stochastic_interpolant_protocol
Construct endpoint-valid stochastic interpolant samples with latent schedules and derivative targets for reduced or full generative 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
- stochastic_interpolant_protocol
- description
- Construct endpoint-valid stochastic interpolant samples with latent schedules and derivative targets for reduced or full generative recovery experiments.
02 资源文件 2
03 SKILL.md 正文
Stochastic Interpolant Protocol
Use this skill when a recovery or implementation needs to construct paper-faithful stochastic-interpolant tuples from endpoint samples. Do not use it as a sampler by itself; it only defines the bridge data used to learn fields.
Inputs
- Endpoint values
x0andx1with the same length. - Time values
tin[0, 1], either one scalar or one value per sample. - Latent noise
zwith the same length as the endpoints. - Schedule name; the included script supports
linear_quadratic_gamma.
Outputs
x_t = (1 - t) x0 + t x1 + gamma(t) z.dot_x_t = x1 - x0 + gamma_dot(t) z.gamma,gamma_dot, and endpoint-validity diagnostics.
Workflow
- Validate matching endpoint and latent vector lengths.
- Broadcast scalar time or validate per-sample time values.
- Use the linear interpolant and
gamma(t)=2t(1-t)unless the caller supplies another validated implementation. - Keep score-related division outside this skill because
gamma=0at endpoints. - Save diagnostics when the output is used as recovery evidence.
Validation
Run:
python scripts/interpolant_protocol.py --demo
python tests/test_interpolant_protocol.py
Limitations
This skill intentionally avoids neural models, endpoint coupling optimization, and source-repository assumptions. It preserves the paper's construction contract and leaves field learning to objective skills.
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).