diffusion_schedule
Build and validate DDPM forward-process schedules and closed-form noisy samples for reduced recovery experiments.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- diffusion_schedule
- description
- Build and validate DDPM forward-process schedules and closed-form noisy samples for reduced recovery experiments.
02 资源文件 2
- 脚本scripts/ddpm_schedule.py3.7 KB
- 脚本tests/test_ddpm_schedule.py890 B
03 SKILL.md 正文
Diffusion Schedule
Use this skill when a recovery or implementation task needs the fixed Gaussian DDPM forward process from Ho et al. 2020. It is appropriate for constructing beta schedules, alpha products, posterior variances, and direct samples from q(x_t | x_0). Do not use it for learned reverse-process parameters or image-quality evaluation.
Inputs
timesteps: positive integer number of diffusion steps.beta_start,beta_end: scalars satisfying0 < beta_start <= beta_end < 1.x0: scalar/list clean data values.epsilon: scalar/list Gaussian noise values matchingx0.t: one-based timestep index.
Outputs
- Schedule dictionary with
betas,alphas,alpha_bars, andposterior_variances. - Noisy values computed as
sqrt(alpha_bar_t) * x0 + sqrt(1-alpha_bar_t) * epsilon. - Validation flags for monotone signal decay and valid variances.
Workflow
- Call
scripts/ddpm_schedule.py build --timesteps T --beta-start B0 --beta-end B1for schedule JSON. - Use
forward_sample(schedule, x0, epsilon, t)or the CLIsamplecommand to constructx_tdirectly. - Keep timestep inputs in one-based paper notation; the script handles internal indexing.
- For deterministic tests or recovery, pass explicit epsilon values rather than random sampling.
Validation
Run python tests/test_ddpm_schedule.py or validate the full skill tree with:
python <distiller>/module-to-skill/scripts/validate_skill_tree.py <skill_dir> --run-tests
Limitations
The schedule helper implements a linear schedule for bounded recovery. It is mechanism-faithful for DDPM equations but not a replacement for full image-scale training hyperparameter search.
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).