CUDA Out of Memory Error on NVIDIA L4 GPU (Google Cloud VM)

[drgom1]

Hey there! I'm running into some CUDA out of memory errors with Deep-Live-Cam on a Google Cloud VM. Here's what's going on:

Setup:

• Google Cloud VM (Windows Server 2022)
• NVIDIA L4 GPU with 23GB VRAM
• NVIDIA Driver 537.13, CUDA 12.2

The Problem: I keep getting CUDA out of memory errors when trying to run Deep-Live-Cam. The weird part is, the GPU has 23GB of VRAM, which should be plenty. When I'm not running the software, the GPU is barely using any memory (about 205MB out of 23034MB).

Error Messages: [Paste your error messages here]

What I've Tried:

• Made sure GPU drivers are up-to-date
• Checked that the GPU is working with nvidia-smi

Questions:

1. Any known issues with Deep-Live-Cam on L4 GPUs in Google Cloud?
2. Are there settings I should tweak for GPU memory?
3. Can I limit how much GPU memory Deep-Live-Cam uses?
4. Anything special I need to do for running this in the cloud?

I'd really appreciate any help or tips you can give me. Let me know if you need any more info – I'm happy to provide whatever might help solve this!

Thanks!