Enable arm64 for Tensorflow CUDA images #1396
Conversation
## Walkthrough
This change updates TensorFlow CUDA Docker images and runtime environments for Python 3.11 and 3.12. It adds HDF5 development libraries to Dockerfiles, upgrades TensorFlow and TensorBoard to version 2.19.0, updates related dependencies and hashes, and extends the ARM64-compatible image target list in the CI build matrix script.
## Changes
| Files/Paths | Change Summary |
|--------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------|
| ci/cached-builds/gen_gha_matrix_jobs.py | Added four new TensorFlow CUDA image targets to the ARM64_COMPATIBLE set for build matrix generation. |
| jupyter/tensorflow/ubi9-python-3.11/Dockerfile.cuda<br>jupyter/tensorflow/ubi9-python-3.12/Dockerfile.cuda<br>runtimes/tensorflow/ubi9-python-3.11/Dockerfile.cuda<br>runtimes/tensorflow/ubi9-python-3.12/Dockerfile.cuda | Added installation of HDF5 development libraries via EPEL to support h5py in CUDA images for Python 3.11/3.12. |
| jupyter/tensorflow/ubi9-python-3.11/Pipfile<br>jupyter/tensorflow/ubi9-python-3.12/Pipfile<br>runtimes/tensorflow/ubi9-python-3.11/Pipfile<br>runtimes/tensorflow/ubi9-python-3.12/Pipfile | Upgraded TensorFlow and TensorBoard version specs from ~2.18.0 to ~2.19.0 in Pipfiles. |
| jupyter/tensorflow/ubi9-python-3.11/requirements.txt<br>runtimes/tensorflow/ubi9-python-3.11/requirements.txt | Upgraded tensorboard and tensorflow[and-cuda] to 2.19.0, nvidia-nccl-cu12 to 2.23.4, and reduced h5py hashes. |
| jupyter/tensorflow/ubi9-python-3.12/requirements.txt<br>runtimes/tensorflow/ubi9-python-3.12/requirements.txt | Upgraded multiple dependencies (tensorboard, tensorflow[and-cuda], nvidia-nccl-cu12, etc.), reduced h5py hashes.|
| manifests/base/jupyter-tensorflow-notebook-imagestream.yaml | Updated TensorFlow and TensorBoard version annotations from 2.18 to 2.19 in image stream metadata. |
## Possibly related PRs
- [opendatahub-io/notebooks#1151](https://github.com/opendatahub-io/notebooks/pull/1151): Introduced a new Jupyter TensorFlow CUDA image for Python 3.12, related to the incremental updates made in this PR.
- [opendatahub-io/notebooks#1359](https://github.com/opendatahub-io/notebooks/pull/1359): Added Python 3.12 beta ImageStream manifests for TensorFlow CUDA notebooks and runtimes, complementing this PR's image build updates.
## Suggested labels
`lgtm`, `approved`, `size/l`
## Suggested reviewers
- jstourac
- dibryant📜 Recent review detailsConfiguration used: .coderabbit.yaml ⛔ Files ignored due to path filters (4)
📒 Files selected for processing (14)
✅ Files skipped from review due to trivial changes (7)
🚧 Files skipped from review as they are similar to previous changes (7)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (9)
✨ Finishing Touches
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
Documentation and Community
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
♻️ Duplicate comments (3)
runtimes/tensorflow/ubi9-python-3.11/Pipfile (1)
13-14: Same wheel-availability concern as the 3.12 variant — please apply the verification & resolution steps suggested in the first comment.jupyter/tensorflow/ubi9-python-3.11/Pipfile (1)
13-14: Same wheel-availability concern as the 3.12 variant — please apply the verification & resolution steps suggested in the first comment.runtimes/tensorflow/ubi9-python-3.12/Pipfile (1)
13-14: Same wheel-availability concern as the 3.12 Jupyter variant — please apply the verification & resolution steps suggested in the first comment.
🧹 Nitpick comments (6)
runtimes/tensorflow/ubi9-python-3.12/Dockerfile.cuda (1)
150-153: hdf5-devel inflates the image; hdf5-libs is enough
h5pyonly needs the runtime libraries (libhdf5.so.*). Pulling the full -devel package drags in headers and static libs, adding ~70 MB uncompressed.-RUN dnf install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm && \ - dnf install -y hdf5-devel && \ +RUN dnf install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm && \ + dnf install -y hdf5 && \ dnf clean allSmaller image, same runtime functionality.
(If you do compile anything against HDF5 inside this image, keep ‑devel; otherwise this is free space.)jupyter/tensorflow/ubi9-python-3.11/Dockerfile.cuda (1)
218-222: Add cache cleanup & disable weak deps to keep the layer lean
dnf clean allleaves/var/cache/yumbehind, and weak-deps bring in >20 MB of docs we never use.
Tack on the extra cleanup flag and remove weak deps in one go.RUN dnf install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm && \ - dnf install -y hdf5-devel && \ - dnf clean all + dnf install -y --setopt=install_weak_deps=False hdf5-devel && \ + dnf clean all && rm -rf /var/cache/yumruntimes/tensorflow/ubi9-python-3.11/Dockerfile.cuda (1)
150-154: Mirror the cache-cleanup pattern used elsewhereOther
dnfsteps end with&& rm -rf /var/cache/yum; this new block doesn’t, so the image gets ~40 MB bigger.RUN dnf install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm && \ - dnf install -y hdf5-devel && \ - dnf clean all + dnf install -y --setopt=install_weak_deps=False hdf5-devel && \ + dnf clean all && rm -rf /var/cache/yumci/cached-builds/gen_gha_matrix_jobs.py (1)
29-32: Maintain alphabetical order inARM64_COMPATIBLEfor easier diffingKeeping the set sorted makes future additions friction-free and avoids noisy diffs.
- "codeserver-ubi9-python-3.12", - "cuda-jupyter-minimal-ubi9-python-3.11", - "cuda-jupyter-minimal-ubi9-python-3.12", - "cuda-jupyter-tensorflow-ubi9-python-3.11", - "cuda-jupyter-tensorflow-ubi9-python-3.12", - "runtime-cuda-tensorflow-ubi9-python-3.11", - "runtime-cuda-tensorflow-ubi9-python-3.12", + "codeserver-ubi9-python-3.12", + "cuda-jupyter-minimal-ubi9-python-3.11", + "cuda-jupyter-minimal-ubi9-python-3.12", + "cuda-jupyter-tensorflow-ubi9-python-3.11", + "cuda-jupyter-tensorflow-ubi9-python-3.12", + "runtime-cuda-tensorflow-ubi9-python-3.11", + "runtime-cuda-tensorflow-ubi9-python-3.12",(Only ordering changed – functional behaviour is identical.)
Optional but keeps reviews pleasant.jupyter/tensorflow/ubi9-python-3.12/Dockerfile.cuda (1)
218-222: Apply the same cleanup optimisation as in the 3.11 variantDuplicate feedback for consistency across images:
RUN dnf install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm && \ - dnf install -y hdf5-devel && \ - dnf clean all + dnf install -y --setopt=install_weak_deps=False hdf5-devel && \ + dnf clean all && rm -rf /var/cache/yumjupyter/tensorflow/ubi9-python-3.11/requirements.txt (1)
2888-2904: Re-consider theand-cudaextra before pushing to multi-archUsing
tensorflow[and-cuda]==2.19.0pulls platform-specific CUDA wheels in addition to the base wheel.
Onlinux/arm64PyPI currently publishes the required CP tags, but:
- The extra still downloads ~1 GiB per build.
- It hard-codes CUDA 12.4 paths; your image installs 12.5 libs.
- Past PRs (#1340) raised cross-platform resolution issues with this extra on macOS and musl systems.
Unless you rely on the pip-supplied libs, you can drop the extra and let TF find the system CUDA stack that the Dockerfile installs, cutting build size and avoiding mismatch.
Diff if you choose to drop the extra:
- tensorflow[and-cuda]==2.19.0; python_version >= '3.9' \ + tensorflow==2.19.0; python_version >= '3.9' \At minimum, run a build for both
linux/amd64andlinux/arm64and launch a quickpython -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"to be sure GPU is detected with the current spec.
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
⛔ Files ignored due to path filters (4)
jupyter/tensorflow/ubi9-python-3.11/Pipfile.lockis excluded by!**/*.lockjupyter/tensorflow/ubi9-python-3.12/Pipfile.lockis excluded by!**/*.lockruntimes/tensorflow/ubi9-python-3.11/Pipfile.lockis excluded by!**/*.lockruntimes/tensorflow/ubi9-python-3.12/Pipfile.lockis excluded by!**/*.lock
📒 Files selected for processing (13)
ci/cached-builds/gen_gha_matrix_jobs.py(1 hunks)jupyter/tensorflow/ubi9-python-3.11/Dockerfile.cuda(1 hunks)jupyter/tensorflow/ubi9-python-3.11/Pipfile(1 hunks)jupyter/tensorflow/ubi9-python-3.11/requirements.txt(3 hunks)jupyter/tensorflow/ubi9-python-3.12/Dockerfile.cuda(1 hunks)jupyter/tensorflow/ubi9-python-3.12/Pipfile(1 hunks)jupyter/tensorflow/ubi9-python-3.12/requirements.txt(8 hunks)runtimes/tensorflow/ubi9-python-3.11/Dockerfile.cuda(1 hunks)runtimes/tensorflow/ubi9-python-3.11/Pipfile(1 hunks)runtimes/tensorflow/ubi9-python-3.11/requirements.txt(3 hunks)runtimes/tensorflow/ubi9-python-3.12/Dockerfile.cuda(1 hunks)runtimes/tensorflow/ubi9-python-3.12/Pipfile(1 hunks)runtimes/tensorflow/ubi9-python-3.12/requirements.txt(5 hunks)
🧰 Additional context used
🧠 Learnings (14)
📓 Common learnings
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Dockerfile.cuda:38-38
Timestamp: 2025-07-08T19:30:20.513Z
Learning: jiridanek requested GitHub issue creation for multi-architecture support in TensorFlow CUDA runtime image during PR #1333 review. Issue was created with comprehensive problem description covering hardcoded NVARCH limitation, multiple solution options using TARGETARCH build argument with architecture mapping, acceptance criteria for multi-architecture builds, implementation guidance with code examples, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/rocm-tensorflow/ubi9-python-3.12/Dockerfile.rocm:50-50
Timestamp: 2025-07-08T19:30:01.738Z
Learning: jiridanek requested GitHub issue creation for multi-architecture support in ROCm TensorFlow image during PR #1333 review. Issue #1346 was created with comprehensive problem description covering hardcoded x86_64 architecture breaking multi-arch support, detailed impact analysis, three solution options (runtime detection, BuildKit TARGETARCH integration, hybrid approach) with pros/cons analysis, comprehensive acceptance criteria covering core requirements and testing, phased implementation guidance, related files identification, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1340 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, testing approach, and implementation timeline, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1345 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, and testing approach, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1320
File: jupyter/pytorch/ubi9-python-3.12/Dockerfile.cuda:42-52
Timestamp: 2025-07-09T12:29:56.162Z
Learning: jiridanek requested GitHub issue creation for OpenShift client architecture mapping problem affecting 29 Dockerfiles during PR #1320 review. Issue was created with comprehensive analysis covering all affected files using $(uname -m) returning 'aarch64' but OpenShift mirror expecting 'arm64', systematic solution using BuildKit TARGETARCH mapping with proper amd64→x86_64 and arm64→arm64 conversion, detailed acceptance criteria, and implementation guidance, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review, affecting both Python 3.11 and 3.12 versions. Issue #1338 was created with comprehensive problem description covering both affected images, repository pattern analysis comparing correct vs incorrect naming conventions, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/kustomize/base/service.yaml:5-15
Timestamp: 2025-07-02T18:59:15.788Z
Learning: jiridanek creates targeted GitHub issues for specific test quality improvements identified during PR reviews in opendatahub-io/notebooks. Issue #1268 demonstrates this by converting a review comment about insufficient tf2onnx conversion test validation into a comprehensive improvement plan with clear acceptance criteria, code examples, and ROCm-specific context.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/rocm-tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-09T08:07:24.937Z
Learning: jiridanek requested GitHub issue creation for tensorflow_rocm Python 3.12 compatibility problem during PR #1333 review. Issue #1354 was successfully created with comprehensive problem description covering missing cp312 wheels causing build failures, three solution options (upstream TensorFlow, Python 3.11 only, custom build), clear acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: atheo89
PR: opendatahub-io/notebooks#1258
File: codeserver/ubi9-python-3.11/Dockerfile.cpu:32-32
Timestamp: 2025-07-07T11:08:48.524Z
Learning: atheo89 requested GitHub issue creation for multi-architecture Dockerfile improvements during PR #1258 review, specifically for enhancing structural consistency across Docker stages, replacing $(uname -m) with ${TARGETARCH} for cross-architecture builds, and adding OCI-compliant metadata labels. Issue #1332 was created with comprehensive problem description, phased implementation approach, detailed acceptance criteria, implementation guidance with code examples, and proper context linking, continuing the established pattern of systematic code quality improvements.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review. Issue was created with comprehensive problem description covering both Python 3.11 and 3.12 versions, repository pattern analysis showing correct vs incorrect naming, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow with runtime ROCm configuration is the recommended alternative approach.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. The tensorflow-rocm upstream project appears abandoned with the last release in 2019. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow 2.18+ with runtime ROCm configuration is the recommended and industry-standard approach, as modern TensorFlow automatically detects and utilizes ROCm when properly installed.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: tensorflow_rocm package versions 2.12.1.570 through 2.14.0.600 do not provide Python 3.12 wheels (cp312) on PyPI, causing Pipfile lock failures when attempting to create Python 3.12-based ROCm TensorFlow notebook images in opendatahub-io/notebooks.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with `extras = ["and-cuda"]` can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-11T11:54:28.202Z
Learning: In opendatahub-io/notebooks, Python 3.12-based images (e.g., runtime-cuda-pytorch-ubi9-python-3.12) may fail container runtime tests with "libcrypt.so.1 => not found" for MySQL SASL2 plugin libraries if `libxcrypt-compat` is missing. The solution is to install `libxcrypt-compat` in the Dockerfile.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-14T15:36:29.788Z
Learning: ROCm is not supported by upstream TensorFlow, unlike CUDA. ROCm support requires the specialized tensorflow_rocm package, which creates a dependency challenge for Python 3.12 environments since tensorflow_rocm lacks Python 3.12 wheels while regular TensorFlow lacks ROCm support.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1247
File: .github/workflows/build-notebooks-TEMPLATE.yaml:50-53
Timestamp: 2025-07-01T14:36:52.852Z
Learning: In the opendatahub-io/notebooks repository, the test runner's Python version (configured in GitHub Actions UV setup) intentionally doesn't need to match the Python version of the container images being tested. jiridanek's team uses Python 3.12 for running tests while images may use different Python versions (like 3.11), and this approach works fine since the test code is separate from the application code running inside the containers.
runtimes/tensorflow/ubi9-python-3.11/Pipfile (10)
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with `extras = ["and-cuda"]` can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: tensorflow_rocm package versions 2.12.1.570 through 2.14.0.600 do not provide Python 3.12 wheels (cp312) on PyPI, causing Pipfile lock failures when attempting to create Python 3.12-based ROCm TensorFlow notebook images in opendatahub-io/notebooks.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1340 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, testing approach, and implementation timeline, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1345 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, and testing approach, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. The tensorflow-rocm upstream project appears abandoned with the last release in 2019. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow 2.18+ with runtime ROCm configuration is the recommended and industry-standard approach, as modern TensorFlow automatically detects and utilizes ROCm when properly installed.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow with runtime ROCm configuration is the recommended alternative approach.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: The jupyter-bokeh package was previously pinned to version 3.0.5 in the TrustyAI notebook image due to compatibility requirements with TrustyAI components, as indicated by the comment "Should be pinned down to this version in order to be compatible with trustyai" that was removed in this update.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: TrustyAI's jupyter-bokeh was pinned to 3.0.5 due to compatibility requirements with TrustyAI's visualization components, but the actual deployed version in requirements.txt shows 3.0.7, indicating incremental testing. The upgrade to 4.0.5 in this PR represents the completion of a gradual migration strategy from the 3.x series after confirming compatibility with Bokeh 3.7.3.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1127
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:20-20
Timestamp: 2025-06-27T07:49:01.198Z
Learning: Transformers v4.50.0 contains only non-breaking changes including documentation redesign, repository maintenance, performance enhancements, and bug fixes, with no API changes that would affect TrustyAI integration.
jupyter/tensorflow/ubi9-python-3.12/Pipfile (11)
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with `extras = ["and-cuda"]` can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: The jupyter-bokeh package was previously pinned to version 3.0.5 in the TrustyAI notebook image due to compatibility requirements with TrustyAI components, as indicated by the comment "Should be pinned down to this version in order to be compatible with trustyai" that was removed in this update.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: tensorflow_rocm package versions 2.12.1.570 through 2.14.0.600 do not provide Python 3.12 wheels (cp312) on PyPI, causing Pipfile lock failures when attempting to create Python 3.12-based ROCm TensorFlow notebook images in opendatahub-io/notebooks.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: The jupyter-bokeh pinning to 3.0.5 in TrustyAI notebook image was not due to TrustyAI code compatibility issues, but because the trustyai package itself explicitly declares jupyter-bokeh~=3.0.5 as a hard dependency, causing pip dependency resolution conflicts when trying to upgrade to jupyter-bokeh 4.x.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: TrustyAI's jupyter-bokeh was pinned to 3.0.5 due to compatibility requirements with TrustyAI's visualization components, but the actual deployed version in requirements.txt shows 3.0.7, indicating incremental testing. The upgrade to 4.0.5 in this PR represents the completion of a gradual migration strategy from the 3.x series after confirming compatibility with Bokeh 3.7.3.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: TrustyAI 0.6.1 (latest version as of June 2025) has a hard dependency constraint on jupyter-bokeh~=3.0.5, preventing upgrades to jupyter-bokeh 4.x in notebook images that include TrustyAI. This requires either waiting for TrustyAI to update their dependency or excluding TrustyAI from jupyter-bokeh upgrades.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1340 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, testing approach, and implementation timeline, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1345 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, and testing approach, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: TrustyAI explicitly declares jupyter-bokeh~=3.0.5 as a hard dependency in both their requirements.txt and pyproject.toml files, with no open issues tracking jupyter-bokeh 4.x compatibility. This creates an unresolvable pip dependency conflict when trying to upgrade jupyter-bokeh to 4.x in notebook images that include TrustyAI.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. The tensorflow-rocm upstream project appears abandoned with the last release in 2019. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow 2.18+ with runtime ROCm configuration is the recommended and industry-standard approach, as modern TensorFlow automatically detects and utilizes ROCm when properly installed.
jupyter/tensorflow/ubi9-python-3.12/Dockerfile.cuda (10)
Learnt from: grdryn
PR: opendatahub-io/notebooks#1320
File: rstudio/rhel9-python-3.11/Dockerfile.cuda:34-35
Timestamp: 2025-07-04T10:41:13.061Z
Learning: In the opendatahub-io/notebooks repository, when adapting NVIDIA CUDA Dockerfiles, the project intentionally maintains consistency with upstream NVIDIA patterns even when it might involve potential risks like empty variable expansions in package installation commands. This is considered acceptable because the containers only run on RHEL 9 with known yum/dnf behavior, and upstream consistency is prioritized over defensive coding practices.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with `extras = ["and-cuda"]` can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-11T11:54:28.202Z
Learning: In opendatahub-io/notebooks, Python 3.12-based images (e.g., runtime-cuda-pytorch-ubi9-python-3.12) may fail container runtime tests with "libcrypt.so.1 => not found" for MySQL SASL2 plugin libraries if `libxcrypt-compat` is missing. The solution is to install `libxcrypt-compat` in the Dockerfile.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review, affecting both Python 3.11 and 3.12 versions. Issue #1338 was created with comprehensive problem description covering both affected images, repository pattern analysis comparing correct vs incorrect naming conventions, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review. Issue was created with comprehensive problem description covering both Python 3.11 and 3.12 versions, repository pattern analysis showing correct vs incorrect naming, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1320
File: jupyter/pytorch/ubi9-python-3.12/Dockerfile.cuda:65-66
Timestamp: 2025-07-09T12:31:02.033Z
Learning: jiridanek requested GitHub issue creation for MSSQL repo file hardcoding problem during PR #1320 review. Issue #1363 was created and updated with comprehensive problem description covering hardcoded x86_64 MSSQL repo files breaking multi-architecture builds across 10 affected Dockerfiles (including datascience, CUDA, ROCm, and TrustyAI variants), detailed root cause analysis, three solution options with code examples, clear acceptance criteria for all image types, implementation guidance following established multi-architecture patterns, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/pytorch/ubi9-python-3.12/Dockerfile.cuda:17-25
Timestamp: 2025-07-09T08:07:30.628Z
Learning: jiridanek requested GitHub issue creation for oc client installation permission problem in PyTorch CUDA runtime Dockerfile during PR #1333 review. Issue #1356 was created with comprehensive problem description covering USER 1001 permission conflicts with root-owned /opt/app-root/bin directory, detailed impact analysis of build failures and non-executable binaries, current problematic code snippet, complete solution with user switching approach, clear acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1345 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, and testing approach, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1340 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, testing approach, and implementation timeline, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
runtimes/tensorflow/ubi9-python-3.12/Pipfile (10)
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with `extras = ["and-cuda"]` can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: tensorflow_rocm package versions 2.12.1.570 through 2.14.0.600 do not provide Python 3.12 wheels (cp312) on PyPI, causing Pipfile lock failures when attempting to create Python 3.12-based ROCm TensorFlow notebook images in opendatahub-io/notebooks.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. The tensorflow-rocm upstream project appears abandoned with the last release in 2019. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow 2.18+ with runtime ROCm configuration is the recommended and industry-standard approach, as modern TensorFlow automatically detects and utilizes ROCm when properly installed.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow with runtime ROCm configuration is the recommended alternative approach.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1340 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, testing approach, and implementation timeline, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1345 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, and testing approach, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: The jupyter-bokeh package was previously pinned to version 3.0.5 in the TrustyAI notebook image due to compatibility requirements with TrustyAI components, as indicated by the comment "Should be pinned down to this version in order to be compatible with trustyai" that was removed in this update.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-14T15:36:29.788Z
Learning: ROCm is not supported by upstream TensorFlow, unlike CUDA. ROCm support requires the specialized tensorflow_rocm package, which creates a dependency challenge for Python 3.12 environments since tensorflow_rocm lacks Python 3.12 wheels while regular TensorFlow lacks ROCm support.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1127
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:20-20
Timestamp: 2025-06-27T07:49:01.198Z
Learning: Transformers v4.50.0 contains only non-breaking changes including documentation redesign, repository maintenance, performance enhancements, and bug fixes, with no API changes that would affect TrustyAI integration.
runtimes/tensorflow/ubi9-python-3.11/Dockerfile.cuda (10)
Learnt from: grdryn
PR: opendatahub-io/notebooks#1320
File: rstudio/rhel9-python-3.11/Dockerfile.cuda:34-35
Timestamp: 2025-07-04T10:41:13.061Z
Learning: In the opendatahub-io/notebooks repository, when adapting NVIDIA CUDA Dockerfiles, the project intentionally maintains consistency with upstream NVIDIA patterns even when it might involve potential risks like empty variable expansions in package installation commands. This is considered acceptable because the containers only run on RHEL 9 with known yum/dnf behavior, and upstream consistency is prioritized over defensive coding practices.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with `extras = ["and-cuda"]` can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-11T11:54:28.202Z
Learning: In opendatahub-io/notebooks, Python 3.12-based images (e.g., runtime-cuda-pytorch-ubi9-python-3.12) may fail container runtime tests with "libcrypt.so.1 => not found" for MySQL SASL2 plugin libraries if `libxcrypt-compat` is missing. The solution is to install `libxcrypt-compat` in the Dockerfile.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-06-16T11:32:09.203Z
Learning: In the opendatahub-io/notebooks repository, there is a known issue with missing `runtimes/rocm/pytorch/ubi9-python-3.11/kustomize/base/kustomization.yaml` file that causes rocm runtime tests to fail with "no such file or directory" error. This is tracked in JIRA RHOAIENG-22044 and was intended to be fixed in PR #1015.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Dockerfile.cuda:38-38
Timestamp: 2025-07-08T19:30:20.513Z
Learning: jiridanek requested GitHub issue creation for multi-architecture support in TensorFlow CUDA runtime image during PR #1333 review. Issue was created with comprehensive problem description covering hardcoded NVARCH limitation, multiple solution options using TARGETARCH build argument with architecture mapping, acceptance criteria for multi-architecture builds, implementation guidance with code examples, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review, affecting both Python 3.11 and 3.12 versions. Issue #1338 was created with comprehensive problem description covering both affected images, repository pattern analysis comparing correct vs incorrect naming conventions, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review. Issue was created with comprehensive problem description covering both Python 3.11 and 3.12 versions, repository pattern analysis showing correct vs incorrect naming, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/pytorch/ubi9-python-3.12/Dockerfile.cuda:17-25
Timestamp: 2025-07-09T08:07:30.628Z
Learning: jiridanek requested GitHub issue creation for oc client installation permission problem in PyTorch CUDA runtime Dockerfile during PR #1333 review. Issue #1356 was created with comprehensive problem description covering USER 1001 permission conflicts with root-owned /opt/app-root/bin directory, detailed impact analysis of build failures and non-executable binaries, current problematic code snippet, complete solution with user switching approach, clear acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1345 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, and testing approach, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
jupyter/tensorflow/ubi9-python-3.11/Dockerfile.cuda (10)
Learnt from: grdryn
PR: opendatahub-io/notebooks#1320
File: rstudio/rhel9-python-3.11/Dockerfile.cuda:34-35
Timestamp: 2025-07-04T10:41:13.061Z
Learning: In the opendatahub-io/notebooks repository, when adapting NVIDIA CUDA Dockerfiles, the project intentionally maintains consistency with upstream NVIDIA patterns even when it might involve potential risks like empty variable expansions in package installation commands. This is considered acceptable because the containers only run on RHEL 9 with known yum/dnf behavior, and upstream consistency is prioritized over defensive coding practices.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with `extras = ["and-cuda"]` can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-11T11:54:28.202Z
Learning: In opendatahub-io/notebooks, Python 3.12-based images (e.g., runtime-cuda-pytorch-ubi9-python-3.12) may fail container runtime tests with "libcrypt.so.1 => not found" for MySQL SASL2 plugin libraries if `libxcrypt-compat` is missing. The solution is to install `libxcrypt-compat` in the Dockerfile.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review, affecting both Python 3.11 and 3.12 versions. Issue #1338 was created with comprehensive problem description covering both affected images, repository pattern analysis comparing correct vs incorrect naming conventions, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review. Issue was created with comprehensive problem description covering both Python 3.11 and 3.12 versions, repository pattern analysis showing correct vs incorrect naming, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1320
File: jupyter/pytorch/ubi9-python-3.12/Dockerfile.cuda:65-66
Timestamp: 2025-07-09T12:31:02.033Z
Learning: jiridanek requested GitHub issue creation for MSSQL repo file hardcoding problem during PR #1320 review. Issue #1363 was created and updated with comprehensive problem description covering hardcoded x86_64 MSSQL repo files breaking multi-architecture builds across 10 affected Dockerfiles (including datascience, CUDA, ROCm, and TrustyAI variants), detailed root cause analysis, three solution options with code examples, clear acceptance criteria for all image types, implementation guidance following established multi-architecture patterns, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/pytorch/ubi9-python-3.12/Dockerfile.cuda:17-25
Timestamp: 2025-07-09T08:07:30.628Z
Learning: jiridanek requested GitHub issue creation for oc client installation permission problem in PyTorch CUDA runtime Dockerfile during PR #1333 review. Issue #1356 was created with comprehensive problem description covering USER 1001 permission conflicts with root-owned /opt/app-root/bin directory, detailed impact analysis of build failures and non-executable binaries, current problematic code snippet, complete solution with user switching approach, clear acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/Dockerfile.cpu:135-136
Timestamp: 2025-07-04T05:52:49.464Z
Learning: jiridanek requested GitHub issue creation for improving fragile sed-based Jupyter kernel display_name modification in jupyter/trustyai/ubi9-python-3.12/Dockerfile.cpu during PR #1306 review. Issue #1321 was created with comprehensive problem description covering JSON corruption risks, greedy regex patterns, maintenance burden, and proposed Python-based JSON parsing solution with detailed acceptance criteria, implementation guidance, and proper context linking, continuing the established pattern of systematic code quality improvements.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1345 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, and testing approach, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
runtimes/tensorflow/ubi9-python-3.12/Dockerfile.cuda (10)
Learnt from: grdryn
PR: opendatahub-io/notebooks#1320
File: rstudio/rhel9-python-3.11/Dockerfile.cuda:34-35
Timestamp: 2025-07-04T10:41:13.061Z
Learning: In the opendatahub-io/notebooks repository, when adapting NVIDIA CUDA Dockerfiles, the project intentionally maintains consistency with upstream NVIDIA patterns even when it might involve potential risks like empty variable expansions in package installation commands. This is considered acceptable because the containers only run on RHEL 9 with known yum/dnf behavior, and upstream consistency is prioritized over defensive coding practices.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with `extras = ["and-cuda"]` can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-11T11:54:28.202Z
Learning: In opendatahub-io/notebooks, Python 3.12-based images (e.g., runtime-cuda-pytorch-ubi9-python-3.12) may fail container runtime tests with "libcrypt.so.1 => not found" for MySQL SASL2 plugin libraries if `libxcrypt-compat` is missing. The solution is to install `libxcrypt-compat` in the Dockerfile.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review, affecting both Python 3.11 and 3.12 versions. Issue #1338 was created with comprehensive problem description covering both affected images, repository pattern analysis comparing correct vs incorrect naming conventions, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Dockerfile.cuda:38-38
Timestamp: 2025-07-08T19:30:20.513Z
Learning: jiridanek requested GitHub issue creation for multi-architecture support in TensorFlow CUDA runtime image during PR #1333 review. Issue was created with comprehensive problem description covering hardcoded NVARCH limitation, multiple solution options using TARGETARCH build argument with architecture mapping, acceptance criteria for multi-architecture builds, implementation guidance with code examples, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/pytorch/ubi9-python-3.12/Dockerfile.cuda:17-25
Timestamp: 2025-07-09T08:07:30.628Z
Learning: jiridanek requested GitHub issue creation for oc client installation permission problem in PyTorch CUDA runtime Dockerfile during PR #1333 review. Issue #1356 was created with comprehensive problem description covering USER 1001 permission conflicts with root-owned /opt/app-root/bin directory, detailed impact analysis of build failures and non-executable binaries, current problematic code snippet, complete solution with user switching approach, clear acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review. Issue was created with comprehensive problem description covering both Python 3.11 and 3.12 versions, repository pattern analysis showing correct vs incorrect naming, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-06-16T11:32:09.203Z
Learning: In the opendatahub-io/notebooks repository, there is a known issue with missing `runtimes/rocm/pytorch/ubi9-python-3.11/kustomize/base/kustomization.yaml` file that causes rocm runtime tests to fail with "no such file or directory" error. This is tracked in JIRA RHOAIENG-22044 and was intended to be fixed in PR #1015.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1340 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, testing approach, and implementation timeline, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
jupyter/tensorflow/ubi9-python-3.11/Pipfile (11)
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with `extras = ["and-cuda"]` can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: The jupyter-bokeh package was previously pinned to version 3.0.5 in the TrustyAI notebook image due to compatibility requirements with TrustyAI components, as indicated by the comment "Should be pinned down to this version in order to be compatible with trustyai" that was removed in this update.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: tensorflow_rocm package versions 2.12.1.570 through 2.14.0.600 do not provide Python 3.12 wheels (cp312) on PyPI, causing Pipfile lock failures when attempting to create Python 3.12-based ROCm TensorFlow notebook images in opendatahub-io/notebooks.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: The jupyter-bokeh pinning to 3.0.5 in TrustyAI notebook image was not due to TrustyAI code compatibility issues, but because the trustyai package itself explicitly declares jupyter-bokeh~=3.0.5 as a hard dependency, causing pip dependency resolution conflicts when trying to upgrade to jupyter-bokeh 4.x.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: TrustyAI's jupyter-bokeh was pinned to 3.0.5 due to compatibility requirements with TrustyAI's visualization components, but the actual deployed version in requirements.txt shows 3.0.7, indicating incremental testing. The upgrade to 4.0.5 in this PR represents the completion of a gradual migration strategy from the 3.x series after confirming compatibility with Bokeh 3.7.3.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: TrustyAI 0.6.1 (latest version as of June 2025) has a hard dependency constraint on jupyter-bokeh~=3.0.5, preventing upgrades to jupyter-bokeh 4.x in notebook images that include TrustyAI. This requires either waiting for TrustyAI to update their dependency or excluding TrustyAI from jupyter-bokeh upgrades.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: TrustyAI explicitly declares jupyter-bokeh~=3.0.5 as a hard dependency in both their requirements.txt and pyproject.toml files, with no open issues tracking jupyter-bokeh 4.x compatibility. This creates an unresolvable pip dependency conflict when trying to upgrade jupyter-bokeh to 4.x in notebook images that include TrustyAI.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1340 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, testing approach, and implementation timeline, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1345 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, and testing approach, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. The tensorflow-rocm upstream project appears abandoned with the last release in 2019. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow 2.18+ with runtime ROCm configuration is the recommended and industry-standard approach, as modern TensorFlow automatically detects and utilizes ROCm when properly installed.
runtimes/tensorflow/ubi9-python-3.12/requirements.txt (16)
undefined
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: tensorflow_rocm package versions 2.12.1.570 through 2.14.0.600 do not provide Python 3.12 wheels (cp312) on PyPI, causing Pipfile lock failures when attempting to create Python 3.12-based ROCm TensorFlow notebook images in opendatahub-io/notebooks.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with extras = ["and-cuda"] can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: TrustyAI's jupyter-bokeh was pinned to 3.0.5 due to compatibility requirements with TrustyAI's visualization components, but the actual deployed version in requirements.txt shows 3.0.7, indicating incremental testing. The upgrade to 4.0.5 in this PR represents the completion of a gradual migration strategy from the 3.x series after confirming compatibility with Bokeh 3.7.3.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1340 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, testing approach, and implementation timeline, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1333
File: runtimes/rocm-tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-09T08:07:24.937Z
Learning: jiridanek requested GitHub issue creation for tensorflow_rocm Python 3.12 compatibility problem during PR #1333 review. Issue #1354 was successfully created with comprehensive problem description covering missing cp312 wheels causing build failures, three solution options (upstream TensorFlow, Python 3.11 only, custom build), clear acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1345 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, and testing approach, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review, affecting both Python 3.11 and 3.12 versions. Issue #1338 was created with comprehensive problem description covering both affected images, repository pattern analysis comparing correct vs incorrect naming conventions, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow with runtime ROCm configuration is the recommended alternative approach.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-14T15:36:29.788Z
Learning: ROCm is not supported by upstream TensorFlow, unlike CUDA. ROCm support requires the specialized tensorflow_rocm package, which creates a dependency challenge for Python 3.12 environments since tensorflow_rocm lacks Python 3.12 wheels while regular TensorFlow lacks ROCm support.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1154
File: manifests/base/jupyter-pytorch-notebook-imagestream.yaml:0-0
Timestamp: 2025-06-16T11:06:33.139Z
Learning: In the opendatahub-io/notebooks repository, N-1 versions of images in manifest files (like imagestream.yaml files) should not be updated regularly. The versions of packages like codeflare-sdk in N-1 images are frozen to what was released when the image was moved from N to N-1 version. N-1 images are only updated for security vulnerabilities of packages, not for regular version bumps. This is why the version of packages in N-1 images may be quite old compared to the latest N version.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: jiridanek's team uses containerized dependency locking for cross-platform compatibility in opendatahub-io/notebooks. They run pipenv lock inside UBI9 containers with specific platform arguments (--platform=linux/amd64 --python-version 3.12) to avoid host OS dependency conflicts when generating Pipfile.lock files.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. The tensorflow-rocm upstream project appears abandoned with the last release in 2019. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow 2.18+ with runtime ROCm configuration is the recommended and industry-standard approach, as modern TensorFlow automatically detects and utilizes ROCm when properly installed.
</retrieved_learning>
<retrieved_learning>
Learnt from: grdryn
PR: #1320
File: rstudio/rhel9-python-3.11/Dockerfile.cuda:34-35
Timestamp: 2025-07-04T10:41:13.061Z
Learning: In the opendatahub-io/notebooks repository, when adapting NVIDIA CUDA Dockerfiles, the project intentionally maintains consistency with upstream NVIDIA patterns even when it might involve potential risks like empty variable expansions in package installation commands. This is considered acceptable because the containers only run on RHEL 9 with known yum/dnf behavior, and upstream consistency is prioritized over defensive coding practices.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: TrustyAI 0.6.1 (latest version as of June 2025) has a hard dependency constraint on jupyter-bokeh~=3.0.5, preventing upgrades to jupyter-bokeh 4.x in notebook images that include TrustyAI. This requires either waiting for TrustyAI to update their dependency or excluding TrustyAI from jupyter-bokeh upgrades.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: TrustyAI explicitly declares jupyter-bokeh~=3.0.5 as a hard dependency in both their requirements.txt and pyproject.toml files, with no open issues tracking jupyter-bokeh 4.x compatibility. This creates an unresolvable pip dependency conflict when trying to upgrade jupyter-bokeh to 4.x in notebook images that include TrustyAI.
</retrieved_learning>
ci/cached-builds/gen_gha_matrix_jobs.py (14)
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/tensorflow/ubi9-python-3.12/Dockerfile.cuda:38-38
Timestamp: 2025-07-08T19:30:20.513Z
Learning: jiridanek requested GitHub issue creation for multi-architecture support in TensorFlow CUDA runtime image during PR #1333 review. Issue was created with comprehensive problem description covering hardcoded NVARCH limitation, multiple solution options using TARGETARCH build argument with architecture mapping, acceptance criteria for multi-architecture builds, implementation guidance with code examples, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1333
File: runtimes/rocm-tensorflow/ubi9-python-3.12/Dockerfile.rocm:50-50
Timestamp: 2025-07-08T19:30:01.738Z
Learning: jiridanek requested GitHub issue creation for multi-architecture support in ROCm TensorFlow image during PR #1333 review. Issue #1346 was created with comprehensive problem description covering hardcoded x86_64 architecture breaking multi-arch support, detailed impact analysis, three solution options (runtime detection, BuildKit TARGETARCH integration, hybrid approach) with pros/cons analysis, comprehensive acceptance criteria covering core requirements and testing, phased implementation guidance, related files identification, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1320
File: jupyter/pytorch/ubi9-python-3.12/Dockerfile.cuda:42-52
Timestamp: 2025-07-09T12:29:56.162Z
Learning: jiridanek requested GitHub issue creation for OpenShift client architecture mapping problem affecting 29 Dockerfiles during PR #1320 review. Issue was created with comprehensive analysis covering all affected files using $(uname -m) returning 'aarch64' but OpenShift mirror expecting 'arm64', systematic solution using BuildKit TARGETARCH mapping with proper amd64→x86_64 and arm64→arm64 conversion, detailed acceptance criteria, and implementation guidance, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review, affecting both Python 3.11 and 3.12 versions. Issue #1338 was created with comprehensive problem description covering both affected images, repository pattern analysis comparing correct vs incorrect naming conventions, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: atheo89
PR: opendatahub-io/notebooks#1258
File: codeserver/ubi9-python-3.11/Dockerfile.cpu:32-32
Timestamp: 2025-07-07T11:08:48.524Z
Learning: atheo89 requested GitHub issue creation for multi-architecture Dockerfile improvements during PR #1258 review, specifically for enhancing structural consistency across Docker stages, replacing $(uname -m) with ${TARGETARCH} for cross-architecture builds, and adding OCI-compliant metadata labels. Issue #1332 was created with comprehensive problem description, phased implementation approach, detailed acceptance criteria, implementation guidance with code examples, and proper context linking, continuing the established pattern of systematic code quality improvements.
Learnt from: grdryn
PR: opendatahub-io/notebooks#1320
File: rstudio/rhel9-python-3.11/Dockerfile.cuda:40-42
Timestamp: 2025-07-04T17:08:02.399Z
Learning: In the opendatahub-io/notebooks repository, when using multi-architecture Dockerfiles with BuildKit, the implicit build argument TARGETARCH is automatically available in the global scope for FROM instructions without explicit declaration. However, if TARGETARCH is used within a build stage, it must be declared explicitly within that stage. The current placement pattern (declaring ARG TARGETARCH after FROM instructions that use it) is correct for modern Docker/Podman/Buildah environments and does not require compatibility with older Docker versions.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review. Issue was created with comprehensive problem description covering both Python 3.11 and 3.12 versions, repository pattern analysis showing correct vs incorrect naming, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1320
File: jupyter/pytorch/ubi9-python-3.12/Dockerfile.cuda:65-66
Timestamp: 2025-07-09T12:31:02.033Z
Learning: jiridanek requested GitHub issue creation for MSSQL repo file hardcoding problem during PR #1320 review. Issue #1363 was created and updated with comprehensive problem description covering hardcoded x86_64 MSSQL repo files breaking multi-architecture builds across 10 affected Dockerfiles (including datascience, CUDA, ROCm, and TrustyAI variants), detailed root cause analysis, three solution options with code examples, clear acceptance criteria for all image types, implementation guidance following established multi-architecture patterns, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-12T20:04:31.021Z
Learning: Issue #862 in opendatahub-io/notebooks reveals that the Docker image tag length fix from issue #631 is mathematically insufficient. While the fix correctly limits branch names to 40 characters, this fails for longer target names like "runtime-minimal-ubi9-python-3.11" (33 chars) which still produces tags exceeding the 128-character Docker limit. The fix needs dynamic calculation based on actual target name length rather than a fixed 40-character assumption.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-11T11:54:28.202Z
Learning: In opendatahub-io/notebooks, Python 3.12-based images (e.g., runtime-cuda-pytorch-ubi9-python-3.12) may fail container runtime tests with "libcrypt.so.1 => not found" for MySQL SASL2 plugin libraries if `libxcrypt-compat` is missing. The solution is to install `libxcrypt-compat` in the Dockerfile.
Learnt from: grdryn
PR: opendatahub-io/notebooks#1320
File: rstudio/rhel9-python-3.11/Dockerfile.cuda:34-35
Timestamp: 2025-07-04T10:41:13.061Z
Learning: In the opendatahub-io/notebooks repository, when adapting NVIDIA CUDA Dockerfiles, the project intentionally maintains consistency with upstream NVIDIA patterns even when it might involve potential risks like empty variable expansions in package installation commands. This is considered acceptable because the containers only run on RHEL 9 with known yum/dnf behavior, and upstream consistency is prioritized over defensive coding practices.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#1197
File: runtimes/minimal/ubi9-python-3.11/requirements.txt:395-405
Timestamp: 2025-06-26T15:28:35.416Z
Learning: psutil version 7.x is compatible with UBI9, CentOS Stream 9, and RHEL 9 platforms in the opendatahub-io/notebooks repository. The upgrade from psutil 5.x to 7.x has been validated for these environments.
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: jiridanek's team uses containerized dependency locking for cross-platform compatibility in opendatahub-io/notebooks. They run `pipenv lock` inside UBI9 containers with specific platform arguments (`--platform=linux/amd64 --python-version 3.12`) to avoid host OS dependency conflicts when generating Pipfile.lock files.
runtimes/tensorflow/ubi9-python-3.11/requirements.txt (12)
undefined
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: tensorflow_rocm package versions 2.12.1.570 through 2.14.0.600 do not provide Python 3.12 wheels (cp312) on PyPI, causing Pipfile lock failures when attempting to create Python 3.12-based ROCm TensorFlow notebook images in opendatahub-io/notebooks.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with extras = ["and-cuda"] can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1340 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, testing approach, and implementation timeline, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: TrustyAI's jupyter-bokeh was pinned to 3.0.5 due to compatibility requirements with TrustyAI's visualization components, but the actual deployed version in requirements.txt shows 3.0.7, indicating incremental testing. The upgrade to 4.0.5 in this PR represents the completion of a gradual migration strategy from the 3.x series after confirming compatibility with Bokeh 3.7.3.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1345 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, and testing approach, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1333
File: runtimes/rocm-tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-09T08:07:24.937Z
Learning: jiridanek requested GitHub issue creation for tensorflow_rocm Python 3.12 compatibility problem during PR #1333 review. Issue #1354 was successfully created with comprehensive problem description covering missing cp312 wheels causing build failures, three solution options (upstream TensorFlow, Python 3.11 only, custom build), clear acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-14T15:36:29.788Z
Learning: ROCm is not supported by upstream TensorFlow, unlike CUDA. ROCm support requires the specialized tensorflow_rocm package, which creates a dependency challenge for Python 3.12 environments since tensorflow_rocm lacks Python 3.12 wheels while regular TensorFlow lacks ROCm support.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow with runtime ROCm configuration is the recommended alternative approach.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review, affecting both Python 3.11 and 3.12 versions. Issue #1338 was created with comprehensive problem description covering both affected images, repository pattern analysis comparing correct vs incorrect naming conventions, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. The tensorflow-rocm upstream project appears abandoned with the last release in 2019. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow 2.18+ with runtime ROCm configuration is the recommended and industry-standard approach, as modern TensorFlow automatically detects and utilizes ROCm when properly installed.
</retrieved_learning>
<retrieved_learning>
Learnt from: grdryn
PR: #1320
File: rstudio/rhel9-python-3.11/Dockerfile.cuda:34-35
Timestamp: 2025-07-04T10:41:13.061Z
Learning: In the opendatahub-io/notebooks repository, when adapting NVIDIA CUDA Dockerfiles, the project intentionally maintains consistency with upstream NVIDIA patterns even when it might involve potential risks like empty variable expansions in package installation commands. This is considered acceptable because the containers only run on RHEL 9 with known yum/dnf behavior, and upstream consistency is prioritized over defensive coding practices.
</retrieved_learning>
jupyter/tensorflow/ubi9-python-3.12/requirements.txt (16)
undefined
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: TrustyAI's jupyter-bokeh was pinned to 3.0.5 due to compatibility requirements with TrustyAI's visualization components, but the actual deployed version in requirements.txt shows 3.0.7, indicating incremental testing. The upgrade to 4.0.5 in this PR represents the completion of a gradual migration strategy from the 3.x series after confirming compatibility with Bokeh 3.7.3.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: tensorflow_rocm package versions 2.12.1.570 through 2.14.0.600 do not provide Python 3.12 wheels (cp312) on PyPI, causing Pipfile lock failures when attempting to create Python 3.12-based ROCm TensorFlow notebook images in opendatahub-io/notebooks.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1333
File: runtimes/datascience/ubi9-python-3.12/Pipfile:33-34
Timestamp: 2025-07-08T19:26:17.140Z
Learning: jiridanek requested GitHub issue creation for jupyter-client dependency pinning inconsistency during PR #1333 review, specifically asking to note the implications of breaking changes in 9.x versions. Issue #1343 was created with comprehensive problem description covering inconsistent pinning style across all Python 3.12 runtime images, detailed breaking changes analysis (kernel protocol, session management, connection security, API changes, async/await modifications), reproducibility and security impact assessment, multiple solution options with code examples, phased acceptance criteria, implementation guidance, testing approach, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: TrustyAI explicitly declares jupyter-bokeh~=3.0.5 as a hard dependency in both their requirements.txt and pyproject.toml files, with no open issues tracking jupyter-bokeh 4.x compatibility. This creates an unresolvable pip dependency conflict when trying to upgrade jupyter-bokeh to 4.x in notebook images that include TrustyAI.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with extras = ["and-cuda"] can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: The jupyter-bokeh package was previously pinned to version 3.0.5 in the TrustyAI notebook image due to compatibility requirements with TrustyAI components, as indicated by the comment "Should be pinned down to this version in order to be compatible with trustyai" that was removed in this update.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1333
File: runtimes/rocm-tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-09T08:07:24.937Z
Learning: jiridanek requested GitHub issue creation for tensorflow_rocm Python 3.12 compatibility problem during PR #1333 review. Issue #1354 was successfully created with comprehensive problem description covering missing cp312 wheels causing build failures, three solution options (upstream TensorFlow, Python 3.11 only, custom build), clear acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review, affecting both Python 3.11 and 3.12 versions. Issue #1338 was created with comprehensive problem description covering both affected images, repository pattern analysis comparing correct vs incorrect naming conventions, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: TrustyAI 0.6.1 (latest version as of June 2025) has a hard dependency constraint on jupyter-bokeh~=3.0.5, preventing upgrades to jupyter-bokeh 4.x in notebook images that include TrustyAI. This requires either waiting for TrustyAI to update their dependency or excluding TrustyAI from jupyter-bokeh upgrades.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1154
File: manifests/base/jupyter-pytorch-notebook-imagestream.yaml:0-0
Timestamp: 2025-06-16T11:06:33.139Z
Learning: In the opendatahub-io/notebooks repository, N-1 versions of images in manifest files (like imagestream.yaml files) should not be updated regularly. The versions of packages like codeflare-sdk in N-1 images are frozen to what was released when the image was moved from N to N-1 version. N-1 images are only updated for security vulnerabilities of packages, not for regular version bumps. This is why the version of packages in N-1 images may be quite old compared to the latest N version.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: jiridanek's team uses containerized dependency locking for cross-platform compatibility in opendatahub-io/notebooks. They run pipenv lock inside UBI9 containers with specific platform arguments (--platform=linux/amd64 --python-version 3.12) to avoid host OS dependency conflicts when generating Pipfile.lock files.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow with runtime ROCm configuration is the recommended alternative approach.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. The tensorflow-rocm upstream project appears abandoned with the last release in 2019. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow 2.18+ with runtime ROCm configuration is the recommended and industry-standard approach, as modern TensorFlow automatically detects and utilizes ROCm when properly installed.
</retrieved_learning>
<retrieved_learning>
Learnt from: grdryn
PR: #1320
File: rstudio/rhel9-python-3.11/Dockerfile.cuda:34-35
Timestamp: 2025-07-04T10:41:13.061Z
Learning: In the opendatahub-io/notebooks repository, when adapting NVIDIA CUDA Dockerfiles, the project intentionally maintains consistency with upstream NVIDIA patterns even when it might involve potential risks like empty variable expansions in package installation commands. This is considered acceptable because the containers only run on RHEL 9 with known yum/dnf behavior, and upstream consistency is prioritized over defensive coding practices.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-14T15:36:29.788Z
Learning: ROCm is not supported by upstream TensorFlow, unlike CUDA. ROCm support requires the specialized tensorflow_rocm package, which creates a dependency challenge for Python 3.12 environments since tensorflow_rocm lacks Python 3.12 wheels while regular TensorFlow lacks ROCm support.
</retrieved_learning>
jupyter/tensorflow/ubi9-python-3.11/requirements.txt (15)
undefined
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: TrustyAI's jupyter-bokeh was pinned to 3.0.5 due to compatibility requirements with TrustyAI's visualization components, but the actual deployed version in requirements.txt shows 3.0.7, indicating incremental testing. The upgrade to 4.0.5 in this PR represents the completion of a gradual migration strategy from the 3.x series after confirming compatibility with Bokeh 3.7.3.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: opendatahub-io/notebooks#0
File: :0-0
Timestamp: 2025-07-01T10:41:56.419Z
Learning: In the opendatahub-io/notebooks repository, TensorFlow packages with extras = ["and-cuda"] can cause build conflicts on macOS due to platform-specific CUDA packages. When the Dockerfile installs CUDA system-wide, removing the extras and letting TensorFlow find CUDA at runtime resolves these conflicts.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:15:41.168Z
Learning: The jupyter-bokeh package was previously pinned to version 3.0.5 in the TrustyAI notebook image due to compatibility requirements with TrustyAI components, as indicated by the comment "Should be pinned down to this version in order to be compatible with trustyai" that was removed in this update.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: The Python 3.11 infrastructure for ROCm TensorFlow images in opendatahub-io/notebooks is already properly configured in the Makefile with both BASE_DIRS entries for Pipfile lock renewals and all-images targets for CI builds, requiring only commenting out the corresponding Python 3.12 entries when downgrading due to wheel availability issues.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: TrustyAI explicitly declares jupyter-bokeh~=3.0.5 as a hard dependency in both their requirements.txt and pyproject.toml files, with no open issues tracking jupyter-bokeh 4.x compatibility. This creates an unresolvable pip dependency conflict when trying to upgrade jupyter-bokeh to 4.x in notebook images that include TrustyAI.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1340 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, testing approach, and implementation timeline, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: TrustyAI 0.6.1 (latest version as of June 2025) has a hard dependency constraint on jupyter-bokeh~=3.0.5, preventing upgrades to jupyter-bokeh 4.x in notebook images that include TrustyAI. This requires either waiting for TrustyAI to update their dependency or excluding TrustyAI from jupyter-bokeh upgrades.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1333
File: runtimes/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-08T19:29:32.006Z
Learning: jiridanek requested GitHub issue creation for investigating TensorFlow "and-cuda" extras usage patterns during PR #1333 review. Issue #1345 was created with comprehensive investigation framework covering platform-specific analysis, deployment scenarios, TensorFlow version compatibility, clear acceptance criteria, and testing approach, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1306
File: jupyter/trustyai/ubi9-python-3.12/kustomize/base/kustomization.yaml:8-12
Timestamp: 2025-07-08T19:09:48.746Z
Learning: jiridanek requested GitHub issue creation for misleading CUDA prefix in TrustyAI image tags during PR #1306 review, affecting both Python 3.11 and 3.12 versions. Issue #1338 was created with comprehensive problem description covering both affected images, repository pattern analysis comparing correct vs incorrect naming conventions, clear solution with code examples, detailed acceptance criteria, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1218
File: jupyter/trustyai/ubi9-python-3.11/Pipfile:49-49
Timestamp: 2025-06-28T14:21:09.429Z
Learning: The jupyter-bokeh pinning to 3.0.5 in TrustyAI notebook image was not due to TrustyAI code compatibility issues, but because the trustyai package itself explicitly declares jupyter-bokeh~=3.0.5 as a hard dependency, causing pip dependency resolution conflicts when trying to upgrade to jupyter-bokeh 4.x.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:03:45.020Z
Learning: tensorflow_rocm package versions 2.12.1.570 through 2.14.0.600 do not provide Python 3.12 wheels (cp312) on PyPI, causing Pipfile lock failures when attempting to create Python 3.12-based ROCm TensorFlow notebook images in opendatahub-io/notebooks.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow with runtime ROCm configuration is the recommended alternative approach.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-03T07:05:33.329Z
Learning: tensorflow_rocm package has no Python 3.12 or 3.13 wheel support as of July 2025, with the latest version 2.14.0.600 only supporting Python 3.9, 3.10, and 3.11. The tensorflow-rocm upstream project appears abandoned with the last release in 2019. For Python 3.12+ ROCm TensorFlow environments, regular TensorFlow 2.18+ with runtime ROCm configuration is the recommended and industry-standard approach, as modern TensorFlow automatically detects and utilizes ROCm when properly installed.
</retrieved_learning>
<retrieved_learning>
Learnt from: grdryn
PR: #1320
File: rstudio/rhel9-python-3.11/Dockerfile.cuda:34-35
Timestamp: 2025-07-04T10:41:13.061Z
Learning: In the opendatahub-io/notebooks repository, when adapting NVIDIA CUDA Dockerfiles, the project intentionally maintains consistency with upstream NVIDIA patterns even when it might involve potential risks like empty variable expansions in package installation commands. This is considered acceptable because the containers only run on RHEL 9 with known yum/dnf behavior, and upstream consistency is prioritized over defensive coding practices.
</retrieved_learning>
<retrieved_learning>
Learnt from: jiridanek
PR: #1259
File: jupyter/rocm/tensorflow/ubi9-python-3.12/Pipfile:13-15
Timestamp: 2025-07-14T15:36:29.788Z
Learning: ROCm is not supported by upstream TensorFlow, unlike CUDA. ROCm support requires the specialized tensorflow_rocm package, which creates a dependency challenge for Python 3.12 environments since tensorflow_rocm lacks Python 3.12 wheels while regular TensorFlow lacks ROCm support.
</retrieved_learning>
🔇 Additional comments (19)
runtimes/tensorflow/ubi9-python-3.12/Dockerfile.cuda (1)
150-153: Arm64 still breaks on OC client download earlier in the fileThis PR’s goal is arm64 support, yet line 21 still uses
$(uname -m)which returnsaarch64, while the mirror path requiresarm64. The failure is immediate (HTTP 404) and was already tracked in issue #1356.Not blocking this exact hunk, but the image won’t build on arm64 until that is fixed.
Consider switching to the mapping that other Dockerfiles now use:ARCH_MAPPING="$( [ "${TARGETARCH}" = "arm64" ] && echo arm64 || echo x86_64 )" curl -L "https://mirror.openshift.com/pub/openshift-v4/${ARCH_MAPPING}/clients/ocp/stable/openshift-client-linux.tar.gz" ...⛔ Skipped due to learnings
Learnt from: jiridanek PR: opendatahub-io/notebooks#1320 File: jupyter/pytorch/ubi9-python-3.12/Dockerfile.cuda:42-52 Timestamp: 2025-07-09T12:29:56.162Z Learning: jiridanek requested GitHub issue creation for OpenShift client architecture mapping problem affecting 29 Dockerfiles during PR #1320 review. Issue was created with comprehensive analysis covering all affected files using $(uname -m) returning 'aarch64' but OpenShift mirror expecting 'arm64', systematic solution using BuildKit TARGETARCH mapping with proper amd64→x86_64 and arm64→arm64 conversion, detailed acceptance criteria, and implementation guidance, continuing the established pattern of systematic code quality improvements through detailed issue tracking.Learnt from: jiridanek PR: opendatahub-io/notebooks#1320 File: jupyter/pytorch/ubi9-python-3.12/Dockerfile.cuda:42-52 Timestamp: 2025-07-09T14:02:39.880Z Learning: grdryn corrected CodeRabbit's false assessment about OpenShift mirror server architecture support during PR #1320 review. The OpenShift mirror server (https://mirror.openshift.com/pub/openshift-v4/) actually supports both aarch64 and arm64 directory structures, making the current code using $(uname -m) work correctly across architectures without modification. This demonstrates the importance of verifying external service capabilities before flagging issues.Learnt from: jiridanek PR: opendatahub-io/notebooks#1320 File: jupyter/pytorch/ubi9-python-3.12/Dockerfile.cuda:42-52 Timestamp: 2025-07-09T14:02:39.880Z Learning: grdryn corrected CodeRabbit's false assessment about OpenShift mirror server architecture support during PR #1320 review. The OpenShift mirror server (https://mirror.openshift.com/pub/openshift-v4/) actually supports both aarch64 and arm64 directory structures, making the current code using $(uname -m) work correctly across architectures without modification. This demonstrates the importance of verifying external service capabilities before flagging issues.Learnt from: atheo89 PR: opendatahub-io/notebooks#1258 File: codeserver/ubi9-python-3.11/Dockerfile.cpu:32-32 Timestamp: 2025-07-07T11:08:48.524Z Learning: atheo89 requested GitHub issue creation for multi-architecture Dockerfile improvements during PR #1258 review, specifically for enhancing structural consistency across Docker stages, replacing $(uname -m) with ${TARGETARCH} for cross-architecture builds, and adding OCI-compliant metadata labels. Issue #1332 was created with comprehensive problem description, phased implementation approach, detailed acceptance criteria, implementation guidance with code examples, and proper context linking, continuing the established pattern of systematic code quality improvements.Learnt from: jiridanek PR: opendatahub-io/notebooks#1333 File: runtimes/rocm-tensorflow/ubi9-python-3.12/Dockerfile.rocm:50-50 Timestamp: 2025-07-08T19:30:01.738Z Learning: jiridanek requested GitHub issue creation for multi-architecture support in ROCm TensorFlow image during PR #1333 review. Issue #1346 was created with comprehensive problem description covering hardcoded x86_64 architecture breaking multi-arch support, detailed impact analysis, three solution options (runtime detection, BuildKit TARGETARCH integration, hybrid approach) with pros/cons analysis, comprehensive acceptance criteria covering core requirements and testing, phased implementation guidance, related files identification, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.Learnt from: jiridanek PR: opendatahub-io/notebooks#1320 File: jupyter/pytorch/ubi9-python-3.12/Dockerfile.cuda:65-66 Timestamp: 2025-07-09T12:31:02.033Z Learning: jiridanek requested GitHub issue creation for MSSQL repo file hardcoding problem during PR #1320 review. Issue #1363 was created and updated with comprehensive problem description covering hardcoded x86_64 MSSQL repo files breaking multi-architecture builds across 10 affected Dockerfiles (including datascience, CUDA, ROCm, and TrustyAI variants), detailed root cause analysis, three solution options with code examples, clear acceptance criteria for all image types, implementation guidance following established multi-architecture patterns, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.Learnt from: jiridanek PR: opendatahub-io/notebooks#1320 File: scripts/buildinputs/dockerfile.go:35-35 Timestamp: 2025-07-10T15:02:13.228Z Learning: jiridanek requested GitHub issue creation for improving architecture detection in buildinputs tool during PR #1320 review. Issue #1364 was created with comprehensive problem description covering qemu-user s390x build failures where runtime.GOARCH returns host architecture instead of target architecture, detailed impact analysis, four solution options (environment variable detection, configuration parameter, platform detection logic, hybrid approach), clear acceptance criteria for multi-architecture support, phased implementation guidance, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.Learnt from: jiridanek PR: opendatahub-io/notebooks#1320 File: scripts/buildinputs/dockerfile.go:35-35 Timestamp: 2025-07-10T15:02:13.228Z Learning: jiridanek requested GitHub issue creation for improving architecture detection in buildinputs tool during PR #1320 review. The automated issue creation failed despite appearing successful, so comprehensive issue content was provided for manual creation covering qemu-user s390x build failures where runtime.GOARCH returns host architecture instead of target architecture, detailed impact analysis, four solution options (environment variable detection, configuration parameter, platform detection logic, hybrid approach), clear acceptance criteria for multi-architecture support, phased implementation guidance, and proper context linking, continuing the established pattern of systematic code quality improvements through detailed issue tracking.Learnt from: grdryn PR: opendatahub-io/notebooks#1320 File: rstudio/rhel9-python-3.11/Dockerfile.cuda:40-42 Timestamp: 2025-07-04T17:08:02.399Z Learning: In the opendatahub-io/notebooks repository, when using multi-architecture Dockerfiles with BuildKit, the implicit build argument TARGETARCH is automatically available in the global scope for FROM instructions without explicit declaration. However, if TARGETARCH is used within a build stage, it must be declared explicitly within that stage. The current placement pattern (declaring ARG TARGETARCH after FROM instructions that use it) is correct for modern Docker/Podman/Buildah environments and does not require compatibility with older Docker versions.Learnt from: jiridanek PR: opendatahub-io/notebooks#1259 File: jupyter/rocm/tensorflow/ubi9-python-3.12/Dockerfile.rocm:34-37 Timestamp: 2025-07-02T18:19:23.024Z Learning: In the opendatahub-io/notebooks repository, issue #1241 "Security: Add checksum verification for downloaded binaries in Python 3.12 images" covers the security concern about verifying downloaded binaries including the oc client. Red Hat does not provide signed RPM packages for the OpenShift CLI (oc) in UBI9 official repositories, making manual binary download with checksum verification the recommended approach.jupyter/tensorflow/ubi9-python-3.11/requirements.txt (2)
1587-1589: Check NCCL 2.23.4 host-runtime compatibility
nvidia-nccl-cu12is bumped to 2.23.4 – good for TF 2.19, but it silently ships its own.sofiles.
Verify that:
- No older NCCL libs are layered in the Docker image (e.g. from CUDA 12.5 base).
- The host driver you target (R470 / R535) accepts this user-space version – older drivers reject ≥2.23.
If either fails you’ll get “unresolved symbol” or “unsupported NCCL version” at runtime.
2880-2883: tensorboard 2.19 upgrade LGTM, but mind the Keras 3 seriesTB 2.19 pairs with TF 2.19 – no objections.
Just note that Keras 3 .x (already at 3.10.0 in this file) removed its TensorBoard callback. Nothing to change here, but keep it in mind when documenting examples.runtimes/tensorflow/ubi9-python-3.11/requirements.txt (2)
2537-2538: TensorBoard bump aligns with TF 2.19 – looks good
No further action required.
1340-1342: Verify NCCL 2.23.4 ↔ CUDA 12.5 CompatibilityThe
fdcheck failed because/usr/local/cudaisn’t available on the host. Please verify from within the built image that/usr/local/cuda/lib/libnccl.so.2is still linked against the CUDA 12.5 runtime. For example:# Replace <image> with your built image name docker run --rm -it <image> ldd /usr/local/cuda/lib/libnccl.so.2 | grep "cuda"Or:
docker run --rm -it <image> bash -c "ldd /usr/local/cuda/lib/libnccl.so.2"This will confirm whether the NCCL 2.23.4 wheel remains ABI-compatible with CUDA 12.5.
runtimes/tensorflow/ubi9-python-3.12/requirements.txt (5)
2519-2520: TensorBoard 2.19 must match TensorFlow runtimeGood to bump
tensorboardin lock-step with TensorFlow, but the package still
ships x86-only native extensions. Please validate an aarch64 wheel exists; if
not, build will fail on ARM.
1337-1339: Check CUDA/NCCL version alignment
nvidia-nccl-cu12==2.23.4targets CUDA 12.4/12.5 runtime. The image installs
CUDA 12.5.82 toolchain – this is fine, but mixing minor lines (e.g. 12.4 vs
12.5) silently breaks collective ops. Double-check the NVIDIA compatibility
matrix or bump to the matching2.23.5-1+cuda12.5build if available.
210-212: Version bumps look saneNo compatibility concerns detected for
certifi 2025.7.14ornarwhals 1.47.0.Also applies to: 1245-1247
751-753: Verify complete h5py wheel hashes for all architecturesOnly two hashes are listed for
h5py==3.14.0, but PyPI provides multiple wheels (e.g. manylinux_2_28_x86_64, manylinux_2_28_aarch64, cp312, etc.). Using--require-hasheswithout the full set will cause installs to fail on ARM64 and other platforms.
- File:
runtimes/tensorflow/ubi9-python-3.12/requirements.txt(lines 751–753)- Restore the complete list of hashes by regenerating this file with
scripts/sync-requirements-txt.shin an ARM64 container.To verify locally:
# List all available h5py 3.14.0 filenames and sha256 hashes curl -s https://pypi.org/pypi/h5py/3.14.0/json \ | jq -r '.releases["3.14.0"][] | .filename + " " + .digests.sha256' # Confirm which hashes are currently in requirements.txt grep -A2 -n "^h5py==3.14.0" runtimes/tensorflow/ubi9-python-3.12/requirements.txt
2525-2541: TensorFlow 2.19.0 cp312 wheels are available for both ARM64 and x86_64
- ARM64: tensorflow-2.19.0-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
- x86_64: tensorflow-2.19.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
Next steps before merging:
- Inspect the wheel’s METADATA or perform a quick import/runtime smoke-test to confirm it’s built against CUDA 12.5 (to match the Dockerfile’s driver/toolkit).
- Reevaluate if the
[and-cuda]extra is still needed—per issue #1340 it may be redundant given the OS-level CUDA installation.jupyter/tensorflow/ubi9-python-3.12/requirements.txt (9)
143-145: Astroid bump is fine, just keep an eye on Pylint↔Astroid couplingAstroid 3.3.11 is the companion release for Pylint 3.3.x, which you already upgraded. No action needed.
257-259: Security-driven Certifi refresh LGTMBumping certifi to 2025-07-14 is a low-risk, security-motivated change.
913-915: Only two hashes left for h5py – verify build reproducibilityPrevious revisions carried hash triples for many platforms. Dropping to two hashes will fail
pip --require-hasheson any build where the resolved wheel’s hash isn’t listed (e.g., musl or future CPython builds).- --hash=... - ... + # At least one hash per expected wheel platform/py-tag is required.Confirm that the two remaining hashes cover all wheels Pipenv installs in CI (amd64 & aarch64, cp311 & cp312).
1482-1484: Narwhals 1.47.0 ships a minor public API tweakThe
Series.to_pyarrow()return type changed fromConversionResult→ArrowOutput. Nothing in our codebase currently calls that, so the bump is safe.
2561-2562: Rope 1.14 resolves the Jupyter-LSP freeze bug – good catch
2881-2882: TensorBoard 2.19 requires protobuf < 4 — already satisfiedNothing else to flag.
1586-1588: CUDA wheels include aarch64 builds, no blockerVerified that both
nvidia-nccl-cu12==2.23.4andnvidia-cublas-cu12==12.5.3.2publish “manylinux2014_aarch64” wheels on PyPI, so the ARM64 build target will succeed.
- Location:
jupyter/tensorflow/ubi9-python-3.12/requirements.txtlines 1586–1588
1643-1657: Stable opentelemetry-exporter-prometheus / semantic-conventions v1.x not available
PyPI shows bothopentelemetry-exporter-prometheusandopentelemetry-semantic-conventionsare still on the 0.56b0 beta line—no stable 1.x release (e.g. “1.35.0”) exists yet. You can’t bump them to a 1.35.0 version that doesn’t exist. To keep your pins consistent, continue using:
opentelemetry-exporter-prometheus==0.56b0opentelemetry-semantic-conventions==0.56b0Revisit this when their stable 1.x releases are officially published.
Likely an incorrect or invalid review comment.
2887-2903: Smoke-test couldn’t run TensorFlow—please verify inside the built imageThe smoke-test script failed with
ModuleNotFoundError: No module named 'tensorflow', indicating it wasn’t executed in the new image where TF 2.19.0 is installed. Once you’ve built the UBI9 Python 3.12 image, please rerun the following inside that container to confirm the upgrade:python - <<'PY' import tensorflow as tf, platform print("TF:", tf.__version__, "CUDA:", tf.sysconfig.get_build_info()["cuda_version"]) print("Platform:", platform.machine()) tf.constant([1.]).cuda() PY
|
@coderabbitai[bot]: changing LGTM is restricted to collaborators In response to this: Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. |
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: coderabbitai[bot] The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
Skipped: Cannot respond to another bot. |
When installing python wheels that contain native bits, the Pipfile.lock will only contain artifact hashes for the architecture that `pip lock` was run against, along with the source archive hash. So when installing on a different architecture, pip will attempt to compile from the source archive, and therefore will need the appropriate development files for the native dependencies that are used by the programs that it's compiling. In this case, the h5py python package is needed in the tensorflow images, and to compile the native shared object files that it contains from source, the libhdf5.so file from hdf5-devel is needed. The compiled object files will be dynamically linked to the .so files from hdf5 (so technically the hdf5-devel is only needed at compile time, then hdf5 at runtime, but since the compilation is only done on _some_ architectures, there isn't a dedicated build stage for these python packages, so to try to make minimal changes, the -devel package is left in place. On the architecture that the Pipfile.lock was generated on (x86_64), the native bits are downloaded pre-compiles as before. This makes things a little weird, as on x86_64 we'll have .so files that are precompiled and link to other .so files downloaded from PyPI, whereas on aarch64 we'll have .so files that were compiled as part of the build and linked to other .so files from hdf5 and other RPMs from the system.
For the tensorflow CUDA images, when trying to build them on aarch64, the `pip install` stage fails because of the lack of support for the version of nvidia-nccl-cu12 (2.21.5) requested by tensorflow 2.18. Since it's a proprietary package, there also isn't a source distribution, so it can't just be compiled at installation time. Updating to tensorflow 2.19.0 pulls in a newer nvidia-nccl-cu12 (2.23.4), which does have wheels available for both x86_64 and aarch64 on PyPI.
grdryn
force-pushed
the
tensorflow-cuda-arm64
branch
from
fb3840e to
7622fc5
Compare
|
@grdryn: The following tests failed, say
Full PR test history. Your PR dashboard. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. I understand the commands that are listed here. |
Description
There are a couple of commits here. See the individual commit message descriptions for more info on each of them (I put as much details as I could into them).
This proposes updating to a new minor version of tensorflow, and I don't know what the consequences of that are. Is this something that you'd be open to, or what are the conditions that need to be met to bump the tensorflow version?
How Has This Been Tested?
Merge criteria:
Summary by CodeRabbit
Summary by CodeRabbit
New Features
Bug Fixes
Chores