mrq/stable-diffusion-webui
1
1
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix #3986 breaking --no-half-vae

Browse Source
This commit is contained in:
AUTOMATIC 2022-11-02 14:41:29 +03:00
parent 675b51ebd3
commit f2a5cbe6f5
1 changed files with 9 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
modules/sd_models.py
View File

@ -183,11 +183,20 @@ def load_model_weights(model, checkpoint_info, vae_file="auto"):
model.to(memory_format=torch.channels_last)
if not shared.cmd_opts.no_half:
vae = model.first_stage_model
# with --no-half-vae, remove VAE from model when doing half() to prevent its weights from being converted to float16
if shared.cmd_opts.no_half_vae:
model.first_stage_model = None
model.half()
model.first_stage_model = vae
devices.dtype = torch.float32 if shared.cmd_opts.no_half else torch.float16
devices.dtype_vae = torch.float32 if shared.cmd_opts.no_half or shared.cmd_opts.no_half_vae else torch.float16
model.first_stage_model.to(devices.dtype_vae)
if shared.opts.sd_checkpoint_cache > 0:
# if PR #4035 were to get merged, restore base VAE first before caching
checkpoints_loaded[checkpoint_key] = model.state_dict().copy()