vpt_prompt_injection
Build shallow or deep visual prompt token insertions for VPT-style Transformer inputs.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- vpt_prompt_injection
- description
- Build shallow or deep visual prompt token insertions for VPT-style Transformer inputs.
02 资源文件 2
03 SKILL.md 正文
vpt_prompt_injection
Use this skill when implementing or checking Visual Prompt Tuning mechanisms from Jia et al. Do not use it for unrelated pixel-prompt or full-backbone fine-tuning experiments.
Inputs
Provide small token vectors, prompt settings, named parameter records, training examples, or prediction/label lists depending on the script. Inputs must make the frozen-backbone boundary explicit.
Outputs
The scripts return prompted token sequences, freeze-audit dictionaries, training traces, or metric reports. Recovery users should save these outputs as experiment evidence.
Workflow
- Preserve the class token and image-patch token order.
- Add continuous prompt parameters in input/token space.
- Freeze all backbone parameters and update only prompt/head records.
- Log loss, parameter changes, accuracy, and storage-efficient parameter ratios.
Validation
Run python -m pytest tests or validate this tree with the Distiller validate_skill_tree.py --run-tests command.
Limitations
These helpers are minimal and deterministic. They are intended for mechanism-faithful bounded recovery, not full VTAB/FGVC reproduction.
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).