First Container
Deploy your first container in <30 minutes.
Quick Start
# 1. First-time setup (run once)
user-setup
# 2. Setup Project (per project)
project-init --guided
# 3a. Launch project (every login)
project launch --guided
#3b OR container-oriented workflow
image-create # Once (then image-update for adding packages)
container-deploy my-project --open # Every login
# 3. You're now inside the container
# Work in /workspace - files here are persistent
# 4. When done
exit
container-retire my-project
Work Inside Container
# You're now inside the container
user@my-project:/workspace$
# Check GPU is available
nvidia-smi
# Start Python
python
>>> import torch
>>> torch.cuda.is_available()
True
Important: Save files in /workspace - this is your persistent storage.
Key Concepts
Files
/workspace/ Your persistent files (always safe)
Everything else Temporary (lost on container removal)
Container States
container-deploy → Container running, GPU allocated
container-retire → Container removed, GPU freed
Your workspace files survive both states.
Getting Help
Every command has built-in help:
<command> --help # Quick reference
<command> --info # Full reference (all options)
<command> --concepts # Learn concepts before running
<command> --guided # Step-by-step with explanations
Example: If you're new to images:
image-create --concepts # Understand what images are
image-create --guided # Create with explanations
--
NB: Container Naming
- Images:
ds01-<user>/<project>:latest - Containers:
<project>._.<user>(AIME convention)
Bonus?
Understand why:
Do more:
Troubleshooting
"No GPUs available/Resource limits reached"
dashboard # Check availability
container-retire old-project # Free your old containers
"Container not found"
container-list # Check what's running
container-list --all # Check all containers (incl those not in running state - advanced)
container-deploy my-project # Recreate
Commands not found
shell-setup
source ~/.bashrc