executing-plans
executing-plans
Use when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available
core-componentsReusable Pyramid-Flow model APIs, Causal Video VAE, diffusion schedulers, and distributed helpers.
Independent GitHub snapshot · Automated classification · Original documentation governs
Use this sub-skill when the task is about Pyramid-Flow's reusable runtime building blocks rather than an end-to-end workflow launcher.
PyramidDiTForVideoGeneration construction, component relationships, text-to-video/image-to-video method signatures, latent decoding, VAE tiling, and CPU offload API semantics.CausalVideoVAE and CausalVideoVAELossWrapper encode/decode/reconstruct APIs, latent shapes, and safe tiny round-trip checks.PyramidFlowMatchEulerDiscreteScheduler and DDPMCosineScheduler initialization, timestep setup, and step() semantics.trainer_misc distributed helpers shared by inference and training: rank/world-size helpers, DDP initialization, sequence-parallel groups, sync-input groups, and FSDP epoch helper entry points.generation-inference.data-preparation.training-workflows.Run the safe component smoke from any working directory once Pyramid-Flow's import roots are importable in the active Python environment:
python PATH_TO_SKILL_TREE/sub-skills/core-components/scripts/smoke_core_components.py --help
python PATH_TO_SKILL_TREE/sub-skills/core-components/scripts/smoke_core_components.py --package-root PATH_TO_PYRAMID_FLOW
The smoke checks imports, package versions, scheduler CPU math, a tiny CPU Causal Video VAE encode/decode round-trip, and optional negative cases without downloading checkpoints or launching generation/training.