AGENT SKILL

python-api

Programmatic Deep Daze API use for Imagine and DeepDaze construction, prompt/image/encoding workflows, progress saving, optimizers, and troubleshooting.

License 许可
MIT
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
python-api
license
MIT
metadata
{ "disco-role": "operating" }
description
Programmatic Deep Daze API use for Imagine and DeepDaze construction, prompt/image/encoding workflows, progress saving, optimizers, and troubleshooting.
disable-model-invocation
true

02 资源文件 4

03 SKILL.md 正文

python-api

Use this sub-skill when a task needs to call Deep Daze from Python rather than assemble shell commands. It covers safe Imagine and DeepDaze construction, prompt/image/custom-encoding choices, story mode, start-image priming, output path helpers, progress saving, optimizers, reproducibility knobs, and API-level failure modes.

Route first

  • For command-line invocation, flags, quoting, overwrite prompts, and shell workflow construction, route to ../cli-workflows/SKILL.md.
  • For installation, CLIP model cache/download/network diagnosis, CUDA/CPU availability, package compatibility, and backend memory diagnosis, route to ../runtime-and-models/SKILL.md.
  • Stay here for Python call structure, constructor arguments, API defaults, helper methods, and code-level troubleshooting after the environment is usable.

Operating sequence

  1. Decide the objective input: exactly one of text-only, image-only, combined text+image, or a precomputed CLIP encoding unless deliberately replacing the target later with set_clip_encoding.
  2. Pick conservative constructor settings before instantiating Imagine; construction loads CLIP and may perform model retrieval through the runtime cache.
  3. Set open_folder=False for noninteractive agents and control the current working directory because images, progress frames, story transitions, GIFs, and videos are written relative to it.
  4. Use exact optimizer names: AdamP, Adam, or DiffGrad.
  5. Run scripts/probe_api_surface.py --verify when a lightweight API check is needed without constructing Imagine or downloading CLIP weights.
  6. Use the references for details:

Minimal Python pattern

from deep_daze import Imagine

imagine = Imagine(
    text="a house in the forest",
    epochs=1,
    iterations=100,
    save_every=25,
    save_progress=True,
    open_folder=False,
)
imagine()

Safety notes

  • Imagine(...) is not a cheap metadata operation: instantiate only when ready for CLIP loading and potential cache use.
  • Regular text prompts must fit CLIP tokenization; use story mode for longer prose but still validate each story segment.
  • DeepDaze is the lower-level SIREN module and requires an already loaded CLIP perceptor, normalization transform, input resolution, total-batch count, and batch settings.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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).

★ 351