메뉴
One-to-three-sentence description of what this data transformation does. Use when the user asks to "REPLACE-WITH-INTENT", or mentions REPLACE-WITH-DOMAIN-TERMS. Include literal trigger phrases the user is likely to say.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Docker conventions for running NVIDIA GPU container workloads — NGC authentication, --gpus flag, mount patterns, env-var passthrough, container inspection, data-root relocation for split-disk hosts, and common error modes. Use when another skill requires running an nvcr.io container or any docker run command on a GPU host. Trigger keywords — docker, docker run, nvcr.io, NGC, --gpus, nvidia-container-toolkit, container image, docker login, docker pull.
One-to-three-sentence description of what the model does and when to use it. Use when the user asks to "fine-tune REPLACE-WITH-NETWORK", "train REPLACE-WITH-NETWORK on REPLACE-WITH-DATA-TYPE", or mentions REPLACE-WITH-DOMAIN-TERMS. Include literal trigger phrases.
Where and how GPU jobs run on this platform. One-to-three-sentence summary. Use when the user asks to "deploy on REPLACE-PLATFORM", "run on REPLACE-PLATFORM", or mentions the platform's distinctive concepts (e.g., resource shape, instance, node group).
Full PAIDF AnomalyGen pipeline — fine-tune on a new anomaly dataset, generate synthetic anomaly images (SDG), evaluate quality (nn_score), and search per-sample (guidance, crop_ratio) parameters. Three modes: full (Phase 0→7: finetune then generate), finetune_only (Phase 0→1: train only), inference_only (Phase 0, 2→7: generate from an existing checkpoint). Use when the user asks to "fine-tune AnomalyGen", "generate anomaly images", "run PAIDF SDG", "evaluate SDG output quality", "run per-sample search", or run any part of the AnomalyGen pipeline, even if they only mention one phase.
One-to-three-sentence description of what the skill does and when to use it. Use when the user asks to "REPLACE-WITH-INTENT-1", "REPLACE-WITH-INTENT-2", or mentions REPLACE-WITH-DOMAIN-TERMS. Include literal trigger phrases the user is likely to say.
Top-level workflow orchestrator for <use case>. Runs an end-to-end pipeline of <N> stages: <stage 1> → <stage 2> → … → <stage N>. Use when the user asks to "REPLACE-WITH-INTENT-1", "REPLACE-WITH-INTENT-2", or mentions REPLACE-WITH-DOMAIN-TERMS. Include literal trigger phrases.
| name | REPLACE-WITH-SKILL-NAME |
| description | One-to-three-sentence description of what this data transformation does. Use when the user asks to "REPLACE-WITH-INTENT", or mentions REPLACE-WITH-DOMAIN-TERMS. Include literal trigger phrases the user is likely to say. |
| license | Apache-2.0 |
| compatibility | REPLACE — typical examples — `Requires docker + nvidia-container-toolkit.` (containerized) or `Requires Python 3.8+ and Pillow.` (local script) |
| metadata | {"author":"REPLACE-WITH-AUTHOR-NAME","version":"0.1"} |
| allowed-tools | Read Bash |
Two-line summary of the transformation. Inputs → outputs.
| Dependency | Purpose | Install |
|---|---|---|
| docker | Run the container | https://docs.docker.com/engine/install/ |
(or for local scripts)
| Dependency | Purpose | Install |
|---|---|---|
| Python 3.8+ | Runtime | System / conda |
| Pillow | Image I/O | pip install pillow |
docker run --gpus all --rm \
-v /path/to/input:/input \
-v /path/to/output:/output \
nvcr.io/nvidian/iva/<image>:<tag> \
<command> --input /input --output /output
python scripts/<script>.py --input-dir <path> --output <path>
for label in bridge tombstone shift; do
python scripts/<script>.py --input-dir data/${label}/ --output ${label}.csv --label ${label}
done
See tao-skill-bank:tao-run-on-docker for docker conventions (when containerized).
| Field | Type | Description |
|---|---|---|
<input-dir> | folder | What's in it, expected format |
<input-parquet> | file | Schema, columns |
| Field | Type | Description |
|---|---|---|
<output-csv> | file | Schema produced |
| Argument | Required | Default | Description |
|---|---|---|---|
--input-dir | Yes | — | Input directory |
--output | No | output.csv | Output path |
--label | No | auto-detect | Force single label for all rows |
Step-by-step description of the transformation:
Optional pipeline diagram:
Input A ─┐
Input B ─┼─→ [processing] ─→ Output
Input C ─┘
| Symptom | Cause | Fix |
|---|---|---|
KeyError: ... | Wrong column name in input parquet | Use --column-name to override |