You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
OutOfMemoryError: CUDA out of memory. Tried to allocate 1.53 GiB. GPU 0 has a total capacity of 11.76 GiB of which 1.12 GiB is free. Process 257405 has 10.62 GiB memory in use. Of the allocated memory 9.90 GiB is allocated by PyTorch, and 588.48 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)
I am using 3060 and i can generate 832×1216 upscale 2x in t2i by pony sdxl models. As a result 1664×2432. (A: 9.37 GB, R: 10.98 GB, Sys: 11.6/11.7627 GB (98.9%))
But when i try to generate smaller image, for example 832×1088 or 832×1024 etc, it happens CUDA out of memory.
I can understand larger image require more vram. but i don't know why smaller image requires more memory.
What is the problem? I can't generate anything other than 832×1216 or 1216×832. i want to change image ratio.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
version: v1.9.4 • python: 3.10.14 • torch: 2.2.2 • xformers: 0.0.25.post1 • gradio: 3.41.2 • checkpoint: 4d2a7da47f
OutOfMemoryError: CUDA out of memory. Tried to allocate 1.53 GiB. GPU 0 has a total capacity of 11.76 GiB of which 1.12 GiB is free. Process 257405 has 10.62 GiB memory in use. Of the allocated memory 9.90 GiB is allocated by PyTorch, and 588.48 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)
소요 시간: 41.1 sec.
A: 10.29 GB, R: 10.93 GB, Sys: 11.5/11.7627 GB (97.9%)
Question about Tiled VAE.
I am using 3060 and i can generate 832×1216 upscale 2x in t2i by pony sdxl models. As a result 1664×2432. (A: 9.37 GB, R: 10.98 GB, Sys: 11.6/11.7627 GB (98.9%))
But when i try to generate smaller image, for example 832×1088 or 832×1024 etc, it happens CUDA out of memory.
I can understand larger image require more vram. but i don't know why smaller image requires more memory.
What is the problem? I can't generate anything other than 832×1216 or 1216×832. i want to change image ratio.
Beta Was this translation helpful? Give feedback.
All reactions