video-sampling
Run, adapt, and troubleshoot standalone SVD, SVD-XT, SV3D, SV4D, and SV4D2 video sampling workflows from bundled generative-models guidance.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- video-sampling
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Run, adapt, and troubleshoot standalone SVD, SVD-XT, SV3D, SV4D, and SV4D2 video sampling workflows from bundled generative-models guidance.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/cli-reference.md8.5 KB
- 参考文档references/model-and-inputs.md6.5 KB
- 参考文档references/troubleshooting.md9.0 KB
- 参考文档references/workflows.md7.4 KB
- 脚本scripts/inspect_video_sampling_cli.py14.6 KB
03 SKILL.md 正文
Video Sampling
Use this sub-skill when a task mentions simple_video_sample, SVD image-to-video, SV3D camera paths, SV4D video-to-4D, SV4D2, 8-view video-to-4D, low-VRAM encoding_t/decoding_t, video frame folders/globs, or background removal for video sampling.
Route First
- Use
references/workflows.mdto choose between SVD/SVD-XT image-to-video, SV3D_u/SV3D_p multi-view image sampling, SV4D, SV4D2, and SV4D2 8-view mode. - Use
references/cli-reference.mdfor parameter defaults, safe command templates, and Fire-style list/boolean argument examples. - Use
references/model-and-inputs.mdfor checkpoint filenames, config names, frame/view counts, camera-list lengths, input formats, and output naming. - Use
references/troubleshooting.mdfor missing checkpoints, CUDA/VRAM failures, invalid inputs, camera-list mismatch,remove_bg/rembg, video write issues, and SV4D/SV4D2-specific failure modes. - Use
scripts/inspect_video_sampling_cli.pyfor static inspection:python <path-to-this-sub-skill>/scripts/inspect_video_sampling_cli.py --script all --json.
Scope
- This sub-skill covers standalone video sampling workflows driven by
scripts/sampling/simple_video_sample.py,scripts/sampling/simple_video_sample_4d.py, andscripts/sampling/simple_video_sample_4d2.py. - Full sampling is checkpoint-bound and normally CUDA/GPU-bound; prefer the bundled static helper for planning and CI-safe inspection.
- For SDXL Python inference APIs, route to
../inference-api/SKILL.md. - For training or config authoring, route to
../training-and-configs/SKILL.md. - For Streamlit/Gradio app operation or watermark/NSFW filtering internals, route to
../demos-and-watermarking/SKILL.md.
Quick Decisions
- Image-to-video from one image or an image folder: use SVD/SVD-XT via
simple_video_sample.py --version svdor--version svd_xt. - 21-frame object orbit from one image: use SV3D_u via
simple_video_sample.py --version sv3d_u. - Custom 21-step elevation/azimuth camera path from one image: use SV3D_p via
simple_video_sample.py --version sv3d_p. - Video-to-4D with SV3D first-frame reference views: use SV4D via
simple_video_sample_4d.py. - Video-to-4D without SV3D reference generation: use SV4D2 via
simple_video_sample_4d2.py. - 8 novel views per temporal chunk: use SV4D2 8-view mode with
--model_path checkpoints/sv4d2_8views.safetensors.
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).