setup-and-checkpoints
Prepares AnyDoor environments, checkpoints, and config paths.
- License 许可
- MIT
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- setup-and-checkpoints
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Prepares AnyDoor environments, checkpoints, and config paths.
- disable-model-invocation
- true
02 资源文件 2
- 参考文档references/checkpoint-checklist.md1.2 KB
- 参考文档references/troubleshooting.md1.4 KB
03 SKILL.md 正文
Setup and Checkpoints
Use this sub-skill when the task is about installing AnyDoor, verifying the runtime environment, locating or patching checkpoint paths, or explaining why a fresh clone still cannot generate images.
What this sub-skill owns
- Environment setup for AnyDoor’s torch/CUDA stack.
- Repo layout checks and import readiness.
- Placeholder checkpoint and DINOv2 path patching.
- Optional dependency awareness for xformers and demo mask refinement.
- The preflight stage before any inference, demo, or training route.
What this sub-skill does not own
- Mask preprocessing for inference inputs.
- Gradio, Cog, or prediction input/output handling.
- Dataset-specific path formats or training recipes.
Route those topics to the sibling sub-skills.
Read first
../../references/environment-and-installation.md../../references/checkpoints-and-configs.md../../references/troubleshooting.mdreferences/checkpoint-checklist.mdreferences/troubleshooting.md
Use these scripts
../../scripts/check_anydoor_environment.pyfor a safe repo-root preflight.../../scripts/patch_anydoor_configs.pyto replace checkpoint placeholders.
Trigger phrases
This is usually the right branch when the user says things like:
- “I just cloned the repo.”
- “Why does import or CUDA fail?”
- “Where do I put the AnyDoor checkpoint?”
- “What should I do about the DINOv2 weight?”
- “Can I skip xformers?”
- “The demo says it needs a model but I only see placeholders.”
Minimum operational checklist
- Confirm the repo root contains
cldm/,ldm/,datasets/,configs/, anddinov2/. - Confirm the Python environment imports torch and sees CUDA when generation is expected.
- Check that the placeholder values in
configs/inference.yaml,configs/demo.yaml, andconfigs/anydoor.yamlhave been replaced. - Verify the DINOv2 checkpoint exists at the configured path.
- If the demo refinement toggle is enabled, verify the optional
isegweight.
Common decisions
- xformers missing: usually record and continue. The attention code has a fallback path.
- share package unavailable: only the source weight-conversion helper is affected; the rest of the repo can still be documented and checked.
- CPU import only: useful for support checks, but not proof that AnyDoor can generate images.
- Placeholder config values remain: stop and patch before routing to any generation workflow.
Troubleshooting focus
This branch owns installation and backend problems such as:
ModuleNotFoundErrorfrom running outside the repo root.- CUDA wheel or driver mismatches.
- Optional dependency gaps like
xformers. - Missing checkpoint or DINOv2 weights.
- Demo refinement toggle failures due to a missing
isegweight. - The stale conversion-script config path issue.
Output expected from a future agent
A future agent should be able to say:
- which paths are ready,
- which placeholders are still unresolved,
- whether CUDA is actually available,
- which optional dependencies were skipped,
- and what remains blocked before any generation route.
Cross-links
When this branch is complete, pass the task to:
inference-and-demofor mask/image generation tasks.data-and-trainingfor dataset and training tasks.
Quality bar
The branch is ready when the future agent can prepare a repo root, patch the configs, and explain any remaining backend or checkpoint limitations without reopening the source files.
04 同仓库更多 Skills
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
python-api
Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP/mAP semantics, and source-style import diagnostics.
file-evaluation
Folder-based PASCAL VOC AP/mAP evaluation from ground-truth and detection text files.
object-detection-metrics
Guides legacy Object-Detection-Metrics PASCAL VOC AP/mAP evaluation from detection text folders and direct Python API classes.
nx-workspace-patterns
Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.