conversion-inference
Convert SketchCode PNG wireframes to GUI DSL and HTML, compile/debug GUI DSL, choose output style, and check inference prerequisites.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- conversion-inference
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Convert SketchCode PNG wireframes to GUI DSL and HTML, compile/debug GUI DSL, choose output style, and check inference prerequisites.
- disable-model-invocation
- true
02 资源文件 7
- 参考文档references/api-reference.md4.9 KB
- 参考文档references/cli-reference.md5.6 KB
- 参考文档references/dsl-and-styles.md5.1 KB
- 参考文档references/troubleshooting.md6.0 KB
- 参考文档references/workflows.md5.3 KB
- 脚本scripts/compile_tiny_dsl.py7.5 KB
- 脚本scripts/run_conversion.py6.4 KB
03 SKILL.md 正文
conversion-inference
Use this sub-skill when the user wants to convert hand-drawn/wireframe PNGs with SketchCode, batch-convert mockups, use pretrained model artifacts, compile or debug generated .gui DSL, or choose between the default, facebook, and airbnb HTML styles.
Route first
- For model training, dataset preparation, vocabulary construction, or image augmentation, route to the
training-datasub-skill. - For BLEU scoring details or standalone evaluation of
.guifiles, route to theevaluationsub-skill. This sub-skill only explains how the conversion CLIs expose optional BLEU flags. - For downloading or checking external pretrained/data assets, route to the parent
sketch-codeasset helper when available. Conversion itself requires a Keras model JSON file and an HDF5 weights file. - Do not treat arbitrary non-PNG images as valid inputs; SketchCode's conversion sampler rejects filenames without
.png.
Minimum operating checklist
- Confirm the user has PNG input: one
*.pngfile for single conversion or a directory containing*.pngfiles for batch conversion. - Confirm both pretrained/model artifacts exist:
model_json.json-style model architecture JSON andweights.h5-style Keras weights. - Choose a style: exactly
default,facebook, orairbnb. - Choose an output folder. The conversion CLIs create it when missing; existing files with the same sample basename may be overwritten.
- If the user asks for BLEU during conversion, require matching original
.guifiles and route metric interpretation toevaluation.
References
- Conversion workflows - safe single-image, batch, and DSL compilation flow.
- CLI reference - public conversion flags, required artifacts, output behavior, and preprocessing facts.
- DSL and styles - supported style names, GUI DSL grammar, known tokens, and compiler behavior.
- API reference - distilled
Sampler,Compiler,Node,SamplerUtils, and BLEU-routing facts. - Troubleshooting - missing assets, invalid PNGs, unsupported styles, parsing errors, imports, and output folder issues.
Bundled helpers
- scripts/run_conversion.py is the bundled, validated replacement for the historical conversion entry points. Use it when the user has a SketchCode runtime checkout or importable
classespackage plus model JSON/weights. - scripts/compile_tiny_dsl.py validates and compiles a tiny
.guitoken string without requiring TensorFlow/Keras or pretrained weights. It can optionally use a user-provided SketchCode project root for the original compiler, but its--helpand fallback compiler are self-contained.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).