input_programming
Construct masked universal input programs for adversarial reprogramming of frozen neural networks.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- input_programming
- description
- Construct masked universal input programs for adversarial reprogramming of frozen neural networks.
02 资源文件 2
- 脚本scripts/input_programming.py1.2 KB
- 脚本tests/test_input_programming.py280 B
03 SKILL.md 正文
Input-Space Adversarial Program Construction
Use this skill when reproducing or testing adversarial reprogramming methods that embed a small task input into a larger target-model input and add a single reusable trainable program. Do not use it for per-example adversarial attacks where every input receives an independently optimized perturbation.
Inputs
task_values: task data values that occupy protected mask positions.mask: flat or nested binary mask with1for task pixels and0for program pixels.program_values: universal program values for non-task positions.
Outputs
- A programmed input with task values preserved and program values inserted elsewhere.
- Metadata indicating task positions, program positions, and preservation status.
Workflow
- Validate the mask has at least one protected task position and one program position.
- Flatten nested lists deterministically.
- Copy task values into mask positions in order.
- Copy program values into non-mask positions in order; a scalar program may be broadcast.
- Verify task values were preserved exactly.
Validation
Run python tests/test_input_programming.py or validate the whole skill tree with the Paper2Skills skill validator.
Limitations
This script is a deterministic data-construction helper. Gradient updates for the program are handled by the program optimization skill.
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).