How do I start a remote job?

Simply prefix your normal command with 'gpu run'. For example, 'gpu run python train.py' will provision a GPU, sync your code, and execute the script remotely.

Can I use this for LLM inference?

Yes, it features dedicated commands for launching LLM inference using Ollama or vLLM, as well as built-in support for ComfyUI workflows.

Does it support persistent storage?

Yes, it includes comprehensive volume management commands to create, list, and migrate network volumes for persistent data across different jobs.

How does file syncing work with this skill?

It uses your .gitignore to determine which files to upload to the remote pod and automatically downloads specified output patterns back to your local machine after the command completes.

GPU CLI is a command-line tool that allows you to run local scripts and commands on remote cloud GPUs like NVIDIA A100s and H100s with automatic file syncing.

GPU Cloud CLI

Name: GPU Cloud CLI
Author: gpu-cli

bygpu-cli

•

Ciencia de Datos y ML

Executes machine learning workloads and local commands on remote NVIDIA GPUs by prefixing them with 'gpu' for instant cloud compute.

GPU CLI streamlines the transition from local development to cloud-scale compute by allowing developers to run any command on remote high-performance NVIDIA GPUs, such as A100s or H100s, as if they were local. It handles the heavy lifting of provisioning pods, syncing source code, streaming logs in real-time, and automatically retrieving output files, making it an essential tool for training models, running LLM inference, or executing ComfyUI workflows without the overhead of manual infrastructure management.