tensorflow: 2.21 legacy tags + gamma flip#6363
Merged
Merged
Conversation
Drops the `platform: "sagemaker"` metadata field and updates
`prod_image` to match the legacy tag shape.
Pairs with the DLContainersReleaseLogicPython change reverting the
tensorflow entry in frameworks.yml from new-format
(release_tag / short_release_tag with {platform_suffix}) back to
legacy PyTorch shape (base_dlc_tag / full_dlc_tag with {cx_suffix}).
- platform: legacy templates ignore it and use {cx_suffix} derived
from customer_type instead. Drop to avoid dead metadata.
- prod_image: was still on the new-format shape
(tensorflow:2.21-cu129-amzn2023-sagemaker). Update to the legacy
shape that will actually be pushed:
tensorflow-training:2.21-gpu-py312-cu129-amzn2023-sagemaker-v1
tensorflow-training:2.21-cpu-py312-amzn2023-sagemaker-v1
Used by resolve-image-uri as fallback in test-only runs.
Release configs stay `release: false` — the gamma flip is a follow-up
PR after the release-logic change lands and gamma ECR repos exist.
Enables autorelease pushes to the DLC gamma ECR (account 028651357192) for validation before flipping to production. Changes per config (both cpu + cuda): - release: false → true (allow the release-gate job to fire) - private_registry: false → true (push to private registry only) - environment: "production" → "gamma" - force_release stays false (respect blocked-day / hours checks) - public_registry stays false (gamma-only, no public gallery) - enable_soci stays false (skip SOCI on gamma; enable for prod later) Prerequisites before this can succeed: 1. DLContainersReleaseLogicPython CR-287814183 must merge (reverts the tensorflow entry in frameworks.yml to legacy shape). 2. Gamma account 028651357192 must have `tensorflow-training` and `tensorflow-inference` ECR repos created.
bhanutejagk
changed the title
sirutBuasai
approved these changes
Jul 8, 2026
bhanutejagk
pushed a commit
that referenced
this pull request
Jul 9, 2026
Promotes TF 2.21 SageMaker configs from gamma to production release after gamma validation. Autorelease will push to the prod ECR (061416347176) and sync to public.ecr.aws. Changes per config (both cpu + cuda): - public_registry: false -> true (publish to public gallery) - enable_soci: false -> true (SOCI indexing enabled for prod) - environment: "gamma" -> "production" Prerequisites before merge: 1. PR #6363 (legacy tags + gamma flip) must have merged to main. 2. At least one clean gamma autorelease cycle must have completed with expected tag shapes verified in the gamma ECR (028651357192.dkr.ecr.<region>.amazonaws.com/tensorflow-training). 3. Optional: gamma image smoke-tested (docker pull + TF import, or SageMaker training job against the gamma image URI).
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
2 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.
Changes
Two commits, per-file:
Commit 1 (
4a274103) — align configs with legacy tag shapeplatform: "sagemaker"metadata field. Legacy templates use{cx_suffix}fromcustomer_type;platformbecomes dead metadata.prod_imageto match the actual legacy tag that will be pushed:tensorflow-training:2.21-gpu-py312-cu129-amzn2023-sagemaker-v1tensorflow-training:2.21-cpu-py312-amzn2023-sagemaker-v1Used by
.github/actions/resolve-image-urias fallback in test-only runs.Commit 2 — flip release configs to gamma
release: false → trueprivate_registry: false → trueenvironment: "production" → "gamma"Expected gamma tags (per legacy shape)
cuda config produces:
2.21.0-gpu-py312-cu129-amzn2023-sagemaker2.21.0-gpu-py312-cu129-amzn2023-sagemaker-v1.02.21.0-gpu-py312-cu129-amzn2023-sagemaker-v1.0-YYYY-MM-DD-HH-MM-SS2.21-gpu-py312-cu129-amzn2023-sagemaker2.21-gpu-py312-cu129-amzn2023-sagemaker-v12.21-gpu-py312(sagemaker_short_tag)2.21.0-gpu-py312(sagemaker_long_tag)cpu config produces the same shape with
-cpu-replacing-gpu-py312-cu129-.