First Ever SDXL Training With Kohya LoRA...
Good afternoon (in CET), Im trying to follow the excellent guide in 'First Ever SDXL Training With Kohya LoRA - Stable Diffusion XL Training' (https://www.youtube.com/watch?v=AY6DMBCIZ3A) but am running into OOM when starting up training.
CUDA out of memory. Tried to allocate 2.00 MiB (GPU 0; 10.00 GiB total capacity; 9.15 GiB already
allocated; 0 bytes free; 9.26 GiB reserved in total by PyTorch).
Im a bit of a noob, buy why dont torch use the already allocated memory to train my lora? Any suggestions are more than welcomed. Im on a windows machine with rtx3080
SECourses
YouTube
First Ever SDXL Training With Kohya LoRA - Stable Diffusion XL Trai...
Updated for SDXL 1.0. How to install #Kohya SS GUI trainer and do #LoRA training with Stable Diffusion XL (#SDXL) this is the video you are looking for. I have shown how to install Kohya from scratch. The best parameters to do LoRA training with SDXL. How to use Kohya SDXL LoRAs with ComfyUI. How to do checkpoint comparison with SDXL LoRAs and m...
1 Reply
I changed to smaller network rank, smaller images, gradient checkpointing, fp16 - same error (but size of ram its trying to allocate smaller)
So I think my main problem could be the already allocated 9.15GiB