---

name: remote description: "Routes LabML remote-project setup, job orchestration, and distributed launch workflows." disable-model-invocation: true metadata: disco-role: operating license: MIT

Remote

Use this subskill for labml_remote: project bootstrap, remote server setup, rsync-based sync, job management, and distributed launch helpers.

Use this when

• The task mentions labml_remote, .remote/configs.yaml, .remote/exclude.txt, init, prepare, run, job-run, job-tail, job-kill, or helper-torch-launch.
• The user wants to set up a remote machine, sync a project, run commands on a server, or launch distributed training jobs over SSH.

Boundaries

Include:

• Remote-project configuration and bootstrap files.
• SSH/rsync command orchestration.
• Job creation, listing, tailing, killing, and syncing.
• Distributed launch helpers that build on the remote job system.

Exclude or route elsewhere:

• Client-side tracking and monitoring → tracking.
• Training-loop utilities and datasets → helpers.
• App backend and monitoring UI deployment → server.

Read next

• references/cli-reference.md for the command set and options.
• references/workflows.md for the remote-project lifecycle and distributed launch recipe.
• references/troubleshooting.md for SSH, conda, rsync, and job-state issues.
• scripts/remote_config_smoke.py for a safe local init/configuration check.

Typical routes

Bootstrap a remote project

Choose this route for labml_remote init, the .remote/ files, and the initial server configuration.

Prepare or sync a server

Choose this route for prepare, setup, rsync, and update-packages.

Run or manage jobs

Choose this route for run, job-run, job-list, job-tail, and job-kill.

Launch distributed training

Choose this route for helper-torch-launch and the environment variables that assemble a multi-node training job.