If you want a better understanding of what I mean, check out this article.PhotoLine: A capable image processing application for photo editing with support for vectorial design Note: as I mentioned before, you should break your prompts up into 5 sections. I get that not everyone will have the capability to create images with these settings, but after making the changes above I have not run into any more CUDA errors, even when changing the to as high as 3 Now I'm able to create images with these settings: Making these changes helped dramatically:Īlso at the end of the FIRST section of my positive prompts (I say first section because you should break your prompts up into 5 sections), I always add "DLSS, Ray Tracing, uncensored, -n_samples 1" Graphics Card Name: MSI GeForce RTX 4070 VENTUS 3X 12G OCīut even with these components, I was still running into fucking CUDA errors. MotherBoard Product: MPG B460I GAMING EDGE (MS-7C86) System Model: MPG B460 Trident AS (MS-B926) System Manufacturer: Micro-Star International Co., Ltd. Ok, so for clarity, I have built a pretty decent machine: See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONFģ.65 GiB already allocated, but it is not allocated before the command runs - especially after a restart and shutting down almost all processes using the GPU (except Windows obviously). Tried to allocate 3.00 GiB (GPU 0 8.00 GiB total capacity 3.65 GiB already allocated 1.18 GiB free 4.30 GiB reserved in total by PyTorch) If reserved memory is > allocated memory try setting max_split_size_mb to avoid fragmentation. > python scripts/txt2img.py -prompt "flying pig" -H 512 -W 512 -seed 27 -n_iter 1 -ddim_steps 100 I already implemented the ideas above (reduce samples, and half the model), but 512 fails: Is 8GB to low for a GPU for this system? I can only make 384x384 work at the most, but would like a higher res image if possible. PyTorch is still taking a lot of memory, and it seems a lot of other GPU memory is taken up by something else while the command runs because the resource monitor shows very little utilized until the command runs.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |