nerdit 0.2.0

CLI client for Nerdit — the local AI orchestration engine for multi-GPU servers

nerdit

CLI client for Nerdit — the local AI orchestration engine for multi-GPU servers.

Submit and manage ML training jobs on remote GPU clusters from your local machine.

Installation

pip install nerdit

Requires Python 3.11+.

Quick Start

# Connect to a remote nerdit daemon
nerdit connect <server-ip> --token <token>

# Submit a job
nerdit run train.py --gpus 2

# Check status
nerdit status

# Follow logs in real-time
nerdit logs <job-id> --follow

Commands

Command	Description
nerdit connect <host> --token <token>	Connect to a remote daemon
nerdit run <script> [--gpus N]	Submit a training job
nerdit status	View GPU and job status
nerdit logs <job-id> [--follow]	View job logs
nerdit stop <job-id>	Stop a running job
nerdit cleanup	Stop all running jobs
nerdit token	Display the current auth token
nerdit init --project	Generate a nerdit.toml template
nerdit check-deps	Check system dependencies

Project Configuration

Create a nerdit.toml in your project root to set default job parameters:

nerdit init --project

[run]
script = "train.py"
gpus = 2
priority = 5
image = "nerdit-runtime:0.1"

[data]
dataset = "/data/imagenet:/data"
output = "/results/{{job_id}}:/output"

CLI flags always override nerdit.toml defaults.

License

Apache License 2.0 — see LICENSE for details.