GPUs
tunnel downThe GPUs show up here once the notebook is running.
Pictures from a borrowed GPU.
ComfyUI on Kaggle's two T4s or the Colab GPU you pick. Run the templates or your own workflows, and keep what they make.
Workflows
What shall we make?
No GPU session is running. Runs wait in the queue until one is ready.
Press Run and the picture develops here.
Runs
Gallery
Everything it made.
GPU session
The GPU is asleep.
Pick where to run, then start.
Now in Colab
- Open the launcher in Colab ↗. It asks for the L4 runtime by itself.
- Press Runtime ▸ Run all (Ctrl+F9). The first time, allow Colab to read the GitHub repository.
- Paste your Colab key when it asks , or save it once as the Colab secret
HALFTONE_KEY.
If Colab can't open it from GitHub, download the launcher and use File ▸ Upload notebook in Colab. This page carries on by itself once the notebook dials in. Stop here hands the runtime back, so it stops using units.
- Session
- —
- Kaggle this week
- —
- Runs waiting
- 0
Live log
Settings
The small print.
Model downloads
Some models are gated: accept the licence on the model's page, then add a token here. The notebook gets them when it starts; they never show on this page again.
Stop by itself
After this many minutes with nothing queued or running, the session stops, so an idle GPU doesn't spend hours or units. 0 means never.
Colab key
The launcher notebook sends this to fetch its session from here. Keep it as the Colab secret HALFTONE_KEY (the key icon in Colab's left bar, with notebook access on) and it never asks.
••••••••••••••••
Storage
- Gallery
- —
- ComfyUI
- —
Past the limit the oldest unstarred items go first. Star what you want to keep.