.sh

@@ -34,8 +34,23 @@ docker_build() {
     docker build "${out[@]}" "$@" -f docker/Dockerfile.xpu .
 }
 
+build_vllm_base() (
+    local vllm_source_dir
+    vllm_source_dir=$(mktemp -d)
+    trap 'rm -rf "${vllm_source_dir}"' EXIT
+
+    git clone --depth 1 --branch "${VLLM_VERSION}" \
+        https://github.com/vllm-project/vllm "${vllm_source_dir}"
+
+    local out=("${export_args[@]/'{{IMAGE}}'/${base_image_name}}")
+    docker build "${out[@]}" \
+        --target vllm-openai \
+        -f "${vllm_source_dir}/docker/Dockerfile.xpu" \
+        "${vllm_source_dir}"
+)
+
 if [ -z "$(docker images -q "${base_image_name}")" ]; then
-    docker_build "${base_image_name}" --target vllm-base --build-arg "VLLM_VERSION=${VLLM_VERSION}"
+    build_vllm_base
 fi
 
 # Try building the docker image

.buildkite/npu/test-npu-nightly.yml

@@ -115,7 +115,7 @@ steps:
           DIFFUSION_ATTENTION_BACKEND: TORCH_SDPA
         commands:
           - export DIFFUSION_BENCHMARK_DIR=tests/dfx/perf/results
-          - export CACHE_DIT_VERSION=1.3.0
+          - export CACHE_DIT_VERSION=1.5.0
           - |
             set +e
             pytest -s -v tests/dfx/perf/scripts/run_diffusion_benchmark.py -m npu --test-config-file tests/dfx/perf/tests/test_qwen_image_edit_2511_vllm_omni.json

.buildkite/npu/test-npu-ready.yml

@@ -17,6 +17,15 @@ steps:
           VLLM_ALLOW_LONG_MAX_MODEL_LEN: "1"
         commands:
           - pytest -s -v tests/e2e/online_serving/test_minicpmo_4_5_duplex.py -m 'core_model and npu and A3 and omni' --run-level 'core_model'
+      - label: "Omni · MiniCPM-o 4.5 Perf Test"
+        timeout_in_minutes: 60
+        mirror_hardwares: a3_npu_2
+        env:
+          VLLM_WORKER_MULTIPROC_METHOD: spawn
+          VLLM_ALLOW_LONG_MAX_MODEL_LEN: "1"
+        commands:
+          - export BENCHMARK_DIR=tests/dfx/perf/results
+          - pytest -s -v tests/dfx/perf/scripts/run_benchmark.py --test-config-file tests/dfx/perf/tests/test_minicpmo_4_5_ready.json -m 'core_model and npu and A3 and omni'
 
   - group: ":card_index_dividers: TTS Model Test"
     key: ready-tts-group

.buildkite/release/release-pipeline.yml

@@ -40,9 +40,11 @@ steps:
     commands:
       - "bash .buildkite/release/scripts/publish-release-wheels-pypi.sh"
 
+  # Auto-skip when NIGHTLY=1 (scheduled); otherwise require manual unblock.
   - block: "Unblock to build release Docker images"
     depends_on: ~
     key: block-build-release-images
+    if: build.env("NIGHTLY") != "1"
 
   - group: "Build release Docker images"
     key: "build-release-images"
@@ -63,7 +65,7 @@ steps:
               --progress plain \
               -f docker/Dockerfile.cuda .
           - "docker push public.ecr.aws/q9t5s3a7/vllm-omni-release-repo:$BUILDKITE_COMMIT-$(uname -m)"
-          # re-tag to default image tag and push, just in case arm64 build fails
+          # Fallback default tag (no arch suffix) if arm64 build fails.
           - "docker tag public.ecr.aws/q9t5s3a7/vllm-omni-release-repo:$BUILDKITE_COMMIT-$(uname -m) public.ecr.aws/q9t5s3a7/vllm-omni-release-repo:$BUILDKITE_COMMIT"
           - "docker push public.ecr.aws/q9t5s3a7/vllm-omni-release-repo:$BUILDKITE_COMMIT"
         env:
@@ -86,6 +88,26 @@ steps:
         env:
           DOCKER_BUILDKIT: "1"
 
+  # NIGHTLY=1 only: promote ECR images to DockerHub nightly tags, then prune.
+  - label: "Publish nightly multi-arch image to DockerHub"
+    depends_on:
+      - build-release-image-x86
+      - build-release-image-arm64
+    if: build.env("NIGHTLY") == "1"
+    key: publish-nightly-images-dockerhub
+    agents:
+      queue: small_cpu_queue_release
+    commands:
+      - "bash .buildkite/release/scripts/publish-release-images.sh nightly"
+      - "bash .buildkite/release/scripts/cleanup-nightly-builds.sh"
+    plugins:
+      - docker-login#v3.0.0:
+          username: vllmbot
+          password-env: DOCKERHUB_TOKEN
+    env:
+      DOCKER_BUILDKIT: "1"
+      DOCKERHUB_USERNAME: "vllmbot"
+
   - block: "Publish release images to DockerHub"
     key: block-publish-release-images
     depends_on:

.buildkite/release/scripts/cleanup-nightly-builds.sh

@@ -0,0 +1,147 @@
+#!/bin/bash
+# SPDX-License-Identifier: Apache-2.0
+#
+# Clean up old nightly builds from DockerHub (vllm/vllm-omni), keeping the
+# newest 14 tags with the nightly- prefix.
+
+set -ex
+
+TAG_PREFIX="nightly-"
+REPO="vllm/vllm-omni"
+REPO_API_URL="https://hub.docker.com/v2/repositories/${REPO}/tags"
+
+echo "Cleaning up tags with prefix: $TAG_PREFIX in repository: $REPO"
+
+if [ -z "$DOCKERHUB_TOKEN" ]; then
+  echo "Error: DOCKERHUB_TOKEN environment variable is not set"
+  exit 1
+fi
+
+if [ -z "$DOCKERHUB_USERNAME" ]; then
+  echo "Error: DOCKERHUB_USERNAME environment variable is not set"
+  exit 1
+fi
+
+echo "Getting DockerHub bearer token..."
+set +x
+BEARER_TOKEN=$(curl -s -X POST \
+  -H "Content-Type: application/json" \
+  -d "{\"username\": \"$DOCKERHUB_USERNAME\", \"password\": \"$DOCKERHUB_TOKEN\"}" \
+  "https://hub.docker.com/v2/users/login" | jq -r '.token')
+set -x
+
+if [ -z "$BEARER_TOKEN" ] || [ "$BEARER_TOKEN" = "null" ]; then
+  echo "Error: Failed to get DockerHub bearer token"
+  exit 1
+fi
+
+get_all_tags() {
+  local page=1
+  local all_tags=""
+  # Hub `name` is a partial match (not a strict prefix). Use it to narrow the
+  # list, then enforce startswith() client-side. Exhaust API pages; do not
+  # stop merely because a page has no matching tags after filtering.
+  local name_query
+  name_query=$(printf '%s' "$TAG_PREFIX" | jq -sRr @uri)
+
+  while true; do
+    set +x
+    local response
+    response=$(curl -s -H "Authorization: Bearer $BEARER_TOKEN" \
+      "$REPO_API_URL?page=$page&page_size=100&name=${name_query}")
+    set -x
+
+    local page_count
+    page_count=$(echo "$response" | jq -r '.results | length')
+    if [ -z "$page_count" ] || [ "$page_count" = "0" ] || [ "$page_count" = "null" ]; then
+      break
+    fi
+
+    local tags
+    tags=$(echo "$response" | jq -r --arg prefix "$TAG_PREFIX" \
+      '.results[] | select(.name | startswith($prefix)) | "\(.last_updated)|\(.name)"')
+
+    if [ -n "$tags" ]; then
+      all_tags="$all_tags$tags"$'\n'
+    fi
+    page=$((page + 1))
+  done
+
+  echo "$all_tags" | sort -r | cut -d'|' -f2
+}
+
+delete_tag() {
+  local tag_name="$1"
+  echo "Deleting tag: $tag_name"
+
+  local delete_url="https://hub.docker.com/v2/repositories/${REPO}/tags/$tag_name"
+  set +x
+  local response http_code body
+  response=$(curl -s -w "\n%{http_code}" -X DELETE \
+    -H "Authorization: Bearer $BEARER_TOKEN" "$delete_url")
+  set -x
+  http_code=$(printf '%s\n' "$response" | tail -n1)
+  body=$(printf '%s\n' "$response" | sed '$d')
+
+  if [ "$http_code" -ge 200 ] && [ "$http_code" -lt 300 ]; then
+    echo "Successfully deleted tag: $tag_name (HTTP ${http_code})"
+    return 0
+  fi
+
+  local detail=""
+  if echo "$body" | jq -e '.detail' > /dev/null 2>&1; then
+    detail=$(echo "$body" | jq -r '.detail')
+  elif [ -n "$body" ]; then
+    detail="$body"
+  else
+    detail="(empty response body)"
+  fi
+  echo "Error: Failed to delete tag $tag_name (HTTP ${http_code}): ${detail}"
+  return 1
+}
+
+echo "Fetching all tags from DockerHub..."
+all_tags=$(get_all_tags)
+
+if [ -z "$all_tags" ]; then
+  echo "No tags found to clean up"
+  exit 0
+fi
+
+total_tags=$(echo "$all_tags" | wc -l)
+echo "Found $total_tags tags"
+
+tags_to_keep=14
+tags_to_delete=$((total_tags - tags_to_keep))
+
+if [ $tags_to_delete -le 0 ]; then
+  echo "No tags need to be deleted (only $total_tags tags found, keeping $tags_to_keep)"
+  exit 0
+fi
+
+echo "Will delete $tags_to_delete old tags, keeping the newest $tags_to_keep"
+
+tags_to_delete_list=$(echo "$all_tags" | tail -n +$((tags_to_keep + 1)))
+
+if [ -z "$tags_to_delete_list" ]; then
+  echo "No tags to delete"
+  exit 0
+fi
+
+echo "Deleting old tags..."
+delete_failures=0
+while IFS= read -r tag; do
+  if [ -n "$tag" ]; then
+    if ! delete_tag "$tag"; then
+      delete_failures=$((delete_failures + 1))
+    fi
+    sleep 1
+  fi
+done <<< "$tags_to_delete_list"
+
+if [ "$delete_failures" -gt 0 ]; then
+  echo "Cleanup finished with ${delete_failures} deletion failure(s)"
+  exit 1
+fi
+
+echo "Cleanup completed successfully"

.buildkite/release/scripts/publish-release-images.sh

@@ -1,50 +1,87 @@
 #!/bin/bash
 # SPDX-License-Identifier: Apache-2.0
 #
-# Publish release Docker images from ECR to DockerHub.
-# Pulls per-arch images, tags with `latest` and versioned tags, pushes them,
-# then creates and pushes a multi-arch manifest.
+# Publish Docker images from ECR staging to DockerHub.
+#
+# Usage:
+#   publish-release-images.sh            # release mode (default)
+#   publish-release-images.sh release
+#   publish-release-images.sh nightly
+#
+# Release tags: latest, v<version>
+# Nightly tags: nightly, nightly-<commit>
 
 set -euo pipefail
 
-RELEASE_VERSION=$(buildkite-agent meta-data get release-version --default "" | sed 's/^v//')
-if [ -z "${RELEASE_VERSION}" ]; then
-  echo "ERROR: release-version metadata not set"
-  exit 1
-fi
+MODE="${1:-release}"
 
-COMMIT="$BUILDKITE_COMMIT"
 DOCKERHUB_REPO="vllm/vllm-omni"
 ECR_REPO="public.ecr.aws/q9t5s3a7/vllm-omni-release-repo"
+COMMIT="$BUILDKITE_COMMIT"
+PRIMARY_TAGS=()
+EXTRA_MANIFEST_TAGS=()
+
+case "${MODE}" in
+  release)
+    RELEASE_VERSION=$(buildkite-agent meta-data get release-version --default "" | sed 's/^v//')
+    if [ -z "${RELEASE_VERSION}" ]; then
+      echo "ERROR: release-version metadata not set"
+      exit 1
+    fi
+    PRIMARY_TAGS=("latest" "v${RELEASE_VERSION}")
+    ;;
+  nightly)
+    # Commit-pinned manifest reuses nightly-{x86_64,aarch64} arch images.
+    PRIMARY_TAGS=("nightly")
+    EXTRA_MANIFEST_TAGS=("nightly-${COMMIT}")
+    ;;
+  *)
+    echo "Usage: $0 {release|nightly}"
+    exit 2
+    ;;
+esac
 
 echo "========================================"
-echo "Publishing release images v${RELEASE_VERSION}"
+echo "Publishing ${MODE} images to ${DOCKERHUB_REPO}"
 echo "  Commit: ${COMMIT}"
+echo "  Primary tags: ${PRIMARY_TAGS[*]}"
+if [ ${#EXTRA_MANIFEST_TAGS[@]} -gt 0 ]; then
+  echo "  Extra manifests: ${EXTRA_MANIFEST_TAGS[*]}"
+fi
 echo "========================================"
 
-# Login to ECR to pull staging images
 aws ecr-public get-login-password --region us-east-1 | \
   docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7
 
-docker pull "${ECR_REPO}:${COMMIT}-x86_64"
-docker pull "${ECR_REPO}:${COMMIT}-aarch64"
+SRC_X86="${ECR_REPO}:${COMMIT}-x86_64"
+SRC_ARM="${ECR_REPO}:${COMMIT}-aarch64"
 
-docker tag "${ECR_REPO}:${COMMIT}-x86_64" "${DOCKERHUB_REPO}:latest-x86_64"
-docker tag "${ECR_REPO}:${COMMIT}-x86_64" "${DOCKERHUB_REPO}:v${RELEASE_VERSION}-x86_64"
-docker push "${DOCKERHUB_REPO}:latest-x86_64"
-docker push "${DOCKERHUB_REPO}:v${RELEASE_VERSION}-x86_64"
+docker pull "${SRC_X86}"
+docker pull "${SRC_ARM}"
 
-docker tag "${ECR_REPO}:${COMMIT}-aarch64" "${DOCKERHUB_REPO}:latest-aarch64"
-docker tag "${ECR_REPO}:${COMMIT}-aarch64" "${DOCKERHUB_REPO}:v${RELEASE_VERSION}-aarch64"
-docker push "${DOCKERHUB_REPO}:latest-aarch64"
-docker push "${DOCKERHUB_REPO}:v${RELEASE_VERSION}-aarch64"
+publish_manifest() {
+  local tag="$1"
+  local arch_tag_base="$2"
+  docker manifest rm "${DOCKERHUB_REPO}:${tag}" || true
+  docker manifest create \
+    "${DOCKERHUB_REPO}:${tag}" \
+    "${DOCKERHUB_REPO}:${arch_tag_base}-x86_64" \
+    "${DOCKERHUB_REPO}:${arch_tag_base}-aarch64"
+  docker manifest push "${DOCKERHUB_REPO}:${tag}"
+}
 
-docker manifest rm "${DOCKERHUB_REPO}:latest" || true
-docker manifest rm "${DOCKERHUB_REPO}:v${RELEASE_VERSION}" || true
-docker manifest create "${DOCKERHUB_REPO}:latest" "${DOCKERHUB_REPO}:latest-x86_64" "${DOCKERHUB_REPO}:latest-aarch64"
-docker manifest create "${DOCKERHUB_REPO}:v${RELEASE_VERSION}" "${DOCKERHUB_REPO}:v${RELEASE_VERSION}-x86_64" "${DOCKERHUB_REPO}:v${RELEASE_VERSION}-aarch64"
-docker manifest push "${DOCKERHUB_REPO}:latest"
-docker manifest push "${DOCKERHUB_REPO}:v${RELEASE_VERSION}"
+for tag in "${PRIMARY_TAGS[@]}"; do
+  docker tag "${SRC_X86}" "${DOCKERHUB_REPO}:${tag}-x86_64"
+  docker tag "${SRC_ARM}" "${DOCKERHUB_REPO}:${tag}-aarch64"
+  docker push "${DOCKERHUB_REPO}:${tag}-x86_64"
+  docker push "${DOCKERHUB_REPO}:${tag}-aarch64"
+  publish_manifest "${tag}" "${tag}"
+done
+
+ARCH_BASE="${PRIMARY_TAGS[0]}"
+for tag in "${EXTRA_MANIFEST_TAGS[@]}"; do
+  publish_manifest "${tag}" "${ARCH_BASE}"
+done
 
 echo ""
-echo "Successfully published release images for v${RELEASE_VERSION}"
+echo "Successfully published ${MODE} images"

.claude/skills/precheck-pr/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: precheck-pr
-description: Self-check your branch before creating a PR — catch dead code, verify accuracy/perf claims, validate PR title format, and confirm merge readiness. Use when the user says "precheck", "self review", "pre-submit check", or "check my PR before I open it." Never posts to GitHub.
+description: Self-check your branch before creating a PR — catch dead code, prevent new model-specific Python examples, verify accuracy/perf claims, validate PR title format, and confirm merge readiness. Use when the user says "precheck", "self review", "pre-submit check", or "check my PR before I open it." Never posts to GitHub.
 ---
 
 # PR Pre-Check
@@ -65,6 +65,8 @@ Ask: "Quick mode or full mode?" Then walk the checklist for the detected PR type
 
 **Also run the Code-Quality sweep on every PR**, regardless of type or mode: the five diff-scoped checks in [references/code-quality.md](references/code-quality.md) — kwargs fragility, broad-except swallow, `Any`/wrong type hints, hot-path `.clone()`/`deepcopy`, and event-loop blocking — plus the advisory conventions (log level, structured logging, synchronization, cleanup, dependencies, naming) in [checklists.md](references/checklists.md). These count only lines the PR *adds* — the pre-existing backlog across the repo is out of scope.
 
+**Also run the Examples-Policy check on every PR** using [references/examples-policy.md](references/examples-policy.md). Inspect only Python paths introduced by the diff. Treat a new model-, checkpoint-, vendor-, or family-specific Python example as blocking; do not report pre-existing example debt.
+
 ### Step 5: Print Report
 
 ```
@@ -77,6 +79,7 @@ Pre-check report for <branch>
   ─────────────────  ──────
   PR title format    ✓
   Code quality       ⚠ 1 broad except, 2 Any hints
+  Examples policy    ✗ new model-specific Python example
   PR desc integrity  ✓
   Registry/config    ✓
   Dead code          ⚠ 2 warnings

.claude/skills/precheck-pr/references/checklists.md

@@ -47,6 +47,19 @@ Diff-scoped sweep for five fragility patterns that are pervasive in the existing
 
 Roll the ⚠/✗ counts into the report as a single **Code quality** dimension row.
 
+---
+
+## All PRs: Examples Policy (run on every PR, quick & full)
+
+Inspect Python files introduced under `examples/` using
+[examples-policy.md](examples-policy.md). This is a diff-scoped ratchet: existing
+model-specific examples may be edited or removed without triggering the check.
+
+- [ ] **No new model-specific Python examples:** the diff adds, copies, or renames no Python file whose path or behavior is specific to one model, checkpoint, vendor, or model family. ✗ for any violation. Route inference through a shared task/protocol entrypoint, model request behavior through `vllm_omni/model_extras`, and commands through task docs or `recipes/`.
+- [ ] **New generic example code is justified:** any new task-neutral Python file under `examples/` is an actual user-facing entrypoint rather than reusable production logic, a benchmark, an app, a setup tool, or a reproducer. ⚠ when the placement is questionable.
+
+Roll the result into the report as a separate **Examples policy** dimension.
+
 ### Conventions (eyeball — no grep)
 
 Judgment calls that don't grep cleanly — apply them while reading the diff, not as a mechanical check:

.claude/skills/precheck-pr/references/examples-policy.md

@@ -0,0 +1,68 @@
+# Examples Python Policy
+
+Prevent new model-specific Python examples without forcing cleanup of the
+existing backlog. Run this check on every PR in quick and full modes.
+
+## Scope
+
+Set the merge base as in the main workflow, then list Python paths introduced
+by the PR, including additions, copies, and renames:
+
+```bash
+BASE="$(git merge-base HEAD origin/main)"
+git diff --name-status --diff-filter=ACR --find-renames --find-copies \
+  "${BASE}"...HEAD -- 'examples/**/*.py'
+```
+
+Inspect destination paths for copies and renames. Existing model-specific files
+that are only modified or deleted are grandfathered and must not be reported.
+
+## Blocking classification
+
+Mark ✗ when a newly introduced Python file does any of the following:
+
+- Names or scopes a path to one model, checkpoint, vendor, or model family.
+- Hard-codes a model identifier, model-specific prompt, request shape, output
+  conversion, or launch configuration behind an otherwise generic filename.
+- Duplicates an existing task/protocol runner to demonstrate one model.
+- Adds a per-model helper below a generic task hub such as `text_to_speech/`.
+
+Do not infer safety from a generic filename alone. For example, a file named
+`text_to_audio.py` is still model-specific if it only implements one model's
+contract.
+
+## Acceptable additions
+
+A new Python example may pass when it is a genuinely model-neutral,
+user-facing task or protocol entrypoint. It must accept model selection through
+configuration and keep model-specific behavior out of the script.
+
+Use ⚠ when the file is model-neutral but reusable logic appears misplaced under
+`examples/`. Route code by purpose:
+
+| Code purpose | Preferred location |
+|---|---|
+| Model prompt, defaults, request/output adaptation | `vllm_omni/model_extras/` or production model modules |
+| Runnable model command and validation evidence | Task documentation or `recipes/` |
+| UI or interactive application | `apps/` |
+| Benchmark or evaluator | `benchmarks/` |
+| Regression reproducer | `tests/` |
+| Download, conversion, or setup utility | `tools/` |
+
+If a model cannot use an existing shared runner, report the missing generic
+capability. Do not solve the gap by adding a model-specific Python example.
+
+## Report format
+
+Add one row to the pre-check report:
+
+```text
+Examples policy    ✓ no new Python example paths
+Examples policy    ⚠ generic helper belongs in tools/
+Examples policy    ✗ examples/offline_inference/new_model/end2end.py is model-specific
+Examples policy    ✗ examples/offline_inference/x_to_y_model_name.py is model-specific
+```
+
+For a blocker, name the path, the model-specific behavior, and the appropriate
+shared runner or destination. Do not propose deleting unrelated existing
+examples as part of the PR.

.claude/skills/review-pr/SKILL.md

@@ -97,6 +97,7 @@ If docs and live code disagree, verify the code/tests and report the drift.
 | [test-quality-evaluation.md](references/checks/test-quality-evaluation.md) | Tests change, are absent for risky code, or may not exercise production behavior. |
 | [tests-docs-checklist.md](references/checks/tests-docs-checklist.md) | Coverage, CI markers, examples, user docs, or PR evidence need review. |
 | [verification.md](references/checks/verification.md) | Hardware, a server, or a runnable affected path is available for active verification. |
+| [examples-policy.md](../precheck-pr/references/examples-policy.md) | The PR adds, copies, or renames Python under `examples/`; apply the canonical policy shared with `precheck-pr`. |
 
 ### Delivery and reviewer coordination
 
@@ -152,6 +153,13 @@ contract they protect or use only the applicable evidence checks.
 Apply every category in [general-checks.md](references/process/general-checks.md) before
 lower-priority comments.
 
+If the diff census contains an added, copied, or renamed Python path under
+`examples/`, read and apply the canonical
+[examples policy](../precheck-pr/references/examples-policy.md). Treat a new
+model-specific Python example as blocking. Do not flag model-specific example
+debt that the PR only modifies or removes, and do not run the rest of the
+author-oriented `precheck-pr` workflow.
+
 For each changed value or behavior, trace:
 
 ```text

.readthedocs.yml

@@ -11,10 +11,15 @@ build:
     post_checkout:
       - git fetch --unshallow || true
     pre_install:
+      # `pre_install` runs before RTD upgrades pip, and the seeded pip 23.1
+      # rejects the `typing_extensions` wheel torch needs (fixed in pip 23.2).
+      - python -m pip install --upgrade pip
       # Pre-install CPU torch so the subsequent `pip install .[docs]` does not
       # pull ~2 GB of CUDA 13 libs (cuBLAS/cuDNN/NCCL/...) that exceed RTD's
-      # 1 GB memory cap and time out the mkdocs build.
-      - python -m pip install --index-url https://download.pytorch.org/whl/cpu torch
+      # 1 GB memory cap and time out the mkdocs build. Wheel-only: `--index-url`
+      # replaces PyPI, so an sdist fallback cannot reach its build deps. The
+      # quotes keep YAML from reading `:all:` as a mapping key.
+      - "python -m pip install --only-binary=:all: --index-url https://download.pytorch.org/whl/cpu torch"
 
 mkdocs:
   configuration: mkdocs.yml

benchmarks/diffusion/quantization_quality.py

@@ -158,7 +158,6 @@ def _build_omni_kwargs(args, quantization=None):
 def _generate_image(omni, args, prompt, seed):
     """Generate a single image and return (PIL.Image, time_seconds, memory_gib)."""
     from vllm_omni.inputs.data import OmniDiffusionSamplingParams
-    from vllm_omni.outputs import OmniRequestOutput
     from vllm_omni.platforms import current_omni_platform
 
     generator = torch.Generator(device=current_omni_platform.device_type).manual_seed(seed)
@@ -176,7 +175,7 @@ def _generate_image(omni, args, prompt, seed):
     elapsed = time.perf_counter() - start
     peak_mem = torch.accelerator.max_memory_allocated() / (1024**3)
 
-    req_out = OmniRequestOutput.unwrap_result(outputs)
+    req_out = outputs[0] if isinstance(outputs, list) else outputs
     if not req_out.images:
         raise ValueError("Could not extract image output from result.")
     img = req_out.images[0]
@@ -206,15 +205,15 @@ def _generate_video(omni, args, prompt, seed):
     elapsed = time.perf_counter() - start
     peak_mem = torch.accelerator.max_memory_allocated() / (1024**3)
 
-    first = outputs[0]
-    if hasattr(first, "request_output") and isinstance(first.request_output, list):
-        inner = first.request_output[0]
-        if isinstance(inner, OmniRequestOutput) and hasattr(inner, "images"):
-            frames = inner.images[0] if inner.images else None
-        else:
-            frames = inner
-    elif hasattr(first, "images") and first.images:
-        frames = first.images[0]
+    if isinstance(outputs, list) and isinstance(outputs[0], OmniRequestOutput):
+        first = outputs[0]
+    elif isinstance(outputs, OmniRequestOutput):
+        first = outputs
+    else:
+        raise ValueError("Could not extract video frames from output.")
+
+    if hasattr(first, "images"):
+        frames = first.images[0] if first.images else None
     else:
         raise ValueError("Could not extract video frames from output.")
 

benchmarks/glm_image/vllm-omni/inference.py

@@ -300,7 +300,7 @@ def benchmark(args: argparse.Namespace) -> None:
         output_idx = 0
         for stage_outputs in omni.generate(all_prompts, sampling_params_list, py_generator=True):
             if stage_outputs.final_output_type == "image":
-                request_output = stage_outputs.request_output
+                request_output = stage_outputs
                 request_id = getattr(request_output, "request_id", "")
 
                 images = getattr(request_output, "images", [])

benchmarks/lingbot_video/dense_pipeline_parity.py

@@ -167,9 +167,11 @@ def main() -> None:
     ]
     native_cmd = [
         sys.executable,
-        "examples/offline_inference/text_to_video/text_to_video_lingbot.py",
+        "examples/offline_inference/text_to_video/text_to_video.py",
         "--model",
         args.model,
+        "--model-class-name",
+        "LingBotVideoPipeline",
         "--prompt",
         args.prompt,
         "--output",

benchmarks/tts/README.md

@@ -139,6 +139,34 @@ python benchmarks/tts/bench_tts.py \
     --output-dir ./results
 ```
 
+#### Fixed IndexTTS 2.5 performance sweep
+
+IndexTTS 2.5 is registered in `model_configs.yaml` like the other TTS models.
+Use `--served-model-name` when the server was launched from a local native
+bundle instead of the registry key. The command below reproduces the acceptance
+workload: Seed-TTS Eval EN, `n=500`, concurrency `4/8/16/32`, five warmups, and
+dataset seed 0.
+
+```bash
+python benchmarks/tts/bench_tts.py \
+    --model IndexTeam/IndexTTS-2.5 \
+    --served-model-name /path/to/indextts-2.5 \
+    --task voice_clone --locale en \
+    --dataset-path /path/to/seedtts_testset \
+    --host 127.0.0.1 --port 8092 \
+    --concurrency 4 8 16 32 \
+    --num-prompts 500 --num-warmups 5 \
+    --output-dir ./results/indextts2_5 \
+    -- \
+    --tokenizer /path/to/indextts-2.5/qwen0.6bemo4-merge \
+    --seed 0 --metric-percentiles 50,95,99 --disable-tqdm --save-detailed
+```
+
+The wrapper benchmarks an already-running server. For reproducible quality
+comparisons, add `--request-seed 42`; production-performance runs should omit
+it. The trailing `--seed 0` controls Seed-TTS row selection rather than model
+sampling.
+
 ### 4. Plot a sweep
 
 ```bash

benchmarks/tts/bench_tts.py

@@ -84,12 +84,17 @@ def build_bench_args(
     result_filename: str | None,
     extra_cli_args: list[str],
     output_len: int | None = None,
+    served_model_name: str | None = None,
+    num_warmups: int = 2,
+    request_seed: int | None = None,
 ) -> list[str]:
     """Build the ``vllm bench serve --omni`` command for one (task, concurrency) run."""
     dataset_name = _TASK_TO_DATASET[task]
     backend: str = model_cfg["backend"]
     endpoint: str = model_cfg["endpoint"]
-    task_extra_body: dict[str, Any] = (model_cfg.get("task_extra_body") or {}).get(task) or {}
+    task_extra_body: dict[str, Any] = dict((model_cfg.get("task_extra_body") or {}).get(task) or {})
+    if request_seed is not None:
+        task_extra_body["seed"] = request_seed
 
     # Resolve dataset path
     if dataset_path:
@@ -109,7 +114,7 @@ def build_bench_args(
         "--port",
         str(port),
         "--model",
-        model,
+        served_model_name or model,
         "--backend",
         backend,
         "--endpoint",
@@ -119,7 +124,7 @@ def build_bench_args(
         "--num-prompts",
         str(num_prompts),
         "--num-warmups",
-        "2",
+        str(num_warmups),
         "--percentile-metrics",
         "ttft,e2el,audio_rtf,audio_ttfp,audio_duration,audio_underrun",
     ]
@@ -152,7 +157,8 @@ def build_bench_args(
         if result_filename:
             cmd += ["--result-filename", result_filename]
 
-    cmd += extra_cli_args
+    passthrough = extra_cli_args[1:] if extra_cli_args[:1] == ["--"] else extra_cli_args
+    cmd += passthrough
     return cmd
 
 
@@ -225,6 +231,12 @@ def main() -> None:
     parser.add_argument(
         "--model", required=True, help="HuggingFace model ID (e.g. Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice)"
     )
+    parser.add_argument(
+        "--served-model-name",
+        default=None,
+        help="Model name/path accepted by the running server. Use this when --model selects a registry entry "
+        "but the server was launched from a local bundle path.",
+    )
     parser.add_argument("--task", default="all", help="Task type: voice_clone | default_voice | voice_design | all")
     parser.add_argument("--locale", default="en", choices=["en", "zh"])
     parser.add_argument("--concurrency", type=int, nargs="+", default=[1, 4], metavar="N")
@@ -236,6 +248,13 @@ def main() -> None:
         metavar="N",
         help="Number of prompts per run. If one value, applied to all concurrency levels.",
     )
+    parser.add_argument("--num-warmups", type=int, default=2, help="Warmup requests before each measured run")
+    parser.add_argument(
+        "--request-seed",
+        type=int,
+        default=None,
+        help="Optional model sampling seed added to every request body for reproducible quality runs",
+    )
     parser.add_argument(
         "--dataset-path", default=None, help="Root of seed-tts-eval dataset (required for voice_clone/default_voice)"
     )
@@ -256,6 +275,9 @@ def main() -> None:
     parser.add_argument("extra", nargs=argparse.REMAINDER, help="Extra args passed directly to vllm bench serve")
     args = parser.parse_args()
 
+    if args.num_warmups < 0:
+        parser.error("--num-warmups cannot be negative")
+
     model_configs = load_model_configs(Path(args.model_configs))
     if args.model not in model_configs:
         known = "\n  ".join(model_configs.keys())
@@ -308,6 +330,9 @@ def main() -> None:
                 result_filename=result_filename,
                 extra_cli_args=args.extra or [],
                 output_len=args.output_len,
+                served_model_name=args.served_model_name,
+                num_warmups=args.num_warmups,
+                request_seed=args.request_seed,
             )
             result = run_one_benchmark(cmd)
             if result is not None:

benchmarks/tts/model_configs.yaml

@@ -7,6 +7,17 @@
 # is tracked here.
 
 models:
+  IndexTeam/IndexTTS-2.5:
+    supported_tasks: [voice_clone]
+    backend: openai-audio-speech
+    endpoint: /v1/audio/speech
+    trust_remote_code: true
+    task_extra_body:
+      voice_clone:
+        extra_params:
+          lang: en
+          text_normalization: true
+
   # -CustomVoice checkpoints lack speaker_encoder weights, so voice_clone is
   # NOT supported (an attempt raises ValueError from _extract_speaker_embedding
   # at model runtime).  Use -Base for voice_clone.

docker/Dockerfile.xpu

@@ -88,7 +88,7 @@ ENV UV_EXTRA_INDEX_URL=${PIP_EXTRA_INDEX_URL}
 ENV UV_INDEX_STRATEGY="unsafe-best-match"
 ENV UV_LINK_MODE="copy"
 
-ARG VLLM_VERSION=v0.26.0
+ARG VLLM_VERSION=v0.27.0
 RUN git clone -b ${VLLM_VERSION} https://github.com/vllm-project/vllm
 WORKDIR /workspace/vllm
 
@@ -108,7 +108,7 @@ ENV VLLM_WORKER_MULTIPROC_METHOD=spawn
 
 RUN --mount=type=cache,target=/root/.cache/uv \
     uv pip uninstall triton triton-xpu && \
-    uv pip install triton-xpu==3.7.1 && \
+    uv pip install triton-xpu==3.7.2 && \
     uv pip install --no-build-isolation --no-deps .
 
 CMD ["/bin/bash"]
@@ -144,12 +144,13 @@ WORKDIR /workspace/vllm-omni
 COPY . .
 
 ENV VLLM_OMNI_TARGET_DEVICE=xpu
+# Keep the upstream oneCCL packages paired with the XPU torch runtime.
 RUN --mount=type=cache,target=/root/.cache/uv \
     uv pip install --no-build-isolation \
     --extra-index-url https://mirrors.ustc.edu.cn/pypi/simple/ \
     ".[dev]" && \
-    uv pip uninstall triton oneccl oneccl-devel && \
-    uv pip install triton-xpu==3.7.1 --reinstall
+    uv pip uninstall triton && \
+    uv pip install triton-xpu==3.7.2 --reinstall
 
 FROM vllm-omni AS vllm-omni-openai
 

docs/.nav.yml

@@ -40,30 +40,13 @@ nav:
   - Models:
     - models/supported_models.md
   - Features:
-    - Sleep Mode: features/sleep_mode.md
-    - Diffusion Features:
-      - Overview: user_guide/diffusion_features.md
-      - Feature Compatibility: user_guide/feature_compatibility.md
-      - Regional Compilation: user_guide/diffusion/regional_compilation.md
-      - Cache Acceleration:
-        - TeaCache: user_guide/diffusion/cache_acceleration/teacache.md
-        - Cache-DiT: user_guide/diffusion/cache_acceleration/cache_dit.md
-      - Attention Backends: user_guide/diffusion/attention_backends.md
-      - Frame Interpolation: user_guide/diffusion/frame_interpolation.md
-      - Parallelism:
-        - Overview: user_guide/diffusion/parallelism/overview.md
-        - CFG Parallel: user_guide/diffusion/parallelism/cfg_parallel.md
-        - Expert Parallel: user_guide/diffusion/parallelism/expert_parallel.md
-        - Hybrid Sharded Data Parallel: user_guide/diffusion/parallelism/hsdp.md
-        - Sequence Parallel: user_guide/diffusion/parallelism/sequence_parallel.md
-        - Tensor Parallel: user_guide/diffusion/parallelism/tensor_parallel.md
-        - VAE Parallelism: user_guide/diffusion/parallelism/vae_parallelism.md
-      - CPU Offloading: user_guide/diffusion/cpu_offload.md
-      - LoRA: user_guide/diffusion/lora.md
-      - Custom Pipeline: features/custom_pipeline.md
-      - Execution Modes: user_guide/diffusion/execution_modes.md
+    - Overview: features/README.md
+    - Runtime and Stage Execution:
+      - Execution Modes and Streaming: user_guide/diffusion/execution_modes.md
+      - Prefill-Decode Disaggregation (experimental): features/pd_disaggregation.md
+      - Sleep Mode: features/sleep_mode.md
     - Quantization:
-      - Quantization: user_guide/quantization/overview.md
+      - Overview: user_guide/quantization/overview.md
       - Online Quantization: user_guide/quantization/online.md
       - Quantized KV Cache: user_guide/quantization/quantized_kvcache.md
       - FP8 W8A8: user_guide/quantization/fp8.md
@@ -72,9 +55,33 @@ nav:
       - MXFP8 W8A8: user_guide/quantization/mxfp8.md
       - MXFP4 W4A4: user_guide/quantization/mxfp4.md
       - ModelOpt: user_guide/quantization/modelopt.md
-      - GGUF: user_guide/quantization/gguf.md
       - AutoRound: user_guide/quantization/autoround.md
       - msModelSlim: user_guide/quantization/msmodelslim.md
+    - Diffusion Acceleration:
+      - Overview and Compatibility:
+        - Overview: user_guide/diffusion_features.md
+        - Feature Compatibility: user_guide/feature_compatibility.md
+      - CPU Offloading: user_guide/diffusion/cpu_offload.md
+      - Cache Acceleration:
+        - TeaCache: user_guide/diffusion/cache_acceleration/teacache.md
+        - Cache-DiT: user_guide/diffusion/cache_acceleration/cache_dit.md
+      - Parallelism:
+        - Overview: user_guide/diffusion/parallelism/overview.md
+        - CFG Parallel: user_guide/diffusion/parallelism/cfg_parallel.md
+        - Expert Parallel: user_guide/diffusion/parallelism/expert_parallel.md
+        - Hybrid Sharded Data Parallel: user_guide/diffusion/parallelism/hsdp.md
+        - Sequence Parallel: user_guide/diffusion/parallelism/sequence_parallel.md
+        - Tensor Parallel: user_guide/diffusion/parallelism/tensor_parallel.md
+        - VAE Parallelism: user_guide/diffusion/parallelism/vae_parallelism.md
+      - Attention Backends: user_guide/diffusion/attention_backends.md
+      - Regional Compilation: user_guide/diffusion/regional_compilation.md
+      - Frame Interpolation: user_guide/diffusion/frame_interpolation.md
+      - Startup and Loading: user_guide/diffusion/startup_and_loading.md
+      - LoRA: user_guide/diffusion/lora.md
+    - Experimental:
+      - Session State Manager: features/session_state_manager.md
+  - Integrations:
+    - Custom Diffusion Pipeline: features/custom_pipeline.md
     - ComfyUI: features/comfyui.md
   - Fault Tolerance:
     - Failure Modes: user_guide/fault_injection_reliability_matrix.md
@@ -108,6 +115,7 @@ nav:
           - Shared Memory Connector: design/feature/omni_connectors/shared_memory_connector.md
           - Yuanrong Store Connector: design/feature/omni_connectors/yuanrong_connector.md
           - Yuanrong Transfer Engine Connector: design/feature/omni_connectors/yuanrong_transfer_engine_connector.md
+      - Quantization: design/feature/quantization.md
       - Diffusion Acceleration:
         - Parallelism:
           - CFG-Parallel: design/feature/cfg_parallel.md
@@ -117,9 +125,8 @@ nav:
           - Sequence Parallel: design/feature/sequence_parallel.md
           - Tensor Parallel: design/feature/tensor_parallel.md
           - VAE Patch Parallelism: design/feature/vae_parallel.md
-        - Attention Backends:
+        - Attention Optimization:
           - Skip-Softmax: design/feature/skip_softmax.md
-        - Quantization: design/feature/quantization.md
         - Cache-DiT: design/feature/cache_dit.md
         - TeaCache: design/feature/teacache.md
         - Diffusion Continuous Batching: design/feature/diffusion_continuous_batching.md

docs/configuration/README.md

@@ -14,10 +14,6 @@ For introduction, please check [Introduction for stage config](./stage_configs.m
 
 - **[GPU Memory Calculation and Configuration](./gpu_memory_utilization.md)** - Guide on how to calculate memory requirements and set up `gpu_memory_utilization` for optimal performance
 
-## Multi-Stage Recipes
-
-- **[Prefill-Decode Disaggregation](./pd_disaggregation.md)** - How to derive a PD-aware Qwen3-Omni stage config from the default config without introducing another bundled YAML
-
 ## Optimization Features
 
 - **[Diffusion Features Overview](../user_guide/diffusion_features.md)** - Complete overview of all diffusion model features and supported models

docs/configuration/pd_disaggregation.md

@@ -1,171 +0,0 @@
-# Prefill-Decode (PD) Disaggregation
-
-PD disaggregation splits the Qwen3-Omni thinker into separate prefill and decode
-stages so prompt processing and token generation can run on different workers.
-
-This is documented as a stage-config recipe instead of a bundled YAML because the
-deployment-specific values usually change per environment:
-
-- GPU placement
-- `tensor_parallel_size`
-- connector backend and connector ports
-- connector IPs or bootstrap addresses
-
-Start from the [default Qwen3-Omni stage config](gh-file:vllm_omni/deploy/qwen3_omni_moe.yaml)
-and copy it to your own file, for example `qwen3_omni_pd.yaml`. Then apply the
-changes below.
-
-## Requirements
-
-- 3+ GPUs for a basic layout: prefill, decode, and talker+code2wav
-- A KV connector supported by vLLM, such as `MooncakeConnector`
-- Matching `tensor_parallel_size` on the prefill and decode thinker stages
-
-## 1. Split the thinker into prefill and decode stages
-
-Replace the original thinker stage with two stages:
-
-```yaml
-stage_args:
-  - stage_id: 0
-    stage_type: llm
-    is_prefill_only: true
-    runtime:
-      devices: "0"
-    engine_args:
-      max_num_seqs: 16
-      model_stage: thinker
-      model_arch: Qwen3OmniMoeForConditionalGeneration
-      worker_type: ar
-      scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler
-      gpu_memory_utilization: 0.9
-      enforce_eager: true
-      trust_remote_code: true
-      engine_output_type: latent
-      distributed_executor_backend: "mp"
-      enable_prefix_caching: false
-      max_num_batched_tokens: 32768
-      hf_config_name: thinker_config
-      tensor_parallel_size: 1
-      kv_transfer_config:
-        kv_connector: "MooncakeConnector"
-        kv_role: "kv_producer"
-        kv_rank: 0
-        kv_parallel_size: 2
-        kv_connector_extra_config:
-          mooncake_bootstrap_port: 25201
-    final_output: false
-    is_comprehension: true
-    default_sampling_params:
-      temperature: 0.4
-      top_p: 0.9
-      top_k: 1
-      max_tokens: 2048
-      seed: 42
-      detokenize: True
-      repetition_penalty: 1.05
-
-  - stage_id: 1
-    stage_type: llm
-    is_decode_only: true
-    runtime:
-      devices: "1"
-    engine_args:
-      max_num_seqs: 64
-      model_stage: thinker
-      model_arch: Qwen3OmniMoeForConditionalGeneration
-      worker_type: ar
-      scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler
-      gpu_memory_utilization: 0.9
-      enforce_eager: true
-      trust_remote_code: true
-      engine_output_type: latent
-      distributed_executor_backend: "mp"
-      enable_prefix_caching: false
-      max_num_batched_tokens: 32768
-      hf_config_name: thinker_config
-      tensor_parallel_size: 1
-      kv_transfer_config:
-        kv_connector: "MooncakeConnector"
-        kv_role: "kv_consumer"
-        kv_rank: 1
-        kv_parallel_size: 2
-        kv_connector_extra_config:
-          mooncake_bootstrap_port: 25202
-    engine_input_source: [0]
-    final_output: true
-    final_output_type: text
-    is_comprehension: true
-    default_sampling_params:
-      temperature: 0.4
-      top_p: 0.9
-      top_k: 1
-      max_tokens: 2048
-      seed: 42
-      detokenize: True
-      repetition_penalty: 1.05
-```
-
-Notes:
-
-- `is_prefill_only: true` marks the thinker stage that only saves KV.
-- `is_decode_only: true` marks the thinker stage that resumes from remote KV.
-- `kv_transfer_config` is required on both stages.
-- The orchestrator forces the prefill stage to run with `max_tokens=1`, so the
-  prefill side only processes the prompt and exports KV.
-
-## 2. Shift the downstream stages by one index
-
-After inserting the extra thinker stage, renumber the remaining stages:
-
-```yaml
-  - stage_id: 2
-    runtime:
-      devices: "2"
-    engine_input_source: [1]
-    sync_process_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.thinker2talker_token_only
-
-  - stage_id: 3
-    runtime:
-      devices: "2"
-    engine_args:
-      max_num_seqs: 1
-    engine_input_source: [2]
-    sync_process_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.talker2code2wav_token_only
-```
-
-Compared with the default Qwen3-Omni config:
-
-- the talker becomes stage `2` instead of stage `1`
-- the code2wav stage becomes stage `3` instead of stage `2`
-- the talker now reads from decode stage `1`
-
-## 3. Add runtime edges for the four-stage pipeline
-
-```yaml
-runtime:
-  enabled: true
-  edges:
-    - from: 0
-      to: 1
-    - from: 1
-      to: 2
-    - from: 2
-      to: 3
-```
-
-## 4. Launch with your custom config
-
-```bash
-vllm serve Qwen/Qwen3-Omni-30B-A3B-Instruct --omni --port 8091 \
-    --deploy-config /path/to/qwen3_omni_pd.yaml
-```
-
-## Operational Notes
-
-- `MooncakeConnector` does not support heterogeneous TP sizes across the PD
-  pair. Keep prefill and decode at the same `tensor_parallel_size`.
-- If the thinker requires TP=2, both thinker stages must use TP=2 and be given
-  separate GPU sets, for example `"0,1"` for prefill and `"2,3"` for decode.
-- Choose connector ports and addresses that match your deployment. The values
-  shown above are examples only.

docs/design/index.md

@@ -11,6 +11,10 @@ in the active navigation.
 
 ## Feature Design Documents
 
+For user-facing configuration and current compatibility, see the
+[Features overview](../features/README.md). A design document defines an
+implementation contract; it is not, by itself, a general support claim.
+
 ### Runtime and stage execution
 
 - [Disaggregated Inference](feature/disaggregated_inference.md)
@@ -30,6 +34,10 @@ in the active navigation.
 - [Yuanrong Store Connector](feature/omni_connectors/yuanrong_connector.md)
 - [Yuanrong Transfer Engine Connector](feature/omni_connectors/yuanrong_transfer_engine_connector.md)
 
+### Quantization
+
+- [Quantization](feature/quantization.md)
+
 ### Diffusion acceleration
 
 #### Parallelism
@@ -42,14 +50,14 @@ in the active navigation.
 - [Tensor Parallel](feature/tensor_parallel.md)
 - [VAE Patch Parallelism](feature/vae_parallel.md)
 
-#### Attention Backends
+#### Attention optimization
+
+The [Diffusion Attention Backends](../user_guide/diffusion/attention_backends.md)
+guide lists all selectable backends and their platform defaults. Skip-Softmax
+is the backend-related optimization with a standalone feature-design contract:
 
 - [Skip-Softmax](feature/skip_softmax.md)
 
-#### Quantization
-
-- [Quantization](feature/quantization.md)
-
 - [Cache-DiT](feature/cache_dit.md)
 - [TeaCache](feature/teacache.md)
 - [Diffusion Continuous Batching](feature/diffusion_continuous_batching.md)

docs/features/README.md

@@ -0,0 +1,73 @@
+# Features
+
+Use this section to choose and configure vLLM-Omni runtime and optimization
+features. The navigation follows the active
+[Feature Design](../design/index.md#feature-design-documents) taxonomy whenever
+an implementation contract has a user-facing workflow.
+
+!!! note "User guides, design documents, and recipes"
+
+    User guides explain how to enable a feature and where it is supported.
+    Design documents define internal contracts and do not, by themselves,
+    imply general or production support. For model-specific launch flags and
+    hardware requirements, use the linked examples and validated recipes from
+    [Supported Models](../models/supported_models.md).
+
+## Runtime and Stage Execution
+
+| Goal | User guide | Related design contract |
+| --- | --- | --- |
+| Choose serial, batched, step-wise, or streaming diffusion execution | [Execution Modes and Streaming](../user_guide/diffusion/execution_modes.md) | [Diffusion Continuous Batching](../design/feature/diffusion_continuous_batching.md), [Async Diffusion Output](../design/feature/async_diffusion_output.md) |
+| Reclaim stage memory without restarting the server | [Sleep Mode](sleep_mode.md) | Runtime lifecycle behavior is documented in the user guide |
+
+The [Prefill-Decode Disaggregation design reference](pd_disaggregation.md)
+describes the experimental Qwen3-Omni runtime topology and its current
+integration boundaries. It is not yet a supported launch recipe.
+
+Some runtime designs are deliberately not promoted as standalone User Guide
+features yet:
+
+- Generic, OmniConnector-based
+  [Disaggregated Inference](../design/feature/disaggregated_inference.md) is not
+  yet a standalone User Guide feature. Its
+  [OmniConnector implementations](../design/index.md#communication) describe
+  transport contracts, while practical connector configuration remains under
+  [Pipeline and deploy configurations](../configuration/stage_configs.md).
+- [Async Chunk](../design/feature/async_chunk.md) and
+  [Async Omni Output Materialization](../design/feature/omni_async_output_materialization.md)
+  are model- and pipeline-dependent. Use the selected model's deploy
+  configuration and recipe for the supported settings.
+- [Automatic Prefix Caching](../design/feature/prefix_caching.md) remains a
+  design-level contract until its user-facing configuration and compatibility
+  surface is consolidated.
+
+## Quantization
+
+Quantization is a cross-model feature rather than a diffusion-only
+optimization. The unified [`quantization_config` guide](../user_guide/quantization/overview.md)
+covers diffusion-only models, multi-stage omni/TTS models, and multi-stage
+diffusion models. Its [design contract](../design/feature/quantization.md)
+defines the shared configuration and backend extension points.
+
+## Diffusion Acceleration
+
+| Goal | User guide | Related design contract |
+| --- | --- | --- |
+| Compare acceleration methods and supported combinations | [Overview](../user_guide/diffusion_features.md), [Feature Compatibility](../user_guide/feature_compatibility.md) | [Diffusion acceleration designs](../design/index.md#diffusion-acceleration) |
+| Move weights between host and device memory | [CPU Offloading](../user_guide/diffusion/cpu_offload.md) | [Distributed Layerwise Offload](../design/feature/distributed_layerwise_offload.md) |
+| Reuse denoising computation | [Cache Acceleration](../user_guide/diffusion/cache_acceleration/cache_dit.md) | [Cache-DiT](../design/feature/cache_dit.md), [TeaCache](../design/feature/teacache.md) |
+| Distribute diffusion work across devices | [Parallelism](../user_guide/diffusion/parallelism/overview.md) | [Parallelism designs](../design/index.md#parallelism) |
+| Select dense, sparse, or quantized attention paths | [Attention Backends](../user_guide/diffusion/attention_backends.md) | [Skip-Softmax](../design/feature/skip_softmax.md) |
+| Compile repeated diffusion regions | [Regional Compilation](../user_guide/diffusion/regional_compilation.md) | User-facing optimization guide |
+| Add generated video frames | [Frame Interpolation](../user_guide/diffusion/frame_interpolation.md) | User-facing extension guide |
+| Reduce diffusion model startup time | [Startup and Loading](../user_guide/diffusion/startup_and_loading.md) | User-facing loading guide |
+| Apply diffusion adapters | [LoRA](../user_guide/diffusion/lora.md) | User-facing extension guide |
+
+The [Pipeline Parallelism guide](../user_guide/diffusion/parallelism/pipeline_parallel.md)
+remains available by direct link, but it is not promoted in the primary User
+Guide navigation until its user-facing support and placement are settled.
+
+## Experimental
+
+[Session State Manager](session_state_manager.md) is opt-in and experimental.
+Its APIs and compatibility may change without notice.

docs/features/pd_disaggregation.md

@@ -0,0 +1,140 @@
+# Prefill-Decode (PD) Disaggregation (experimental)
+
+!!! warning "Experimental design reference"
+    PD disaggregation is only partially integrated on the current `main` branch.
+    The runtime contains PD detection and routing scaffolding, but there is no
+    supported deploy overlay or end-to-end validated launch recipe yet. Treat this
+    page as a description of the current design, not as production guidance.
+
+Prefill-decode disaggregation splits the Qwen3-Omni Thinker into two logical
+stages: a prefill worker that processes the prompt and produces KV cache, and a
+decode worker that imports that cache and generates tokens. The Talker and
+Code2Wav stages remain downstream; PD does not disaggregate the entire
+Qwen3-Omni pipeline.
+
+## Current design
+
+### Runtime topology
+
+```mermaid
+flowchart LR
+    client["Client request"] --> orchestrator["Omni Orchestrator"]
+    orchestrator -->|"original prompt<br/>max_tokens = 1"| prefill["Thinker prefill<br/>KV producer"]
+    prefill -->|"KV blocks"| connector[("vLLM KV connector")]
+    prefill -->|"KV routing metadata"| orchestrator
+    orchestrator -->|"original prompt<br/>remote KV parameters"| decode["Thinker decode<br/>KV consumer"]
+    connector -->|"remote KV load"| decode
+    decode -->|"generated text"| client
+```
+
+The orchestrator owns the request lifecycle, while the vLLM KV connector owns
+the bulk KV transfer. The original prompt is submitted to both Thinker stages:
+the decode worker still needs the request tokens and multimodal features to
+construct its request, while the imported KV cache prevents it from repeating
+the completed prefill computation.
+
+For the full Qwen3-Omni pipeline, splitting the Thinker adds one logical stage:
+
+| Stage | Role | Output or handoff |
+| --- | --- | --- |
+| 0 | Thinker prefill | Saves prompt KV and prompt-side multimodal output |
+| 1 | Thinker decode | Loads remote KV, generates text, and produces decode-side conditioning |
+| 2 | Talker | Converts combined Thinker conditioning into codec codes |
+| 3 | Code2Wav | Converts codec codes into audio |
+
+Text remains a final output of the decode stage, and audio remains a final
+output of Code2Wav.
+
+### Request lifecycle
+
+1. At startup, PD detection looks for one stage marked `is_prefill_only` and
+   one downstream stage marked `is_decode_only`. The decode stage must list the
+   prefill stage as an input source.
+2. The orchestrator clones the Thinker sampling parameters for prefill, forces
+   `max_tokens=1`, clears stop conditions, and sets producer-side
+   `kv_transfer_params`. This makes prefill finish by length after exporting KV.
+3. When prefill finishes, the orchestrator captures its KV routing metadata and
+   Qwen multimodal output under the original request ID.
+4. The orchestrator submits the original prompt to decode with consumer-side
+   `kv_transfer_params`. Those parameters identify the transfer, prefill engine,
+   bootstrap endpoint, and, when required by the connector, remote request.
+5. The decode engine imports the KV cache and continues generation. The normal
+   Omni stage path then carries Thinker conditioning through Talker and
+   Code2Wav.
+
+### Control and data paths
+
+PD uses two different connector layers. They solve different handoffs and are
+not interchangeable.
+
+| Path | Payload | Owner |
+| --- | --- | --- |
+| PD control path | Request ID, original prompt, `transfer_id`, bootstrap address, engine ID, and connector-specific request metadata | Omni Orchestrator |
+| PD data path | Thinker attention KV blocks | vLLM KV connector |
+| Downstream model path | Thinker hidden states and embeddings, Talker codec codes, and audio tensors | vLLM-Omni stage connectors and processors |
+
+### Detection and validation
+
+The current runtime accepts one logical prefill/decode pair per pipeline. Each
+stage may use tensor parallelism, but both sides of the pair must satisfy these
+invariants:
+
+- both stages define `kv_transfer_config`
+- prefill uses `kv_producer` or `kv_both`
+- decode uses `kv_consumer` or `kv_both`
+- both stages use the same KV connector
+- `kv_buffer_device` and `kv_buffer_size` agree when set on both stages
+- both stages use the same `tensor_parallel_size`
+
+These checks happen before request processing so an inconsistent PD topology
+fails during engine initialization.
+
+## What is implemented today
+
+| Area | Current state |
+| --- | --- |
+| Pair detection and topology validation | Implemented in the shared PD helper |
+| Prefill sampling preparation | Implemented for sync and async entrypoints |
+| Prefill-to-decode routing | Implemented in the current Orchestrator |
+| Deploy-based configuration | Not yet migrated; the supported `stages:` deploy schema cannot currently expand Qwen3-Omni into the four-stage PD topology |
+| Qwen3 Thinker-to-Talker handoff | Bridge state and embedding-merge helpers exist, but the merge is not wired into the live stage processor |
+| Connector compatibility | Incomplete; the Orchestrator currently requires `remote_request_id`, while the pinned Mooncake connector completes prefill without returning that metadata |
+| Validation coverage | PD entrypoint tests and the PD deploy overlay in online end-to-end tests are temporarily disabled during migration |
+| Performance | No non-regression or throughput/latency result is established for the current path |
+
+Because these pieces are still converging, the previous legacy `stage_args` YAML
+and `vllm serve` command have been removed. The serve CLI now accepts the
+deploy-based `stages:` schema, so retaining the legacy recipe would imply a
+launch path that current `main` rejects.
+
+## Configuration contract under development
+
+The intended deploy integration starts from the
+[default Qwen3-Omni deploy config](gh-file:vllm_omni/deploy/qwen3_omni_moe.yaml)
+and expands its frozen three-stage pipeline into the four stages shown above.
+A complete integration must:
+
+- split the Thinker definition without duplicating the Talker or Code2Wav model
+  contracts
+- remap downstream stage IDs, input sources, and stage connectors
+- inject producer and consumer `kv_transfer_config` values into the two Thinker
+  stages
+- allocate separate device sets for prefill and decode
+- preserve both final outputs: decode text and Code2Wav audio
+- provide connector-specific routing metadata without hard-coding one connector
+  contract
+
+The required number of GPUs depends on tensor-parallel sizes and whether Talker
+and Code2Wav share a device. A basic TP=1 layout uses one device for prefill, one
+for decode, and one for the two downstream audio stages.
+
+## Implementation map
+
+- PD detection, validation, and sampling-parameter preparation:
+  [`PDDisaggregationMixin`](gh-file:vllm_omni/entrypoints/pd_utils.py)
+- startup detection and connector metadata:
+  [`AsyncOmniEngine`](gh-file:vllm_omni/engine/async_omni_engine.py)
+- request-scoped prefill-to-decode routing:
+  [`Orchestrator`](gh-file:vllm_omni/engine/orchestrator.py)
+- Qwen3 downstream conditioning helpers:
+  [Qwen3-Omni stage input processor](gh-file:vllm_omni/model_executor/stage_input_processors/qwen3_omni.py)

docs/models/supported_models.md

@@ -80,6 +80,7 @@ th {
 | `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` | ✅︎ | | | | — |
+| `IndexTTS25TalkerForConditionalGeneration` | IndexTTS-2.5 | native `checkpoints/` bundle | ✅︎ | | | | [Repository](https://github.com/vllm-project/vllm-omni/blob/main/recipes/IndexTeam/IndexTTS-2_5.md) |
 | `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` | ✅︎ | ✅︎ | | | — |

docs/user_guide/diffusion/startup_and_loading.md

@@ -0,0 +1,60 @@
+# Diffusion Startup and Loading
+
+Large diffusion models can take several minutes to load at startup. vLLM-Omni
+loads safetensors shards in parallel to reduce this initialization time.
+
+Multi-thread weight loading is enabled by default with four threads. No
+configuration is needed for the default behavior.
+
+## Configuration
+
+| Parameter | CLI flag | Default | Description |
+| --- | --- | --- | --- |
+| `enable_multithread_weight_load` | `--disable-multithread-weight-load` | `True` | Pass the flag to disable multi-thread loading |
+| `num_weight_load_threads` | `--num-weight-load-threads` | `4` | Number of parallel weight-loading threads |
+
+!!! tip
+
+    The default balances startup speed and disk I/O contention. Fast NVMe
+    storage may benefit from more threads, while network storage or hard disks
+    may not.
+
+## Online Serving
+
+```bash
+# Default: multi-thread loading with four threads
+vllm serve Qwen/Qwen-Image --omni --port 8091
+
+# Increase the thread count
+vllm serve Wan-AI/Wan2.2-I2V-A14B-Diffusers --omni \
+  --num-weight-load-threads 8
+
+# Disable multi-thread loading
+vllm serve Qwen/Qwen-Image --omni --disable-multithread-weight-load
+```
+
+## Offline Inference
+
+```python
+from vllm_omni import Omni
+
+# Default: multi-thread loading with four threads
+omni = Omni(model="Qwen/Qwen-Image")
+
+# Increase the thread count
+omni = Omni(
+    model="Wan-AI/Wan2.2-I2V-A14B-Diffusers",
+    num_weight_load_threads=8,
+)
+```
+
+## Reference Benchmarks
+
+The following measurements were collected on NVIDIA H800 hardware. Treat them
+as reference results rather than a guarantee for other storage or hardware
+configurations.
+
+| Model | Sequential loading | Multi-thread loading | Speedup |
+| --- | ---: | ---: | ---: |
+| **Qwen/Qwen-Image** (53.7 GiB) | 168 s | 27 s | **6.2x** |
+| **Wan-AI/Wan2.2-I2V-A14B-Diffusers** (64.5 GiB) | 283 s | 56 s | **5.1x** |

docs/user_guide/diffusion_features.md

@@ -49,7 +49,7 @@ Parallelism methods distribute computation across GPUs without quality loss (mat
 
 | Method | Description | Best For |
 |--------|-------------|----------|
-| **[Multi-Thread Weight Loading](#multi-thread-weight-loading)** | Loads safetensors shards in parallel using a thread pool | All diffusion models; reduces startup from minutes to seconds |
+| **[Multi-Thread Weight Loading](diffusion/startup_and_loading.md)** | Loads safetensors shards in parallel using a thread pool | All diffusion models; reduces startup from minutes to seconds |
 
 **Note:** Some acceleration methods can be combined together for optimized performance. See [Feature Compatibility Table](#feature-compatibility) and [Feature Compatibility Tutorial](feature_compatibility.md) for detailed configuration examples.
 
@@ -100,7 +100,9 @@ The following tables show which models support each feature:
 
 - **🔀SP (Ulysses & Ring)**: Includes both Ulysses-SP and Ring-Attention methods
 - ✅ = Fully supported
+- ✅* = Supported with the constraint listed below the table
 - ❌ = Not supported
+- ❓ = Not verified; not recommended
 
 > Notes:
 
@@ -119,7 +121,7 @@ The following tables show which models support each feature:
 | **FLUX.2-dev**           |     ✅     |     ✅      |           ✅           |       ✅        |         ✅         |          ❌          |   ✅    |             ✅             |          ✅           |       ❌        |        ❌         |
 | **GLM-Image**            |     ❌     |     ❌      |           ❌           |       ✅        |         ✅         |          ❌          |   ✅    |             ❌             |          ❌           |       ❌        |        ❌         |
 | **Hidream-I1-Full**        |     ❌     |     ❌      |           ❌           |       ❌        |         ✅         |          ❌          |   ❌    |             ❌             |          ❌           |       ❌        |        ❌         |
-| **HunyuanImage3**        |     ❌     |     ✅      |           ❌           |       ❌        |         ✅         |          ❌          |   ❌    |             ❌             |          ❌           |       ✅        |        ❌         |
+| **HunyuanImage3**        |     ❌     |     ✅      |           ❌           |       ❌        |         ✅         |          ❌          |   ❌    |             ❌             |          ❌           |       ✅        |        ✅*        |
 | **Krea 2**               |     ❌     |     ✅      |           ❌           |       ❌        |         ❌         |          ❌          |   ✅    |             ✅             |      ✅ (decode)      |       ❌        |        ❌         |
 | **LongCat-Image**        |     ✅     |     ✅      |           ✅           |       ✅        |         ✅         |          ❌          |   ❌    |             ✅             |          ❌           |       ❌        |        ❌         |
 | **LongCat-Image-Edit**   |     ✅     |     ✅      |           ✅           |       ✅        |         ✅         |          ❌          |   ❌    |             ✅             |          ❌           |       ❌        |        ❌         |
@@ -145,6 +147,7 @@ The following tables show which models support each feature:
 > 2. `Tongyi-MAI/Z-Image-Turbo` and `SII-GAIR/daVinci-MagiHuman-Base-1080p` are distilled models with minimal NFEs; CFG-Parallel is not necessary.
 > 3. Cosmos3 T2I uses `Cosmos3OmniDiffusersPipeline` with `modalities=["image"]`. Model-level CPU offload swaps the nested UND reasoner and GEN generator pathways; layerwise offload remains available for blockwise GEN/UND offload.
 > 4. Krea 2 currently supports single-GPU inference plus LoRA, Cache-DiT, HSDP, CPU/layerwise offload, and VAE-patch-parallel (decode). TP/SP/CFG-Parallel are not yet wired. The few-step distilled (Turbo) checkpoint uses `is_distilled=true` (fixed timestep shift `mu=1.15`); generate at 2048x2048 by default with `num_inference_steps≈8` and `guidance_scale=0`. The Raw checkpoint uses 1024x1024, `num_inference_steps=28`, and `guidance_scale=4.5`.
+> 5. HunyuanImage3 supports step execution. Multi-request step execution requires `TORCH_SDPA`; see [Diffusion Execution Modes](diffusion/execution_modes.md#step-execution).
 
 ### VideoGen
 
@@ -155,13 +158,16 @@ The following tables show which models support each feature:
 | **Wan2.1-VACE**              |     ❌     |     ✅      |           ✅           |       ✅        |         ✅         |         ❌         |   ✅    |             ✅             |      ✅ (decode)      |       ❌        |        ❌         |
 | **LTX-2**                    |     ❌     |     ✅      |           ✅           |       ✅        |         ✅         |         ❌         |   ✅    |             ✅             |      ✅ (decode)      |       ❌        |        ❌         |
 | **LTX-2.3**                  |     ❌     |     ✅      |           ✅           |       ✅        |         ✅         |         ❌         |   ✅    |             ✅             |      ✅ (decode)      |       ❌        |        ❌         |
-| **Helios**                   |     ❌     |     ✅      |           ✅           |       ✅        |         ✅         |         ❌         |   ✅    |             ✅             |          ❌           |       ❌        |        ❌         |
+| **Helios**                   |     ❌     |     ✅      |           ✅           |       ✅        |         ✅         |         ❌         |   ✅    |             ✅             |          ❌           |       ❌        |        ✅*        |
 | **HunyuanVideo-1.5 T2V I2V** |     ❌     |     ✅      |           ✅           |       ✅        |         ✅         |         ❌         |   ✅    |             ✅             |  ✅ (encode/decode)   |       ✅        |        ❌         |
 | **DreamID-Omni**             |     ❌     |     ❌      |           ❌           |       ✅        |         ❌         |         ❌         |   ✅    |             ✅             |          ❌           |       ❌        |        ❌         |
 | **Cosmos3**                  |     ❌     |     ✅      |           ✅           |       ✅        |         ✅         |         ❌         |   ✅    |             ✅             |  ✅ (encode/decode)   |       ✅        |        ❌         |
 | **LongCat-Video-Avatar-1.5** |     ❌     |     ❌      |           ❌           |       ❌        |         ❌         |         ❌         |   ❌    |             ❌             |          ❌           |       ❌        |        ❌         |
 | **MiniMax-H3**               | ✅ (FL2VA) |     ✅      |           ✅           |       ❌        |       ✅ (DiT/TE)  |         ❌         |   ✅    |             ✅             |       ✅ (tile)       |      ✅ (DiT)      |        ❌         |
 
+> **Step execution note:** Helios supports single-request step execution only;
+> use `max_num_seqs=1`.
+
 **Frame Interpolation Support**
 
 - **Supported**: Wan2.2 text-to-video, image-to-video, and TI2V pipelines
@@ -190,7 +196,7 @@ The following tables show which models support each feature:
 | **🔀Ring-Attn** | ✅ | ✅ | ✅ | | | | | | | | | | | |
 | **🔀CFG-Parallel** | ✅ | ✅ | ✅ | ✅ | | | | | | | | | | |
 | **🔀Tensor Parallel** | ✅ | ✅ | ✅ | ✅ | ✅ | | | | | | | | | |
-| **🔀HSDP** | ❓ | ❓ | ❓ | ❓ | ❓ | ❌ | | | | | | | | |
+| **🔀HSDP** | ❓ | ❓ | ✅ | ❓ | ✅ | ❌ | | | | | | | | |
 | **🔀Expert Parallel** | ❓ | ❓ | ❓ | ❓ | ❓ | ❓ | ❓ | | | | | | | |
 | **💾CPU Offloading (Layerwise)** | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | | | | | | |
 | **💾CPU Offloading (Module-wise)** | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❓ | ❓ | ❌ | | | | | |
@@ -201,97 +207,43 @@ The following tables show which models support each feature:
 
 !!! info
 
-    1. Tensor Parallel and HSDP are not compatible.
+    1. HSDP can be combined with Ulysses-SP or CFG-Parallel. Tensor Parallel
+       and HSDP are not compatible; other HSDP combinations in the table
+       remain unverified.
     2. TeaCache and Cache-DiT are not compatible.
     3. CPU Offloading (Layerwise) and CPU Offloading (Module-wise) are not compatible.
-    4. CPU Offloading (Layerwise) supports single-card for now.
-    5. Using FP8-Quant as an example of qunatization methods.
+    4. The CPU Offloading (Layerwise) row describes local layerwise offload.
+       Multi-device Distributed Layerwise Offload has a separate topology and
+       compatibility matrix in the [CPU Offloading guide](diffusion/cpu_offload.md#distributed-layerwise-offloading).
+    5. The compatibility matrix uses FP8 as the representative quantization method.
     6. Step Execution is not compatible with any diffusion cache backend. LoRA is supported, but each scheduled batch must use a single adapter (requests with different `lora_request` or `lora_scale` are kept in separate batches).
 
 
 ## Multi-Thread Weight Loading
 
-Large diffusion models can take several minutes to load weights at startup (e.g., ~3 min for Qwen-Image, ~5 min for Wan2.2 I2V 14B). Multi-thread weight loading speeds up this process by loading safetensors shards in parallel using a thread pool instead of sequentially.
-
-This optimization is **enabled by default** with 4 threads. No configuration is needed for the default behavior.
-
-### Configuration
-
-| Parameter | CLI Flag | Default | Description |
-|-----------|----------|---------|-------------|
-| `enable_multithread_weight_load` | `--disable-multithread-weight-load` | `True` (enabled) | Pass the flag to disable multi-thread loading |
-| `num_weight_load_threads` | `--num-weight-load-threads` | `4` | Number of threads for parallel weight loading |
-
-!!! tip
-    The default of 4 threads balances speed and disk I/O contention. On fast NVMe storage you may benefit from more threads (e.g., 8). On HDD or network storage, the default of 4 avoids saturating I/O bandwidth.
-
-### Online Serving
-
-```bash
-# Default (multi-thread enabled, 4 threads)
-vllm serve Qwen/Qwen-Image --omni --port 8091
-
-# Custom thread count
-vllm serve Wan-AI/Wan2.2-I2V-A14B-Diffusers --omni --num-weight-load-threads 8
-
-# Disable multi-thread loading
-vllm serve Qwen/Qwen-Image --omni --disable-multithread-weight-load
-```
-
-### Offline Inference
-
-```python
-from vllm_omni import Omni
-
-# Default (multi-thread enabled, 4 threads)
-omni = Omni(model="Qwen/Qwen-Image")
-
-# Custom thread count
-omni = Omni(
-    model="Wan-AI/Wan2.2-I2V-A14B-Diffusers",
-    num_weight_load_threads=8,
-)
-```
-
-### Benchmarks
-
-Measured on NVIDIA H800:
-
-| Model | Before | After | Speedup |
-|-------|--------|-------|---------|
-| **Qwen/Qwen-Image** (53.7 GiB) | 168s | 27s | **6.2x** |
-| **Wan-AI/Wan2.2-I2V-A14B-Diffusers** (64.5 GiB) | 283s | 56s | **5.1x** |
+The loading guide now lives at [Diffusion Startup and
+Loading](diffusion/startup_and_loading.md). This heading remains so existing
+links to this section continue to work.
 
 ## Learn More
 
-**Cache Acceleration:**
+The Diffusion Acceleration navigation groups the remaining guides as follows:
 
-- **[TeaCache Configuration Guide](diffusion/cache_acceleration/teacache.md)** - Parameter tuning, performance tips, troubleshooting
-- **[Cache-DiT Advanced Guide](diffusion/cache_acceleration/cache_dit.md)** - DBCache, TaylorSeer, SCM techniques and optimization
+| Area | Guide |
+| --- | --- |
+| Compatibility | [Feature Compatibility](feature_compatibility.md) |
+| CPU offloading | [CPU Offloading](diffusion/cpu_offload.md) |
+| Cache acceleration | [TeaCache](diffusion/cache_acceleration/teacache.md), [Cache-DiT](diffusion/cache_acceleration/cache_dit.md) |
+| Parallelism | [Parallelism Overview](diffusion/parallelism/overview.md) |
+| Attention | [Attention Backends](diffusion/attention_backends.md) |
+| Compilation | [Regional Compilation](diffusion/regional_compilation.md) |
+| Video extension | [Frame Interpolation](diffusion/frame_interpolation.md) |
+| Startup | [Startup and Loading](diffusion/startup_and_loading.md) |
+| Adapters | [LoRA](diffusion/lora.md) |
 
-**Parallelism Methods:**
+Related cross-model and runtime features are documented separately:
 
-- **[Parallelism Overview](diffusion/parallelism/overview.md)** - Tensor Parallelism, Sequence Parallelism, CFG Parallelism, Pipeline Parallelism, HSDP, and Expert Parallelism
-
-**Memory Optimization:**
-
-- **[CPU Offload Guide](diffusion/cpu_offload.md)** - Offload model components to CPU, reduce GPU memory usage
-- **[VAE Parallelism Guide](diffusion/parallelism/vae_parallelism.md)** - Distribute VAE decode work across GPUs for high-resolution images and videos
-- **[Quantization](quantization/overview.md)** - Quantization methods for diffusion, multi-stage omni/TTS, and multi-stage diffusion models
-
-**Extensions:**
-
-- **[LoRA Inference Guide](diffusion/lora.md)** - Low-Rank Adaptation for style customization and fine-tuning
-- **[Frame Interpolation Guide](diffusion/frame_interpolation.md)** - Worker-side post-generation video frame interpolation for smoother motion
-
-**Execution Modes:**
-
-- **[Diffusion Execution Modes](diffusion/execution_modes.md)** - Configure request batching, step execution, continuous batching, and streaming output
-
-**Startup Optimization:**
-
-- **[Multi-Thread Weight Loading](#multi-thread-weight-loading)** - Speed up model startup by loading safetensors shards in parallel
-
-**Advanced Topics:**
-
-- **[Feature Compatibility](feature_compatibility.md)** - How to combine multiple features for maximum performance
+- [Quantization](quantization/overview.md) covers diffusion-only models,
+  multi-stage omni/TTS models, and multi-stage diffusion models.
+- [Execution Modes and Streaming](diffusion/execution_modes.md) covers the
+  diffusion runtime, including batching, step execution, and streaming output.

docs/user_guide/examples/offline_inference/text_to_video.md

@@ -12,6 +12,7 @@ For backend selection and SageAttention usage, see the [Diffusion Attention Back
 | Model | Default Resolution | Default Frames | Default Steps | Guidance | VRAM (BF16) |
 |---|---|---|---|---|---|
 | `Wan-AI/Wan2.2-T2V-A14B-Diffusers` | 720x1280 | 81 | 40 | 4.0 | ~60 GiB |
+| `robbyant/lingbot-video-dense-1.3b` / `robbyant/lingbot-video-moe-30b-a3b` | 192x320 | 9 | 2 | 3.0 | ~68 GiB (MoE smoke) |
 | `Lightricks/LTX-2` | 512x768 | 121 | 40 | video 3.0 / audio 7.0 | Model-dependent |
 | `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-480p_t2v` | 480x832 | 121 | 50 | 6.0 | 1×A100 80GB |
 | `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-720p_t2v` | 720x1280 | 121 | 50 | 6.0 | FP8 + VAE tiling required |
@@ -35,6 +36,25 @@ python text_to_video.py \
   --output t2v_out.mp4
 ```
 
+### LingBot-Video
+
+The shared runner recognizes both official LingBot checkpoint IDs and selects
+the native `LingBotVideoPipeline` automatically:
+
+```bash
+python text_to_video.py \
+  --model robbyant/lingbot-video-dense-1.3b \
+  --prompt "a robotic arm picks up a red block" \
+  --height 192 --width 320 --num-frames 9 --num-inference-steps 2 \
+  --guidance-scale 3.0 --flow-shift 3.0 --fps 24 \
+  --output lingbot_t2v.mp4
+```
+
+Use `robbyant/lingbot-video-moe-30b-a3b` for the MoE checkpoint. For local
+paths without `lingbot` in the directory name, add
+`--model-class-name LingBotVideoPipeline`. Model-only options use
+`--extra-body`, for example `'{"batch_cfg": true, "output_type": "np"}'`.
+
 ### LTX-2
 
 ```bash

docs/user_guide/quantization/overview.md

@@ -12,20 +12,20 @@ For the internal architecture and backend extension points, see the
 
 | Mode | Guide | Description | Methods |
 |------|-------|-------------|---------|
-| Online quantization | [Online Quantization](online.md) | vLLM-Omni computes quantized weights and scales while loading the model. | FP8 W8A8, Int8 W8A8, MXFP8 W8A8, MXFP4 W4A4 |
+| Online quantization | [Online Quantization](online.md) | vLLM-Omni computes quantized weights and scales while loading the model. | FP8 W8A8, Int8 W8A8, BitsAndBytes W4, MXFP8 W8A8, MXFP4 W4A4 |
 | Runtime attention quantization | [Quantized KV Cache](quantized_kvcache.md) | vLLM-Omni dynamically quantizes eligible diffusion Flash Attention tensors during inference. | FP8 FA |
-| Pre-quantized checkpoints | Method-specific guides | The checkpoint or an offline quantizer provides quantized weights and scales before serving. | ModelOpt, GGUF, AutoRound, msModelSlim, serialized Int8, offline MXFP8, offline MXFP4 DualScale |
+| Pre-quantized checkpoints | Method-specific guides | The checkpoint or an offline quantizer provides quantized weights and scales before serving. | ModelOpt, AutoRound, msModelSlim, serialized Int8, offline MXFP8, offline MXFP4 DualScale |
 
 ## Hardware Support
 
-| Device | FP8 W8A8 | Int8 W8A8 | ModelOpt | MXFP8 W8A8 | MXFP4 W4A4 | GGUF | AutoRound | msModelSlim |
-|--------|----------|-----------|----------|------------|------------|------|-----------|-------------|
-| NVIDIA Blackwell GPU (SM 100+) | ✅ | ✅ | ✅ | ⭕ | ⭕ | ✅ | ✅ | ❌ |
-| NVIDIA Ada/Hopper GPU (SM 89+) | ✅ | ✅ | ✅ | ⭕ | ⭕ | ✅ | ✅ | ❌ |
-| NVIDIA Ampere GPU (SM 80+) | ✅ | ✅ | ⭕ | ⭕ | ⭕ | ✅ | ✅ | ❌ |
-| AMD ROCm | ⭕ | ⭕ | ⭕ | ⭕ | ✅ | ⭕ | ⭕ | ❌ |
-| Intel XPU | ⭕ | ⭕ | ⭕ | ⭕ | ⭕ | ⭕ | ✅ | ❌ |
-| Ascend NPU | ❌ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ |
+| Device | FP8 W8A8 | Int8 W8A8 | BitsAndBytes W4 | ModelOpt | MXFP8 W8A8 | MXFP4 W4A4 | AutoRound | msModelSlim |
+|--------|----------|-----------|------------------|----------|------------|------------|-----------|-------------|
+| NVIDIA Blackwell GPU (SM 100+) | ✅ | ✅ | ✅ | ✅ | ⭕ | ⭕ | ✅ | ❌ |
+| NVIDIA Ada/Hopper GPU (SM 89+) | ✅ | ✅ | ✅ | ✅ | ⭕ | ⭕ | ✅ | ❌ |
+| NVIDIA Ampere GPU (SM 80+) | ✅ | ✅ | ✅ | ⭕ | ⭕ | ⭕ | ✅ | ❌ |
+| AMD ROCm | ⭕ | ⭕ | ❌ | ⭕ | ⭕ | ✅ | ⭕ | ❌ |
+| Intel XPU | ⭕ | ⭕ | ❌ | ⭕ | ⭕ | ⭕ | ✅ | ❌ |
+| Ascend NPU | ❌ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ✅ |
 
 Legend: `✅` supported, `❌` unsupported, `⭕` not verified in this
 guide. FP8 on Ampere may use a weight-only path where available.
@@ -43,10 +43,10 @@ otherwise.
 |--------|-------|------|----------------|--------|
 | FP8 W8A8 | [FP8](fp8.md) | Online W8A8 or checkpoint FP8 | Qwen-Image; Wan2.2 is not validated | Validated for Qwen-Image family and other DiT models |
 | Int8 W8A8 | [Int8](int8.md) | Online or serialized W8A8 | Qwen-Image; Wan2.2 is not validated | Validated for Qwen-Image and Z-Image |
+| BitsAndBytes W4 | [BitsAndBytes](bitsandbytes.md) | Online W4 weight-only | Z-Image-Turbo; Qwen-Image/Wan2.2 not validated | Validated for Z-Image-Turbo |
 | ModelOpt | [ModelOpt](modelopt.md) | Pre-quantized FP8 checkpoints | Qwen-Image, Z-Image, FLUX.2, HunyuanImage-3.0 | Validated for ModelOpt FP8 diffusion checkpoints |
 | MXFP8 W8A8 | [MXFP8](mxfp8.md) | Online W8A8 or offline pre-quantized | Wan2.2-T2V-A14B, I2V-A14B, TI2V-5B | Ascend NPU only; validated for Wan2.2 |
 | MXFP4 W4A4 | [MXFP4](mxfp4.md) | `mxfp4`: online single-scale only; `mxfp4_dualscale`: online or offline dual-scale (offline recommended) | Wan2.2-T2V-A14B, I2V-A14B | Ascend NPU only; validated for Wan2.2 A14B cascade models; TI2V-5B not supported; offline `mxfp4_dualscale` uses calibrated `mul_scale` for best accuracy |
-| GGUF | [GGUF](gguf.md) | Pre-quantized transformer weights | Qwen-Image | Validated where a model-specific GGUF adapter exists |
 | AutoRound | [AutoRound](autoround.md) | Pre-quantized W4A16 checkpoints | FLUX.1-dev; Qwen-Image/Wan2.2 not validated | Checkpoint-driven |
 | msModelSlim | [msModelSlim](msmodelslim.md) | Pre-quantized Ascend checkpoints | Wan2.2 recipe; HunyuanImage-3.0 inference target | Ascend/NPU path |
 
@@ -61,9 +61,9 @@ in BF16 unless the model guide explicitly adds support.
 |--------|-------|-------|----------------|--------|
 | ModelOpt | [ModelOpt](modelopt.md) | Thinker or language-model checkpoint config | Qwen3-Omni thinker | ModelOpt checkpoint path |
 | Int8 | [Int8](int8.md) | Not currently validated for omni/TTS stages | Qwen3-Omni, Qwen3-TTS | Not validated |
+| BitsAndBytes | [BitsAndBytes](bitsandbytes.md) | Not currently validated for omni/TTS stages | Qwen3-Omni, Qwen3-TTS | Not validated |
 | MXFP8 | [MXFP8](mxfp8.md) | Not currently validated for omni/TTS stages | Qwen3-Omni, Qwen3-TTS | Not validated |
 | MXFP4 | [MXFP4](mxfp4.md) | Not currently validated for omni/TTS stages | Qwen3-Omni, Qwen3-TTS | Not validated |
-| GGUF | [GGUF](gguf.md) | Not currently validated for omni/TTS stages | Qwen3-Omni, Qwen3-TTS | Not validated |
 | AutoRound | [AutoRound](autoround.md) | Thinker or language-model checkpoint config | Qwen2.5-Omni, Qwen3-Omni | Supported through AutoRound checkpoints |
 | msModelSlim | [msModelSlim](msmodelslim.md) | Not currently validated for omni/TTS stages | Qwen3-Omni, Qwen3-TTS | Not validated |
 
@@ -76,10 +76,10 @@ attached to the intended stage rather than applied globally.
 |--------|-------|-------|----------------|--------|
 | FP8 | [FP8](fp8.md) | Stage-specific DiT or transformer module | BAGEL, GLM-Image | Requires model-specific validation |
 | Int8 | [Int8](int8.md) | Stage-specific DiT or transformer module | BAGEL, GLM-Image | Requires model-specific validation |
+| BitsAndBytes | [BitsAndBytes](bitsandbytes.md) | Stage-specific transformer or DiT module | BAGEL, GLM-Image | Not validated |
 | ModelOpt | [ModelOpt](modelopt.md) | Checkpoint-defined diffusion stage | BAGEL, GLM-Image | Requires model-specific validation |
 | MXFP8 | [MXFP8](mxfp8.md) | Stage-specific DiT or transformer module | BAGEL, GLM-Image | Not validated |
 | MXFP4 | [MXFP4](mxfp4.md) | Stage-specific DiT or transformer module | BAGEL, GLM-Image | Not validated |
-| GGUF | [GGUF](gguf.md) | Stage-specific transformer weights | BAGEL, GLM-Image | No validated adapter listed |
 | AutoRound | [AutoRound](autoround.md) | Checkpoint-defined stage | BAGEL, GLM-Image | No validated checkpoint listed |
 | msModelSlim | [msModelSlim](msmodelslim.md) | Ascend-generated stage weights | GLM-Image | Requires model-specific adaptation |
 
@@ -106,7 +106,7 @@ config = build_quant_config({
 
 | Component | Default quantized? | Notes |
 |-----------|--------------------|-------|
-| Diffusion transformer | Yes | Primary target for FP8, Int8, ModelOpt, MXFP8, MXFP4, GGUF, AutoRound, and msModelSlim |
+| Diffusion transformer | Yes | Primary target for FP8, Int8, BitsAndBytes, ModelOpt, MXFP8, MXFP4, AutoRound, and msModelSlim |
 | Text encoder | No | Keep BF16 unless a method-specific guide documents support |
 | VAE | No | Keep BF16; storage-only paths are method-specific |
 | Scheduler/tokenizer | No | Loaded from the base model repository |
@@ -139,8 +139,8 @@ from vllm_omni.quantization import build_quant_config
 
 build_quant_config("fp8")
 build_quant_config({"method": "fp8", "activation_scheme": "static"})
+build_quant_config("bitsandbytes")
 build_quant_config("auto-round", bits=4, group_size=128)
-build_quant_config({"method": "gguf", "gguf_model": "/path/to/model.gguf"})
 build_quant_config({"transformer": {"method": "fp8"}, "vae": None})
 build_quant_config(None)
 ```

examples/offline_inference/aura_omni/end2end.py

@@ -131,7 +131,7 @@ def build_sampling_params() -> list[SamplingParams]:
 
 
 def save_stage_output(stage_outputs, output_dir: str) -> None:
-    output = stage_outputs.request_output
+    output = stage_outputs
     request_id = output.request_id
     if stage_outputs.final_output_type == "text":
         text_output = output.outputs[0].text

examples/offline_inference/covo_audio/end2end.py

@@ -104,7 +104,7 @@ def main(args):
     os.makedirs(output_dir, exist_ok=True)
 
     for stage_outputs in omni_outputs:
-        output = stage_outputs.request_output
+        output = stage_outputs
         if stage_outputs.final_output_type == "text":
             request_id = output.request_id
             text_output = output.outputs[0].text

examples/offline_inference/custom_pipeline/image_to_image/image_edit.py

@@ -228,7 +228,7 @@ async def main():
     if not outputs:
         raise ValueError("No output produced from omni.generate()")
 
-    first_out = outputs[0].request_output
+    first_out = outputs[0]
     req_out: OmniRequestOutput = first_out
 
     # Verify trajectory data (from custom pipeline)

examples/offline_inference/dynin_omni/end2end.py

@@ -741,7 +741,7 @@ def make_mmu_prompt(
 
 def iter_mm_outputs(outputs: list[Any]):
     for omni_out in outputs:
-        req_out = getattr(omni_out, "request_output", None)
+        req_out = omni_out
         req_list = req_out if isinstance(req_out, list) else [req_out]
         for item in req_list:
             if item is None:

examples/offline_inference/hunyuan_image3/end2end.py

@@ -303,7 +303,7 @@ def main():
     )
     img_idx = 0
     for req_output in omni_outputs:
-        ro = getattr(req_output, "request_output", None)
+        ro = req_output
         stage_id = getattr(req_output, "stage_id", None)
 
         # AR stage text — each CompletionOutput.text is already a delta when

examples/offline_inference/image_to_image/image_edit.py

@@ -621,7 +621,7 @@ def main():
         images = getattr(output, "images", None)
         if images:
             break
-        req_out = getattr(output, "request_output", None)
+        req_out = output
         images = getattr(req_out, "images", None) if req_out is not None else None
         if images:
             break

examples/offline_inference/image_to_video/image_to_video.py

@@ -63,7 +63,9 @@ from vllm_omni.inputs.data import OmniDiffusionSamplingParams
 from vllm_omni.lora.request import LoRARequest
 from vllm_omni.lora.utils import stable_lora_int_id
 from vllm_omni.model_extras import (
-    build_image_to_video_prompt,
+    build_image_to_video_prompt as build_model_image_to_video_prompt,
+)
+from vllm_omni.model_extras import (
     get_extra_body_params,
     get_model_class_name,
 )
@@ -85,6 +87,22 @@ def parse_json_object(value: str, flag_name: str = "argument") -> dict[str, Any]
 parse_profiler_config = functools.partial(parse_json_object, flag_name="--profiler-config")
 
 
+def build_image_to_video_prompt(
+    prompt: str,
+    negative_prompt: str | None,
+    media_inputs: dict[str, Any],
+) -> dict[str, Any]:
+    """Build the canonical request envelope for the shared I2V example."""
+    result: dict[str, Any] = {
+        "prompt": prompt,
+        "modalities": ["video"],
+        "multi_modal_data": media_inputs,
+    }
+    if negative_prompt is not None:
+        result["negative_prompt"] = negative_prompt
+    return result
+
+
 def parse_args() -> argparse.Namespace:
     parser = argparse.ArgumentParser(
         description=(
@@ -566,10 +584,13 @@ def main():
         # Preserve the historical empty-prompt behavior for non-LTX examples.
         negative_prompt = ""
     prompt_dict = build_image_to_video_prompt(
-        model_class_name=model_class_name,
         prompt=args.prompt,
         negative_prompt=negative_prompt,
         media_inputs=media_inputs,
+    )
+    prompt_dict = build_model_image_to_video_prompt(
+        model_class_name=model_class_name,
+        prompt=prompt_dict,
         height=height,
         width=width,
         num_frames=num_frames,
@@ -625,8 +646,8 @@ def main():
         if frames.multimodal_output and "audio" in frames.multimodal_output:
             audio = frames.multimodal_output["audio"]
             audio_sample_rate = frames.multimodal_output.get("audio_sample_rate", audio_sample_rate)
-        if frames.is_pipeline_output and frames.request_output is not None:
-            inner_output = frames.request_output
+        if frames.is_pipeline_output and frames is not None:
+            inner_output = frames
             if isinstance(inner_output, OmniRequestOutput):
                 if inner_output.multimodal_output and "audio" in inner_output.multimodal_output:
                     audio = inner_output.multimodal_output["audio"]

examples/offline_inference/mammothmodal2_preview/run_mammothmoda2_image_summarize.py

@@ -0,0 +1,117 @@
+"""
+Offline inference example: MammothModa2 image summarization (single AR stage).
+
+Example:
+  uv run python examples/offline_inference/mammothmodal2_preview/run_mammothmoda2_image_summarize.py \
+    --model path/to/MammothModa2-Preview \
+    --deploy-config vllm_omni/deploy/mammoth_moda2_ar.yaml \
+    --image /path/to/input.jpg \
+    --question "Please summarize the content of this image."
+"""
+
+from __future__ import annotations
+
+import argparse
+import os
+
+from PIL import Image
+from vllm import SamplingParams
+from vllm.multimodal.image import convert_image_mode
+
+from vllm_omni import Omni
+
+DEFAULT_SYSTEM = "You are a helpful assistant."
+DEFAULT_QUESTION = "Please summarize the content of this image."
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description="MammothModa2 image summarization (offline, AR only).")
+    parser.add_argument("--model", type=str, required=True, help="Path to model directory or model id.")
+    parser.add_argument("--deploy-config", type=str, required=True, help="Path to deploy config yaml.")
+    parser.add_argument("--image", type=str, required=True, help="Path to input image.")
+    parser.add_argument("--question", type=str, default=DEFAULT_QUESTION, help="Question/instruction for the model.")
+    parser.add_argument("--system", type=str, default=DEFAULT_SYSTEM, help="System prompt.")
+    parser.add_argument(
+        "--max-tokens",
+        type=int,
+        default=512,
+        help="Max new tokens to generate.",
+    )
+    parser.add_argument("--temperature", type=float, default=0.2)
+    parser.add_argument("--top-p", type=float, default=0.9)
+    parser.add_argument("--seed", type=int, default=42)
+    parser.add_argument("--trust-remote-code", action="store_true")
+    parser.add_argument(
+        "--enable-diffusion-pipeline-profiler",
+        action="store_true",
+        help="Enable diffusion pipeline profiler to display stage durations.",
+    )
+    return parser.parse_args()
+
+
+def build_prompt(system: str, question: str) -> str:
+    return (
+        f"<|im_start|>system\n{system}<|im_end|>\n"
+        "<|im_start|>user\n"
+        "<|vision_start|><|image_pad|><|vision_end|>"
+        f"{question}<|im_end|>\n"
+        "<|im_start|>assistant\n"
+    )
+
+
+def main() -> None:
+    args = parse_args()
+
+    if not os.path.exists(args.image):
+        raise FileNotFoundError(f"Image file not found: {args.image}")
+
+    pil_image = Image.open(args.image)
+    image_data = convert_image_mode(pil_image, "RGB")
+    prompt = build_prompt(args.system, args.question)
+
+    omni = Omni(
+        model=args.model,
+        deploy_config=args.deploy_config,
+        trust_remote_code=args.trust_remote_code,
+        enable_diffusion_pipeline_profiler=args.enable_diffusion_pipeline_profiler,
+    )
+    try:
+        sp = SamplingParams(
+            temperature=float(args.temperature),
+            top_p=float(args.top_p),
+            top_k=-1,
+            max_tokens=int(args.max_tokens),
+            seed=int(args.seed),
+            detokenize=True,
+        )
+        # NOTE: omni.generate() returns a Generator[OmniRequestOutput, None, None].
+        # Consume it inside the try block so the worker isn't closed early.
+        outputs = list(
+            omni.generate(
+                [
+                    {
+                        "prompt": prompt,
+                        "multi_modal_data": {"image": image_data},
+                        "additional_information": {"omni_task": ["chat"]},
+                    }
+                ],
+                [sp],
+            )
+        )
+    finally:
+        omni.close()
+
+    lines: list[str] = []
+    for stage_outputs in outputs:
+        ro = stage_outputs
+        text = ro.outputs[0].text if getattr(ro, "outputs", None) else str(ro)
+        lines.append(f"request_id: {getattr(ro, 'request_id', 'unknown')}\n")
+        lines.append("answer:\n")
+        lines.append(text.strip() + "\n")
+        lines.append("\n")
+
+    print("\n".join(lines))
+
+
+if __name__ == "__main__":
+    main()

examples/offline_inference/mimo_audio/end2end.py

@@ -297,7 +297,7 @@ def main(args):
     os.makedirs(output_dir, exist_ok=True)
 
     for stage_outputs in omni_outputs:
-        output = stage_outputs.request_output
+        output = stage_outputs
         if stage_outputs.final_output_type == "text":
             request_id = output.request_id
             text_output = output.outputs[0].text

examples/offline_inference/ming_flash_omni/end2end.py

@@ -347,7 +347,7 @@ def main(args):
         omni.start_profile(stages=args.profiler_stages)
 
     for stage_outputs in omni.generate(prompts, sampling_params_list):
-        output = stage_outputs.request_output
+        output = stage_outputs
         if stage_outputs.final_output_type == "text":
             request_id = output.request_id
             text_output = output.outputs[0].text

examples/offline_inference/nemotron_voicechat/end2end.py

@@ -0,0 +1,229 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Offline speech-to-speech with NVIDIA-NemotronLabs-VoiceChat-11B.
+
+Feeds a 16 kHz user utterance (plus a spoken-style system prompt) through the
+3-stage nemotron_voicechat pipeline and writes the agent's reply as text and a
+22.05 kHz WAV.
+
+Usage:
+    python end2end.py \
+        --checkpoint /path/to/NVIDIA-NemotronLabs-VoiceChat-11B \
+        --wav /path/to/user_question_16k.wav \
+        --output-dir results/nemotron_voicechat
+
+The system prompt defaults to the NeMo reference default. The text tokenizer
+resolves from the ``nvidia/NVIDIA-Nemotron-Nano-9B-v2`` HF id (or the
+``NEMOTRON_VOICECHAT_LLM_PATH`` env var for air-gapped runs).
+
+Sizing the input WAV: the timeline is frame-locked, so the reply budget IS the
+input duration — the acoustic channel trails the text channel, and a reply
+that has not finished when the frames run out is truncated SILENTLY (no error;
+ASR of the WAV will just disagree with the emitted text). Leave generous
+trailing silence after the question (a question ending at ~4.5 s truncated in
+an 8 s WAV but completed cleanly in a 16 s one). The script warns when the
+text channel is still emitting non-PAD tokens near the last frame.
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+from pathlib import Path
+
+import numpy as np
+import soundfile as sf
+
+# Warn if the text channel is still speaking within this many final frames
+# (12 frames = ~1 s) — the frame-locked reply likely got cut off.
+TRUNCATION_GUARD_FRAMES = 12
+
+# The NeMo reference offline script's default system prompt.
+DEFAULT_SYSTEM_PROMPT = (
+    "You are an AI voice assistant developed by NVIDIA. "
+    "Your name is NVIDIA Voice Chat. "
+    "Answer in a spoken, conversational style rather than a written one. "
+    "Do not repeat the same sentence over and over again. "
+    "Start the conversation by greeting the user."
+)
+
+
+def load_wav_16k_mono(path: str) -> np.ndarray:
+    wav, sr = sf.read(path, dtype="float32")
+    if wav.ndim > 1:
+        wav = wav.mean(axis=1)
+    if sr != 16000:
+        from vllm.multimodal.audio import resample_audio_scipy
+
+        wav = resample_audio_scipy(wav, orig_sr=sr, target_sr=16000)
+    return wav.astype(np.float32)
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--checkpoint", required=True, help="NVIDIA-NemotronLabs-VoiceChat-11B directory")
+    parser.add_argument("--wav", required=True, help="user speech (any sr; resampled to 16 kHz mono)")
+    parser.add_argument("--output-dir", default="results/nemotron_voicechat")
+    parser.add_argument("--system-prompt", default=DEFAULT_SYSTEM_PROMPT)
+    parser.add_argument(
+        "--stage-configs-path",
+        default=None,
+        help=(
+            "Deploy yaml override. Pass vllm_omni/deploy/nemotron_labs_voicechat_streaming.yaml "
+            "for the async-chunk streaming pipeline (stages overlap; audio starts decoding "
+            "before the thinker finishes). Default: the bundled non-streaming yaml (bit-exact "
+            "parity path)."
+        ),
+    )
+    args = parser.parse_args()
+
+    if not Path(args.wav).is_file():
+        raise FileNotFoundError(f"Input WAV not found: {args.wav}")
+    ckpt = Path(args.checkpoint)
+    raw_cfg = json.loads((ckpt / "config.json").read_text())
+    stt_cfg = raw_cfg["model"]["stt"]["model"]
+
+    from transformers import AutoTokenizer
+    from vllm.sampling_params import SamplingParams
+
+    from vllm_omni.entrypoints.omni import Omni
+    from vllm_omni.model_executor.models.nemotron_voicechat.nemotron_voicechat_thinker import (
+        compute_acoustic_frame_count,
+    )
+
+    tok_ref = os.environ.get("NEMOTRON_VOICECHAT_LLM_PATH") or stt_cfg.get(
+        "pretrained_llm", "nvidia/NVIDIA-Nemotron-Nano-9B-v2"
+    )
+    tokenizer = AutoTokenizer.from_pretrained(tok_ref, trust_remote_code=False)
+    bos_id = tokenizer.convert_tokens_to_ids(stt_cfg.get("bos_token", "<s>"))
+    eos_id = tokenizer.convert_tokens_to_ids(stt_cfg.get("eos_token", "</s>"))
+    pad_id = tokenizer.convert_tokens_to_ids(stt_cfg.get("pad_token", "<SPECIAL_12>"))
+    # NeMo encode_system_prompt: [bos] + tokens + [eos].
+    prompt_ids = [bos_id] + tokenizer.encode(args.system_prompt, add_special_tokens=False) + [eos_id]
+
+    wav = load_wav_16k_mono(args.wav)
+    n_frames = compute_acoustic_frame_count(stt_cfg, int(wav.shape[0]))
+    print(f"input: {wav.shape[0]} samples @16kHz -> {n_frames} acoustic frames; prompt {len(prompt_ids)} tokens")
+
+    inputs = {
+        # vLLM prompt = system prompt + one placeholder position (acoustic
+        # frame 0); the thinker replaces every position with fused embeddings.
+        "prompt_token_ids": prompt_ids + [pad_id],
+        "additional_information": {
+            # Plain float list: raw ndarrays do not survive the engine-core
+            # message serialization.
+            "nvc_audio": wav.astype(np.float32).tolist(),
+            "nvc_sr": 16000,
+            "nvc_prompt_token_ids": prompt_ids,
+            "nvc_expected_frames": n_frames,
+        },
+    }
+    sampling_params_list = [
+        # thinker: greedy, frame-locked generation length. NO min_tokens: the
+        # tokenizer's eos_token is <SPECIAL_12> — the frame-locked PAD/silence
+        # token — and vLLM's min_tokens masks "EOS" to -inf, which would forbid
+        # silence and force the agent to babble through the whole utterance.
+        SamplingParams(temperature=0.0, max_tokens=n_frames, ignore_eos=True, seed=0),
+        # talker: placeholder loop, stops on the stage's stop token. 16383:
+        # the 1-token placeholder prompt + max_tokens must fit the stage's
+        # max_model_len (16384).
+        SamplingParams(temperature=0.0, max_tokens=16383, stop_token_ids=[1], detokenize=False, seed=0),
+        # code2wav: single decode step.
+        SamplingParams(temperature=0.0, max_tokens=1, detokenize=False, seed=0),
+    ]
+
+    omni_kwargs: dict = {}
+    if args.stage_configs_path:
+        omni_kwargs["stage_configs_path"] = args.stage_configs_path
+    omni = Omni(model=str(ckpt), trust_remote_code=True, **omni_kwargs)
+    outputs = omni.generate(inputs, sampling_params_list)
+
+    out_dir = Path(args.output_dir)
+    out_dir.mkdir(parents=True, exist_ok=True)
+    stem = Path(args.wav).stem
+
+    agent_text = None
+    audio = None
+    sr = 22050
+    for stage_outputs in outputs:
+        # One entry per final stage (text from the thinker, audio from code2wav);
+        # mirrors the qwen2_5_omni offline example's collection loop.
+        output = getattr(stage_outputs, "request_output", stage_outputs)
+        final_type = getattr(stage_outputs, "final_output_type", None)
+        completions = getattr(output, "outputs", None) or []
+        if not completions:
+            continue
+        completion = completions[0]
+        if final_type in (None, "text"):
+            text = getattr(completion, "text", None)
+            if text:
+                agent_text = text
+            token_ids = getattr(completion, "token_ids", None)
+            if token_ids:
+                (out_dir / f"{stem}_text_tokens.json").write_text(json.dumps(list(map(int, token_ids))))
+                # Frame lock: the reply budget IS the input length. If the text
+                # channel is still mid-sentence when the frames run out, the
+                # spoken answer is silently truncated — warn instead.
+                tail = list(map(int, token_ids))[-TRUNCATION_GUARD_FRAMES:]
+                if any(t != pad_id for t in tail):
+                    print(
+                        f"WARNING: the text channel emits non-PAD tokens within the last "
+                        f"{TRUNCATION_GUARD_FRAMES} frames ({TRUNCATION_GUARD_FRAMES * 80} ms) — the reply "
+                        "likely ran out of acoustic frames and is cut off mid-sentence. "
+                        "Pad the input WAV with more trailing silence and rerun."
+                    )
+            # Best-effort: dump the function-channel debug timeline if the
+            # engine surfaced the request info on this output.
+            info = getattr(output, "additional_information", None)
+            func_timeline = info.get("nvc_function_tokens") if isinstance(info, dict) else None
+            if func_timeline is not None:
+                values = func_timeline.tolist() if hasattr(func_timeline, "tolist") else list(func_timeline)
+                (out_dir / f"{stem}_function_tokens.json").write_text(json.dumps([int(v) for v in values]))
+                print(f"function-channel timeline: {len(values)} entries")
+        if final_type in (None, "audio"):
+            # multimodal_output is a MultimodalPayload (Mapping) — duck-type it.
+            mm = getattr(completion, "multimodal_output", None)
+            get = getattr(mm, "get", None)
+            if not callable(get):
+                continue
+            candidate = None
+            for key in ("audio", "model_outputs", "wav", "waveform"):
+                value = get(key)
+                if value is not None:
+                    candidate = value
+                    break
+            if isinstance(candidate, list):
+                candidate = candidate[0] if candidate else None
+            if candidate is not None:
+                arr = np.asarray(
+                    candidate.float().cpu() if hasattr(candidate, "cpu") else candidate, dtype=np.float32
+                ).reshape(-1)
+                if arr.size > 0:
+                    audio = arr
+                sr_val = get("sr")
+                if sr_val is not None:
+                    sr_item = sr_val[0] if isinstance(sr_val, list) else sr_val
+                    sr = int(sr_item.item() if hasattr(sr_item, "item") else sr_item)
+
+    if agent_text is not None:
+        # The frame-locked text channel emits PAD tokens on silent frames;
+        # strip them (NeMo's post-inference detokenizer does the same).
+        pad_token = stt_cfg.get("pad_token", "<SPECIAL_12>")
+        agent_text = " ".join(agent_text.replace(pad_token, " ").split())
+        text_path = out_dir / f"{stem}_output.txt"
+        text_path.write_text(agent_text + "\n")
+        print(f"agent text -> {text_path}\n{agent_text}")
+    else:
+        print("WARNING: no agent text in outputs")
+
+    if audio is not None and audio.size > 0:
+        wav_path = out_dir / f"{stem}_output.wav"
+        sf.write(wav_path, audio, sr)
+        print(f"agent audio -> {wav_path} ({audio.size / sr:.2f}s @ {sr} Hz)")
+    else:
+        raise RuntimeError("No audio produced by the code2wav stage.")
+
+
+if __name__ == "__main__":
+    main()

examples/offline_inference/qwen2_5_omni/end2end.py

@@ -392,7 +392,7 @@ def main(args):
     total_requests = len(prompts)
     processed_count = 0
     for stage_outputs in omni_generator:
-        output = stage_outputs.request_output
+        output = stage_outputs
         if stage_outputs.final_output_type == "text":
             request_id = output.request_id
             text_output = output.outputs[0].text

examples/offline_inference/qwen3_omni/end2end.py

@@ -367,7 +367,7 @@ def main(args):
     print(f"query type: {args.query_type}")
 
     for stage_outputs in omni_generator:
-        output = stage_outputs.request_output
+        output = stage_outputs
         if stage_outputs.final_output_type == "text":
             request_id = output.request_id
             text_output = output.outputs[0].text

examples/offline_inference/qwen3_omni/end2end_async_chunk.py

@@ -237,7 +237,7 @@ async def run_single_request(
             sampling_params_list=sampling_params_list,
             output_modalities=output_modalities,
         ):
-            output = omni_output.request_output
+            output = omni_output
             if omni_output.final_output_type == "text":
                 if stage_0_first_output_ts is None:
                     stage_0_first_output_ts = time.perf_counter()

examples/offline_inference/speech_to_video/speech_to_video.py

@@ -26,7 +26,6 @@ import torch
 from vllm_omni.diffusion.data import DiffusionParallelConfig
 from vllm_omni.entrypoints.omni import Omni
 from vllm_omni.inputs.data import OmniDiffusionSamplingParams
-from vllm_omni.outputs import OmniRequestOutput
 from vllm_omni.platforms import current_omni_platform
 
 
@@ -352,8 +351,8 @@ def main():
         else:
             print("[Profiler] No valid profiling data returned.")
 
-    # Extract output from result
-    output = OmniRequestOutput.unwrap_result(result)
+    # omni.generate() returns a list for sync calls; unwrap single-result list.
+    output = result[0] if isinstance(result, list) else result
 
     if not output.images:
         raise ValueError("No video frames found in OmniRequestOutput.")

examples/offline_inference/step_audio2/end2end.py

@@ -345,7 +345,7 @@ def main(args):
     # Process outputs from each stage
     for stage_idx, stage_outputs in enumerate(omni_outputs):
         # request_output may be a single object or a list
-        raw_output = stage_outputs.request_output
+        raw_output = stage_outputs
         outputs_list = raw_output if isinstance(raw_output, list) else [raw_output]
 
         if stage_outputs.final_output_type == "text":

examples/offline_inference/text_to_audio/text_to_audio.py

@@ -43,6 +43,7 @@ from vllm_omni.diffusion.utils.param_utils import apply_declared_extra_args
 from vllm_omni.entrypoints.omni import Omni
 from vllm_omni.inputs.data import OmniDiffusionSamplingParams
 from vllm_omni.model_extras import get_extra_body_params, get_model_class_name
+from vllm_omni.outputs import OmniRequestOutput
 from vllm_omni.platforms import current_omni_platform
 
 AUDIOX_TASKS = ("t2a", "t2m", "v2a", "v2m", "tv2a", "tv2m")
@@ -388,9 +389,9 @@ def main():
         raise ValueError("No output generated from omni.generate()")
 
     output = outputs[0]
-    if not hasattr(output, "request_output") or not output.request_output:
+    if not isinstance(output, OmniRequestOutput) or not output:
         raise ValueError("No request_output found in OmniRequestOutput")
-    request_output = output.request_output
+    request_output = output
     if not hasattr(request_output, "multimodal_output"):
         raise ValueError("No multimodal_output found in request_output")
 

examples/offline_inference/text_to_image/README.md

@@ -57,8 +57,9 @@ python examples/offline_inference/text_to_image/text_to_image.py \
   --output lingbot_t2i.png
 ```
 
-The LingBot prompt builder selects the `image` output modality and requests one
-frame. Passing an input image is not supported in LingBot T2I mode.
+The shared T2I entry point selects the `image` output modality. The LingBot
+pipeline validates that T2I requests produce one frame and do not include an
+input image.
 
 ## Quick Start
 

examples/offline_inference/text_to_image/gradio_demo.py

@@ -115,7 +115,7 @@ def build_demo(args: argparse.Namespace) -> gr.Blocks:
         )
         images_outputs = []
         for output in outputs:
-            req_out = output.request_output
+            req_out = output
             if not isinstance(req_out, OmniRequestOutput) or not hasattr(req_out, "images"):
                 raise ValueError("Invalid request_output structure or missing 'images' key")
             images = req_out.images

examples/offline_inference/text_to_image/text_to_image.py

@@ -21,7 +21,9 @@ from vllm_omni.inputs.data import OmniDiffusionSamplingParams
 from vllm_omni.lora.request import LoRARequest
 from vllm_omni.lora.utils import stable_lora_int_id
 from vllm_omni.model_extras import (
-    build_text_to_image_prompt,
+    build_text_to_image_prompt as build_model_text_to_image_prompt,
+)
+from vllm_omni.model_extras import (
     get_extra_body_params,
     get_model_class_name,
     should_init_extra_args_for_non_diffusion_stages,
@@ -56,6 +58,17 @@ def parse_json_object(value: str, flag_name: str = "argument") -> dict[str, Any]
 parse_profiler_config = functools.partial(parse_json_object, flag_name="--profiler-config")
 
 
+def build_text_to_image_prompt(prompt: str, negative_prompt: str | None) -> dict[str, Any]:
+    """Build the canonical request envelope for the shared T2I example."""
+    result: dict[str, Any] = {
+        "prompt": prompt,
+        "modalities": ["image"],
+    }
+    if negative_prompt is not None:
+        result["negative_prompt"] = negative_prompt
+    return result
+
+
 def _normalize_images_for_save(images: list[Any]) -> list[Any]:
     """Convert NumPy diffusion outputs to PIL images before saving."""
     normalized = []
@@ -523,9 +536,12 @@ def main():
     generation_start = time.perf_counter()
 
     prompt_dict = build_text_to_image_prompt(
-        model_class_name=model_class_name,
         prompt=args.prompt,
         negative_prompt=args.negative_prompt,
+    )
+    prompt_dict = build_model_text_to_image_prompt(
+        model_class_name=model_class_name,
+        prompt=prompt_dict,
         height=args.height,
         width=args.width,
     )
@@ -640,7 +656,7 @@ def main():
         images = getattr(output, "images", None)
         if images:
             break
-        req_out = getattr(output, "request_output", None)
+        req_out = output
         images = getattr(req_out, "images", None) if req_out is not None else None
         if images:
             break

examples/offline_inference/text_to_speech/README.md

@@ -24,6 +24,7 @@ list of supported architectures across all modalities, see
 | Qwen3-TTS | `Qwen/Qwen3-TTS-12Hz-1.7B-{CustomVoice,VoiceDesign,Base}` | 2 (talker + code2wav) | ✓ (Base) | ✓ | 3 task variants | 24 kHz |
 | VoxCPM2 | `openbmb/VoxCPM2` | single (native AR) | ✓ | ✓ (online) | continuation | 48 kHz |
 | IndexTTS-2 | `IndexTeam/IndexTTS-2` | 2 (AR talker + S2Mel DiT + BigVGAN) | ✓ (required) | — | emotion control (`--emo-audio`, `--emo-text`, `--emo-vector`) | 22.05 kHz |
+| IndexTTS-2.5 | native `checkpoints/` bundle | 2 (AR talker + EnhancedCodec + S2Mel DiT + BigVGAN) | ✓ (required) | — | multilingual (`--lang`) + emotion control | 22.05 kHz |
 | Voxtral TTS | `mistralai/Voxtral-4B-TTS-2603` | varies | ✓ | ✓ | voice presets | 24 kHz |
 
 ## Common Quick Start
@@ -371,18 +372,10 @@ python examples/offline_inference/text_to_speech/qwen3_tts/end2end.py \
 Streaming requires `async_chunk: true` in the stage config.
 
 ### Word Timestamps
-Generate a WAV offline and a JSON sidecar with word-level timestamps from
-`Qwen/Qwen3-ForcedAligner-0.6B`:
-```bash
-python examples/offline_inference/text_to_speech/qwen3_tts/word_timestamps.py \
-    --model Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice \
-    --forced-aligner Qwen/Qwen3-ForcedAligner-0.6B \
-    --text "Hello world." \
-    --output-dir /tmp/qwen3_tts_timestamps
-```
-The script writes `qwen3_tts_word_timestamps.wav` and
-`qwen3_tts_word_timestamps.json`. On machines without a local CUDA toolkit,
-set `VLLM_USE_FLASHINFER_SAMPLER=0` to avoid FlashInfer sampler JIT.
+Word-level timestamps are currently a serving-path feature: launch
+`vllm-omni serve` with `--forced-aligner` and request `word_timestamps`
+(see `examples/online_serving/text_to_speech/README.md`). An offline
+example is not provided in this release.
 
 ### Batched decoding
 The Code2Wav stage supports batched decoding through the SpeechTokenizer. Pass multiple prompts via `--txt-prompts` and set `--batch-size` accordingly. To raise `max_num_seqs` on either stage, point `--stage-configs-path` at a stage configs YAML with the desired values (see `vllm_omni/model_executor/stage_configs/` for templates):
@@ -437,11 +430,14 @@ Streaming is exposed through the online OpenAI Speech API (`stream=true`). See [
 
 ---
 
-## IndexTTS-2
+## IndexTTS-2 and IndexTTS-2.5
 
-2-stage TTS pipeline (GPT AR talker + S2Mel CFM DiT + BigVGAN vocoder) at 22.05 kHz. Every request requires reference audio for zero-shot voice cloning. Supports emotion conditioning via audio, text, or 8-dim vector.
+Both versions use a 2-stage TTS pipeline and produce 22.05 kHz mono audio. Stage 0 is a GPT AR talker. IndexTTS-2 Stage 1 uses RepCodec semantic embeddings, GPT latent, S2Mel CFM/DiT, and BigVGAN. IndexTTS-2.5 replaces RepCodec with EnhancedCodec and uses the official code-only path (`use_gpt_latent=false`). Every request requires reference audio for zero-shot voice cloning.
 
 ### Quick start
+
+IndexTTS-2:
+
 ```bash
 python examples/offline_inference/text_to_speech/indextts2/end2end.py \
     --model IndexTeam/IndexTTS-2 \
@@ -449,6 +445,24 @@ python examples/offline_inference/text_to_speech/indextts2/end2end.py \
     --ref-audio /path/to/reference.wav
 ```
 
+IndexTTS-2 uses its existing frontend and does not require the IndexTTS-2.5 text dependencies. Before running IndexTTS-2.5, install its optional frontend dependencies:
+
+```bash
+pip install 'vllm-omni[indextts2]'
+```
+
+IndexTTS-2.5 with a native repository/lab bundle whose model root contains `checkpoints/`:
+
+```bash
+python examples/offline_inference/text_to_speech/indextts2/end2end.py \
+    --model /path/to/indextts-2.5 \
+    --model-version 2.5 \
+    --lang ja \
+    --text "こんにちは、音声合成のテストです。" \
+    --ref-audio /path/to/reference.wav
+
+```
+
 ### Emotion control
 ```bash
 # Emotion from reference audio
@@ -474,10 +488,13 @@ python examples/offline_inference/text_to_speech/indextts2/end2end.py \
 ```
 
 ### Notes
-- `--ref-audio` is **required** — IndexTTS-2 does not support text-only synthesis.
-- Stage 0 (AR Talker): GPT-2 generates mel codes from text + reference audio.
-- Stage 1 (S2Mel + BigVGAN): CFM DiT converts mel codes to waveform at 22.05 kHz.
-- Deploy config: `vllm_omni/deploy/indextts2.yaml`. Stage 1 runs with `enforce_eager: true` (DiT has dynamic shapes).
+- `--ref-audio` is **required** — neither version provides text-only synthesis.
+- IndexTTS-2.5 Stage 0 (AR Talker) generates mel codes from text + reference audio using plain vLLM sampling. It does not reproduce the official default `num_beams=3` beam search. For parity comparisons, run upstream with `num_beams=1`; output quality can differ from the official beam-search result.
+- IndexTTS-2.5 accepts language codes such as `zh`, `en`, `zhen` (mixed Chinese/English), `ja`, and `yue`. `Mandarin` is a vLLM-Omni convenience alias for `zh`. Japanese (`ja`) uses `fugashi` tokenization and produces audio, but it does not automatically expand numbers, dates, or percentages; write those inputs as readable Japanese text before calling the model. Use `--no-text-normalization` only when the input is already normalized.
+- A seed controls Stage 0 AR sampling and per-request CFM noise. Different concurrent batch compositions do not guarantee a bit-identical waveform.
+- Stage 1 (semantic codec + S2Mel + BigVGAN): CFM/DiT converts the complete semantic-code sequence to waveform.
+- Deploy configs: `vllm_omni/deploy/indextts2.yaml` and `indextts2_5.yaml`.
+- `async_chunk=false` is intentional for IndexTTS-2.5 correctness: S2Mel consumes the completed semantic sequence.
 
 ---
 

examples/offline_inference/text_to_speech/cosyvoice3/end2end.py

@@ -168,7 +168,7 @@ def run_e2e():
     print(f"Received {len(outputs)} outputs.")
     for i, output in enumerate(outputs):
         try:
-            ro = output.request_output
+            ro = output
             if ro is None:
                 print("No request_output found.")
                 continue

examples/offline_inference/text_to_speech/fish_speech/end2end.py

@@ -179,7 +179,7 @@ def main(args):
 
     t_start = time.perf_counter()
     for stage_outputs in omni.generate(inputs):
-        request_output = stage_outputs.request_output
+        request_output = stage_outputs
         if request_output is None or not request_output.outputs:
             continue
         _save_wav(
@@ -215,7 +215,7 @@ async def main_streaming(args):
     sample_rate = None
 
     async for stage_output in omni.generate(prompt, request_id=request_id):
-        mm = stage_output.request_output.outputs[0].multimodal_output
+        mm = stage_output.outputs[0].multimodal_output
         if not stage_output.finished:
             t_now = time.perf_counter()
             sample_rate, n = _collect_audio_chunk(mm, all_audio_chunks, sample_rate)

examples/offline_inference/text_to_speech/indextts2/end2end.py

@@ -1,13 +1,16 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-"""Offline inference example for IndexTTS2 via vLLM-Omni.
+"""Offline inference example for IndexTTS 2.0 and 2.5 via vLLM-Omni.
 
-Two-stage pipeline: GPT AR (Stage 0) → S2Mel + BigVGAN (Stage 1).
+Two-stage pipeline: GPT AR (Stage 0) → semantic codec + S2Mel + BigVGAN
+(Stage 1).
 Output is 22050 Hz mono WAV.
 
 Usage:
   python end2end.py \
-    --model /path/to/IndexTeam/IndexTTS-2 \
+    --model /path/to/indextts-2.5 \
+    --model-version 2.5 \
+    --lang zh \
     --text "你好，这是一个语音合成测试。" \
     --ref-audio /path/to/ref.wav
 
@@ -47,6 +50,10 @@ def build_request(
     emo_alpha: float | None = None,
     use_emo_text: bool = False,
     use_random: bool = False,
+    model_type: str = "indextts2",
+    lang: str = "zh",
+    text_normalization: bool = True,
+    tokenizer_file: str | None = None,
 ) -> dict:
     additional: dict = {"text": [text]}
     if ref_audio_path:
@@ -63,12 +70,44 @@ def build_request(
         additional["use_emo_text"] = [True]
     if use_random:
         additional["use_random"] = [True]
+    if model_type == "indextts2_5":
+        additional["lang"] = [lang]
+        additional["text_normalization"] = [text_normalization]
+    prompt_kwargs = {
+        "model_type": model_type,
+        "lang": lang,
+        "text_normalization": text_normalization,
+    }
+    if tokenizer_file is not None:
+        prompt_kwargs["tokenizer_file"] = tokenizer_file
     return {
-        "prompt_token_ids": build_indextts2_prefill_prompt_ids(model, text),
+        "prompt_token_ids": build_indextts2_prefill_prompt_ids(
+            model,
+            text,
+            **prompt_kwargs,
+        ),
         "additional_information": additional,
     }
 
 
+def _get_stage0_tokenizer_file(omni: Omni) -> str | None:
+    """Read the effective tokenizer override from the resolved Stage 0 config."""
+    stage_configs = omni.stage_configs
+    if not stage_configs:
+        return None
+    engine_args = getattr(stage_configs[0], "engine_args", None)
+    if engine_args is None:
+        return None
+    hf_overrides = getattr(engine_args, "hf_overrides", None)
+    if hf_overrides is None and isinstance(engine_args, dict):
+        hf_overrides = engine_args.get("hf_overrides")
+    if hf_overrides is None:
+        return None
+    if hasattr(hf_overrides, "get"):
+        return hf_overrides.get("tokenizer_file")
+    return getattr(hf_overrides, "tokenizer_file", None)
+
+
 def save_audio(waveform: torch.Tensor, path: str, sample_rate: int = 22050) -> None:
     audio_np = waveform.float().numpy()
     sf.write(path, audio_np, sample_rate)
@@ -95,9 +134,14 @@ def extract_audio(mm: dict) -> tuple[torch.Tensor | None, int]:
 
 
 def main(args) -> None:
+    model_type = "indextts2_5" if args.model_version == "2.5" else "indextts2"
+    deploy_config = args.deploy_config
+    if deploy_config is None and args.model_version == "2.5":
+        repo_root = Path(__file__).resolve().parents[4]
+        deploy_config = str(repo_root / "vllm_omni" / "deploy" / "indextts2_5.yaml")
     omni = Omni(
         model=args.model,
-        deploy_config=args.deploy_config,
+        deploy_config=deploy_config,
         stage_init_timeout=args.stage_init_timeout,
     )
 
@@ -126,7 +170,7 @@ def main(args) -> None:
     print(f"Synthesizing: {args.text!r}")
     if args.ref_audio:
         print(f"  ref_audio: {args.ref_audio}")
-    inputs = build_request(
+    request_kwargs = dict(
         model=args.model,
         text=args.text,
         ref_audio_path=args.ref_audio,
@@ -136,7 +180,17 @@ def main(args) -> None:
         emo_alpha=args.emo_alpha,
         use_emo_text=args.use_emo_text,
         use_random=args.use_random,
+        model_type=model_type,
+        lang=args.lang,
+        text_normalization=args.text_normalization,
     )
+    if model_type == "indextts2_5":
+        tokenizer_file = _get_stage0_tokenizer_file(omni)
+        if tokenizer_file is not None:
+            request_kwargs["tokenizer_file"] = tokenizer_file
+    inputs = build_request(**request_kwargs)
+    if model_type == "indextts2_5":
+        inputs["additional_information"]["duration_factor"] = [1.0 / getattr(args, "speed", 1.0)]
 
     for i, omni_out in enumerate(omni.generate(inputs, sampling_params_list=sampling_params)):
         mm = omni_out.multimodal_output
@@ -154,8 +208,37 @@ def main(args) -> None:
 
 
 def parse_args():
-    parser = FlexibleArgumentParser(description="IndexTTS2 offline inference")
-    parser.add_argument("--model", required=True, help="HF model path for IndexTTS2.")
+    parser = FlexibleArgumentParser(description="IndexTTS 2.0/2.5 offline inference")
+    parser.add_argument(
+        "--model",
+        required=True,
+        help="HF model path or native bundle containing checkpoints/.",
+    )
+    parser.add_argument(
+        "--model-version",
+        choices=("2.0", "2.5"),
+        default="2.0",
+    )
+    parser.add_argument(
+        "--lang",
+        default="zh",
+        help=(
+            "IndexTTS 2.5 language code, for example zh/en/zhen/ja/yue; "
+            "zhen is mixed Chinese/English and Mandarin is a vLLM-Omni alias for zh."
+        ),
+    )
+    parser.add_argument(
+        "--no-text-normalization",
+        action="store_false",
+        dest="text_normalization",
+        help="Disable IndexTTS 2.5 text normalization.",
+    )
+    parser.add_argument(
+        "--speed",
+        type=float,
+        default=1.0,
+        help="IndexTTS 2.5 native synthesis speed in [0.5, 2.0]; 2.0 is faster.",
+    )
     parser.add_argument("--text", default="你好，这是IndexTTS2语音合成测试。")
     parser.add_argument("--ref-audio", required=True, help="Reference audio for voice cloning.")
     parser.add_argument("--emo-audio", default=None, help="Emotion reference audio.")
@@ -170,7 +253,15 @@ def parse_args():
     parser.add_argument("--emo-alpha", type=float, default=None, help="Emotion weight in [0, 1].")
     parser.add_argument("--use-emo-text", action="store_true", help="Infer emotion vector from emo-text or text.")
     parser.add_argument("--use-random", action="store_true", help="Use random emotion prototypes.")
-    parser.add_argument("--seed", type=int, default=None)
+    parser.add_argument(
+        "--seed",
+        type=int,
+        default=None,
+        help=(
+            "Seed Stage 0 AR sampling and per-request CFM noise; changing the "
+            "concurrent batch composition may still change the final waveform."
+        ),
+    )
     parser.add_argument(
         "--output-dir",
         default=os.path.join(
@@ -180,7 +271,10 @@ def parse_args():
     )
     parser.add_argument("--deploy-config", default=None)
     parser.add_argument("--stage-init-timeout", type=int, default=600)
-    return parser.parse_args()
+    args = parser.parse_args()
+    if args.model_version == "2.5" and not 0.5 <= args.speed <= 2.0:
+        parser.error("IndexTTS 2.5 --speed must be between 0.5 and 2.0")
+    return args
 
 
 if __name__ == "__main__":

examples/offline_inference/text_to_speech/moss_tts_nano/end2end.py

@@ -137,7 +137,7 @@ def main(args) -> None:
     params_list = sampling_params
 
     for stage_outputs in omni.generate(inputs, params_list):
-        for i, req_output in enumerate(stage_outputs.request_output):
+        for i, req_output in enumerate(stage_outputs):
             for j, out in enumerate(req_output.outputs):
                 mm = out.multimodal_output
                 if mm is None:

examples/offline_inference/text_to_speech/omnivoice/end2end.py

@@ -135,7 +135,7 @@ def run_e2e():
     print(f"Received {len(outputs)} outputs.")
     for i, output in enumerate(outputs):
         try:
-            ro = output.request_output
+            ro = output
             if ro is None:
                 print("No request_output found.")
                 continue

examples/offline_inference/text_to_speech/qwen3_tts/end2end.py

@@ -391,7 +391,7 @@ def main(args):
     for batch_start in range(0, len(inputs), batch_size):
         batch = inputs[batch_start : batch_start + batch_size]
         for stage_outputs in omni.generate(batch):
-            output = stage_outputs.request_output
+            output = stage_outputs
             _save_wav(output_dir, output.request_id, output.outputs[0].multimodal_output)
 
 
@@ -412,7 +412,7 @@ async def main_streaming(args):
         t_prev = t_start
         chunk_idx = 0
         async for stage_output in omni.generate(prompt, request_id=request_id):
-            mm = stage_output.request_output.outputs[0].multimodal_output
+            mm = stage_output.outputs[0].multimodal_output
             if not stage_output.finished:
                 t_now = time.perf_counter()
                 audio = mm.get("audio")

examples/offline_inference/text_to_speech/qwen3_tts/word_timestamps.py

@@ -116,7 +116,7 @@ def main(args: Any) -> None:
     prompt = _build_custom_voice_input(args)
     final_output = None
     for stage_outputs in omni.generate([prompt]):
-        final_output = stage_outputs.request_output
+        final_output = stage_outputs
     if final_output is None:
         raise RuntimeError("Qwen3-TTS did not produce an output.")
 

examples/offline_inference/text_to_speech/soulxsinger/benchmark.py

@@ -38,7 +38,7 @@ def _stage_durations_from_output(outputs) -> dict[str, float]:
     durations = getattr(omni_out, "stage_durations", None) or {}
     if durations:
         return dict(durations)
-    ro = getattr(omni_out, "request_output", None)
+    ro = omni_out
     if ro is not None:
         return dict(getattr(ro, "stage_durations", None) or {})
     return {}

examples/offline_inference/text_to_speech/soulxsinger/end2end.py

@@ -99,7 +99,7 @@ def extract_audio(outputs) -> tuple[np.ndarray, int]:
     out = outputs[0]
     mm = getattr(out, "multimodal_output", None)
     if mm is None:
-        ro = getattr(out, "request_output", None)
+        ro = out
         if ro is not None:
             mm = getattr(ro, "multimodal_output", None)
             if mm is None and ro.outputs:

examples/offline_inference/text_to_video/text_to_video.md

@@ -8,6 +8,7 @@ A unified script for text-to-video generation. Supports multiple models with mod
 |---|---|---|---|---|---|
 | `Wan-AI/Wan2.1-VACE-1.3B-diffusers` | 480x832 | 81 | 30 | 5.0 | ~20 GiB (RTX 5090, VAE tiling) |
 | `Wan-AI/Wan2.2-T2V-A14B-Diffusers` | 720x1280 | 81 | 40 | 4.0 | ~60 GiB |
+| `robbyant/lingbot-video-dense-1.3b` / `robbyant/lingbot-video-moe-30b-a3b` | 192x320 | 9 | 2 | 3.0 | ~68 GiB (MoE smoke) |
 | `Lightricks/LTX-2` | 512x768 | 121 | 40 | video 3.0 / audio 7.0 | Model-dependent |
 | `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-480p_t2v` | 480x832 | 121 | 50 | 6.0 | 1×A100 80GB |
 | `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-720p_t2v` | 720x1280 | 121 | 50 | 6.0 | FP8 + VAE tiling required |
@@ -57,18 +58,31 @@ python text_to_video.py \
 
 ### LingBot-Video
 
-The dedicated LingBot example explicitly selects the video output modality:
+The shared runner recognizes the official dense and MoE checkpoint IDs, selects
+`LingBotVideoPipeline`, and builds the canonical video request envelope:
 
 ```bash
-python examples/offline_inference/text_to_video/text_to_video_lingbot.py \
+python text_to_video.py \
   --model robbyant/lingbot-video-dense-1.3b \
+  --prompt "a robotic arm picks up a red block" \
   --height 192 --width 320 --num-frames 9 --num-inference-steps 2 \
+  --guidance-scale 3.0 --flow-shift 3.0 --fps 24 \
   --output lingbot_t2v.mp4
 ```
 
 Use the same command with `robbyant/lingbot-video-moe-30b-a3b` for the MoE
 checkpoint. Requested frame counts are rounded upward to the causal
-VAE's `4n+1` grid.
+VAE's `4n+1` grid. For a local checkpoint path whose name does not contain
+`lingbot`, pass `--model-class-name LingBotVideoPipeline`.
+
+LingBot-only options such as `batch_cfg` and `output_type` use the shared
+model-extra channel, for example:
+
+```bash
+python text_to_video.py \
+  --model robbyant/lingbot-video-dense-1.3b \
+  --extra-body '{"batch_cfg": true, "output_type": "np"}'
+```
 
 ### LTX-2
 

examples/offline_inference/text_to_video/text_to_video.py

@@ -16,7 +16,7 @@ from vllm_omni.entrypoints.omni import Omni
 from vllm_omni.inputs.data import OmniDiffusionSamplingParams
 from vllm_omni.lora.request import LoRARequest
 from vllm_omni.lora.utils import stable_lora_int_id
-from vllm_omni.model_extras import get_extra_body_params, get_model_class_name
+from vllm_omni.model_extras import get_extra_body_params, get_model_class_name, get_output_tensor_range
 from vllm_omni.outputs import OmniRequestOutput
 from vllm_omni.platforms import current_omni_platform
 
@@ -80,6 +80,17 @@ _MODEL_PRESETS = {
         "fps": 16,
         "output": "helios_output.mp4",
     },
+    "lingbot": {
+        "model_class_name": "LingBotVideoPipeline",
+        "height": 192,
+        "width": 320,
+        "num_frames": 9,
+        "num_inference_steps": 2,
+        "guidance_scale": 3.0,
+        "fps": 24,
+        "flow_shift": 3.0,
+        "output": "lingbot_video_output.mp4",
+    },
     "ltx2": {
         "height": 512,
         "width": 768,
@@ -110,27 +121,51 @@ _MODEL_PRESETS = {
 def _detect_preset(model: str, model_class_name: str | None = None) -> dict:
     model_lower = model.lower()
     class_lower = (model_class_name or "").lower()
-    if "distilled" in class_lower or "distilled" in model_lower:
-        return _MODEL_PRESETS["ltx2_distilled"]
-    if "ltx23" in class_lower or "ltx-2.3" in model_lower or "ltx_2.3" in model_lower:
-        return _MODEL_PRESETS["ltx23"]
-    if "ltx2" in class_lower or "ltx-2" in model_lower or "ltx_2" in model_lower:
+    if "lingbot" in model_lower or "lingbotvideo" in class_lower:
+        return _MODEL_PRESETS["lingbot"]
+    if "ltx" in class_lower or "ltx" in model_lower:
+        if "distilled" in class_lower or "distilled" in model_lower:
+            return _MODEL_PRESETS["ltx2_distilled"]
+        if "ltx23" in class_lower or "ltx-2.3" in model_lower or "ltx_2.3" in model_lower:
+            return _MODEL_PRESETS["ltx23"]
         return _MODEL_PRESETS["ltx2"]
-    if "vace" in model_lower:
+    if "vace" in model_lower or "vace" in class_lower:
         return _MODEL_PRESETS["vace"]
     # Edge must be matched before the generic cosmos branch (its "cosmos" substring would
     # otherwise pick up the Nano/Super 720p / gs 6.0 / flow_shift 10.0 preset).
     if ("cosmos" in model_lower or "cosmos" in class_lower) and ("edge" in model_lower or "edge" in class_lower):
         return _MODEL_PRESETS["cosmos3_edge"]
-    if "cosmos" in model_lower:
+    if "cosmos" in model_lower or "cosmos" in class_lower:
         return _MODEL_PRESETS["cosmos"]
-    if "hunyuan" in model_lower:
+    if "hunyuan" in model_lower or "hunyuan" in class_lower:
         return _MODEL_PRESETS["hunyuan"]
-    if "helios" in model_lower:
+    if "helios" in model_lower or "helios" in class_lower:
         return _MODEL_PRESETS["helios"]
     return _MODEL_PRESETS["wan"]
 
 
+def build_text_to_video_prompt(prompt: str, negative_prompt: str | None) -> dict[str, Any]:
+    """Build the canonical request envelope for the shared T2V example."""
+    result: dict[str, Any] = {
+        "prompt": prompt,
+        "modalities": ["video"],
+    }
+    if negative_prompt is not None:
+        result["negative_prompt"] = negative_prompt
+    return result
+
+
+def _normalize_float_tensor(tensor: torch.Tensor, source_range: str) -> torch.Tensor:
+    """Normalize decoded frames according to the pipeline's output contract."""
+    if not tensor.is_floating_point():
+        return tensor
+    if source_range == "negative_one_to_one":
+        return tensor.clamp(-1, 1) * 0.5 + 0.5
+    if source_range == "zero_to_one":
+        return tensor.clamp(0, 1)
+    raise ValueError(f"Unsupported floating-point tensor range: {source_range!r}")
+
+
 def parse_profiler_config(value: str) -> dict[str, Any]:
     try:
         config = json.loads(value)
@@ -160,7 +195,7 @@ def parse_extra_body(value: str) -> dict[str, Any]:
 def parse_args() -> argparse.Namespace:
     parser = argparse.ArgumentParser(
         description="Generate a video from a text prompt. "
-        "Supports Wan2.2, HunyuanVideo-1.5, Helios, and other text-to-video models."
+        "Supports Wan2.2, HunyuanVideo-1.5, Helios, LingBot-Video, and other text-to-video models."
     )
     parser.add_argument(
         "--model",
@@ -390,7 +425,7 @@ def _extract_peak_memory_mb(result: Any) -> float:
         return 0.0
     val = getattr(result, "peak_memory_mb", 0.0)
     if not val:
-        inner = getattr(result, "request_output", None)
+        inner = result
         if isinstance(inner, list):
             inner = inner[0] if inner else None
         val = getattr(inner, "peak_memory_mb", 0.0)
@@ -408,6 +443,7 @@ def main():
     for key, default_val in preset.items():
         if getattr(args, key.replace("-", "_"), None) is None:
             setattr(args, key.replace("-", "_"), default_val)
+    model_class_name = args.model_class_name
 
     generator = torch.Generator(device=current_omni_platform.device_type).manual_seed(args.seed)
     # Cache-dit config (Wan2.2 only)
@@ -480,6 +516,7 @@ def main():
     omni = Omni(**omni_kwargs)
     model_class_name = get_model_class_name(omni) or model_class_name
     declared_extra_body_params = get_extra_body_params(model_class_name)
+    output_tensor_range = get_output_tensor_range(model_class_name)
 
     if profiler_enabled:
         print("[Profiler] Starting profiling...")
@@ -513,12 +550,11 @@ def main():
             lora_path=lora_path,
         )
 
-    prompt_dict = {"prompt": args.prompt}
-    if args.negative_prompt is not None:
-        prompt_dict["negative_prompt"] = args.negative_prompt
-    elif preset not in (_MODEL_PRESETS["ltx2"], _MODEL_PRESETS["ltx23"]):
+    negative_prompt = args.negative_prompt
+    if negative_prompt is None and all(preset is not _MODEL_PRESETS[name] for name in ("lingbot", "ltx2", "ltx23")):
         # Preserve the historical empty-prompt behavior for non-LTX examples.
-        prompt_dict["negative_prompt"] = ""
+        negative_prompt = ""
+    prompt_dict = build_text_to_video_prompt(args.prompt, negative_prompt)
 
     extra_args = {}
     if lora_request:
@@ -575,8 +611,8 @@ def main():
         if frames.multimodal_output and "audio" in frames.multimodal_output:
             audio = frames.multimodal_output["audio"]
             audio_sample_rate = frames.multimodal_output.get("audio_sample_rate", audio_sample_rate)
-        if frames.is_pipeline_output and frames.request_output is not None:
-            inner_output = frames.request_output
+        if frames.is_pipeline_output and frames is not None:
+            inner_output = frames
             if isinstance(inner_output, OmniRequestOutput):
                 if inner_output.multimodal_output and "audio" in inner_output.multimodal_output:
                     audio = inner_output.multimodal_output["audio"]
@@ -630,8 +666,7 @@ def main():
                 frame_tensor = frame_tensor[0]
             if frame_tensor.dim() == 3 and frame_tensor.shape[0] in (3, 4):
                 frame_tensor = frame_tensor.permute(1, 2, 0)
-            if frame_tensor.is_floating_point():
-                frame_tensor = frame_tensor.clamp(-1, 1) * 0.5 + 0.5
+            frame_tensor = _normalize_float_tensor(frame_tensor, output_tensor_range)
             return frame_tensor.float().numpy()
         if isinstance(frame, np.ndarray):
             frame_array = frame
@@ -683,8 +718,7 @@ def main():
                 video_tensor = video_tensor[0]
         elif video_tensor.dim() == 4 and video_tensor.shape[0] in (3, 4):
             video_tensor = video_tensor.permute(1, 2, 3, 0)
-        if video_tensor.is_floating_point():
-            video_tensor = video_tensor.clamp(-1, 1) * 0.5 + 0.5
+        video_tensor = _normalize_float_tensor(video_tensor, output_tensor_range)
         video_array = video_tensor.float().numpy()
     elif isinstance(frames, np.ndarray):
         video_array = frames

examples/offline_inference/text_to_video/text_to_video_lingbot.py

@@ -1,131 +0,0 @@
-# SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-
-from __future__ import annotations
-
-import argparse
-import time
-from pathlib import Path
-from typing import Any
-
-import numpy as np
-import torch
-
-from vllm_omni.diffusion.data import DiffusionParallelConfig
-from vllm_omni.entrypoints.omni import Omni
-from vllm_omni.inputs.data import OmniDiffusionSamplingParams
-from vllm_omni.outputs import OmniRequestOutput
-from vllm_omni.platforms import current_omni_platform
-
-
-def parse_args() -> argparse.Namespace:
-    parser = argparse.ArgumentParser(description="Generate text-to-video with a LingBot-Video checkpoint.")
-    parser.add_argument("--model", default="robbyant/lingbot-video-dense-1.3b")
-    parser.add_argument("--prompt", default="a robotic arm picks up a red block")
-    parser.add_argument("--negative-prompt", default=None)
-    parser.add_argument("--output", default="lingbot_video_output.mp4")
-    parser.add_argument("--height", type=int, default=192)
-    parser.add_argument("--width", type=int, default=320)
-    parser.add_argument("--num-frames", type=int, default=9)
-    parser.add_argument("--num-inference-steps", type=int, default=2)
-    parser.add_argument("--guidance-scale", type=float, default=3.0)
-    parser.add_argument("--flow-shift", type=float, default=3.0)
-    parser.add_argument("--seed", type=int, default=42)
-    parser.add_argument("--fps", type=int, default=24)
-    parser.add_argument("--output-type", choices=["pt", "np"], default="pt")
-    parser.add_argument("--batch-cfg", action="store_true")
-    return parser.parse_args()
-
-
-def _peak_memory_mb(output: Any) -> float:
-    result = output[0] if isinstance(output, list) and output else output
-    return float(getattr(result, "peak_memory_mb", 0.0) or 0.0)
-
-
-def _extract_video(output: Any) -> Any:
-    if isinstance(output, list):
-        output = output[0] if output else None
-    if isinstance(output, OmniRequestOutput):
-        if output.final_output_type != "image":
-            raise ValueError(f"Unexpected output type {output.final_output_type!r}.")
-        if not output.images:
-            raise ValueError("No video frames found in OmniRequestOutput.")
-        return output.images[0] if len(output.images) == 1 else output.images
-    return output
-
-
-def _to_frame_array(video: Any) -> np.ndarray:
-    if isinstance(video, torch.Tensor):
-        tensor = video.detach().cpu()
-        if tensor.dim() == 5:
-            tensor = tensor[0]
-        if tensor.dim() == 4 and tensor.shape[-1] not in (3, 4):
-            if tensor.shape[0] in (3, 4):
-                tensor = tensor.permute(1, 2, 3, 0)
-            elif tensor.shape[1] in (3, 4):
-                tensor = tensor.permute(0, 2, 3, 1)
-        if tensor.is_floating_point():
-            if float(tensor.min()) < 0.0:
-                tensor = tensor.clamp(-1, 1) * 0.5 + 0.5
-            else:
-                tensor = tensor.clamp(0, 1)
-        return tensor.float().numpy()
-    if isinstance(video, np.ndarray):
-        array = video[0] if video.ndim == 5 else video
-        if np.issubdtype(array.dtype, np.integer):
-            return array.astype(np.float32) / 255.0
-        return array.astype(np.float32, copy=False)
-    if isinstance(video, list):
-        return np.stack([_to_frame_array(frame) for frame in video], axis=0)
-    return np.asarray(video, dtype=np.float32)
-
-
-def main() -> None:
-    args = parse_args()
-    generator = torch.Generator(device=current_omni_platform.device_type).manual_seed(args.seed)
-    omni = Omni(
-        model=args.model,
-        model_class_name="LingBotVideoPipeline",
-        flow_shift=args.flow_shift,
-        parallel_config=DiffusionParallelConfig(),
-    )
-
-    prompt: dict[str, Any] = {
-        "prompt": args.prompt,
-        "modalities": ["video"],
-        "num_frames": args.num_frames,
-    }
-    if args.negative_prompt is not None:
-        prompt["negative_prompt"] = args.negative_prompt
-
-    sampling_params = OmniDiffusionSamplingParams(
-        height=args.height,
-        width=args.width,
-        num_frames=args.num_frames,
-        num_inference_steps=args.num_inference_steps,
-        guidance_scale=args.guidance_scale,
-        generator=generator,
-        output_type=args.output_type,
-        extra_args={"batch_cfg": args.batch_cfg, "shift": args.flow_shift},
-    )
-
-    start = time.perf_counter()
-    output = omni.generate(prompt, sampling_params)
-    elapsed = time.perf_counter() - start
-    print(f"Total generation time: {elapsed:.4f} seconds ({elapsed * 1000:.2f} ms)")
-    peak_mb = _peak_memory_mb(output)
-    if peak_mb:
-        print(f"Worker peak GPU memory (reserved): {peak_mb:.2f} MiB ({peak_mb / 1024:.2f} GiB)")
-
-    video = _to_frame_array(_extract_video(output))
-    output_path = Path(args.output)
-    output_path.parent.mkdir(parents=True, exist_ok=True)
-    from diffusers.utils import export_to_video
-
-    export_to_video(list(video), str(output_path), fps=args.fps)
-    print(f"Saved generated video to {output_path}")
-    omni.close()
-
-
-if __name__ == "__main__":
-    main()

examples/offline_inference/x_to_video_audio/x_to_video_audio.py

@@ -138,7 +138,7 @@ def _extract_peak_memory_mb(result: Any) -> float:
         return 0.0
     val = getattr(result, "peak_memory_mb", 0.0)
     if not val:
-        inner = getattr(result, "request_output", None)
+        inner = result
         if isinstance(inner, list):
             inner = inner[0] if inner else None
         val = getattr(inner, "peak_memory_mb", 0.0)

examples/online_serving/image_to_video/README.md

@@ -1,6 +1,9 @@
 # Image-To-Video
 
-This example demonstrates how to deploy the Wan2.2 image-to-video model for online video generation using vLLM-Omni.
+This example demonstrates online image-to-video generation with vLLM-Omni.
+The startup script defaults to Wan2.2, but `MODEL` can select any supported
+image-to-video model. The existing curl helper remains a concrete Wan2.2
+request through the standard Videos API.
 
 ## Start Server
 
@@ -20,9 +23,7 @@ bash run_server.sh
 
 The script allows overriding:
 - `MODEL` (default: `Wan-AI/Wan2.2-I2V-A14B-Diffusers`)
-- `PORT` (default: `8091`)
-- `BOUNDARY_RATIO` (default: `0.875`)
-- `FLOW_SHIFT` (default: `12.0`)
+- `PORT` (default: `8099`)
 - `CACHE_BACKEND` (default: `none`)
 - `ENABLE_CACHE_DIT_SUMMARY` (default: `0`)
 

examples/online_serving/image_to_video/run_server.sh

@@ -1,12 +1,12 @@
 #!/bin/bash
-# Wan2.2 image-to-video server start script
+# Shared image-to-video server start script
 
 MODEL="${MODEL:-Wan-AI/Wan2.2-I2V-A14B-Diffusers}"
 PORT="${PORT:-8099}"
 CACHE_BACKEND="${CACHE_BACKEND:-none}"
 ENABLE_CACHE_DIT_SUMMARY="${ENABLE_CACHE_DIT_SUMMARY:-0}"
 
-echo "Starting Wan2.2 I2V server..."
+echo "Starting image-to-video server..."
 echo "Model: $MODEL"
 echo "Port: $PORT"
 echo "Cache backend: $CACHE_BACKEND"

examples/online_serving/text_to_image/README.md

@@ -1,6 +1,9 @@
 # Text-To-Image
 
-This example demonstrates how to deploy Qwen-Image model for online image generation service using vLLM-Omni.
+This example demonstrates online text-to-image generation with vLLM-Omni. The
+startup script and Python clients default to Qwen-Image, but `MODEL` can select
+any supported text-to-image model. The existing curl helper remains a concrete
+Qwen-Image request through the standard Images API.
 
 ## Start Server
 
@@ -242,8 +245,8 @@ count, use `size` and `n` rather than `height`, `width`, or
 | `height`                 | int   | None    | Image height in pixels         |
 | `width`                  | int   | None    | Image width in pixels          |
 | `size`                   | str   | None    | Image size (e.g., "1024x1024") |
-| `num_inference_steps`    | int   | 50      | Number of denoising steps      |
-| `true_cfg_scale`         | float | 4.0     | Qwen-Image CFG scale           |
+| `num_inference_steps`    | int   | model   | Number of denoising steps      |
+| `true_cfg_scale`         | float | model   | True CFG scale when supported  |
 | `seed`                   | int   | None    | Random seed (reproducible)     |
 | `negative_prompt`        | str   | None    | Negative prompt                |
 | `num_outputs_per_prompt` | int   | 1       | Number of images to generate   |

examples/online_serving/text_to_image/gradio_demo.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python3
 """
-Qwen-Image Gradio Demo for online serving.
+Shared text-to-image Gradio demo for online serving.
 
 Usage:
     python gradio_demo.py [--server http://localhost:8091] [--port 7860]
@@ -77,9 +77,9 @@ def generate_image(
 def create_demo(server_url: str):
     """Create Gradio demo interface."""
 
-    with gr.Blocks(title="Qwen-Image Demo") as demo:
-        gr.Markdown("# Qwen-Image Online Generation")
-        gr.Markdown("Generate images using Qwen-Image model")
+    with gr.Blocks(title="Text-to-Image Demo") as demo:
+        gr.Markdown("# Text-to-Image Online Generation")
+        gr.Markdown("Generate images using the model configured on the server")
 
         with gr.Row():
             with gr.Column(scale=1):
@@ -115,7 +115,6 @@ def create_demo(server_url: str):
                         label="Inference Steps",
                         minimum=10,
                         maximum=100,
-                        # Default steps aligned with run_curl_text_to_image.sh to 100
                         value=100,
                         step=5,
                     )
@@ -173,7 +172,7 @@ def create_demo(server_url: str):
 
 
 def main():
-    parser = argparse.ArgumentParser(description="Qwen-Image Gradio Demo")
+    parser = argparse.ArgumentParser(description="Shared text-to-image Gradio demo")
     parser.add_argument("--server", default="http://localhost:8091", help="Server URL")
     parser.add_argument("--port", type=int, default=7860, help="Gradio port")
     parser.add_argument("--share", action="store_true", help="Create public link")

examples/online_serving/text_to_image/openai_chat_client.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python3
 """
-Qwen-Image OpenAI-compatible image generation client.
+Shared OpenAI-compatible text-to-image generation client.
 
 Usage:
     python openai_chat_client.py --prompt "A beautiful landscape" --output output.png
@@ -39,7 +39,7 @@ def generate_image(
         height: Image height in pixels
         width: Image width in pixels
         steps: Number of diffusion steps
-        true_cfg_scale: Qwen-Image CFG scale
+        true_cfg_scale: True CFG scale for models that support it
         seed: Random seed
         negative_prompt: Negative prompt
         num_outputs_per_prompt: Number of images to generate
@@ -70,7 +70,7 @@ def generate_image(
         payload["num_inference_steps"] = steps
     if true_cfg_scale is not None:
         payload["true_cfg_scale"] = true_cfg_scale
-    if negative_prompt:
+    if negative_prompt is not None:
         payload["negative_prompt"] = negative_prompt
     if seed is not None:
         payload["seed"] = seed
@@ -115,15 +115,15 @@ def generate_image(
 
 
 def main():
-    parser = argparse.ArgumentParser(description="Qwen-Image chat client")
+    parser = argparse.ArgumentParser(description="Shared text-to-image client")
     parser.add_argument("--prompt", "-p", default="a cup of coffee on the table", help="Text prompt")
-    parser.add_argument("--output", "-o", default="qwen_image_output.png", help="Output file")
+    parser.add_argument("--output", "-o", default="text_to_image_output.png", help="Output file")
     parser.add_argument("--server", "-s", default="http://localhost:8091", help="Server URL")
-    parser.add_argument("--height", type=int, default=1024, help="Image height")
-    parser.add_argument("--width", type=int, default=1024, help="Image width")
-    parser.add_argument("--steps", type=int, default=50, help="Inference steps")
-    parser.add_argument("--cfg-scale", type=float, default=4.0, help="True CFG scale")
-    parser.add_argument("--seed", type=int, default=0, help="Random seed")
+    parser.add_argument("--height", type=int, help="Image height; defaults to the model setting")
+    parser.add_argument("--width", type=int, help="Image width; defaults to the model setting")
+    parser.add_argument("--steps", type=int, help="Inference steps; defaults to the model setting")
+    parser.add_argument("--cfg-scale", type=float, help="True CFG scale when supported")
+    parser.add_argument("--seed", type=int, help="Random seed")
     parser.add_argument("--negative", help="Negative prompt")
 
     parser.add_argument("--lora-path", default=None, help="Server-local LoRA adapter folder (PEFT format)")

examples/online_serving/text_to_image/run_server.sh

@@ -1,10 +1,10 @@
 #!/bin/bash
-# Qwen-Image online serving startup script
+# Shared text-to-image online serving startup script
 
 MODEL="${MODEL:-Qwen/Qwen-Image}"
 PORT="${PORT:-8091}"
 
-echo "Starting Qwen-Image server..."
+echo "Starting text-to-image server..."
 echo "Model: $MODEL"
 echo "Port: $PORT"
 

examples/online_serving/text_to_speech/README.md

@@ -16,7 +16,8 @@ For the full list of supported architectures across all modalities, see
 |---|---|---|---|---|---|
 | Fish Speech S2 Pro | `fishaudio/s2-pro` | ✓ (`ref_audio`+`ref_text`) | ✓ (PCM stream) | — | ✓ |
 | GLM-TTS | `zai-org/GLM-TTS` | ✓ (`ref_audio`+`ref_text`, required) | ✓ (PCM stream) | — | ✓ |
-| IndexTTS-2 | `IndexTeam/IndexTTS-2` | ✓ (`ref_audio` or uploaded `voice`) | compat only, non-chunk | uploaded audio voice only; no presets | — |
+| IndexTTS-2 | `IndexTeam/IndexTTS-2` | ✓ (`ref_audio` or uploaded `voice`) | `stream=true` response, non-chunk | uploaded audio voice only; no presets | — |
+| IndexTTS-2.5 | native `checkpoints/` bundle | ✓ (`ref_audio` or uploaded `voice`) | `stream=true` response, non-chunk | uploaded audio voice only; no presets | — |
 | Ming-omni-tts | `inclusionAI/Ming-omni-tts-0.5B` | ✓ (`ref_audio` / `speaker_embedding`) | ✓ (PCM stream) | IP labels + structured `instructions` | — |
 | Ming-flash-omni-TTS | `Jonathan1909/Ming-flash-omni-2.0` | — (caption-controlled) | — | caption fields (`instructions`) | — |
 | MOSS-TTS-Nano | `OpenMOSS-Team/MOSS-TTS-Nano` | ✓ (`ref_audio` required) | ✓ (PCM stream) | — | ✓ |
@@ -138,15 +139,29 @@ bash examples/online_serving/text_to_speech/glm_tts/run_gradio_demo.sh
 
 ---
 
-## IndexTTS-2
+## IndexTTS-2 and IndexTTS-2.5
 
-2-stage TTS (GPT AR + S2Mel CFM DiT + BigVGAN) at 22.05 kHz. Requests use `ref_audio` for voice cloning, or an uploaded audio `voice` from `/v1/audio/voices`. Supports emotion conditioning via `emo_audio`, `emo_text`, or `emo_vector` passed in `extra_params`.
+2-stage TTS at 22.05 kHz. Requests use `ref_audio` for voice cloning, or an uploaded audio `voice` from `/v1/audio/voices`. IndexTTS-2.5 uses a multilingual tokenizer, CAMPPlus speaker projection, EnhancedCodec, and code-only Stage 0→1 transfer by default. Both versions support emotion conditioning via `emo_audio`, `emo_text`, or `emo_vector` passed in `extra_params`.
+
+### Prerequisites
+
+IndexTTS-2 uses its existing frontend and does not require the IndexTTS-2.5 text dependencies. Before launching IndexTTS-2.5, install its optional frontend dependencies:
+
+```bash
+pip install 'vllm-omni[indextts2]'
+```
 
 ### Launch
 ```bash
 vllm serve IndexTeam/IndexTTS-2 --omni --trust-remote-code --port 8092
 # or, to pass the bundled deploy config explicitly:
 bash examples/online_serving/text_to_speech/indextts2/run_server.sh
+
+# IndexTTS-2.5 default: use_gpt_latent=false
+MODEL_VERSION=2.5 \
+MODEL=/path/to/indextts-2.5 \
+bash examples/online_serving/text_to_speech/indextts2/run_server.sh
+
 ```
 
 ### Sending requests
@@ -161,14 +176,35 @@ python examples/online_serving/text_to_speech/indextts2/speech_client.py \
     --text "今天心情很好！" \
     --ref-audio /path/to/ref.wav \
     --emo-audio /path/to/happy.wav
+
+# IndexTTS-2.5 Japanese request
+python examples/online_serving/text_to_speech/indextts2/speech_client.py \
+    --model-version 2.5 \
+    --model /path/to/indextts-2.5 \
+    --lang ja \
+    --text "こんにちは、音声合成のテストです。" \
+    --ref-audio /path/to/ref.wav
+
+# IndexTTS-2.5 native speed control
+python examples/online_serving/text_to_speech/indextts2/speech_client.py \
+    --model-version 2.5 \
+    --model /path/to/indextts-2.5 \
+    --speed 2.0 \
+    --text "这是两倍速度的语音合成测试。" \
+    --ref-audio /path/to/ref.wav
 ```
 
 ### Notes
 - Output: 22.05 kHz mono WAV.
-- Provide `ref_audio` on the documented raw request path, or pass `voice` only when it names an uploaded audio voice; IndexTTS-2 does not provide a built-in text-only preset voice.
+- Provide `ref_audio` on the documented raw request path, or pass `voice` only when it names an uploaded audio voice; neither version provides a built-in text-only preset voice.
+- IndexTTS-2.5 request controls `lang` and `text_normalization` are carried in `extra_params`; the client does this for `--model-version 2.5`.
+- IndexTTS-2.5 accepts the public `speed` request field in `[0.5, 2.0]`: `2.0` generates shorter, faster speech and `0.5` generates longer, slower speech. This is native Stage 1 duration control, so serving does not apply the generic playback-speed adjustment a second time. IndexTTS-2 does not use this control.
+- IndexTTS-2.5 accepts language codes such as `zh`, `en`, `zhen` (mixed Chinese/English), `ja`, and `yue`. `Mandarin` is a vLLM-Omni convenience alias for `zh`. Japanese (`ja`) uses `fugashi` tokenization and produces audio, but it does not automatically expand numbers, dates, or percentages; callers should first write those inputs as readable Japanese text.
+- A request `seed` controls Stage 0 AR sampling and per-request CFM noise. Different concurrent batch compositions do not guarantee a bit-identical waveform.
+- IndexTTS-2.5 Stage 0 uses plain vLLM sampling and does not reproduce the official default `num_beams=3` beam search. For parity comparisons, run upstream with `num_beams=1`; output quality can differ from the official beam-search result.
 - Emotion params (`emo_audio`, `emo_text`, `emo_vector`, `emo_alpha`, `use_emo_text`, `use_random`) are passed via the `extra_params` field. Official precedence is `use_emo_text` > `emo_vector` > `emo_audio` > same emotion as the speaker reference.
-- `stream=true` is accepted as an OpenAI-compatible response path, but IndexTTS-2 is not async-chunk streaming; audio is produced after S2Mel receives the full mel-code sequence.
-- Deploy config: `vllm_omni/deploy/indextts2.yaml` (auto-loaded).
+- IndexTTS-2.5 uses `vllm_omni/deploy/indextts2_5.yaml` with the official code-only Stage 0 to Stage 1 contract.
+- For IndexTTS-2.5, set `MODEL` to the local native bundle; asset discovery accepts its nested `checkpoints/` layout.
 
 ---
 
@@ -569,6 +605,13 @@ python qwen3_tts/streaming_speech_client.py \
 The client writes one PCM file per sentence and a matching
 `sentence_XXX_timestamps.json` sidecar.
 
+Non-streaming requests can also ask for timestamps: pass
+`"word_timestamps": true` to `POST /v1/audio/speech` and read the
+`X-Word-Timestamps` response header (JSON list of `{word, start_ms, end_ms}`,
+ASCII-escaped). Past 4 KB the header is replaced by
+`X-Word-Timestamps-Omitted: oversize; bytes=<n>; limit=4096` and the audio
+still returns — use the WebSocket path for long transcripts.
+
 To *see* the alignment instead of reading a JSON sidecar, run the
 word-timestamp Gradio demo (server must be launched with `--forced-aligner`):
 ```bash

examples/online_serving/text_to_speech/indextts2/run_server.sh

@@ -1,20 +1,37 @@
 #!/bin/bash
-# Launch vLLM-Omni server for IndexTTS2
+# Launch vLLM-Omni server for IndexTTS 2.0 or 2.5.
 #
 # Usage from repository root:
 #   examples/online_serving/text_to_speech/indextts2/run_server.sh
-#   CUDA_VISIBLE_DEVICES=0 PORT=8092 MODEL=/path/to/IndexTeam/IndexTTS-2 examples/online_serving/text_to_speech/indextts2/run_server.sh
+#   MODEL_VERSION=2.5 MODEL=/path/to/native/bundle \
+#     examples/online_serving/text_to_speech/indextts2/run_server.sh
 
 set -e
 
 SCRIPT_DIR="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)"
 ROOT_DIR="$(cd -- "$SCRIPT_DIR/../../../.." && pwd)"
 
-MODEL="${MODEL:-IndexTeam/IndexTTS-2}"
+MODEL_VERSION="${MODEL_VERSION:-2.0}"
 PORT="${PORT:-8092}"
-DEPLOY_CONFIG="${DEPLOY_CONFIG:-$ROOT_DIR/vllm_omni/deploy/indextts2.yaml}"
 
-echo "Starting IndexTTS2 server with model: $MODEL"
+if [[ "$MODEL_VERSION" == "2.5" ]]; then
+    if [[ -z "${MODEL:-}" ]]; then
+        echo "Usage: MODEL_VERSION=2.5 MODEL=/path/to/native/bundle bash $0" >&2
+        exit 2
+    fi
+    DEFAULT_DEPLOY_CONFIG="$ROOT_DIR/vllm_omni/deploy/indextts2_5.yaml"
+elif [[ "$MODEL_VERSION" == "2.0" ]]; then
+    MODEL="${MODEL:-IndexTeam/IndexTTS-2}"
+    DEFAULT_DEPLOY_CONFIG="$ROOT_DIR/vllm_omni/deploy/indextts2.yaml"
+else
+    echo "MODEL_VERSION must be 2.0 or 2.5" >&2
+    exit 2
+fi
+
+DEPLOY_CONFIG="${DEPLOY_CONFIG:-$DEFAULT_DEPLOY_CONFIG}"
+
+echo "Starting IndexTTS $MODEL_VERSION server with model: $MODEL"
+echo "Deploy config: $DEPLOY_CONFIG"
 
 FLASHINFER_DISABLE_VERSION_CHECK=1 \
 vllm serve "$MODEL" \

examples/online_serving/text_to_speech/indextts2/speech_client.py

@@ -1,6 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-"""OpenAI-compatible client for IndexTTS2 TTS via /v1/audio/speech endpoint.
+"""OpenAI-compatible client for IndexTTS 2.0/2.5 via /v1/audio/speech.
 
 Examples:
     # With reference audio for voice cloning
@@ -54,13 +54,44 @@ def main() -> None:
     parser.add_argument("--emo-alpha", type=float, default=None, help="Emotion weight in [0, 1]")
     parser.add_argument("--use-emo-text", action="store_true", help="Infer emotion vector from emo-text or text")
     parser.add_argument("--use-random", action="store_true", help="Use random emotion prototypes")
-    parser.add_argument("--model", type=str, default="IndexTeam/IndexTTS-2")
+    parser.add_argument(
+        "--model-version",
+        choices=("2.0", "2.5"),
+        default="2.0",
+    )
+    parser.add_argument("--model", type=str, default=None)
+    parser.add_argument(
+        "--lang",
+        default="zh",
+        help=(
+            "IndexTTS 2.5 language code, for example zh/en/zhen/ja/yue; "
+            "zhen is mixed Chinese/English and Mandarin is a vLLM-Omni alias for zh"
+        ),
+    )
+    parser.add_argument(
+        "--no-text-normalization",
+        action="store_false",
+        dest="text_normalization",
+        help="Disable IndexTTS 2.5 text normalization",
+    )
+    parser.add_argument(
+        "--speed",
+        type=float,
+        default=1.0,
+        help="IndexTTS 2.5 native synthesis speed in [0.5, 2.0]; 2.0 is faster",
+    )
     parser.add_argument("--voice", type=str, default=None, help="Uploaded voice name to use instead of --ref-audio")
     parser.add_argument("--output", type=str, default="output.wav")
     parser.add_argument("--api-base", type=str, default=DEFAULT_API_BASE)
     parser.add_argument("--api-key", type=str, default=DEFAULT_API_KEY)
     parser.add_argument("--response-format", type=str, default="wav")
     args = parser.parse_args()
+    if args.model is None:
+        if args.model_version == "2.5":
+            parser.error("--model is required for IndexTTS 2.5")
+        args.model = "IndexTeam/IndexTTS-2"
+    if args.model_version == "2.5" and not 0.5 <= args.speed <= 2.0:
+        parser.error("IndexTTS 2.5 --speed must be between 0.5 and 2.0")
 
     if not args.ref_audio and not args.voice:
         parser.error("IndexTTS2 requires --ref-audio or --voice for voice cloning")
@@ -97,6 +128,10 @@ def main() -> None:
         extra_params["use_emo_text"] = True
     if args.use_random:
         extra_params["use_random"] = True
+    if args.model_version == "2.5":
+        payload["speed"] = args.speed
+        extra_params["lang"] = args.lang
+        extra_params["text_normalization"] = args.text_normalization
     if extra_params:
         payload["extra_params"] = extra_params
 

mkdocs.yml

@@ -77,6 +77,7 @@ plugins:
         user_guide/diffusion/step_execution.md: user_guide/diffusion/execution_modes.md
         user_guide/diffusion/cpu_offload_diffusion.md: user_guide/diffusion/cpu_offload.md
         user_guide/quantization/index.md: user_guide/quantization/overview.md
+        configuration/pd_disaggregation.md: features/pd_disaggregation.md
         design/feature/diffusion_request_level_batching.md: design/feature/diffusion_continuous_batching.md
         design/feature/diffusion_step_execution.md: design/feature/diffusion_continuous_batching.md
   - glightbox
@@ -111,6 +112,7 @@ plugins:
         - "vllm_omni.entrypoints.openai"  # avoid importing vllm in mkdocs building
         - "vllm_omni.entrypoints.openai.protocol"  # avoid importing vllm in mkdocs building
         - "vllm_omni.experimental"  # avoid importing optional serving deps (openai, fastapi, uvicorn, httpx) in mkdocs building
+        - "vllm_omni.model_executor.models.nemotron_voicechat.nemo_vendored"  # vendored NeMo code; upstream docstrings trip griffe strict warnings
   - mkdocstrings:
       handlers:
         python:

pyproject.toml

@@ -84,6 +84,13 @@ forced-aligner = [
     "qwen-asr>=0.0.6",  # faithful Japanese/Korean word segmentation for the TTS forced aligner
 ]
 
+indextts2 = [
+    "fugashi>=1.2.0",
+    "unidic-lite>=1.0.0",
+    "wetext>=0.0.9; sys_platform != 'linux'",
+    "WeTextProcessing; sys_platform == 'linux'",
+]
+
 # Local-test-only extras
 local = [
     "FlagEmbedding",
@@ -199,6 +206,12 @@ ignore = [
 # These files keep upstream conventions (star imports, long lines, math-symbol
 # argument names) so they stay easy to diff against upstream releases.
 "vllm_omni/model_executor/models/minicpmo_4_5/minicpmo_4_5_omni_llm.py" = ["E402", "E501", "F403", "F405", "N803"]
+# NemotronVoiceChat inference modules vendored from NVIDIA-NeMo/Speech (branch
+# nemotron-labs-voicechat). Numerics-bearing code is kept byte-close to
+# upstream to stay diffable and parity-safe, so upstream style (long lines,
+# legacy typing, whitespace, unused-import shims) is exempted. TID251 and the
+# F-correctness rules still apply.
+"vllm_omni/model_executor/models/nemotron_voicechat/nemo_vendored/**" = ["E501", "E721", "E741", "F401", "F841", "I001", "N803", "N814", "UP006", "UP007", "UP008", "UP011", "UP032", "UP035", "W291", "W293"]
 "vllm_omni/experimental/fullduplex/joyvl/decision/prompts.py" = ["E501"]  # System/summarizer prompts are long prose lines
 
 [tool.ruff.lint.flake8-tidy-imports.banned-api]
@@ -307,6 +320,10 @@ extend-ignore-identifiers-re = [
     ".*[Nn][Dd]3.*",
     ".*updator.*",
     "byt5.*",
+    # NemotronVoiceChat vendored triton kernel: dout/DOUT abbreviate d_out
+    # (output feature dim), not "doubt".
+    ".*dout.*",
+    ".*DOUT.*",
 ]
 extend-ignore-re = [
     "\\.\\.\\.n,d->\\.\\.\\.nd",
@@ -327,3 +344,10 @@ byt5 = "byt5"
 # "seeked" is the standard HTML media element DOM event name (fires when a
 # seek completes), used by the word-timestamps Gradio demo. Not a typo.
 seeked = "seeked"
+# Ryan Leary: author credited in the NemotronVoiceChat vendored NeMo
+# features.py copyright header. A surname, not a typo of "leery".
+Leary = "Leary"
+# IndexTTS 2.5 language codes and speech-emotion-recognition event token.
+fo = "fo"
+ba = "ba"
+SER = "SER"

recipes/IndexTeam/IndexTTS-2_5.md

@@ -0,0 +1,308 @@
+# IndexTTS-2.5 for multilingual voice-cloned TTS on 1x GPU
+
+## Summary
+
+- Vendor: IndexTeam
+- Model: `IndexTeam/IndexTTS-2.5`
+- Task: Multilingual text-to-speech with voice cloning, emotion control, and
+  native speed control
+- Mode: Online serving with the OpenAI-compatible `/v1/audio/speech` API, or
+  offline inference through `Omni`
+- Maintainer: Community
+
+## When to use this recipe
+
+Use this recipe to serve IndexTTS-2.5 as a two-stage TTS system. Stage 0 is an
+autoregressive talker; Stage 1 uses EnhancedCodec, S2Mel CFM/DiT, and BigVGAN
+to produce 22.05 kHz mono speech. Each request supplies synthesis text and
+reference audio, or an uploaded audio voice, for zero-shot voice cloning.
+
+IndexTTS-2.5 adds multilingual text processing and model-native speed control
+to the IndexTTS-2 serving contract.
+
+## References
+
+- Upstream model:
+  [IndexTeam/IndexTTS-2.5 on Hugging Face](https://huggingface.co/IndexTeam/IndexTTS-2.5)
+- Online serving example:
+  [`examples/online_serving/text_to_speech/README.md#indextts-2-and-indextts-25`](../../examples/online_serving/text_to_speech/README.md#indextts-2-and-indextts-25)
+- Offline inference example:
+  [`examples/offline_inference/text_to_speech/README.md#indextts-2-and-indextts-25`](../../examples/offline_inference/text_to_speech/README.md#indextts-2-and-indextts-25)
+- OpenAI-compatible client:
+  [`examples/online_serving/text_to_speech/indextts2/speech_client.py`](../../examples/online_serving/text_to_speech/indextts2/speech_client.py)
+- Standard deploy config:
+  [`vllm_omni/deploy/indextts2_5.yaml`](../../vllm_omni/deploy/indextts2_5.yaml)
+
+## Hardware Support
+
+### GPU
+
+### 1x NVIDIA H20 96GB
+
+#### Environment
+
+- OS: Linux
+- Python: 3.10+
+- Driver / runtime: NVIDIA CUDA environment
+- vLLM version: Match the repository requirements for your checkout
+- vLLM-Omni version or commit: Use the commit you are deploying from
+
+Install vLLM-Omni with the IndexTTS text-processing dependencies:
+
+```bash
+pip install 'vllm-omni[indextts2]'
+```
+
+Obtain the native IndexTTS-2.5 bundle and point `MODEL` at its root. The model
+loader accepts the upstream nested `checkpoints/` layout.
+
+#### Command
+
+Start the standard code-only server from the repository root:
+
+```bash
+MODEL_VERSION=2.5 \
+MODEL=/path/to/indextts-2.5 \
+bash examples/online_serving/text_to_speech/indextts2/run_server.sh
+```
+
+This selects `vllm_omni/deploy/indextts2_5.yaml`, which uses
+`use_gpt_latent=false`. To launch directly:
+
+```bash
+vllm serve /path/to/indextts-2.5 \
+  --omni \
+  --trust-remote-code \
+  --port 8092 \
+  --deploy-config vllm_omni/deploy/indextts2_5.yaml
+```
+
+#### Optional: NVIDIA MPS for higher single-GPU throughput
+
+Stage 0 and Stage 1 run in separate processes on the same GPU. Enabling
+NVIDIA Multi-Process Service (MPS) can improve steady-state throughput by
+allowing work from both stages to overlap more effectively. MPS is an opt-in
+deployment optimization: this recipe does not start it automatically, and the
+gain depends on the request mix and concurrency, so benchmark it with the
+intended workload before enabling it in production.
+
+Pay attention to CUDA device renumbering. For example, if the MPS control
+daemon is started with physical `CUDA_VISIBLE_DEVICES=1`, that GPU is exposed
+to MPS clients as logical device `0`; launch the server with
+`CUDA_VISIBLE_DEVICES=0`, not `1`. Use a dedicated MPS pipe/log directory when
+the host is shared with other services, and stop only the MPS daemon owned by
+this deployment.
+
+#### Verification
+
+Send a multilingual voice-cloning request with the bundled client. Local audio
+paths are converted to base64 data URLs before transmission:
+
+```bash
+python examples/online_serving/text_to_speech/indextts2/speech_client.py \
+  --api-base http://localhost:8092 \
+  --model-version 2.5 \
+  --model /path/to/indextts-2.5 \
+  --lang zh \
+  --text "你好，这是 IndexTTS-2.5 语音合成测试。" \
+  --ref-audio /path/to/reference.wav \
+  --output indextts2_5.wav
+```
+
+##### Native speed control
+
+The public `speed` field is handled natively. Its accepted range is
+`[0.5, 2.0]`; values above `1.0` produce shorter, faster speech. The serving
+adapter converts the API convention to the model-native control with
+`duration_factor = 1.0 / speed`, so `speed=1.25` uses
+`duration_factor=0.8`:
+
+```bash
+python examples/online_serving/text_to_speech/indextts2/speech_client.py \
+  --api-base http://localhost:8092 \
+  --model-version 2.5 \
+  --model /path/to/indextts-2.5 \
+  --speed 1.25 \
+  --text "这是使用模型原生语速控制的测试。" \
+  --ref-audio /path/to/reference.wav \
+  --output indextts2_5_speed.wav
+```
+
+##### Upload and reuse a named voice
+
+Upload a reference recording once when callers should reuse the same voice
+without sending `ref_audio` on every request. `consent` is required by the
+voice-storage API:
+
+```bash
+curl -X POST http://localhost:8092/v1/audio/voices \
+  -F "audio_sample=@/path/to/reference.wav" \
+  -F "consent=user-consent-id" \
+  -F "name=indextts_demo_voice" \
+  -F "speaker_description=IndexTTS-2.5 demonstration voice"
+```
+
+Confirm that the voice was stored, then synthesize with its name instead of a
+reference-audio payload:
+
+```bash
+curl http://localhost:8092/v1/audio/voices
+
+python examples/online_serving/text_to_speech/indextts2/speech_client.py \
+  --api-base http://localhost:8092 \
+  --model-version 2.5 \
+  --model /path/to/indextts-2.5 \
+  --voice indextts_demo_voice \
+  --lang zh \
+  --text "你好，这是复用已上传音色的语音合成测试。" \
+  --output indextts2_5_named_voice.wav
+```
+
+There are no built-in text-only preset voices. Each named IndexTTS voice must
+first be created from an uploaded reference recording.
+
+##### Multilingual synthesis and text normalization
+
+The client accepts `zh` (Mandarin), `en` (English), `zhen` (mixed Chinese and
+English), `ja` (Japanese), and `yue` (Cantonese). The following commands use
+the same reference voice across all supported language modes:
+
+```bash
+CLIENT=examples/online_serving/text_to_speech/indextts2/speech_client.py
+MODEL=/path/to/indextts-2.5
+REF_AUDIO=/path/to/reference.wav
+
+python "$CLIENT" --model-version 2.5 --model "$MODEL" \
+  --ref-audio "$REF_AUDIO" --lang zh \
+  --text "你好，这是中文语音合成测试。" --output indextts2_5_zh.wav
+
+python "$CLIENT" --model-version 2.5 --model "$MODEL" \
+  --ref-audio "$REF_AUDIO" --lang en \
+  --text "Hello, this is an English speech synthesis test." \
+  --output indextts2_5_en.wav
+
+python "$CLIENT" --model-version 2.5 --model "$MODEL" \
+  --ref-audio "$REF_AUDIO" --lang zhen \
+  --text "Hello，欢迎使用 IndexTTS 二点五。" --output indextts2_5_zhen.wav
+
+python "$CLIENT" --model-version 2.5 --model "$MODEL" \
+  --ref-audio "$REF_AUDIO" --lang ja \
+  --text "こんにちは、これは日本語の音声合成テストです。" \
+  --output indextts2_5_ja.wav
+
+python "$CLIENT" --model-version 2.5 --model "$MODEL" \
+  --ref-audio "$REF_AUDIO" --lang yue \
+  --text "你好，呢段係粵語語音合成測試。" --output indextts2_5_yue.wav
+```
+
+Text normalization is enabled by default. Disable it only when the input is
+already written exactly as it should be spoken:
+
+```bash
+python examples/online_serving/text_to_speech/indextts2/speech_client.py \
+  --model-version 2.5 --model /path/to/indextts-2.5 \
+  --ref-audio /path/to/reference.wav --lang en --no-text-normalization \
+  --text "The temperature is twenty five degrees." \
+  --output indextts2_5_pre_normalized.wav
+```
+
+For raw HTTP requests, put the same language and text-normalization controls
+under `extra_params`:
+
+```bash
+curl http://localhost:8092/v1/audio/speech \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": "/path/to/indextts-2.5",
+    "input": "This is an IndexTTS-2.5 test.",
+    "response_format": "wav",
+    "speed": 1.0,
+    "ref_audio": "data:audio/wav;base64,<BASE64_ENCODED_AUDIO>",
+    "extra_params": {
+      "lang": "en",
+      "text_normalization": true
+    }
+  }' \
+  --output indextts2_5_en.wav
+```
+
+##### Emotion control
+
+IndexTTS-2.5 supports an explicit eight-value emotion vector, an emotion text
+description, or a separate emotion-reference recording. Use one source at a
+time. Vector values are ordered as `happy`, `angry`, `sad`, `afraid`,
+`disgusted`, `melancholic`, `surprised`, and `calm`; each value must be in
+`[0, 1.2]`. `emo_alpha` controls the strength in `[0, 1]`.
+
+Use an explicit emotion vector for deterministic control:
+
+```bash
+python examples/online_serving/text_to_speech/indextts2/speech_client.py \
+  --model-version 2.5 --model /path/to/indextts-2.5 \
+  --ref-audio /path/to/reference.wav --lang zh \
+  --text "今天真是令人开心的一天！" \
+  --emo-vector 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 \
+  --emo-alpha 0.8 --output indextts2_5_happy_vector.wav
+```
+
+Infer the emotion from a natural-language description with `use_emo_text`:
+
+```bash
+python examples/online_serving/text_to_speech/indextts2/speech_client.py \
+  --model-version 2.5 --model /path/to/indextts-2.5 \
+  --ref-audio /path/to/reference.wav --lang zh \
+  --text "我们终于完成了这项工作！" \
+  --use-emo-text --emo-text "开心、兴奋而且充满活力" \
+  --emo-alpha 0.8 --output indextts2_5_happy_text.wav
+```
+
+Transfer emotion from a recording while retaining the speaker identity from
+`ref_audio`:
+
+```bash
+python examples/online_serving/text_to_speech/indextts2/speech_client.py \
+  --model-version 2.5 --model /path/to/indextts-2.5 \
+  --ref-audio /path/to/reference.wav --lang zh \
+  --text "请用情绪参考音频中的表达方式朗读这句话。" \
+  --emo-audio /path/to/emotion_reference.wav \
+  --emo-alpha 0.8 --output indextts2_5_emotion_audio.wav
+```
+
+If multiple emotion sources are supplied, the official precedence is
+`use_emo_text` > `emo_vector` > `emo_audio` > the emotion in the speaker
+reference. `--use-random` is also available for exploratory random emotion
+prototypes, but should not be used when reproducible conditioning is required.
+
+For offline inference:
+
+```bash
+python examples/offline_inference/text_to_speech/indextts2/end2end.py \
+  --model /path/to/indextts-2.5 \
+  --model-version 2.5 \
+  --lang zh \
+  --speed 1.0 \
+  --text "你好，这是离线语音合成测试。" \
+  --ref-audio /path/to/reference.wav
+```
+
+#### Notes
+
+- Hardware scope: the standard recipe has been exercised on one NVIDIA H200.
+  It does not claim an unverified throughput or quality result.
+- Audio output: 22.05 kHz mono WAV.
+- Voice cloning: reference audio is required on the documented raw request
+  path. Alternatively, `voice` may name an uploaded audio voice; there is no
+  built-in text-only preset voice.
+- Native speed: serving maps `speed` to the model's duration factor and skips
+  generic waveform speed adjustment, so speed is applied exactly once.
+- Languages: common codes include `zh`, `en`, `zhen` (mixed Chinese/English),
+  `ja`, and `yue`; `Mandarin` is accepted as an alias for `zh`.
+- Japanese normalization: `ja` tokenization does not automatically expand
+  numbers, dates, or percentages. Write these inputs as readable Japanese text
+  before inference.
+- Emotion controls: `use_emo_text`, `emo_vector`, and `emo_audio` are
+  alternative conditioning modes. Their precedence is `use_emo_text` >
+  `emo_vector` > `emo_audio` > the speaker-reference emotion.
+- Sampling difference: Stage 0 uses plain vLLM sampling, not the upstream
+  default `num_beams=3` beam search. Use upstream `num_beams=1` for parity
+  comparisons; output may differ from the official beam-search result.

recipes/MiniMaxAI/MiniMax-H3-RTX-PRO-5000.md

@@ -0,0 +1,198 @@
+# MiniMax-H3 on RTX PRO 5000 Blackwell GPUs
+
+This recipe runs MiniMax-H3 in BF16 on 72 GiB RTX PRO 5000 Blackwell GPUs. It
+contains the validated two-GPU DLO configuration and the recommended resident
+configurations: TP1 x Ulysses2 with 20 resident layers on two GPUs, TP2 x
+Ulysses2 on four GPUs, and TP4 x Ulysses2 on eight GPUs. The four- and
+eight-GPU routes do not require offload.
+
+## Capacity requirements
+
+| Resource | Requirement |
+| --- | ---: |
+| GPUs | 2, 4, or 8 x RTX PRO 5000 Blackwell |
+| GPU HBM | 72 GiB per GPU |
+| Checkpoint storage | 135 GiB per partition |
+| Available system RAM | 200 GiB minimum |
+| Recommended system RAM | 384 GiB |
+
+`FL2VA` and `Ref2VA` are separate checkpoint partitions. Start one server at a
+time on a host sized for the minimum system-memory requirement.
+
+## PCIe topology and GPU order
+
+RTX PRO 5000 does not provide NVLink. Before starting the server, identify the
+closest PCIe pairs and their NUMA affinity:
+
+```bash
+nvidia-smi topo -m
+nvidia-smi nvlink -s
+```
+
+For two GPUs, select the closest PCIe pair on one NUMA node. For four GPUs,
+select all cards from one NUMA node. On the validated host,
+physical GPUs `(0,1)` and `(2,3)` are the two `PXB` pairs on NUMA node 0. The
+order `CUDA_VISIBLE_DEVICES=0,2,1,3` maps the Ulysses groups to those local
+pairs. For eight GPUs, the validated order is `0,4,1,5,2,6,3,7`, with host
+memory interleaved across NUMA nodes 0 and 1. Do not copy these IDs blindly:
+reproduce the same PCIe and NUMA relationships on the target host.
+
+## Recommended serving configurations
+
+### Two GPUs
+
+Two 72 GiB cards require distributed layerwise offload. The validated route
+uses TP1 x Ulysses2, keeps 20 leading DiT layers resident, and streams
+rank-local weights without AllGather. Eager execution avoids regional-compile
+instability on this offload path.
+
+```bash
+export MODEL_ROOT=/path/to/MiniMax-H3
+export MODEL="${MODEL_ROOT}/FL2VA"
+export PORT=8091
+
+CUDA_VISIBLE_DEVICES=0,1 \
+VLLM_WORKER_MULTIPROC_METHOD=spawn \
+VLLM_OMNI_VIDEO_SYNC_TIMEOUT=1800 \
+numactl --cpunodebind=0 --membind=0 \
+vllm serve "${MODEL}" \
+  --omni \
+  --host 0.0.0.0 \
+  --port "${PORT}" \
+  --trust-remote-code \
+  --num-gpus 2 \
+  --tensor-parallel-size 1 \
+  --usp 2 \
+  --ring 1 \
+  --text-encoder-tp-size 2 \
+  --vae-patch-parallel-size 2 \
+  --vae-parallel-mode tile \
+  --vae-use-tiling \
+  --diffusion-attention-backend CUDNN_ATTN \
+  --enable-distributed-layerwise-offload \
+  --dlo-no-use-allgather \
+  --dlo-resident-layers 20 \
+  --enforce-eager
+```
+
+### Four GPUs
+
+The validated baseline uses TP2 x Ulysses2, text-encoder TP4, VAE patch
+parallelism 4, and explicit cuDNN BF16 attention. Selecting the backend
+explicitly keeps the recipe independent of platform-default backend changes.
+
+```bash
+export MODEL_ROOT=/path/to/MiniMax-H3
+export MODEL="${MODEL_ROOT}/FL2VA"
+export PORT=8091
+
+CUDA_VISIBLE_DEVICES=0,2,1,3 \
+VLLM_WORKER_MULTIPROC_METHOD=spawn \
+VLLM_OMNI_VIDEO_SYNC_TIMEOUT=1800 \
+numactl --cpunodebind=0 --membind=0 \
+vllm serve "${MODEL}" \
+  --omni \
+  --host 0.0.0.0 \
+  --port "${PORT}" \
+  --trust-remote-code \
+  --num-gpus 4 \
+  --tensor-parallel-size 2 \
+  --usp 2 \
+  --ring 1 \
+  --text-encoder-tp-size 4 \
+  --vae-patch-parallel-size 4 \
+  --vae-parallel-mode tile \
+  --vae-use-tiling \
+  --diffusion-attention-backend CUDNN_ATTN
+```
+
+### Eight GPUs
+
+The recommended eight-GPU route uses TP4 x Ulysses2, text-encoder TP8, VAE
+patch parallelism 8, and host-memory interleaving across both NUMA nodes.
+
+```bash
+export MODEL_ROOT=/path/to/MiniMax-H3
+export MODEL="${MODEL_ROOT}/FL2VA"
+export PORT=8091
+
+CUDA_VISIBLE_DEVICES=0,4,1,5,2,6,3,7 \
+VLLM_WORKER_MULTIPROC_METHOD=spawn \
+VLLM_OMNI_VIDEO_SYNC_TIMEOUT=1800 \
+numactl --interleave=0,1 \
+vllm serve "${MODEL}" \
+  --omni \
+  --host 0.0.0.0 \
+  --port "${PORT}" \
+  --trust-remote-code \
+  --num-gpus 8 \
+  --tensor-parallel-size 4 \
+  --usp 2 \
+  --ring 1 \
+  --text-encoder-tp-size 8 \
+  --vae-patch-parallel-size 8 \
+  --vae-parallel-mode tile \
+  --vae-use-tiling \
+  --diffusion-attention-backend CUDNN_ATTN
+```
+
+For the four- and eight-GPU resident routes, do not add `--enforce-eager`.
+Warm the server once before measuring so regional compilation is outside the
+measured request. The two-GPU DLO route intentionally remains eager.
+
+For Ref2VA, stop the FL2VA server and restart the same command with
+`MODEL="${MODEL_ROOT}/Ref2VA"`.
+
+## Target-hardware validation
+
+All three configurations were exercised on a PCIe-only, dual-socket host with
+eight RTX PRO 5000 GPUs. The run used PyTorch 2.11.0+cu130, CUDA 13.0, driver
+580.95.05, 1344x768 output, 124 frames, and two warmups.
+
+The four-GPU route also received a five-step profiling validation for GPU
+balance:
+
+| Measurement | Result |
+| --- | ---: |
+| Maximum externally sampled peak | 69,219 MiB (67.60 GiB) per GPU |
+| T2VA worker-reported peak | 65,314 MiB |
+| First-frame FL2VA worker-reported peak | 66,780 MiB |
+| T2VA maximum GPU kernel-time deviation | 0.56% |
+| First-frame FL2VA maximum GPU kernel-time deviation | 0.09% |
+
+The recommended routes produced the following 50-step results. MiniMax-H3
+requested 50 denoise steps and executed 49 denoise updates, so per-step latency
+is `denoise / 49`.
+
+| GPUs | Workload | Parallelism | E2E (s) | Text encode (s) | Denoise (s) | VAE decode (s) | Per step (ms) | Peak memory (GiB) |
+| ---: | --- | --- | ---: | ---: | ---: | ---: | ---: | ---: |
+| 2 | T2VA | TP1 x Ulysses2, DLO resident 20 | 515.57 | 1.19 | 504.69 | 8.79 | 10,300 | 36.38 |
+| 2 | FL2VA first-frame I2VA | TP1 x Ulysses2, DLO resident 20 | 553.45 | 1.27 | 541.90 | 8.76 | 11,059 | 36.38 |
+| 4 | T2VA | TP2 x Ulysses2 | 284.01 | 0.03 | 278.73 | 4.36 | 5,688 | 67.58 |
+| 4 | FL2VA first-frame I2VA | TP2 x Ulysses2 | 305.72 | 0.28 | 299.85 | 4.36 | 6,119 | 67.58 |
+| 8 | T2VA | TP4 x Ulysses2 | 163.20 | 0.04 | 159.40 | 2.58 | 3,253 | 45.83 |
+| 8 | FL2VA first-frame I2VA | TP4 x Ulysses2 | 171.62 | 0.27 | 167.25 | 2.57 | 3,413 | 45.83 |
+
+Peak memory is the maximum per-GPU value sampled externally with
+`nvidia-smi`. The two-GPU route leaves about 34.7 GiB below the reported
+73,415 MiB device capacity, the four-GPU route leaves about 4.1 GiB, and the
+eight-GPU route leaves about 25.9 GiB. Re-measure memory for longer reference
+inputs, concurrency greater than one, or a different output shape.
+
+## T2VA request example
+
+```bash
+export API_URL="http://127.0.0.1:${PORT}/v1/videos/sync"
+
+curl -sS --max-time 1800 -X POST "${API_URL}" \
+  -F 'prompt=At night, three cats march into a bedroom playing tiny brass instruments, then abruptly file out, with synchronized room ambience.' \
+  -F 'width=1344' \
+  -F 'height=768' \
+  -F 'aspect_ratio=16:9' \
+  -F 'fps=24' \
+  -F 'num_inference_steps=50' \
+  -F 'flow_shift=12' \
+  -F 'seed=1101' \
+  -F 'extra_params={"task":"t2va","duration":5.0,"audio_flow_shift":3.0}' \
+  -o t2va.mp4
+```

recipes/NVIDIA/NemotronLabs-VoiceChat.md

@@ -0,0 +1,113 @@
+# NemotronLabs VoiceChat 11B
+
+> Offline speech-to-speech: a 16 kHz user utterance in, the agent's spoken reply
+> (text + 22.05 kHz WAV) out, on a 3-stage vLLM-Omni pipeline.
+
+## Summary
+
+- Vendor: NVIDIA
+- Model: [`nvidia/NVIDIA-NemotronLabs-VoiceChat-11B`](https://huggingface.co/nvidia/NVIDIA-NemotronLabs-VoiceChat-11B)
+- Task: Speech-to-speech voice chat. The model runs a frame-locked 12.5 Hz
+  timeline: a Conformer + NemotronH hybrid-Mamba thinker emits one text token
+  per acoustic frame, a Gemma3-1B EAR-TTS talker turns the text timeline into
+  31-quantizer RVQ code stacks, and an RVQ-VAE codec decodes them to audio.
+- Mode: Offline single-turn inference (batch=1). Online/duplex serving,
+  function calling, and batch>1 are documented follow-ups.
+- Maintainer: [`@yuekaizhang`](https://github.com/yuekaizhang)
+
+## When to use this recipe
+
+Use this recipe to run a single-turn voice-chat exchange with
+`NVIDIA-NemotronLabs-VoiceChat-11B` on one GPU: you provide a user utterance as
+a WAV file (any sample rate; it is resampled to 16 kHz mono) plus an optional
+spoken-style system prompt, and get back the agent's reply as text and a
+22.05 kHz WAV. The integration is NeMo-free at runtime — the perception
+Conformer, EAR-TTS talker, and RVQ-VAE codec are vendored, dependency-stripped
+NeMo modules (`nemo_vendored/`), so no `nemo_toolkit` install is needed.
+
+## References
+
+- Offline example:
+  [`examples/offline_inference/nemotron_voicechat/end2end.py`](../../examples/offline_inference/nemotron_voicechat/end2end.py)
+- Model modules (thinker / talker / code2wav / vendored NeMo):
+  [`vllm_omni/model_executor/models/nemotron_voicechat/`](../../vllm_omni/model_executor/models/nemotron_voicechat/)
+- Staged pipeline config:
+  [`vllm_omni/deploy/nemotron_labs_voicechat.yaml`](../../vllm_omni/deploy/nemotron_labs_voicechat.yaml)
+- Upstream model card:
+  [`nvidia/NVIDIA-NemotronLabs-VoiceChat-11B`](https://huggingface.co/nvidia/NVIDIA-NemotronLabs-VoiceChat-11B)
+- Reference implementation: NVIDIA-NeMo/Speech, branch `nemotron-labs-voicechat`
+
+## Pipeline
+
+| stage | arch | dtype | role |
+|---|---|---|---|
+| 0 thinker | `NemotronVoiceChatThinkerForConditionalGeneration` (LLM_AR) | fp32 | WAV + system prompt -> frame-locked text-token timeline (+ function channel) |
+| 1 talker | `NemotronVoiceChatTalker` (LLM_AR) | fp32 | text timeline -> 31-quantizer RVQ code stacks (one per 80 ms frame) |
+| 2 code2wav | `NemotronVoiceChatCode2Wav` (LLM_GENERATION) | fp32 | RVQ-VAE decode -> 22.05 kHz PCM |
+
+Stages 0/1 default to fp32 for exact parity with the NeMo reference
+implementation (greedy decoding matches it token for token on the acceptance
+fixture). The deploy yaml documents a ~2x-faster bf16 thinker option whose
+output stayed within one word of the reference in testing.
+
+## Hardware Support
+
+## GPU
+
+### 1x H100 80GB
+
+#### Environment
+
+- OS: Linux
+- Python: 3.12
+- vLLM version: 0.26.0
+- vLLM-Omni version or commit: this PR / current `main`
+
+#### Command
+
+```bash
+# Tokenizer: the checkpoint ships no HF tokenizer; it resolves from the
+# nvidia/NVIDIA-Nemotron-Nano-9B-v2 HF id automatically. For air-gapped runs,
+# point NEMOTRON_VOICECHAT_LLM_PATH at a local snapshot of that repo instead.
+python examples/offline_inference/nemotron_voicechat/end2end.py \
+    --checkpoint /path/to/NVIDIA-NemotronLabs-VoiceChat-11B \
+    --wav /path/to/user_question.wav \
+    --output-dir results/nemotron_voicechat
+```
+
+#### Verification
+
+```bash
+ls results/nemotron_voicechat
+# <stem>_output.txt          the agent reply as text
+# <stem>_output.wav          the agent reply as 22.05 kHz audio
+# <stem>_text_tokens.json    the frame-locked text-token timeline
+```
+
+The reply text should read as a coherent spoken-style answer to the question in
+the input WAV, and the WAV should transcribe to (approximately) the same text
+with any ASR model.
+
+#### Notes
+
+- Memory usage: the shipped yaml runs all three stages on one GPU
+  (`gpu_memory_utilization` 0.62 / 0.12 / 0.06); peak usage is dominated by the
+  fp32 thinker. The fp32 default has a hard floor of roughly 43 GB of thinker
+  weights alone (9B backbone + 587M `embed_tokens` + 587M `function_head` +
+  0.6B Conformer), so 48 GB cards cannot run it — use the bf16 thinker option
+  documented in the deploy yaml on anything smaller than an 80 GB part.
+- Input sizing: the timeline is frame-locked, so the reply budget IS the input
+  duration. The acoustic channel trails the text channel; if the WAV does not
+  carry enough trailing silence for the reply to finish, the spoken answer is
+  truncated silently. Leave generous trailing silence (a question ending at
+  ~4.5 s truncated in an 8 s WAV but completed cleanly in 16 s); the offline
+  example warns when the text channel is still speaking near the last frame.
+- Key flags: sampling is greedy end to end. The thinker is frame-locked —
+  `max_tokens` equals the acoustic frame count with `ignore_eos=True`. Do NOT
+  set `min_tokens` on the thinker: the tokenizer's EOS token is also the
+  frame-locked PAD/silence token, so masking it forces the model to babble
+  instead of pausing.
+- The talker's `max_tokens` is 16383 (its stage prompt is one placeholder
+  token, and the stage context is 16384).
+- Known limitations: batch=1 only; offline single turn (no online/duplex
+  serving yet); the function-call channel is decoded but not yet acted on.

recipes/README.md

@@ -50,6 +50,7 @@ recipes/
 | [`inclusionAI/Ming-flash-omni-2.0.md`](./inclusionAI/Ming-flash-omni-2.0.md) | Online serving for multimodal chat + standalone TTS | 4x H100 / 1x H100 80GB |
 | [`inclusionAI/Ming-omni-tts.md`](./inclusionAI/Ming-omni-tts.md) | Offline + online dense Ming TTS/audio generation | 1x H100 80GB / 1x AMD MI300X (ROCm 7.2) |
 | [`IndexTeam/IndexTTS-2.md`](./IndexTeam/IndexTTS-2.md) | Online serving for voice-cloned TTS with optional emotion control | 1x L4 24GB or larger CUDA GPU |
+| [`IndexTeam/IndexTTS-2_5.md`](./IndexTeam/IndexTTS-2_5.md) | Online + offline multilingual voice-cloned TTS with native speed and emotion control | 1x NVIDIA H20 96GB |
 | [`MiniMaxAI/MiniMax-H3.md`](./MiniMaxAI/MiniMax-H3.md) | T2VA, FL2VA, image+audio Ref2VA, and multi-video Ref2VA serving | 1x GPU with CPU offload / 4x B300 |
 | [`MiniMaxAI/MiniMax-H3-MUSA.md`](./MiniMaxAI/MiniMax-H3-MUSA.md) | MiniMax H3 T2VA, FL2VA, and Ref2VA on MUSA | 4x MTT S5000 (TP4 Ref2VA profile) |
 | [`krea/Krea-2.md`](./krea/Krea-2.md) | Text-to-image (Turbo + Raw), offline + online, with LoRA | 1x H100 80GB |
@@ -65,6 +66,7 @@ recipes/
 | [`OpenBMB/VoxCPM2.md`](./OpenBMB/VoxCPM2.md) | Online + offline TTS with native AR pipeline (48 kHz, 30+ languages) | 1x RTX 4090 24GB |
 | [`OpenMOSS/MOSS-TTS.md`](./OpenMOSS/MOSS-TTS.md) | Online + offline multilingual TTS (MOSS-TTS family, 8B) | 1x H100 80GB |
 | [`NVIDIA/Nemotron-Labs-Audex.md`](./NVIDIA/Nemotron-Labs-Audex.md) | TTS / text-to-audio / audio understanding / cascaded S2S (Audex 2B + 30B-A3B) | 1x H100 80GB |
+| [`NVIDIA/NemotronLabs-VoiceChat.md`](./NVIDIA/NemotronLabs-VoiceChat.md) | Offline speech-to-speech voice chat (11B, frame-locked 12.5 Hz, 3-stage thinker/talker/code2wav) | 1x H100 80GB |
 | [`Qwen/Qwen-Image.md`](./Qwen/Qwen-Image.md) | Text-to-image serving with step-wise continuous batching replay and ModelOpt mixed FP8/NVFP4 | 1x A100 80GB / 2x B200 |
 | [`Qwen/Qwen-Image-2512.md`](./Qwen/Qwen-Image-2512.md) | Text-to-image serving with step-wise continuous batching replay and ModelOpt FP8 / mixed FP8/NVFP4 | 1x A800 80GB / 2x B200 |
 | [`Qwen/Qwen-Image-Edit.md`](./Qwen/Qwen-Image-Edit.md) | Text-guided single-image editing | 1x or 2x H200 141GB |

recipes/Robbyant/LingBot-Video.md

@@ -27,7 +27,7 @@ parallel, cache, quantized, or expert-kernel backends.
 - Dense checkpoint: <https://huggingface.co/robbyant/lingbot-video-dense-1.3b>
 - MoE checkpoint: <https://huggingface.co/robbyant/lingbot-video-moe-30b-a3b>
 - Upstream project: <https://github.com/Robbyant/lingbot-video>
-- Related offline example: [`examples/offline_inference/text_to_video/text_to_video_lingbot.py`](../../examples/offline_inference/text_to_video/text_to_video_lingbot.py)
+- Related offline example: [`examples/offline_inference/text_to_video/text_to_video.py`](../../examples/offline_inference/text_to_video/text_to_video.py)
 - Related online video API docs: [`docs/serving/videos_api.md`](../../docs/serving/videos_api.md)
 
 ## Hardware Support
@@ -52,7 +52,7 @@ The MoE path is validated with BF16 expert weights.
 
 ```bash
 CUDA_VISIBLE_DEVICES=0 \
-python examples/offline_inference/text_to_video/text_to_video_lingbot.py \
+python examples/offline_inference/text_to_video/text_to_video.py \
   --model robbyant/lingbot-video-dense-1.3b \
   --prompt "a robotic arm picks up a red block" \
   --output lingbot_t2v.mp4 \
@@ -70,7 +70,7 @@ python examples/offline_inference/text_to_video/text_to_video_lingbot.py \
 
 ```bash
 CUDA_VISIBLE_DEVICES=0 \
-python examples/offline_inference/text_to_video/text_to_video_lingbot.py \
+python examples/offline_inference/text_to_video/text_to_video.py \
   --model robbyant/lingbot-video-moe-30b-a3b \
   --prompt "a robotic arm picks up a red block" \
   --output lingbot_moe_t2v.mp4 \

recipes/Tencent/HunyuanImage-3.0-Instruct.md

@@ -82,7 +82,7 @@ example and the DiT sections below.
 - Optional environment variables:
 
 ```bash
-export CACHE_DIT_VERSION=1.3.0
+export CACHE_DIT_VERSION=1.5.0
 ```
 
 HunyuanImage-3.0 sets the diffusion attention backend to `TORCH_SDPA`

requirements/common.txt

@@ -16,7 +16,7 @@ diffusers==0.38.0
 safetensors>=0.8.0
 accelerate==1.12.0
 soundfile>=0.13.1
-cache-dit==1.3.0
+cache-dit==1.5.0
 tqdm>=4.66.0
 torchsde>=0.2.6
 openai-whisper>=20250625
@@ -34,3 +34,5 @@ pydantic-settings>=2.0.0
 # base image does not ship it, so declare it explicitly. The loader also guards
 # this import so a missing gguf no longer breaks all diffusion model loading.
 gguf>=0.10.0
+# Required for running the Cosmos3 model with guardrails enabled (default behavior).
+cosmos-guardrail>=0.3.1

tests/benchmarks/test_bench_tts_cli.py

@@ -51,6 +51,14 @@ def test_load_model_configs(model_configs_path: Path) -> None:
     assert configs["test/ModelA"]["supported_tasks"] == ["voice_clone", "default_voice"]
 
 
+def test_indextts25_is_registered_in_shared_model_configs() -> None:
+    config_path = Path(bench_tts.__file__).with_name("model_configs.yaml")
+    config = bench_tts.load_model_configs(config_path)["IndexTeam/IndexTTS-2.5"]
+
+    assert config["supported_tasks"] == ["voice_clone"]
+    assert config["task_extra_body"]["voice_clone"]["extra_params"]["lang"] == "en"
+
+
 def test_build_bench_args_voice_clone(model_configs_path: Path) -> None:
     configs = bench_tts.load_model_configs(model_configs_path)
     cmd = bench_tts.build_bench_args(
@@ -120,6 +128,34 @@ def test_build_bench_args_wer_eval_adds_flag(model_configs_path: Path) -> None:
     assert "--seed-tts-wer-eval" in cmd
 
 
+def test_build_bench_args_supports_local_model_and_shared_sweep_options(model_configs_path: Path) -> None:
+    configs = bench_tts.load_model_configs(model_configs_path)
+    cmd = bench_tts.build_bench_args(
+        host="localhost",
+        port=8092,
+        model="test/ModelA",
+        served_model_name="/models/indextts25",
+        task="voice_clone",
+        model_cfg=configs["test/ModelA"],
+        locale="en",
+        num_prompts=500,
+        num_warmups=5,
+        request_seed=42,
+        concurrency=8,
+        dataset_path="/data/seed-tts",
+        wer_eval=False,
+        output_dir=None,
+        result_filename=None,
+        extra_cli_args=["--", "--tokenizer", "/models/indextts25/qwen0.6bemo4-merge"],
+    )
+
+    assert cmd[cmd.index("--model") + 1] == "/models/indextts25"
+    assert cmd[cmd.index("--num-warmups") + 1] == "5"
+    assert cmd[-2:] == ["--tokenizer", "/models/indextts25/qwen0.6bemo4-merge"]
+    extra_body = json.loads(cmd[cmd.index("--extra-body") + 1])
+    assert extra_body == {"task_type": "Base", "seed": 42}
+
+
 def test_unsupported_task_exits(model_configs_path: Path, capsys: pytest.CaptureFixture, mocker) -> None:
     # ModelB does not support voice_design
     mocker.patch.object(

tests/benchmarks/test_daily_omni_local_source.py

@@ -0,0 +1,92 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Offline Daily-Omni sources: local mirrors must not fall back to the Hub id."""
+
+import json
+import os
+import tarfile
+from pathlib import Path
+
+import pytest
+
+from vllm_omni.benchmarks.data_modules.daily_omni_dataset import (
+    daily_omni_local_qa_json,
+    daily_omni_local_videos_dir,
+    resolve_daily_omni_local_root,
+)
+
+pytestmark = [pytest.mark.core_model, pytest.mark.cpu, pytest.mark.benchmark]
+
+
+def _write_qa_json(root: Path) -> Path:
+    qa = root / "qa.json"
+    qa.write_text(json.dumps([{"Question": "q", "Answer": "A", "video_id": "vid0"}]), encoding="utf-8")
+    return qa
+
+
+def test_resolve_local_root_returns_none_for_hub_id() -> None:
+    assert resolve_daily_omni_local_root("liarliar/Daily-Omni") is None
+    assert resolve_daily_omni_local_root("") is None
+    assert resolve_daily_omni_local_root(None) is None
+
+
+def test_plain_local_mirror(tmp_path: Path) -> None:
+    _write_qa_json(tmp_path)
+    (tmp_path / "Videos" / "vid0").mkdir(parents=True)
+
+    root = resolve_daily_omni_local_root(str(tmp_path))
+    assert root == tmp_path.resolve()
+    assert daily_omni_local_qa_json(root) == (tmp_path / "qa.json").resolve()
+    assert daily_omni_local_videos_dir(root) == root / "Videos"
+
+
+def test_hf_cache_dir_resolves_to_snapshot(tmp_path: Path) -> None:
+    cache_dir = tmp_path / "datasets--liarliar--Daily-Omni"
+    snapshot = cache_dir / "snapshots" / "deadbeef"
+    snapshot.mkdir(parents=True)
+    _write_qa_json(snapshot)
+    (cache_dir / "refs").mkdir()
+    (cache_dir / "refs" / "main").write_text("deadbeef", encoding="utf-8")
+
+    root = resolve_daily_omni_local_root(str(cache_dir))
+    assert root == snapshot.resolve()
+    assert daily_omni_local_qa_json(root) == (snapshot / "qa.json").resolve()
+
+
+def test_hf_cache_dir_without_refs_picks_newest_snapshot(tmp_path: Path) -> None:
+    cache_dir = tmp_path / "datasets--liarliar--Daily-Omni"
+    older = cache_dir / "snapshots" / "aaa"
+    newer = cache_dir / "snapshots" / "bbb"
+    older.mkdir(parents=True)
+    newer.mkdir(parents=True)
+    os.utime(older, (1, 1))
+    os.utime(newer, (10, 10))
+
+    assert resolve_daily_omni_local_root(str(cache_dir)) == newer.resolve()
+
+
+def test_videos_tar_is_extracted(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setenv("HF_HOME", str(tmp_path / "hf_home"))
+    root = tmp_path / "mirror"
+    (root / "Videos" / "vid0").mkdir(parents=True)
+    (root / "Videos" / "vid0" / "vid0_video.mp4").write_bytes(b"fake")
+    with tarfile.open(root / "Videos.tar", "w") as tf:
+        tf.add(root / "Videos", arcname="Videos")
+    # Only the tarball is left in the mirror.
+    for child in (root / "Videos").rglob("*"):
+        if child.is_file():
+            child.unlink()
+    (root / "Videos" / "vid0").rmdir()
+    (root / "Videos").rmdir()
+
+    videos_dir = daily_omni_local_videos_dir(root)
+    assert videos_dir is not None
+    assert (videos_dir / "vid0" / "vid0_video.mp4").is_file()
+
+    # Second call reuses the extraction instead of redoing it.
+    assert daily_omni_local_videos_dir(root) == videos_dir
+
+
+def test_missing_qa_json_reports_none(tmp_path: Path) -> None:
+    root = resolve_daily_omni_local_root(str(tmp_path))
+    assert daily_omni_local_qa_json(root) is None
+    assert daily_omni_local_videos_dir(root) is None

tests/benchmarks/test_daily_omni_pack_modes.py

@@ -2,9 +2,8 @@
 
 The ``minicpm-interleave`` mode is an accuracy-critical protocol: MiniCPM-o only reaches
 OpenBMB's reported Daily-Omni score when frames and 1s audio segments arrive strictly
-interleaved, with ``max_slice_nums=1`` / ``use_image_id=False`` and an empty MiniCPM
-system message (not the Qwen system prompt). These tests pin that contract so a refactor
-cannot silently regress the score.
+interleaved, with ``max_slice_nums=1`` / ``use_image_id=False`` and no Qwen system prompt.
+These tests pin that contract so a refactor cannot silently regress the score.
 
 vllm stubs are installed by tests/benchmarks/conftest.py before collection.
 """
@@ -99,22 +98,13 @@ def _stub_extract(monkeypatch, dataset: DailyOmniDataset, counter: list[int]) ->
 # ---------------------------------------------------------------------------
 
 
-@pytest.mark.parametrize(("n", "k"), [(10, 4), (100, 64), (1000, 64), (7, 7), (3, 5), (1, 1)])
-def test_uniform_sample_indices_matches_openbmb_midpoint(n: int, k: int) -> None:
-    """Pin OpenBMB midpoint-bin sampling (not np.linspace endpoints)."""
+@pytest.mark.parametrize(("n", "k"), [(10, 4), (100, 64), (7, 7), (3, 5), (1, 1)])
+def test_uniform_sample_indices_matches_numpy_linspace(n: int, k: int) -> None:
     got = _uniform_sample_indices(n, k)
     if n <= k:
         assert got == list(range(n))
     else:
-        gap = n / k
-        expected = [int(i * gap + gap / 2) for i in range(k)]
-        assert got == expected
-        # Endpoints differ from linspace for typical Daily-Omni downsample sizes.
-        linspace = [int(i) for i in np.linspace(0, n - 1, k, dtype=int)]
-        if n == 1000 and k == 64:
-            assert got[0] == 7 and got[-1] == 992
-            assert linspace[0] == 0 and linspace[-1] == 999
-            assert got != linspace
+        assert got == [int(i) for i in np.linspace(0, n - 1, k, dtype=int)]
 
 
 def test_uniform_sample_indices_non_positive_k() -> None:
@@ -223,16 +213,14 @@ def test_minicpm_interleave_missing_local_video_returns_none(qa_json, tmp_path)
 # ---------------------------------------------------------------------------
 
 
-def test_minicpm_interleave_keeps_empty_system_message(monkeypatch, qa_json, video_dir) -> None:
-    """Official MiniCPM omni mode still sends role=system with empty content."""
+def test_minicpm_interleave_omits_qwen_system_message(monkeypatch, qa_json, video_dir) -> None:
     ds = _make_dataset(qa_json, video_dir, pack_mode="minicpm-interleave", input_mode="all")
     _stub_extract(monkeypatch, ds, [])
     parts, _extra, _position = ds._compose_daily_omni_multimodal(_VIDEO_ID, None)
 
     messages = ds._build_daily_omni_openai_messages(parts, "q?", {"A": "a"})
 
-    assert [m["role"] for m in messages] == ["system", "user"]
-    assert messages[0]["content"][0]["text"] == ""
+    assert [m["role"] for m in messages] == ["user"]
 
 
 def test_qwen_pack_mode_keeps_system_message(qa_json, video_dir) -> None:

tests/config/test_config_factory.py

@@ -2276,6 +2276,16 @@ class TestPlatformOverrides:
         # Stage 2 unaffected fields stay at base
         assert deploy.stages[2].enforce_eager is False
 
+    def test_qwen2_5_omni_xpu_uses_eager_ar_stages(self):
+        deploy_path = Path(get_deploy_config_path("qwen2_5_omni.yaml"))
+
+        deploy = load_deploy_config(deploy_path)
+        deploy = _apply_platform_overrides(deploy, platform="xpu")
+
+        assert deploy.stages[0].enforce_eager is True
+        assert deploy.stages[1].enforce_eager is True
+        assert deploy.stages[2].enforce_eager is True
+
     def test_xpu_overrides(self):
         deploy_path = Path(get_deploy_config_path("qwen3_omni_moe.yaml"))
         if not deploy_path.exists():

tests/config/test_forced_aligner_injection.py

@@ -0,0 +1,71 @@
+import pytest
+
+import vllm_omni.config.pipeline_registry  # noqa: F401  (populate registry)
+from vllm_omni.config.pipeline_registry import OMNI_PIPELINES as _PIPELINE_REGISTRY
+from vllm_omni.config.stage_config import (
+    DeployConfig,
+    StageExecutionType,
+)
+from vllm_omni.model_executor.stage_input_processors.forced_aligner import (
+    POOLING_OUTPUT_DECODER_PATH,
+    TIMESTAMPS_MODALITY,
+)
+from vllm_omni.utils.forced_aligner import _DEFAULT_CONFIG_PATH, inject_forced_aligner_stage
+
+pytestmark = [pytest.mark.core_model, pytest.mark.cpu]
+
+
+def test_inject_appends_pooling_aligner_stage():
+    pipeline = _PIPELINE_REGISTRY["qwen3_tts"]
+    deploy = DeployConfig()
+    n0 = len(pipeline.stages)
+
+    ext_pipeline, ext_deploy = inject_forced_aligner_stage(
+        pipeline, deploy, {"forced_aligner": "/models/Qwen3-ForcedAligner-0.6B"}
+    )
+
+    # A new pooling stage is appended; the original pipeline is unchanged.
+    assert len(pipeline.stages) == n0
+    assert len(ext_pipeline.stages) == n0 + 1
+    # The original deploy config is not mutated either (double-injection safe).
+    assert len(deploy.stages) == 0
+    assert len(ext_deploy.stages) == 1
+    aligner = ext_pipeline.stages[-1]
+    # Pooling stage reuses LLM_AR (+ runner=pooling), no dedicated LLM_POOLING.
+    assert aligner.execution_type == StageExecutionType.LLM_AR
+    assert aligner.final_output is True
+    assert aligner.final_output_type == TIMESTAMPS_MODALITY
+    assert aligner.input_sources == (n0 - 1,)
+    assert aligner.requires_multimodal_data is True
+    assert aligner.owns_tokenizer is True
+
+    # The deploy stage carries the separate aligner checkpoint + decoder hook.
+    ds = ext_deploy.stages[-1]
+    assert ds.engine_extras["model"] == "/models/Qwen3-ForcedAligner-0.6B"
+    assert ds.engine_extras["runner"] == "pooling"
+    assert ds.engine_extras["pooling_output_decoder"] == POOLING_OUTPUT_DECODER_PATH
+
+
+def test_inject_from_config_only():
+    # --forced-aligner-config alone (model in the YAML) must also inject.
+    pipeline = _PIPELINE_REGISTRY["qwen3_tts"]
+    deploy = DeployConfig()
+    n0 = len(pipeline.stages)
+
+    ext_pipeline, ext_deploy = inject_forced_aligner_stage(
+        pipeline, deploy, {"forced_aligner": None, "forced_aligner_config": str(_DEFAULT_CONFIG_PATH)}
+    )
+
+    assert len(ext_pipeline.stages) == n0 + 1
+    assert ext_pipeline.stages[-1].final_output_type == TIMESTAMPS_MODALITY
+    assert ext_deploy.stages[-1].engine_extras["model"] == "Qwen/Qwen3-ForcedAligner-0.6B"
+
+
+def test_inject_noop_without_forced_aligner():
+    pipeline = _PIPELINE_REGISTRY["qwen3_tts"]
+    deploy = DeployConfig()
+
+    ext_pipeline, ext_deploy = inject_forced_aligner_stage(pipeline, deploy, {})
+
+    assert ext_pipeline is pipeline
+    assert len(ext_deploy.stages) == 0

tests/config/test_omni_config.py

@@ -513,6 +513,7 @@ def test_sub_config_fields_match_structured_scopes():
         "enable_prefix_caching",
         "disable_hybrid_kv_cache_manager",
         "mm_processor_cache_gb",
+        "mamba_ssm_cache_dtype",
     }
     assert {f.name for f in fields(OmniStageSchedulerConfig)} == {
         "max_num_seqs",

tests/core/sched/test_omni_ar_scheduler_logprobs.py

@@ -27,6 +27,7 @@ _MIXIN_UPDATE_HELPERS = (
     "_attach_finished_request_sets",
     "_attach_scheduler_stats",
     "_capture_omni_connector_output",
+    "_maybe_decode_pooling_output",
 )
 
 
@@ -138,6 +139,7 @@ def _make_scheduler_stub(requests: list[_Request]) -> SimpleNamespace:
         pending_stop_after_extraction=set(),
         waiting_for_transfer_free=set(),
         _new_prompt_len_snapshot={},
+        _pooling_output_decoder=None,
         finished_req_ids=set(),
         finished_req_ids_dict=defaultdict(set),
         kv_cache_manager=SimpleNamespace(take_events=lambda: None),
@@ -267,3 +269,72 @@ def test_invalid_logprobs_finish_only_the_affected_scheduler_request() -> None:
     assert output_by_id["bad"].new_token_ids == []
     assert output_by_id["good"].new_token_ids == [8]
     np.testing.assert_array_equal(output_by_id["good"].new_logprobs.logprob_token_ids[:, 0], [8])
+
+
+def _pooling_model_runner_output(pooler_tensor) -> SimpleNamespace:
+    return SimpleNamespace(
+        sampled_token_ids=[[]],
+        logprobs=None,
+        prompt_logprobs_dict={},
+        pooler_output=[pooler_tensor],
+        num_nans_in_logits=None,
+        kv_connector_output=None,
+        cudagraph_stats=None,
+        req_id_to_index={"req": 0},
+        routed_experts=None,
+    )
+
+
+def _make_pooling_stub(decoder) -> tuple[_Request, SimpleNamespace]:
+    request = _Request("req")
+    request.sampling_params = SimpleNamespace(num_logprobs=None)
+    request.pooling_params = SimpleNamespace()
+    scheduler = _make_scheduler_stub([request])
+    scheduler._pooling_output_decoder = decoder
+    scheduler.vllm_config = SimpleNamespace(model_config=SimpleNamespace(hf_config=None))
+    _bind_request_lifecycle(scheduler, update_request=lambda req, tokens: (tokens, False))
+    return request, scheduler
+
+
+def test_pooling_decode_success_emits_decoded_payload() -> None:
+    import torch
+
+    decoded = torch.tensor([[1, 2]], dtype=torch.int32)
+    request, scheduler = _make_pooling_stub(lambda logits, req, hf: decoded)
+
+    scheduler_output = SimpleNamespace(
+        num_scheduled_tokens={"req": 1},
+        scheduled_spec_decode_tokens={},
+        num_invalid_spec_tokens=0,
+    )
+    outputs = OmniARScheduler.update_from_output(
+        scheduler, scheduler_output, _pooling_model_runner_output(torch.zeros(2, 4))
+    )
+    (output,) = outputs[0].outputs
+
+    assert output.finish_reason is FinishReason.STOP
+    assert output.pooling_output is decoded
+
+
+def test_pooling_decode_failure_finishes_request_with_error() -> None:
+    import torch
+
+    def raising_decoder(logits, req, hf):
+        raise ValueError("boom")
+
+    request, scheduler = _make_pooling_stub(raising_decoder)
+
+    scheduler_output = SimpleNamespace(
+        num_scheduled_tokens={"req": 1},
+        scheduled_spec_decode_tokens={},
+        num_invalid_spec_tokens=0,
+    )
+    outputs = OmniARScheduler.update_from_output(
+        scheduler, scheduler_output, _pooling_model_runner_output(torch.zeros(2, 4))
+    )
+    (output,) = outputs[0].outputs
+
+    # Decoder failure is a request error (500), not an empty success.
+    assert output.finish_reason is FinishReason.ERROR
+    assert output.pooling_output is None
+    assert "pooling output decode failed" in str(request.stop_reason)

tests/core/sched/test_omni_scheduling_coordinator.py

@@ -120,12 +120,16 @@ class TestFullPayloadCoordinatorSelection(unittest.TestCase):
             ("MiMoAudioModel", "code2wav"),
             ("Qwen3TTSCode2Wav", "code2wav"),
             ("MiniCPMO45Code2Wav", "code2wav"),
+            # nemotron_voicechat: sync-path RVQ-VAE decode waits for the
+            # talker's full [frames, 31] code-stack payload.
+            ("NemotronVoiceChatCode2Wav", "code2wav"),
             ("CosyVoice3Model", "cosyvoice3_code2wav"),
             # audex: sync-path TTS decoder (async_chunk: false deploys) and
             # the TTA XCodec1 stage (always sync full-payload).
             ("AudexCode2Wav", "audex_code2wav"),
             ("AudexXCodec1", "audex_xcodec"),
             ("IndexTTS2S2MelDecoder", "indextts2_s2mel_decoder"),
+            ("IndexTTS25S2MelDecoder", "indextts2_5_s2mel_decoder"),
             ("DyninOmniForConditionalGeneration", "token2image"),
             ("DyninOmniForConditionalGeneration", "token2audio"),
         }

tests/dfx/perf/tests/test_minicpmo_4_5_ready.json

@@ -0,0 +1,80 @@
+[
+  {
+    "test_name": "test_minicpmo_4_5_ready",
+    "mark": [
+      {
+        "hardware_marks": {
+          "res": {
+            "cuda": "H100",
+            "npu": "A3"
+          },
+          "num_cards": 1
+        }
+      },
+      "core_model",
+      "omni"
+    ],
+    "server_params": {
+      "model": "openbmb/MiniCPM-o-4_5",
+      "extra_cli_args": [
+        "--deploy-config",
+        "vllm_omni/deploy/minicpmo_4_5.yaml",
+        "--trust-remote-code"
+      ]
+    },
+    "benchmark_params": [
+      {
+        "dataset_name": "seed-tts",
+        "dataset_path": "zhaochenyang20/seed-tts-eval",
+        "backend": "openai-chat-omni",
+        "endpoint": "/v1/chat/completions",
+        "num_prompts": [
+          10,
+          40
+        ],
+        "max_concurrency": [
+          1,
+          4
+        ],
+        "no_oversample": true,
+        "trust_remote_code": true,
+        "seed_tts_locale": "en",
+        "extra_body": {
+          "modalities": [
+            "text",
+            "audio"
+          ],
+          "chat_template_kwargs": {
+            "enable_thinking": false,
+            "use_tts_template": true
+          }
+        },
+        "percentile-metrics": "ttft,e2el,audio_ttfp,audio_rtf,audio_duration",
+        "baseline": {
+          "H100": {
+            "request_throughput": [
+              0.7803,
+              1.1008
+            ],
+            "mean_ttft_ms": [
+              166.9764,
+              307.7133
+            ],
+            "mean_e2el_ms": [
+              1361.1682,
+              3800.6332
+            ],
+            "mean_audio_ttfp_ms": [
+              635.8099,
+              1615.6591
+            ],
+            "mean_audio_rtf": [
+              0.3151,
+              0.8568
+            ]
+          }
+        }
+      }
+    ]
+  }
+]

tests/diffusion/batching/test_diffusion_batching.py

@@ -125,7 +125,7 @@ def _extract_images(output: OmniRequestOutput) -> list:
         return output.images
     # When the output comes from the orchestrator pipeline, images may be
     # nested inside request_output.
-    inner = getattr(output, "request_output", None)
+    inner = output
     if inner is not None and hasattr(inner, "images") and inner.images:
         return inner.images
     return []

tests/diffusion/cache/test_teacache.py

@@ -56,10 +56,10 @@ def test_teacache(model_name: str):
         # Extract images from request_output['images']
         first_output = outputs[0]
         assert first_output.final_output_type == "image"
-        if not hasattr(first_output, "request_output") or not first_output.request_output:
+        if not isinstance(first_output, OmniRequestOutput) or not first_output:
             raise ValueError("No request_output found in OmniRequestOutput")
 
-        req_out = first_output.request_output
+        req_out = first_output
         if not isinstance(req_out, OmniRequestOutput) or not hasattr(req_out, "images"):
             raise ValueError("Invalid request_output structure or missing 'images' key")
 

tests/diffusion/distributed/test_sequence_parallel.py

@@ -131,7 +131,7 @@ def _run_inference(
             elapsed_ms = (time.time() - start) * 1000
 
             return InferenceResult(
-                images=outputs[0].request_output.images,
+                images=outputs[0].images,
                 elapsed_ms=elapsed_ms,
             )
     finally:

tests/diffusion/distributed/test_tensor_parallel.py

@@ -49,10 +49,10 @@ def _diff_metrics(a: Image.Image, b: Image.Image) -> tuple[float, float]:
 def _extract_single_image(outputs) -> Image.Image:
     first_output = outputs[0]
     assert first_output.final_output_type == "image"
-    if not hasattr(first_output, "request_output") or not first_output.request_output:
+    if not isinstance(first_output, OmniRequestOutput) or not first_output:
         raise ValueError("No request_output found in OmniRequestOutput")
 
-    req_out = first_output.request_output
+    req_out = first_output
     if not isinstance(req_out, OmniRequestOutput) or not hasattr(req_out, "images"):
         raise ValueError("Invalid request_output structure or missing 'images' key")
 

tests/diffusion/distributed/test_wan_spatial_shard.py

@@ -231,6 +231,128 @@ def test_reshard_from_trimmed_width_pads_invalid_columns(monkeypatch: pytest.Mon
     assert torch.equal(out.flatten(), torch.tensor([4.0, 0.0]))
 
 
+@pytest.mark.core_model
+@pytest.mark.cpu
+@pytest.mark.parametrize("split_dim", ["height", "width"])
+@pytest.mark.parametrize(
+    ("full_extent", "world_size"),
+    [
+        pytest.param(5, 4, id="four-ranks-empty-tail"),
+        pytest.param(20, 8, id="production-eight-ranks-empty-tail"),
+        pytest.param(39, 2, id="two-ranks-partial-tail"),
+    ],
+)
+def test_reshard_from_trimmed_extent_clamps_and_reconstructs(
+    monkeypatch: pytest.MonkeyPatch,
+    split_dim: str,
+    full_extent: int,
+    world_size: int,
+):
+    local_extent = (full_extent + world_size - 1) // world_size
+    shape = (1, 1, 1, full_extent, 1) if split_dim == "height" else (1, 1, 1, 1, full_extent)
+    x = torch.arange(full_extent, dtype=torch.float32).reshape(shape)
+    outputs = []
+
+    for rank in range(world_size):
+        monkeypatch.setattr(wan_spatial_shard, "_rank_world", lambda group, rank=rank: (rank, world_size))
+        token = wan_spatial_shard._SPATIAL_SHARD_CONTEXT.set(
+            wan_spatial_shard.SpatialShardContext(
+                input_extent=full_extent,
+                local_input_extent=local_extent,
+                split_dim=split_dim,
+                rank=rank,
+                world_size=world_size,
+            )
+        )
+        try:
+            out = wan_spatial_shard.reshard_from_trimmed_extent(
+                x,
+                local_extent=local_extent,
+                split_dim=split_dim,
+                group=object(),
+            )
+        finally:
+            wan_spatial_shard._SPATIAL_SHARD_CONTEXT.reset(token)
+
+        start = rank * local_extent
+        valid_extent = min(local_extent, max(0, full_extent - start))
+        expected = torch.zeros(local_extent, dtype=torch.float32)
+        if valid_extent:
+            expected[:valid_extent] = torch.arange(start, start + valid_extent, dtype=torch.float32)
+        assert out.shape[_spatial_dim_for_test(split_dim)] == local_extent
+        assert torch.equal(out.flatten(), expected)
+        outputs.append(out)
+
+    reconstructed = torch.cat(outputs, dim=_spatial_dim_for_test(split_dim))
+    reconstructed = reconstructed.narrow(_spatial_dim_for_test(split_dim), 0, full_extent)
+    assert torch.equal(reconstructed, x)
+
+
+def _spatial_dim_for_test(split_dim: str) -> int:
+    return -2 if split_dim == "height" else -1
+
+
+@pytest.mark.core_model
+@pytest.mark.cpu
+def test_reshard_from_trimmed_extent_rejects_rank_mismatch(monkeypatch: pytest.MonkeyPatch):
+    monkeypatch.setattr(wan_spatial_shard, "_rank_world", lambda group: (1, 2))
+    x = torch.zeros((1, 1, 1, 1, 4), dtype=torch.float32)
+    token = wan_spatial_shard._SPATIAL_SHARD_CONTEXT.set(
+        wan_spatial_shard.SpatialShardContext(
+            input_extent=4,
+            local_input_extent=2,
+            split_dim="width",
+            rank=0,
+            world_size=2,
+        )
+    )
+    try:
+        with pytest.raises(RuntimeError, match="group_rank=1, context_rank=0"):
+            wan_spatial_shard.reshard_from_trimmed_extent(
+                x,
+                local_extent=2,
+                split_dim="width",
+                group=object(),
+            )
+    finally:
+        wan_spatial_shard._SPATIAL_SHARD_CONTEXT.reset(token)
+
+
+@pytest.mark.core_model
+@pytest.mark.cpu
+def test_attention_wrapper_rejects_spatial_extent_change(monkeypatch: pytest.MonkeyPatch):
+    class ShrinkingWanAttentionBlock(torch.nn.Module):
+        def forward(self, x):
+            return x[..., :-1]
+
+    monkeypatch.setattr(wan_spatial_shard, "_rank_world", lambda group: (0, 2))
+    monkeypatch.setattr(
+        wan_spatial_shard,
+        "all_gather_along_dim",
+        lambda x, *, group, dim, dst=None: torch.cat([x, x], dim=dim),
+    )
+    module = ShrinkingWanAttentionBlock()
+    wan_spatial_shard._patch_attention_block(module, group=object(), split_dim="width")
+    x = torch.zeros((1, 1, 1, 1, 2), dtype=torch.float32)
+    token = wan_spatial_shard._SPATIAL_SHARD_CONTEXT.set(
+        wan_spatial_shard.SpatialShardContext(
+            input_extent=4,
+            local_input_extent=2,
+            split_dim="width",
+            rank=0,
+            world_size=2,
+        )
+    )
+    try:
+        with pytest.raises(
+            RuntimeError,
+            match=r"local_extent=2, gathered_extent=4, trimmed_extent=4, output_extent=3",
+        ):
+            module(x)
+    finally:
+        wan_spatial_shard._SPATIAL_SHARD_CONTEXT.reset(token)
+
+
 @pytest.mark.core_model
 @pytest.mark.cpu
 def test_halo_exchange_single_rank_noop(monkeypatch: pytest.MonkeyPatch):

tests/diffusion/lora/test_diffusion_lora.py

@@ -29,10 +29,10 @@ def test_diffusion_model(tmp_path: Path, omni_runner):
             raise ValueError("Empty outputs from Omni.generate()")
         first_output = outputs[0]
         assert first_output.final_output_type == "image"
-        if not hasattr(first_output, "request_output") or not first_output.request_output:
+        if not isinstance(first_output, OmniRequestOutput) or not first_output:
             raise ValueError("No request_output found in OmniRequestOutput")
 
-        req_out = first_output.request_output
+        req_out = first_output
         if not isinstance(req_out, OmniRequestOutput) or not hasattr(req_out, "images"):
             raise ValueError("Invalid request_output structure or missing 'images' key")
         return req_out.images

tests/diffusion/models/minimax_h3/test_minimax_h3_rmsnorm.py

@@ -0,0 +1,26 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import torch
+
+from vllm_omni.diffusion.layers.norm import RMSNorm
+from vllm_omni.diffusion.models.minimax_h3.encoder import MiniMaxH3Qwen3VLRMSNorm
+
+
+def test_qwen3_vl_rmsnorm_uses_common_fused_rmsnorm_contract() -> None:
+    """Qwen keeps BF16 gamma while native fallback accumulates in FP32."""
+    eps = 1e-6
+    norm = MiniMaxH3Qwen3VLRMSNorm(hidden_size=4, eps=eps, dtype=torch.bfloat16)
+    norm.weight.data.copy_(torch.tensor([1.0, 0.5, 1.5, 2.0], dtype=torch.bfloat16))
+    x = torch.tensor(
+        [[1.0, -2.0, 3.0, -4.0], [-5.0, 6.0, -7.0, 8.0]],
+        dtype=torch.bfloat16,
+    )
+
+    x_fp32 = x.float()
+    expected = (x_fp32 * torch.rsqrt(x_fp32.square().mean(-1, keepdim=True) + eps) * norm.weight.float()).to(x.dtype)
+
+    assert isinstance(norm, RMSNorm)
+    assert norm.weight.dtype == torch.bfloat16
+    assert set(norm.state_dict()) == {"weight"}
+    torch.testing.assert_close(norm.forward_native(x), expected, atol=0, rtol=0)