chore(post-release): advance main to 26.08 upstreams

[mc-nv]

What does the PR do?

• Moves main onto the 26.08 upstream containers now that the release has shipped to NGC.
• build.py map: upstream_container_version 26.07 -> 26.08, ORT 1.28.0, OpenVINO 2026.3.0, DCGM 4.6.1-1. release_version 2.73.0dev and triton_container_version 26.09dev are kept - they are the next dev train, set at code freeze.
• Picks up the manylinux migration (build(before-code-freeze): Build against latest upstream container #8918) and the 26.08 compatibility rows (docs: add 26.08 rows to the release compatibility matrix #8935).

Checklist

• PR title reflects the change and is of format <commit_type>: <Title>
• Changes are described in the pull request.
• Related issues are referenced.
• Populated github labels field
• Added test plan and verified test passes.
• Verified that the PR passes existing CI.
• Verified copyright is correct on all changed files.
• Added succinct git squash message before merging ref.
• All template sections are filled out.
• Optional: Additional screenshots for behavior/output changes with before/after.

Commit Type:

Check the conventional commit type
box here and add the label to the github PR.

• build
• ci
• docs
• feat
• fix
• perf
• refactor
• revert
• style
• test

Related PRs:

• build(deps): Bump Go example gRPC client dependencies (#919) client#920
• build(before-code-freeze): Align CMake configuration with the other repos (#162) common#164
• fix: reserve completion-queue slot at enqueue to restore preserve_ordering (#520) core#523
• build(before-code-freeze): Prepare the ONNX Runtime backend for the manylinux base (#352) onnxruntime_backend#358
• build(before-code-freeze): Export the stub's dynamic symbols when libpython is static (#451) python_backend#453
• build: Fix manylinux Python version (#204) pytorch_backend#209
• chore(versions): align tutorials examples with 26.08 release train (#167) tutorials#168

Where should the reviewer start?

build.py DEFAULT_TRITON_VERSION_MAP — the split between adopted and deliberately-kept fields.

Test plan:

• Full CI run on the internal pipeline at TRITON_CONTAINER_VERSION=26.09dev_user, with TRITON_NIGHTLY=1 and TRITON_SBSA=1.

• Local verification: per-repo pre-commit run --files <changed> passes with no auto-fixes.

• CI Pipeline ID:

Caveats:

• Do not merge before the model-repository mount point is rotated: TRITON_SERVER_MODEL_REPO_VERSION derives from TRITON_CONTAINER_VERSION, so 26.09dev must exist on the share first.

Background

Post-release step of the 26.08 train: the default branches move onto the same upstream containers as the just-released r26.08, and pick up the fixes that landed only on the release branch while it stabilised.

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

• Relates to: TRI-1691

[greptile-apps]

Greptile Summary

The PR advances main-branch release references and upstream dependencies to the 26.08 train while retaining the 26.09 development container and 2.73.0dev server versions.

• Updates build, SDK, deployment, documentation, and QA defaults to the 26.08 upstream containers.
• Migrates RHEL builds from a pyenv-installed interpreter to the manylinux Python layout and restores the static embedding library during backend builds.
• Advances ONNX Runtime, OpenVINO, DCGM, QA model-generation versions, and compatibility documentation.
• Removes obsolete pre-Python-3.12 backend test setup and corresponding build-preset options.

Confidence Score: 5/5

The PR appears safe to merge based on the reviewed version alignment, build migration, deployment defaults, and QA updates.

The changed release references are consistently advanced, and no concrete build, runtime, deployment, or QA failure remains in the final review.

Important Files Changed

Filename	Overview
build.py	Advances upstream component versions and replaces RHEL pyenv setup with the manylinux Python toolchain and static embedding library.
qa/common/gen_qa_model_repository	Aligns generated QA repositories with the 26.09dev server train, 26.08 upstreams, ONNX 1.22.0, and opset 26.
qa/common/gen_qa_model_repository.py	Keeps Python model-generation defaults aligned with the shell driver and central build version map.
qa/L0_backend_python/test.sh	Removes obsolete alternate-interpreter setup while retaining Python 3.12 validation and default-version environment coverage.
compose.py	Updates the DCGM fallback and cleans up regex escaping without changing matching semantics.
tools/build/build_presets.py	Removes mapping for the retired RHEL Python-version build option.
deploy/gke-marketplace-app/server-deployer/chart/triton/values.yaml	Advances the marketplace image, published version, and sample model repository to the 26.08 release.
docs/introduction/compatibility.md	Adds the 26.08 vLLM container and ONNX Runtime compatibility rows.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
    A[26.08 upstream containers] --> B[Build and SDK images]
    A --> C[Deployment defaults]
    A --> D[Documentation examples]
    E[Manylinux Python] --> B
    F[ORT 1.28 and ONNX opset 26] --> G[QA model generation]
    B --> H[26.09dev Triton development image]
    G --> H

Loading

_{Reviews (1): Last reviewed commit: "docs: add 26.08 rows to the release comp..." | Re-trigger Greptile}