chore(versions): align tutorials examples with 26.06 release train#151
Merged
Conversation
Bumps in tutorials/ examples to patched versions:
Collapses an over-wrapped multi-line expression onto a single line to satisfy black. Picked up by pre-commit run --all-files; not otherwise related to TRI-1421.
Aligns container references and pinned client versions with the 26.06 release (release_version 2.70.0, triton_container_version 26.06, upstream_container_version 26.05 per server/build.py): - nvcr.io/nvidia/tritonserver XX.XX-* -> 26.06-* (covers -py3, -py3-sdk, -vllm-python-py3, -trtllm-python-py3 across Conceptual_Guide, Quick_Deploy, Popular_Models_Guide, Feature_Guide, Deployment/Kubernetes, Triton_Inference_Server_Python_API) - nvcr.io/nvidia/pytorch 23.05-py3 -> 26.05-py3 - nvcr.io/nvidia/tensorflow 24.04-tf2-py3 -> 26.05-tf2-py3 - tritonclient 2.47.0 -> 2.70.0 (kafka-io example) - Internal image tag triton-python-api:r24.08 -> r26.06 in Triton_Inference_Server_Python_API/README.md Placeholder references (yy.mm, xx.yy, <yy.mm>) intentionally left unchanged. Examples that span multiple release trains may need follow-up code adjustments to keep working against the new SDK.
The 26.06 server build reports version 2.70.0 but the tritonclient wheel for 2.70.0 has not been published to PyPI yet (latest published is 2.69.0). Pinning the kafka-io example to 2.69.0 so pip install -r requirements.txt resolves today. Bump to 2.70.0 once the wheel ships.
The 26.06-py3 tritonserver image ships Python 3.12. sentencepiece 0.1.99 has no cp312 wheel and pip falls back to building from source, which fails inside the container. 0.2.1 has a cp312 wheel and installs cleanly. Verified by building the Dockerfile against the staged tritonserver:26.06-py3 image.
Caught by the staged-image test pass: build.sh hardcoded BASE_IMAGE_TAG_IDENTITY/DIFFUSION, the default TAG slug, and the StableDiffusion sub-build references at 24.08. docker/Dockerfile ARG BASE_IMAGE_TAG also defaulted to 24.08-py3. Aligned all to 26.06. The earlier sweep missed these because the regex looked for nvcr.io / tritonserver: prefixes; the literals lived as bare shell-variable values.
mc-nv
changed the title
nv-rinig
approved these changes
Jun 24, 2026
whoisj
approved these changes
Jun 24, 2026
mc-nv
added a commit
that referenced
this pull request
Jun 29, 2026
) (#152) * fix(deps): bump transformers/torch/ray to mitigate CVEs (TRI-1421) Bumps in tutorials/ examples to patched versions: * style: black reformat for Quick_Deploy/PyTorch/export.py Collapses an over-wrapped multi-line expression onto a single line to satisfy black. Picked up by pre-commit run --all-files; not otherwise related to TRI-1421. * chore(versions): bump tutorials examples to 26.06 release Aligns container references and pinned client versions with the 26.06 release (release_version 2.70.0, triton_container_version 26.06, upstream_container_version 26.05 per server/build.py): - nvcr.io/nvidia/tritonserver XX.XX-* -> 26.06-* (covers -py3, -py3-sdk, -vllm-python-py3, -trtllm-python-py3 across Conceptual_Guide, Quick_Deploy, Popular_Models_Guide, Feature_Guide, Deployment/Kubernetes, Triton_Inference_Server_Python_API) - nvcr.io/nvidia/pytorch 23.05-py3 -> 26.05-py3 - nvcr.io/nvidia/tensorflow 24.04-tf2-py3 -> 26.05-tf2-py3 - tritonclient 2.47.0 -> 2.70.0 (kafka-io example) - Internal image tag triton-python-api:r24.08 -> r26.06 in Triton_Inference_Server_Python_API/README.md Placeholder references (yy.mm, xx.yy, <yy.mm>) intentionally left unchanged. Examples that span multiple release trains may need follow-up code adjustments to keep working against the new SDK. * fix(deps): correct tritonclient pin to 2.69.0 (latest on PyPI) The 26.06 server build reports version 2.70.0 but the tritonclient wheel for 2.70.0 has not been published to PyPI yet (latest published is 2.69.0). Pinning the kafka-io example to 2.69.0 so pip install -r requirements.txt resolves today. Bump to 2.70.0 once the wheel ships. * fix(deps): bump sentencepiece pin to 0.2.1 for Python 3.12 wheel The 26.06-py3 tritonserver image ships Python 3.12. sentencepiece 0.1.99 has no cp312 wheel and pip falls back to building from source, which fails inside the container. 0.2.1 has a cp312 wheel and installs cleanly. Verified by building the Dockerfile against the staged tritonserver:26.06-py3 image. * fix(versions): bump 24.08 -> 26.06 in Python_API build.sh / Dockerfile Caught by the staged-image test pass: build.sh hardcoded BASE_IMAGE_TAG_IDENTITY/DIFFUSION, the default TAG slug, and the StableDiffusion sub-build references at 24.08. docker/Dockerfile ARG BASE_IMAGE_TAG also defaulted to 24.08-py3. Aligned all to 26.06. The earlier sweep missed these because the regex looked for nvcr.io / tritonserver: prefixes; the literals lived as bare shell-variable values.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Align tutorials examples with the 26.06 release ecosystem:
.containerfiles, and K8s deployment YAMLs to the26.06train (tritonserver:26.06-*,pytorch:26.05-py3,tensorflow:26.05-tf2-py3).transformers4.42.4 -> 5.12.1,torch2.3.1 -> 2.12.1,ray/ray[serve]/ray[all]2.32.0 / 2.36.0 -> 2.55.1,tritonclient2.47.0 -> 2.69.0 (latest published wheel).transformers4.34.0 -> 5.12.1 andsentencepiece0.1.99 -> 0.2.1 (the older sentencepiece pin has no Python 3.12 wheel; the 26.06 container ships Python 3.12).Triton_Inference_Server_Python_API/build.shand itsdocker/Dockerfileto default to26.06-py3(previously24.08-py3).Quick_Deploy/PyTorch/export.pyso pre-commit is clean against the tree.Why
Routine refresh to keep examples reproducible against the 26.06 container train and the current published Python wheels. Several of the existing pins (e.g.
sentencepiece==0.1.99,transformers==4.34.0) predate Python 3.12 wheel publication, so users following the README on a 26.06 base hit pip build-from-source failures. Bumping to currently-published versions removes that friction.Test plan
Verified end-to-end against staged 26.06 images:
-py3,-py3-sdk,-vllm-python-py3,-trtllm-python-py3) launch and respond totritonserver --help.pip install --dry-run -r requirements.txtforkafka-ioresolves cleanly inside the stagedtritonserver:26.06-py3image.Quick_Deploy/HuggingFaceTransformers/Dockerfilebuilds clean with the stagedtritonserver:26.06-py3base (FROM swap via sed-pipe; file references unchanged).Triton_Inference_Server_Python_API/build.shbuilds clean with--base gitlab-master.nvidia.com:5005/dl/dgx/tritonserver --base-image-tag 26.06-py3-stage.Quick_Deploy/ONNXwalkthrough: densenet model loads on staged server, client returns matching top-5 classification (92, 14, 95, 17, 88) within FP precision drift from the README's expected output.Notes
tritonclient==2.69.0is the latest published wheel at the time of this PR. Once 2.70.0 ships to PyPI, the kafka-io requirements pin should be advanced.transformersmajor-version bump (4.x -> 5.x) may surface API-deprecation issues in examples that exercise the tokenizer / pipeline surface; reach out if any example needs follow-up adjustments.Related Issues:
Related PRs: