JupyterLab Quick Start for ML

Overview

JupyterLab is pre-wired on your CloudDock desktop. Use the Launcher to open it — no SSH, no terminal. Sessions start in your workspace (e.g., /workspace) with GPU-enabled kernels when available.

• One click: open from Launcher; a browser tab appears automatically.
• GPU-ready: verify with a single cell (see below).
• Kernels: choose the right Python kernel; keep environments organized.

Launch & Workspace

1. Open the Launcher inside your desktop.
2. Click JupyterLab. A browser tab opens automatically.
3. You land in your workspace (default: /workspace unless your admin customized it).

Launcher → JupyterLab (one click).

JupyterLab Home — starts in your workspace.

Kernels & Packages

Pick a Python kernel from the top-right status bar or the Command Palette. Install extra packages from inside a notebook using %pip (preferred) or !pip.

Select the kernel that matches your project.

Use %pip install inside the running kernel.

Minimal package commands

# list packages
%pip list


# install packages into the current kernel environment
%pip install pandas matplotlib scikit-learn


# pin exact versions for reproducibility
!pip freeze > requirements.txt

GPU check (PyTorch)

Run this cell — if cuda is available, you’re on GPU:

import torch, platform
print("torch:", torch.__version__)
print("python:", platform.python_version())
print("cuda available:", torch.cuda.is_available())
if torch.cuda.is_available():
    print("device:", torch.cuda.get_device_name(0))
    print("memory allocated (MB):", round(torch.cuda.memory_allocated()/1024**2, 1))

Optional: a quick system snapshot:

!nvidia-smi --query-gpu=name,memory.total,memory.used,utilization.gpu --format=csv

Files & Paths

Keep data, notebooks, and outputs tidy under your workspace. A simple layout:

/workspace/
  data/            # input datasets (read-only or original copies)
  notebooks/       # .ipynb files
  outputs/         # models, CSVs, plots
  runs/            # experiment logs & artifacts
  requirements.txt # pinned dependencies for this project

A clean layout saves time later.

Notebook → Save and Export As…

Save & Export

• Save: Ctrl+S (or menu).
• Export: File → Save and Export As… → .py, .html, etc.
• Freeze deps: !pip freeze > requirements.txt and keep it in the project root.

Troubleshooting

Notebook says “kernel dead”

• Restart the kernel; close unused notebooks.
• Check VRAM on Launcher → System; free memory if close to the limit.

CUDA not available

• Confirm you launched a GPU-backed session.
• Use the PyTorch GPU check cell above; if still false, try a fresh session.

Package import errors

• Install into the current kernel with %pip install ....
• Reopen the notebook if the kernel changed.