#I got an OOM Error from
trail arch
formal nebulaBOT
Hi @trail arch. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:
The Out of Memory (OOM) error you're encountering is likely due to your system running out of available memory. This can happen when the processes you're running, such as AI model training or inference, require more memory than your system has available.
In one of the Discord support threads (https://discord.com/channels/1020123559063990373/1165694937220796498), a community member experienced a similar issue where their application crashed due to high memory usage. The Trusted Support Staff suggested reducing the size of the RAM cache to mitigate the issue.
In a Github issue (https://github.com/invoke-ai/InvokeAI/issues/5761), a user encountered an OOM error when trying to generate more than one image at 1280x720. The error message suggested that the GPU was out of memory and recommended setting max_split_size_mb to avoid fragmentation.
If you're using Invoke AI and encounter any technical issues, you can reach out to the support team by emailing email protected], opening a ticket in the help desk, or contacting your dedicated account manager if you're an Enterprise Customer. Be sure to provide a clear description of the issue you're facing, any error messages you've encountered, and any relevant screenshots to help the team resolve your issue as quickly as possible. You can find more information on how to contact support [here (https://support.invoke.ai/support/solutions/articles/151000096654-i-m-experiencing-technical-issues-with-the-platform-how-can-i-contact-support-for-help-).
Remember to confirm that you are experiencing issues with default settings. Invoke AI offers a significant amount of flexibility in the generative AI processing, and some settings can cause image qualities that aren’t quite what humans would expect. The AI is powerful, but ultimately relies on us for taste!
Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:
undone acorn
how much vram cache do you have in your invokeai.yaml config file ?
trail arch
my invokeai.yaml
my invokeai.yaml file just says
`# Internal metadata - do not edit:
schema_version: 4.0.1
Put user settings here - see https://invoke-ai.github.io/InvokeAI/features/CONFIGURATION/:`
my sd_xl_base.yaml says:
undone acorn
so each model/controlnet needs space in the vram, how big is the model do you use ?
trail arch
I'm using this model https://civitai.com/models/133005?modelVersionId=456194
undone acorn
6,6 GB vram for the model. so there is not much for the generation of the image, but have you tried without the canny controlnet ?
undone acorn
the canny model is like 2.5 GB alone if it is this one: https://huggingface.co/diffusers/controlnet-canny-sdxl-1.0/tree/main
trail arch
I think I'm using that controlnet, I downloaded it from the starter models menu in Invokeai interface.
is there a smaller version if that one gives me OOM errors?
undone acorn
#🌏invoke-chat message
look here, maybe this can be a solution
trail arch
I will try out T2I Adapters.