Repository navigation
Conversation
…loses #66) The `compileiq` package on PyPI (v0.0.0a0) is an empty placeholder with no CLI or functional code. Both pilot v1 (PR #66, Python 3.14 blocking theory) and pilot v2 (CUDA 13.3 + Ubuntu 26.04 + Python 3.14 confirmed) are abandoned. Decision recorded in ADR-0742. Recommended next step for filter1d.cu tuning: /profile-hotpath cuda vif with ncu. Deliverables: - docs/adr/0742-cuda-compileiq-filter1d-pilot-v2.md (ADR, Accepted) - docs/adr/README.md (index row added) - docs/research/research-0742-cuda-compileiq-filter1d-pilot-v2.md - core/AGENTS.md (ACF provenance invariant — re-tune on toolkit bump) - changelog.d/added/cuda-compileiq-pilot-filter1d-v2.md - docs/rebase-notes.md (no rebase impact: doc-only) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This was referenced May 28, 2026
Contributor
Author
|
Research-only abandon report (ADR-0742 documented). CompileIQ v2 also abandoned — package is empty PyPI placeholder. The ADR + research digest stand as the record of the investigation. Closing. |
lusoris
added a commit
that referenced
this pull request
Sep 6, 2026
… run (#1349) The Containerfile's own layout comment has always described compat/ as "pip-installed via the python/ shim", and the Netflix golden YUV fetch exists, in its own words, "so that `pytest python/test/` passes inside the container without any host-side manual provisioning". Nothing ever performed that install, so the command died at import with ModuleNotFoundError: No module named 'slugify', and the fixture fetch served no reachable purpose. Found while running the epic #1246 GPU smoke, which had to fall back to the CLI because the suite could not be collected. Installs python/ (package name vmaf, re-exporting compat/python-vmaf per ADR-0700) into /opt/vmaf-venv, first among the editable packages so the ai/ and vmaf-tune pins resolve last and win on shared dependencies. Adds a build-time assertion that the harness imports and the golden suite collects, because the failure mode is silent: the build stayed green for a long time while the command those fixtures exist for could not run at all. --collect-only keeps it to an import plus collection check; the suite itself is a CI gate, not a container build step. The interpreter is spelled /opt/vmaf-venv/bin/python explicitly because ENV PATH is set below this layer, so a bare python3 here would be the system interpreter. VERIFIED by a full dev-mcp build: #72 harness import OK: /build/vmaf/compat/python-vmaf/__init__.py #72 62 tests collected in 0.89s #73 naming to docker.io/library/vmaf-harness-test:local That build needed the nv-codec-headers mirror fallback from #1348 applied locally, because code.ffmpeg.org has been unreachable all day; the fallback was NOT committed here and is not part of this change. Once #1348 lands this branch builds on its own. no docs needed: dev/Containerfile is build infrastructure; the user-visible effect is that a documented command now works, and docs/development/dev-mcp.md already tells operators to run the suite in the container no state delta: closes no tracked bug id -- the gap was between the Containerfile's stated intent and its contents, not a defect in shipped code Co-authored-by: Lusoris <lusoris@pm.me> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
compileiqon PyPI (pypi.org/project/compileiq) is version0.0.0a0— an empty placeholder package with no CLI, norunsubcommand, and no functional code. Thepip install compileiqsucceeds with exit 0 but installs a 3 KB package whose__init__.pycontains only__version__ = "0.0.0a0"and the docstring"Empty package compileiq.".vmaf-dev-mcp:cuda13.3are confirmed working. No code changes needed.Environment verified
vmaf-dev-mcp:cuda13.3(Ubuntu 26.04 LTS, 50.9 GB)release 13.3, V13.3.33(vianvcc --version)compileiqPyPIWhat was tried
--entrypoint bashto bypass SYCL/HIP entrypoint probes (~5 min combined).pip install compileiqin Python 3.14 venv — exit 0, nocompileiqbinary.Empty package compileiq.with no module contents.python3.13for Ubuntu 26.04 — but irrelevant since the package is empty regardless of Python version.Deliverables (CLAUDE §12 r11)
docs/research/research-0742-cuda-compileiq-filter1d-pilot-v2.mddocs/adr/0742-cuda-compileiq-filter1d-pilot-v2.mdcore/AGENTS.mdinvariant: ACF provenance notedocker run --rm --gpus all --entrypoint bash -v $(pwd):/workspace vmaf-dev-mcp:cuda13.3 -c 'python3 -m venv /tmp/civ && /tmp/civ/bin/pip install --verbose compileiq 2>&1 | grep -E "version|Empty|entry"'changelog.d/added/cuda-compileiq-pilot-filter1d-v2.mddocs/rebase-notes.mdentry (no rebase impact: doc-only)Next steps for filter1d.cu tuning
Per ADR-0742, the recommended path is
/profile-hotpath cuda vifwithncu, or a grid-search wrapper oversubprocess.run(nvcc ...)+ meson reconfigure.Closes #66
🤖 Generated with Claude Code