Initialization-Related Errors

Error	Suggested Fix
Network issues	Please verify your internet connectivity and ensure that it is not intermittent.
Insufficient permissions to run microk8s	Please ensure that the user has the right permissions to run `microk8s`.
Invalid workspace / secret key combination	Please check that you have provided the correct workspace key, as well as the correct secret key. If you have forgotten your secret key, you can regenerate one here. Additionally, please ensure that you are using a workspace secret key, not a project secret key.
Microk8s error	The `microk8s` instance errored during GPU plugin installation, please ensure that the appropriate NVIDIA drivers are properly installed. If the error persists, please contact us for support.
Missing NVIDIA drivers / insufficient permissions	Please ensure that you have the appropriate NVIDIA drivers installed, and that you have the right permissions to access your drivers. For more information on how to install the right NVIDIA drivers, please refer to NVIDIA's driver installation guide.

Training-Related Errors

Error	Suggested Fix
Out of memory	The existing GPUs available to the run do not have sufficient memory to perform the training. Try reducing the batch size or model size, or use a GPU with more memory.
Failed to start training	Please contact us for support.
Error occurred	Please contact us for support.

Common Questions

Which OS Should I Use?

We recommend using Linux (Ubuntu ≥ 20.04).

What GPUs Are Supported?

Currently, only NVIDIA GPUs are supported.

Why does the Runner Installation Require Sudo Permissions?

We require microk8s to be installed to handle the model training and infrastructure, which requires sudo. If you do not wish to use sudo, ensure that the user running the initialization script has access permissions to microk8s.

What Models are Supported for Training?

All Datature models (other than Paligemma) are supported. However, note that some trainings require GPUs with sufficient memory size. Ensure that your hardware resources are sufficient to run the selected training.