Merge branch 'main' into yadan-v2

Author: Yadan-Wei

.github/workflows/pr-vllm.yml

@@ -21,11 +21,11 @@ env:
   OS_VERSION: "ubuntu22.04"
 
   # vLLM EC2
-  VLLM_VERSION: 0.14.0
-  PROD_EC2_IMAGE: vllm:0.14-gpu-py312-ec2
+  VLLM_VERSION: 0.15.0
+  PROD_EC2_IMAGE: vllm:0.15-gpu-py312-ec2
 
   # vLLM SageMaker
-  PROD_SAGEMAKER_IMAGE: vllm:0.14-gpu-py312
+  PROD_SAGEMAKER_IMAGE: vllm:0.15-gpu-py312
 
   # vLLM RayServe
   VLLM_RAYSERVE_VERSION: 0.10.2
@@ -192,7 +192,7 @@ jobs:
 
       - name: Setup for vLLM tests
         run: |
-          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_14_0_test_setup.sh
+          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_15_0_test_setup.sh
 
       - name: Run vLLM tests
         run: |
@@ -238,7 +238,7 @@ jobs:
 
       - name: Setup for vLLM tests
         run: |
-          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_14_0_test_setup.sh
+          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_15_0_test_setup.sh
 
       - name: Run vLLM tests
         run: |
@@ -284,7 +284,7 @@ jobs:
 
       - name: Setup for vLLM tests
         run: |
-          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_14_0_test_setup.sh
+          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_15_0_test_setup.sh
 
       - name: Run vLLM tests
         run: |
@@ -594,7 +594,7 @@ jobs:
 
       - name: Setup for vLLM tests
         run: |
-          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_14_0_test_setup.sh
+          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_15_0_test_setup.sh
 
       - name: Run vLLM tests
         run: |
@@ -640,7 +640,7 @@ jobs:
 
       - name: Setup for vLLM tests
         run: |
-          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_14_0_test_setup.sh
+          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_15_0_test_setup.sh
 
       - name: Run vLLM tests
         run: |
@@ -686,7 +686,7 @@ jobs:
 
       - name: Setup for vLLM tests
         run: |
-          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_14_0_test_setup.sh
+          docker exec ${CONTAINER_ID} scripts/vllm/vllm_0_15_0_test_setup.sh
 
       - name: Run vLLM tests
         run: |

.gitignore

@@ -8,6 +8,6 @@ __pycache__
 
 # Generated docs
 docs/tutorials
-# docs/releasenotes/*/
+docs/releasenotes/*/
 docs/reference/available_images.md
 docs/reference/support_policy.md

.kiro/steering/docs.md

@@ -172,6 +172,9 @@ from utils import render_table
 # Build a complete row (uses img.get_display() internally)
 row = build_image_row(img, columns)  # ["PyTorch 2.9", "py312", ...]
 
+# Build row with field overrides (e.g., for version consolidation in support_policy)
+row = build_image_row(img, columns, overrides={"version": "2.9"})
+
 # Render markdown table
 table = render_table(headers, rows)
 ```
@@ -216,7 +219,7 @@ To generate release notes for an image, add the following fields to the image co
 
 ```yaml
 # Required for release notes generation (both must be present)
-announcement:
+announcements:
   - "Introduced containers for PyTorch 2.9 for Training"
   - "Added Python 3.12 support"
 
@@ -236,14 +239,14 @@ optional:
 **Required fields for release notes** are defined in `constants.py`:
 
 ```python
-RELEASE_NOTES_REQUIRED_FIELDS = ["announcement", "packages"]
+RELEASE_NOTES_REQUIRED_FIELDS = ["announcements", "packages"]
 ```
 
 The `packages` field uses keys that map to display names in `global.yml` under `display_names`.
 
 ### Release Notes Generation
 
-Release notes are automatically generated for images that have all required fields (`announcement` and `packages`).
+Release notes are automatically generated for images that have all required fields (`announcements` and `packages`).
 
 **Output structure:**
 
@@ -275,7 +278,7 @@ This separation uses the `ImageConfig.is_supported` property.
 
 **Generated release note sections:**
 
-1. **Announcement** - Bullet list from `announcement` field
+1. **Announcements** - Bullet list from `announcements` field
 1. **Core Packages** - Table from `packages` field (keys mapped via `display_names`)
 1. **Security Advisory** - Hardcoded section with link to AWS Security Bulletin
 1. **Reference** - Docker image URIs (private ECR + public ECR if `public_registry: true`) and links to available_images.md and support_policy.md
@@ -357,11 +360,16 @@ table_order:
 
 The `framework_groups` configuration consolidates support policy rows by framework. Repositories in the same group are combined into a single row using the framework name (e.g., "PyTorch").
 
+**Version Display:**
+
+- Images with the same major.minor version (e.g., `2.6.0` and `2.6.1`) are consolidated into a single row displayed as `2.6` if they have identical GA/EOP dates
+- If patch versions have different GA/EOP dates, each is displayed separately with full version (e.g., `2.6.0`, `2.6.1`) and a warning is logged
+
 **Requirements:**
 
-- All repositories in a group that have a given version must have identical GA/EOP dates
+- All repositories in a group that have a given full version (X.Y.Z) must have identical GA/EOP dates
 - Missing versions in some repositories are allowed (only present repos are consolidated)
-- A `ValueError` is raised if dates differ within a group for the same version
+- A `ValueError` is raised if dates differ within a group for the same full version
 
 To add a new framework group, add an entry to `framework_groups` with the framework name as key and list of repositories as value.
 

docker/vllm/Dockerfile

@@ -1,6 +1,6 @@
 # Declare the argument as default to use as input
 # base image: https://hub.docker.com/r/vllm/vllm-openai/tags
-ARG BASE_IMAGE=vllm/vllm-openai:v0.14.0
+ARG BASE_IMAGE=vllm/vllm-openai:v0.15.0
 
 # Use input argument as base image
 FROM $BASE_IMAGE AS base
@@ -19,7 +19,7 @@ ENV DEBIAN_FRONTEND=noninteractive \
   PYTHONDONTWRITEBYTECODE=1 \
   PYTHONUNBUFFERED=1 \
   PYTHONIOENCODING=UTF-8 \
-  LD_LIBRARY_PATH="/usr/local/lib:/opt/amazon/ofi-nccl/lib:/opt/amazon/openmpi/lib:/opt/amazon/efa/lib:/usr/local/cuda/lib64:${LD_LIBRARY_PATH}" \
+  LD_LIBRARY_PATH="/opt/amazon/ofi-nccl/lib:/opt/amazon/openmpi/lib:/opt/amazon/efa/lib:/usr/local/cuda/lib64:${LD_LIBRARY_PATH}" \
   PATH="/opt/amazon/openmpi/bin:/opt/amazon/efa/bin:/usr/local/cuda/bin:${PATH}"
 
 WORKDIR /
@@ -65,6 +65,11 @@ RUN bash install_efa.sh ${EFA_VERSION} \
   && rm -rf /usr/local/cuda/bin/cuobjdump* \
   && rm -rf /usr/local/cuda/bin/nvdisasm*
 
+# TODO: remove this code when issue is fixed:
+# issue introduced by https://github.com/vllm-project/vllm/pull/30784
+# until https://github.com/vllm-project/vllm/pull/32377 is merged
+RUN rm /etc/ld.so.conf.d/00-cuda-compat.conf || true
+
 # ====================== ec2 =========================================
 FROM base AS vllm-ec2
 

docs/DEVELOPMENT.md

@@ -15,30 +15,30 @@ cd docs/src && python main.py --verbose
 ```
 docs/
 ├── src/
-│   ├── data/                    # Per-image configuration files
+│   ├── data/                                           # Per-image configuration files
 │   │   ├── template/
-│   │   │   └── image-template.yml  # Template with all fields documented
+│   │   │   └── image-template.yml                      # Template with all fields documented
 │   │   ├── pytorch-training/
-│   │   │   └── <version>-<accelerator>-<platform>.yml
+│   │   │   └── <version>-<accelerator>-<platform>.yml  # Naming is for organization only
 │   │   └── ...
-│   ├── legacy/                  # Historical support data
+│   ├── legacy/                                         # Historical support data
 │   │   └── legacy_support.yml
-│   ├── tables/                  # Table column configurations
+│   ├── tables/                                         # Table column configurations
 │   │   └── <repository>.yml
 │   ├── templates/
-│   │   ├── reference/           # Reference page templates
-│   │   └── releasenotes/        # Release notes templates
-│   ├── constants.py             # Path constants and GLOBAL_CONFIG
-│   ├── generate.py              # Generation logic
-│   ├── global.yml               # Shared terminology and configuration
-│   ├── hooks.py                 # MkDocs hooks
-│   ├── image_config.py          # ImageConfig class
-│   ├── macros.py                # MkDocs macros plugin
-│   ├── main.py                  # CLI entry point
-│   ├── sorter.py                # Sorting tiebreaker functions
-│   └── utils.py                 # Utility functions
-├── reference/                   # Generated reference pages
-├── releasenotes/                # Generated release notes
+│   │   ├── reference/                                  # Reference page templates
+│   │   └── releasenotes/                               # Release notes templates
+│   ├── constants.py                                    # Path constants and GLOBAL_CONFIG
+│   ├── generate.py                                     # Generation logic
+│   ├── global.yml                                      # Shared terminology and configuration
+│   ├── hooks.py                                        # MkDocs hooks
+│   ├── image_config.py                                 # ImageConfig class
+│   ├── macros.py                                       # MkDocs macros plugin
+│   ├── main.py                                         # CLI entry point
+│   ├── sorter.py                                       # Sorting tiebreaker functions
+│   └── utils.py                                        # Utility functions
+├── reference/                                          # Generated reference pages
+├── releasenotes/                                       # Generated release notes
 └── mkdocs.yml
 ```
 
@@ -55,17 +55,19 @@ Create `docs/src/data/<repository>/<version>-<accelerator>-<platform>.yml`:
 framework: PyTorch
 version: "2.9"
 accelerator: gpu              # gpu, cpu, or neuronx
-python: py312
 platform: ec2                 # ec2 or sagemaker
 tags:
   - "2.9.0-gpu-py312-cu130-ubuntu22.04-ec2"
 
 # Optional metadata
+python: py312
 cuda: cu130
 os: ubuntu22.04
 public_registry: true
 ```
 
+The YAML file name is for organizational purposes only. However, make sure that the image configuration file lives in the correct repository directory.
+
 See `docs/src/data/template/image-template.yml` for all available fields.
 
 ### Step 2: Regenerate
@@ -85,7 +87,12 @@ ga: "2025-10-15"    # General Availability date
 eop: "2026-10-15"   # End of Patch date
 ```
 
-**Validation:** All images in the same framework group with the same version must have identical GA/EOP dates.
+**Version Consolidation:**
+
+- Images with the same major.minor version (e.g., `2.6.0` and `2.6.1`) are consolidated into a single row displayed as `2.6` if they have identical GA/EOP dates
+- If patch versions have different GA/EOP dates, each is displayed separately with full version (e.g., `2.6.0`, `2.6.1`) and a warning is logged
+
+**Validation:** All images in the same framework group with the same full version (X.Y.Z) must have identical GA/EOP dates.
 
 ______________________________________________________________________
 
@@ -132,6 +139,10 @@ Release notes are generated automatically for images with `announcements` and `p
 
 Sections render in YAML order as bullet lists.
 
+Section headers in optional sections are rendered via the section key.
+To format your optional section headers, add a new field in `docs/src/global.yml` under `display_names` section.
+Eg: deprecation_notice section will render its header as `## deprecation_notice` unless a formatted string is provided in `docs/src/global.yml`.
+
 ______________________________________________________________________
 
 ## Adding a New Repository
@@ -176,6 +187,8 @@ columns:
 ```
 
 **Available fields:** `framework_version`, `python`, `cuda`, `sdk`, `accelerator`, `platform`, `os`, `example_url`, `version`, `ga`, `eop`, `framework_group`, `repository`, `release_note_link`
+To add additional fields, ensure that the image configuration YAML file contains said field of the same name.
+Additionally, if you require the field to be formatted, add an additional attribute in `ImageConfig` class of `display_<field_name>` to grab the formatted field.
 
 ______________________________________________________________________
 
@@ -190,6 +203,8 @@ pytorch:
     eop: "2025-10-29"
 ```
 
+Generally, this is only required if an image configuration file does not already exist and the image is already past its support.
+
 ______________________________________________________________________
 
 ## Global Configuration
@@ -199,14 +214,17 @@ ______________________________________________________________________
 - **Terminology:** `aws`, `dlc_long`, `sagemaker`, etc.
 - **display_names:** Repository and package display names
 - **framework_groups:** Support policy consolidation groups
-- **table_order:** Order of tables in available_images.md
+- **table_order:** Order of tables displayed within the documentations website (eg: available_images.md and support_policy.md)
 - **platforms/accelerators:** Display mappings
 
 ______________________________________________________________________
 
 ## Running Generation
 
 ```bash
+# Help
+python main.py --help
+
 # Full generation
 python main.py --verbose