Check these settings before deploying
- For HearmemanAI templates, turn off global networking in Edit Pod.
- In the advanced GPU filters, choose CUDA 12.8 or newer for these templates.
- Use one network volume per template. Sharing a volume across different templates can cause their files to overlap.
If first boot is taking a long time
First boot downloads the models you enabled. Check the logs for progress before restarting; larger model selections take longer.
If downloads stay stalled, try a pod in another region. Save your logs and files before terminating the existing pod.
If ComfyUI stays on Starting Up
- Open My Pods, select the pod and open Logs.
- Keep a copy of the startup errors for support.
- If you attached your own network volume, test a fresh deployment without that volume. Keep the existing volume intact so you can recover its files.
If the pod runs out of memory
Check My Pods for an OOM notice, then restart the pod. If it happens again, check whether your model, resolution and batch size fit the available GPU memory before retrying.
If images are black
Save the workflow and settings, then check the generation logs. If you test another GPU, keep the inputs unchanged so you can compare results.
Send useful information to support
Include the template name, GPU model, workflow and relevant logs. Say whether you attached a network volume and which model flags you enabled. Remove tokens and private data before sharing logs.