Routes 3D ResNets PyTorch video action-recognition workflows across training, inference, and data preparation.
中文简介正在准备中。当前显示原文说明。
VectorSpaceLab/AREX-Skill3392026年9月3日 更新
免费在 GitHub 公开
denoising-diffusion-pytorch
Use denoising-diffusion-pytorch for PyTorch DDPM/DDIM image diffusion, 1D sequence diffusion, conditioning and guidance, and advanced diffusion variants.
中文简介正在准备中。当前显示原文说明。
查看安装方法包含的文件（27）
- SKILL.md2.9 KB
- references/package-overview.md2.4 KB
- references/repo-provenance.md1.4 KB
- references/repo-routing-metadata.json540 B
- references/troubleshooting.md1.8 KB
- scripts/check_env.py2.8 KB
- sub-skills/advanced-variants/references/api-reference.md3.7 KB
- sub-skills/advanced-variants/references/compatibility.md3.5 KB
- sub-skills/advanced-variants/references/troubleshooting.md2.9 KB
- sub-skills/advanced-variants/references/workflows.md4.9 KB
- sub-skills/advanced-variants/scripts/smoke_advanced_variants.py5.0 KB
- sub-skills/advanced-variants/SKILL.md2.7 KB
- sub-skills/conditioning-guidance/references/api-reference.md4.3 KB
- sub-skills/conditioning-guidance/references/troubleshooting.md2.8 KB
- sub-skills/conditioning-guidance/references/workflows.md3.1 KB
- sub-skills/conditioning-guidance/scripts/smoke_conditioning_guidance.py4.7 KB
- sub-skills/conditioning-guidance/SKILL.md2.7 KB
- sub-skills/image-diffusion/references/api-reference.md5.0 KB
- sub-skills/image-diffusion/references/troubleshooting.md3.0 KB
- sub-skills/image-diffusion/references/workflows.md4.5 KB
- sub-skills/image-diffusion/scripts/smoke_image_diffusion.py4.1 KB
- sub-skills/image-diffusion/SKILL.md2.4 KB
- sub-skills/sequence-diffusion/references/api-reference.md3.5 KB
- sub-skills/sequence-diffusion/references/troubleshooting.md2.2 KB
- sub-skills/sequence-diffusion/references/workflows.md3.5 KB
- sub-skills/sequence-diffusion/scripts/smoke_sequence_diffusion.py5.6 KB
- sub-skills/sequence-diffusion/SKILL.md2.5 KB
SKILL.md（原文）
安装前可以查看交给 AI 代理的指令内容。
denoising-diffusion-pytorch
Use this repo skill when a Researcher needs to operate the public denoising-diffusion-pytorch package (denoising_diffusion_pytorch import) from self-contained guidance. It covers the inspected public API surface at version 2.3.1.
Install and sanity check
python -m pip install denoising-diffusion-pytorch
python - <<'PYCODE'
from denoising_diffusion_pytorch import Unet, GaussianDiffusion, Unet1D, GaussianDiffusion1D
print('denoising_diffusion_pytorch imports ok')
PYCODE
For a no-training environment check, run scripts/check_env.py. It verifies package metadata, public imports, a tiny CPU tensor operation, and optional CUDA visibility.
Route by task
| User task | Read |
|---|---|
2D image DDPM/DDIM, Unet, GaussianDiffusion, image folder Trainer, FID, RePaint/inpainting, image tensor shape errors | sub-skills/image-diffusion/SKILL.md |
1D/time-series/sequence tensors, Unet1D, GaussianDiffusion1D, Dataset1D, Trainer1D, channel-first vs channel-last layout | sub-skills/sequence-diffusion/SKILL.md |
classifier-free guidance, class labels, cond_scale, CFG++, external classifier cond_fn, XMWrapper multi-candidate loss | sub-skills/conditioning-guidance/SKILL.md |
| Karras UNets, video-shaped Karras 3D, continuous-time / v-param / EDM / simple diffusion, learned variance, weighted objective, flash/SDPA compatibility | sub-skills/advanced-variants/SKILL.md |
Shared references
- references/package-overview.md maps public exports, dependencies, and sub-skill ownership.
- references/troubleshooting.md covers cross-cutting install/import, PyTorch/CUDA, Accelerate, optional FID, and smoke-test failures.
- references/repo-provenance.md records the source snapshot and refresh baseline.
Operating rules
- Use the bundled scripts for smoke tests; do not run long training or FID for routine verification.
- Keep data tensors normalized to
[0, 1]unlessauto_normalize=Falseis explicitly chosen. - CPU is the default verification backend; CUDA, flash attention, FID, and Accelerate multi-GPU are optional accelerators unless a user explicitly asks for them.
- The package has no public CLI entry points. Workflows are Python API workflows.
- If a task asks to edit this repository rather than use the package, this repo skill can explain APIs, but repository-maintenance policy is outside the selected operating graph.
评价
还没有评价。欢迎分享你的使用感受。
评论
无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
同一仓库的技能
3ddfa
免费Guide 3DDFA Python inference, geometry rendering, training/evaluation, and optional C++ ONNX workflows for 3D dense face alignment.
中文简介正在准备中。当前显示原文说明。
VectorSpaceLab/AREX-Skill3392026年9月3日 更新
3ddfa-v2
免费Routes 3DDFA_V2 face-alignment setup, still-image demos, video tracking, and ONNX benchmarking workflows.
中文简介正在准备中。当前显示原文说明。
VectorSpaceLab/AREX-Skill3392026年9月3日 更新
ab3dmot
免费Operate AB3DMOT 3D multi-object tracking workflows for KITTI and nuScenes data, tracking, evaluation, and visualization.
中文简介正在准备中。当前显示原文说明。
VectorSpaceLab/AREX-Skill3392026年9月3日 更新
Use Hugging Face Accelerate for PyTorch training-loop migration, distributed launch/configuration, DeepSpeed/FSDP/TPU backend setup, big-model inference/offload, checkpointing, tracking, and troubleshooting.
中文简介正在准备中。当前显示原文说明。
VectorSpaceLab/AREX-Skill3392026年9月3日 更新
acme
免费Route Acme reinforcement-learning framework tasks across core loops, replay/data, JAX agents, and TensorFlow/Sonnet agents.
中文简介正在准备中。当前显示原文说明。
VectorSpaceLab/AREX-Skill3392026年9月3日 更新