mirror of
https://github.com/kuhyx/testsAndMisc.git
synced 2026-11-10 20:07:04 +01:00
Krzysztof kuhy Rudnicki
e6e0ba5ab3
Closes the remaining bake-off cells. The phase's real result was not a route
score but a MISSING GATE.
All twelve SDXL icons measured 100% opaque coverage -- square tiles with no
sprite silhouette -- and passed every gate in the harness. Gate-pass was
measuring pipeline hygiene, not usability. MARGIN_TOO_SMALL inverts the table
from a meaningless 12/12/12 tie to:
diffusion 0/12 (every output full-bleed)
vector 10/12
procedural 11/12
The two non-diffusion failures were true positives in MY shapes, not gate
error: the procedural sword pommel ran off the bottom edge and the vector meat
bone clipped the top. Both fixed; both routes back to 12/12.
Diffusion recovers to 12/12 with flood-fill background removal, but visual
inspection shows the fill eats into the art (bone reduced to a fragment, sword
nearly erased). Reported as inadequate rather than shipped -- proper matting is
needed before route (b) is a usable pipeline.
Blender exits 0 even when its embedded script raises: BLENDER_EEVEE_NEXT does
not exist in 5.2, and the failure produced a clean exit and no file. render()
now asserts the output exists, or the route would report phantom successes.
Model gating verified at execution time, never recalled: SDXL base and
nerijs/pixel-art-xl are ungated, FLUX.1-dev is gated=auto, and
runwayml/stable-diffusion-v1-5 now 307-redirects.
The finishing stage now generalises across three continuous-tone sources
(SVG raster, Blender render, SDXL): each is 0/N raw and N/N finished.
Verified: 114 tests, 100% branch coverage on 13 modules, zero suppressions,
all hooks pass. Note: a local pytest run masked a missing branch that the
pre-commit hook caught -- the hook's run is authoritative.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01T9rQ12yZ7sJxiPQTCS5dUF
21 lines
2.3 KiB
JSON
21 lines
2.3 KiB
JSON
{
|
|
"title": "artgate Phase 4: diffusion route, 3D cell, full vector coverage, and the margin gate",
|
|
"objective": "Close the remaining bake-off cells - route (b) local diffusion on the 3090, route (d) 3D via headless Blender, and the other seven vector subjects - so all four routes are measured on the same twelve subjects. The phase's real outcome was a missing gate: diffusion output passed everything while being unusable, which forced MARGIN_TOO_SMALL into existence and inverted the decision table.",
|
|
"acceptance_criteria": [
|
|
"diffusers runs on the GPU in the isolated uv 3.12 venv, with model gating status verified at execution time rather than recalled.",
|
|
"Route (b) generates all twelve subjects locally with no gated download on the critical path.",
|
|
"Route (d) covers all twelve vector subjects, making its number comparable to route (a)'s.",
|
|
"Route (d) 3D renders via headless Blender, with a phantom-success guard: Blender exits 0 even when its embedded script raises.",
|
|
"A margin gate rejects full-bleed assets, because an item icon must be a sprite with a silhouette, not a square tile.",
|
|
"Every gate failure observed on real output is verified as a true positive before any threshold or shape is changed.",
|
|
"100% branch coverage, zero lint suppressions, all pre-commit hooks pass."
|
|
],
|
|
"out_of_scope": [
|
|
"Deciding which route WINS. That needs the human contact-sheet verdicts, which are the user's call; this change only produces the machine column and the sheets.",
|
|
"A production matting solution for diffusion output. The flood-fill background removal demonstrated recoverability but visibly damages the art, and is reported as inadequate rather than shipped as the answer.",
|
|
"Modelling all twelve subjects in 3D. Five meshes answer whether render->finish clears the gates; twelve would be modelling work, not a route comparison.",
|
|
"Gated model families (FLUX.1-dev, SD3.x). Verified gated at execution time and deliberately kept off the critical path."
|
|
],
|
|
"verifier": "python -m pytest python_pkg/artgate/tests/ --cov=python_pkg.artgate --cov-branch --cov-fail-under=100 (114 tests, 100% branch on 13 modules); pre-commit run over all artgate files; plus end-to-end generation and gating of all four routes against styles/pixel_icon.toml, with each observed failure traced to a specific measurable defect."
|
|
}
|