- name
SKILL.md
技能檔案保留在發布者的儲存庫中。
在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add fcakyon/labrat --skill launch -a cursor -y在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add fcakyon/labrat --skill launch -a claude-code -y在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add fcakyon/labrat --skill launch -a codex -y在終端機中貼上複製的指令。它會將這個 SKILL.md 安裝到對應的代理中。
npx skills add fcakyon/labrat --skill launch -a grok -y在 Grok Bot 聊天中傳送複製的訊息。Bot 會從 GitHub 取得這個 SKILL.md,並加入你所有 Bot 共用的技能中。
Install this skill: https://github.com/fcakyon/labrat/tree/HEAD/plugin/skills/launch來自發布者
Long training jobs are expensive to fail. A 12-hour run that crashes on epoch 3 from a missing dataset path or a default workers=8 against an NFS mount is a full day lost. This skill walks six quick checks before you commit the GPUs.
The agentic Stop hook in this plugin will route here from reason when an assistant tries to launch a run without going through the checklist.
The user just asked to:
Or the user is about to run any of: python train.py, accelerate launch, torchrun, deepspeed, sbatch train.sh, tmux new-session ... python ... train, wandb sweep.
The most expensive failure is launching with the wrong knobs. Before starting:
find configs/ recipes/ experiments/ -maxdepth 3 \( -name '*.yaml' -o -name '*.yml' -o -name '*.json' -o -name '*.toml' \) -mtime -30 2> /dev/null | head
Pick the most-recently-modified config that resembles the intended run (same model family, same task). Diff against the intended config:
diff -u configs/baseline_v1.yaml configs/intended.yaml
Walk every diff line. For each, ask: is this difference intentional and motivated, or is it a stale default I forgot to set? Common silent regressors:
num_workers / dataloader workers (default in many repos is 8: wrong on NFS)batch_size (per-device vs global mismatch under DDP)learning_rate (linearly scaled with batch size; if batch changed, lr should too)optimizer betas / weight decay (paper-default vs framework-default)mixed_precision (fp16 vs bf16 matters for some models)gradient_accumulation_stepsseed (still set if you care about reproducibility)If no reference exists in this project, ask the user to point at one. Do not launch with framework defaults alone.
The run name will live in wandb / neptune / checkpoint dirs / status reports for the rest of its life. It must describe the experiment in plain English without internal codes:
run-1, exp-b, final-v2, phase2-internalimagenet-resnet50-bs256-lr0.1-cosine, coco-baseline-bs256-lr3e-4, swin-t-imagenet-distill-from-vit-lThe pattern: <dataset/task>-<model>-<key-config>-<distinctive-recipe-piece>. If you can't describe the experiment from the name in one sentence, the name is wrong.
Before launching, every path the run depends on must be confirmed to exist:
# Dataset path
ls -la /path/to/dataset | head
# Pretrained checkpoint (if loading)
ls -la /path/to/checkpoint.pt
# Output directory parent (must exist; the run dir will be created)
ls -la /path/to/runs/
# Config file
cat configs/intended.yaml | head
Never trust a path that was recalled from memory. A run started with a nonexistent dataset path crashes 30 minutes in instead of immediately.
Auto-detect the experiment tracker:
WANDB_API_KEY set or wandb import in the launcher → wandbNEPTUNE_API_TOKEN set → neptuneMLFLOW_TRACKING_URI set or mlflow in launcher → mlflowruns/ or lightning_logs/ → tensorboardConfirm the run will appear under the right project / entity / experiment-name. Confirm any tags / groups for cohort comparison are set.
Estimate wall-clock duration: epochs × seconds-per-epoch / 3600 = hours. State the ETA in your local TZ (the system's TZ). If the run will straddle a meeting / sleep / OOO window, decide whether to defer or split.
On the machine that runs the job, in the repo it runs from:
git status --porcelain # must print nothing
git fetch -q && git rev-parse HEAD @{u} # both hashes must match
Uncommitted or unpushed changes mean nobody can tell later which code produced the run. If either check fails, edit locally, commit, push, and git pull on the host before launching. Record the commit hash with the run.
If this is a restart of a previously-failed run, or a kill before launching a replacement, purge stale artifacts in this exact order:
rm -rf /local/runs/<run-name> (verify path first).rm -rf /remote/runs/<run-name> (or equivalent).wandb api.run(...).delete(), neptune run.stop() + delete via UI, etc.). Stale tracker runs corrupt later comparisons.scancel <jobid>), the lambda labs reservation, the cron entry, etc. Runs that "killed but the GPUs are still allocated" are a recurring waste.Skipping any of these creates ghost state that will confuse the next launch or the next comparison.
When the user invokes this skill, walk the six checks (or three checks + cleanup, if killing) and report which passed and which failed. Block the launch on any failure unless the user explicitly waives the check.
For a clean launch, end with the launch command itself in a fenced block, ready to copy.
Pre-flight checklist for long-running ML training jobs covering config diff, run naming, path verification, monitoring setup, and restart-cleanup. Use when the user asks to launch, kick off, start, re
---name: launchdescription: Pre-flight checklist for long-running ML training jobs covering config diff, run naming, path verification…---# plugin/skills/launch/SKILL.md總覽
Pre-flight checklist for long-running ML training jobs covering config diff, run naming, path verification, monitoring setup, and restart-cleanup. Use when the user asks to launch, kick off, start, restart, or kill a training run, or mentions launching a multi-hour or multi-day GPU job (python train, accelerate launch, torchrun, deepspeed, sbatch, tmux training).
常見問題
從發布者下載 plugin/skills/launch/SKILL.md,或複製路徑後開啟儲存庫。PlugMCP 只收錄這項技能,不代管檔案。
發布者將它存放在 https://github.com/fcakyon/labrat。
Claude、Cursor、Codex 和 Grok Build 可以讀取你從發布者儲存庫加入的 SKILL.md。在聊天中傳送技能的 GitHub 資料夾連結,Grok Bot 就會安裝它。