AGENT SKILL

data-preparation

Prepare, validate, and troubleshoot sd-scripts training data before LoRA, fine-tuning, Textual Inversion, ControlNet, or inpainting training.

License 许可
Apache-2.0
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
data-preparation
license
Apache 2.0
metadata
{ "disco-role": "operating" }
description
Prepare, validate, and troubleshoot sd-scripts training data before LoRA, fine-tuning, Textual Inversion, ControlNet, or inpainting training.
disable-model-invocation
true

02 资源文件 4

03 SKILL.md 正文

Data Preparation

Use this sub-skill when a user needs to build or debug sd-scripts dataset inputs before launching training. It covers dataset TOML files, DreamBooth directory subsets, fine-tuning metadata JSON/JSONL, captions/tags, validation splits, bucketing/resolution choices, cached latents/text-encoder outputs, and safe data sanity checks.

Route actual training commands to ../training. Route generation prompt files to ../generation. Route LoRA merge/convert or checkpoint utilities to ../model-utilities.

Fast Path

  1. Decide the dataset style per [[datasets]] block:
    • DreamBooth: every subset has image_dir and no metadata_file.
    • Fine-tuning: every subset has metadata_file; image_dir is required when metadata paths are relative.
    • ControlNet-style paired data: every subset has image_dir and conditioning_data_dir.
  2. Put shared defaults in [general], dataset-level shape/batch/bucket settings in [[datasets]], and directory/metadata-specific settings in [[datasets.subsets]].
  3. Validate the TOML and metadata before training:
python skills/sd-scripts/sub-skills/data-preparation/scripts/validate_dataset_inputs.py --dataset-config dataset_config.toml
  1. If the config passes, use ../training to construct the matching accelerate launch ... --dataset_config dataset_config.toml command.

References

  • references/data-formats.md: TOML structure, metadata formats, caption/tag rules, and validation split precedence.
  • references/workflows.md: Practical data prep flows for DreamBooth, fine-tuning metadata, ControlNet/masked loss, resize/bucketing, and cache helpers.
  • references/troubleshooting.md: Common errors and corrective actions.
  • scripts/validate_dataset_inputs.py: Read-only local preflight for TOML, JSON/JSONL, directories, metadata paths, image presence, mixed subset modes, and likely cache-only entries.

Safety Notes

  • The bundled validator is read-only and does not load sd-scripts, model weights, VAE, CLIP, T5, BLIP, WD14, ONNX, TensorFlow, or torch.
  • Cache and caption/tag generation tools can write many files or download models. Treat them as explicit user-approved preprocessing steps, not as automatic validation.
  • Do not mix DreamBooth and fine-tuning subsets inside one [[datasets]]. Split them into separate [[datasets]] blocks if a training script supports both.
  • Inpainting training with --train_inpainting needs source images at training time and is incompatible with latent caching flags.

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