Open pedx78 opened 1 year ago
On Auto111 running on cloud, HD-18 loading caused out of memory failure. But HD-17 works well without issue.
Can I suggest converting .ckpt to .safetensors ? This is more memory efficient, and enables model to load faster in Auto111. https://github.com/huggingface/safetensors
Also, have you tried pruning the model ? URPM reduced the size from 8gb to 1.6gb using below script.
https://github.com/saftle/stable-diffusion-prune (-eca command).
On Auto111 running on cloud, HD-18 loading caused out of memory failure. But HD-17 works well without issue.
Can I suggest converting .ckpt to .safetensors ? This is more memory efficient, and enables model to load faster in Auto111. https://github.com/huggingface/safetensors
Also, have you tried pruning the model ? URPM reduced the size from 8gb to 1.6gb using below script.
https://github.com/saftle/stable-diffusion-prune (-eca command).