PyPi Package Deployment + Remove Banshee Dept #154
Conversation
…y into pr/package-deployment
|
Warning Rate limit exceeded
⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. 📝 WalkthroughWalkthroughSwitches Banshee from source builds to prebuilt downloads, removes Rust toolchain bits from images and toolchain, adds a GitHub Actions Publish workflow for PyPI/TestPyPI, updates Makefile/CMake to use BANSHEE_INSTALL_DIR, adjusts container Dockerfiles, relaxes an ONNX dependency, and adds releasing docs and a constant-tensor coercion before ONNX export. Changes
Sequence Diagram(s)mermaid Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
Suggested reviewers
🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches🧪 Generate unit tests (beta)
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. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 5
🤖 Fix all issues with AI agents
In @.github/workflows/_select-env.yml:
- Line 37: Replace the fallback container image value assigned to IMAGE
(currently set to "ghcr.io/victor-jung/deeploy:package-deployment") with the
org-owned registry image "ghcr.io/pulp-platform/deeploy:devel" so CI does not
depend on a personal registry; update the IMAGE assignment accordingly wherever
that fallback is defined.
In `@Deeploy/DeeployTypes.py`:
- Around line 3106-3109: The current loop in DeeployTypes.py unconditionally
casts every gs.Constant's tensor.values to float32 (using tensor.values.astype),
which will corrupt integer/bool/quantized constants; change the logic that
builds or iterates constTensors (from self.graph.tensors().values() and
gs.Constant) to only perform the cast when the constant's dtype is already a
floating type (use numpy's type check, e.g., np.issubdtype(tensor.values.dtype,
np.floating)) and skip casting for integer/bool/quantized tensors so only true
float tensors are converted to np.float32 before export.
In `@docs/releasing.md`:
- Line 17: Clarify that tags are not created automatically on merge to main;
update the sentence about committing/merging to state who must create the
release tag and how (e.g., manually with git tag <version> && git push origin
<tag> or via your release tooling), and note that the release workflow is
triggered by tags matching the pattern "v*" so the tag must be created/pushed to
trigger the workflow; reference the branch name "main" and the tag pattern "v*"
in the new line for clarity.
In `@Makefile`:
- Around line 55-70: The Darwin branch currently hardcodes TARGET :=
aarch64-apple-darwin which mistargets Intel Macs; update the conditional logic
that uses OS and ARCH to check ARCH when OS is Darwin (similar to the Linux
block) and set TARGET to x86_64-apple-darwin for ARCH == x86_64 and
aarch64-apple-darwin for ARCH == aarch64, otherwise emit an unsupported
architecture error; ensure you modify the Makefile's OS/ARCH conditional section
and preserve the existing Linux checks (TARGET, OS, ARCH symbols) while removing
any unfounded special-casing about arm64 on Linux.
- Around line 520-524: The Makefile rule that populates ${BANSHEE_INSTALL_DIR}
currently downloads prebuilt banshee via the curl/tar pipeline using the
$(TARGET) variable without any integrity checks; update this rule to verify the
downloaded artifact before extracting by either fetching a corresponding
checksum or signature and validating it (e.g., download a .sha256 or .asc
alongside the banshee-0.5.0-$(TARGET).tar.gz and run sha256sum -c or gpg
--verify), or if no machine-readable checksum exists, pin and document an
expected SHA256 value and compare it after download; ensure the build fails on
mismatch and reference the same symbols (${BANSHEE_INSTALL_DIR}, TARGET, the
curl download of banshee-0.5.0-$(TARGET).tar.gz) so the change is applied to the
right Makefile target.
🧹 Nitpick comments (3)
pyproject.toml (1)
6-7: Consider adding an upper bound for onnx-graphsurgeon.Relaxing to
>=0.5.8may allow future breaking changes. If you only support a tested major/minor series, add an upper bound and verify compatibility with the latest allowed version.📌 Suggested constraint
-'onnx-graphsurgeon>=0.5.8', +'onnx-graphsurgeon>=0.5.8,<0.6',Also applies to: 25-25
cmake/simulation.cmake (1)
45-52: Validate the banshee executable path.Line 48 sets
BANSHEE_EXECUTABLEbut doesn’t verify it exists. A quick existence check gives a clearer error when the env var is mis-set.✅ Suggested validation
set(BANSHEE_EXECUTABLE "$ENV{BANSHEE_INSTALL_DIR}/banshee") +if(NOT EXISTS "${BANSHEE_EXECUTABLE}") + message(FATAL_ERROR "Banshee executable not found at ${BANSHEE_EXECUTABLE}") +endif()Container/Dockerfile.deeploy (1)
17-17: Use a directory in PATH (or drop this line).
/app/benderlooks like a binary path, not a directory. PATH entries should be directories, otherwise the entry is a no-op. Consider using/app(if the binary resides there) or removing this line since you copybenderinto/binlater.
| IMAGE="ghcr.io/pulp-platform/deeploy:main" | ||
| else | ||
| IMAGE="ghcr.io/pulp-platform/deeploy:devel" | ||
| IMAGE="ghcr.io/victor-jung/deeploy:package-deployment" |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
# First, let's examine the workflow file structure and context around line 37
cat -n .github/workflows/_select-env.yml | head -50Repository: pulp-platform/Deeploy
Length of output: 1858
🏁 Script executed:
# Check git history to see what the previous value was
git log --oneline -p .github/workflows/_select-env.yml | head -200Repository: pulp-platform/Deeploy
Length of output: 1781
🏁 Script executed:
# Search for other usages of these registry images in the codebase
rg "ghcr.io/(victor-jung|pulp-platform)" --max-count=20Repository: pulp-platform/Deeploy
Length of output: 667
🏁 Script executed:
# Search for any comments or documentation about the personal image choice
rg "victor-jung|package-deployment" --context=5Repository: pulp-platform/Deeploy
Length of output: 269
🏁 Script executed:
# Check what tags are available in the registry or documented
rg "deeploy:" .github/ -t yaml | head -20Repository: pulp-platform/Deeploy
Length of output: 1958
🏁 Script executed:
# Check which workflows call this _select-env.yml
rg "select-env" .github/workflows/*.ymlRepository: pulp-platform/Deeploy
Length of output: 11056
🏁 Script executed:
# Look more closely at the logic flow - see how docker_image_deeploy input is used
rg -A 3 "docker_image_deeploy" .github/workflows/ | head -40Repository: pulp-platform/Deeploy
Length of output: 2603
Default image should use org-owned registry.
Line 37 uses ghcr.io/victor-jung/deeploy:package-deployment as a fallback, creating a dependency on a personal registry. This can break CI for forks or future runs if the image becomes inaccessible. Change to the org-owned image ghcr.io/pulp-platform/deeploy:devel, which is already used as the default in all calling workflows.
🔁 Suggested fix
- IMAGE="ghcr.io/victor-jung/deeploy:package-deployment"
+ IMAGE="ghcr.io/pulp-platform/deeploy:devel"📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| IMAGE="ghcr.io/victor-jung/deeploy:package-deployment" | |
| IMAGE="ghcr.io/pulp-platform/deeploy:devel" |
🤖 Prompt for AI Agents
In @.github/workflows/_select-env.yml at line 37, Replace the fallback container
image value assigned to IMAGE (currently set to
"ghcr.io/victor-jung/deeploy:package-deployment") with the org-owned registry
image "ghcr.io/pulp-platform/deeploy:devel" so CI does not depend on a personal
registry; update the IMAGE assignment accordingly wherever that fallback is
defined.
1 participant
Added
publish.ymlaction to build a branch and push it to PyPi. The action is automatically triggered when a tag with the "v*" format is emitted.Makefilenow pulls that release depending on the platform._export_graphassigns their export type to the tensors before export.TODO: Update readme with new pyindex
TODO: Update changelogs
Changed
PULP-Deeploytpdeeploy-pulp.Fixed
install.mdto remove rust mention and fix test command.README.mdto remove reference to NVIDIA's PyPi index.PR Merge Checklist
develcommit and pointing todevel.CHANGELOG.mdfile has been updated.