GLKSA

Same issue in Win10 with 12Gb Graphics RAM RuntimeError: CUDA out of memory. Tried to allocate 1.50 GiB (GPU 0; 12.00 GiB total capacity; 8.62 GiB already allocated; 967.06 MiB...

README says at least 10Gib VRAM required - I have 12 Gib VRAM. I set the system environment variable PYTORCH_CUDA_ALLOC_CONF to 100 mb but it was not honoured in the...

OK, Success on an RTX 3060, win10. Thanks to everyone who contributed to the project!! Solution: I reduced the n-samples to 1, ( still got 2 samples ) here is...

https://www.youtube.com/watch?v=z99WBrs1D3g Noted that was the program is very heavy on system RAM and my spinny disk.

Hi MartinAbilev, just cut down the image size until you get a clean run. Then build it back up to find the sweet spot. e..g. my 12Gb card can run...

BTW the environment variable does not seem to work on Windows 10. You are allocating in 1Gb chunks, would be nice to reduce this, but I can;t find anywhere PYTORCH_CUDA_ALLOC_CONF...

On win 10 I had no luck at all using virtual environments due to incompatible usage of environment variables. I downloaded base versions of everything (Python, Miniconda, git) and let...

also you need "conda activate ldm" everytime you open the miniconda window.

Hi reecelightning, please post your error. also confirm you are using "base" installs of the 3 software ( installed on your actual OS), not virtual environments except the one referenced...