Stable-textual-inversion_win
Stable-textual-inversion_win copied to clipboard
nicolai256
how much vram do I need?
I have 8gb of vram, and im running out of memory trying to run this on 700 images. it says I need 30 but is that a strict requirement?
I'm training on 24gb vram, so 30gb vram is not a requirement atm it doesn't work for 8gb vram, maybe someone who has knowledge of optimisation will change the script a little, but for now it's not possible, maybe try the colab :)
maybe try the colab :)
there's a colab notebook for textual inversion on stable diffusion?
I have 8gb of vram, and im running out of memory trying to run this on 700 images. it says I need 30 but is that a strict requirement?
I am using the basujinda repo. It let me run the 768x768 150 steps. This repo let me run on my rtx 3070 base 512x512 resolution
go to v1-finetune.yaml file and change batch size to 1 , that should slove it, also change number of workers to half of whats now.
File "E:\ModelTraining\ldm\modules\attention.py", line 180, in forward sim = einsum('b i d, b j d -> b i j', q, k) * self.scale RuntimeError: CUDA out of memory. Tried to allocate 512.00 MiB (GPU 0; 8.00 GiB total capacity; 6.82 GiB already allocated; 0 bytes free; 7.00 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
File "E:\ModelTraining\ldm\modules\attention.py", line 180, in forward sim = einsum('b i d, b j d -> b i j', q, k) * self.scale RuntimeError: CUDA out of memory. Tried to allocate 512.00 MiB (GPU 0; 8.00 GiB total capacity; 6.82 GiB already allocated; 0 bytes free; 7.00 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
8gb might not be enough, u can use free colab to train tho
