Windows
Sets up the model on this computer or on a separate GPU server over SSH.
Size 53 MB
SHA-256 checksum
3dffa11dbaa6e698acef72dc117fe64692fd5d9b6cb319ab346dbeb107656bf4
Setup program
Set up your model on your own hardware
This program sets up your finished model on your own hardware. It is identical for every customer and contains no customer data; you load your model into it with the activation code you receive by email once it is ready.
Note for macOS
The model needs an NVIDIA graphics card, which Mac computers do not have. The macOS version therefore does not set up a model on the Mac itself; it only controls a separate Windows or Linux server with a suitable GPU over SSH.
Linux
Sets up the model on this computer or on a separate GPU server over SSH.
Size 68 MB
SHA-256 checksum
6d171ed411122ecba5e02a76138a41af406a08b1034e0526f52f83eeb95b2227
macOS
Only controls a separate GPU server over SSH; no model runs locally on the Mac.
Size 48 MB
SHA-256 checksum
5728adae29d4c14623a41dbfff8b31ee7b635a0f3b241c7256eadefa9706c950
What you need before setting up
- A server or computer with a supported NVIDIA GPU; for Windows and Linux either this computer itself or a separate target over SSH, for macOS only a separate target. GPUs of the Blackwell generation use the faster FP4 mode, every other supported GPU runs automatically in BF16 mode.
- A current NVIDIA driver plus Docker with Compose v2 and the NVIDIA Container Toolkit on the target system; if Docker is missing, the program offers a one-click install.
- Your IonKon account (email and password) and the activation code from the email you receive once your model is ready.
- Internet access for sign-in, activation and the one-time model download; operation runs offline afterwards. The full instructions are included with your model.
Go to your account
No account yet? Create an account