AGENT SKILL

install-and-cli

Covers Anomalib installation choices, the install subcommand, CLI help behavior, config loading, and command selection for fit, validate, test, train, predict, export, and benchmark.

License 许可
Apache-2.0
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
install-and-cli
license
Apache 2.0
metadata
{ "disco-role": "operating" }
description
Covers Anomalib installation choices, the install subcommand, CLI help behavior, config loading, and command selection for fit, validate, test, train, predict, export, and benchmark.
disable-model-invocation
true

02 资源文件 5

03 SKILL.md 正文

Install and CLI

Use this sub-skill when the user wants to install Anomalib, understand the anomalib install subcommand, inspect CLI help, load configs, or choose between fit, validate, test, train, predict, export, and benchmark.

Read first

Safe operating boundary

This sub-skill stays at the command, installation, and routing layer.

It covers:

  • package installation choices for CPU and OpenVINO support;
  • anomalib install options and when they are appropriate;
  • top-level help output, verbose help formatting, and config-file loading;
  • command selection for fit, validate, test, train, predict, export, and benchmark.

It does not cover:

  • model math, dataset schemas, export internals, or training loops;
  • benchmark pipeline mechanics beyond choosing the benchmark command;
  • Studio or UI application code.

Quick routing

User need Route here? Read or run
Smallest CPU plus OpenVINO install Yes Read references/installation.md, then print commands with scripts/cli_recipes.sh
Decide between anomalib install and pip/uv Yes Read references/installation.md
Explain why anomalib -h only shows install Yes Read references/troubleshooting.md
Check top-level help, install help, and verbose help behavior Yes Run scripts/check_cli_help.sh
Choose fit vs train vs validate vs test Yes Read references/cli-reference.md
Choose predict or recover from old --data_path examples Yes Read references/cli-reference.md and references/troubleshooting.md
Diagnose export flags, especially OpenVINO options Yes Read references/cli-reference.md and references/troubleshooting.md
Understand benchmark routing or benchmark config semantics Partly Route the command choice here, then hand off to the sibling pipelines-and-benchmarks sub-skill for workflow details

Operating notes

  • Use direct package install when the user is bootstrapping a fresh environment or needs a specific backend wheel.
  • Use anomalib install only after Anomalib is already installed and the user wants to add optional bundles such as openvino, loggers, or notebooks.
  • Treat cpu as a package extra, not as an anomalib install --option value.
  • Prefer the current --data syntax for prediction examples; older --data_path shell snippets are legacy and should be translated before reuse.
  • If a config file names the right subcommand but the wrong model or data class path, diagnose it as a command/config mismatch rather than a parser failure.
  • If the benchmark command is missing, check whether the pipeline module is importable before assuming the CLI is broken.

Recommended workflow

  1. Confirm whether the user needs a fresh install, an editable source install, or an add-on bundle inside an existing environment.
  2. Read the installation reference and pick the smallest command that satisfies the backend.
  3. Run the CLI smoke helper after installation or dependency changes.
  4. Use the CLI reference to choose the right subcommand and flags.
  5. Escalate to troubleshooting when help output, config loading, or export flags do not match the expected command shape.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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).

★ 351