Mirrors
/
stable-diffusion-webui
mirror of https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Revert "Merge pull request #16078 from huchenlei/fix_sd2" 

This reverts commit 4cc3add770, reversing
changes made to 50514ce414.
This commit is contained in:
AUTOMATIC1111 2024-07-06 10:40:48 +03:00
parent 0a6628bad0
commit ffead92d4e
2 changed files with 1 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
modules/sd_hijack_unet.py
View File

@ -138,7 +138,6 @@ CondFunc('ldm.models.diffusion.ddpm.LatentDiffusion.decode_first_stage', first_s
CondFunc('ldm.models.diffusion.ddpm.LatentDiffusion.encode_first_stage', first_stage_sub, first_stage_cond)
CondFunc('ldm.models.diffusion.ddpm.LatentDiffusion.get_first_stage_encoding', lambda orig_func, *args, **kwargs: orig_func(*args, **kwargs).float(), first_stage_cond)
# Always make sure inputs to unet are in correct dtype
CondFunc('ldm.models.diffusion.ddpm.LatentDiffusion.apply_model', apply_model)
CondFunc('sgm.modules.diffusionmodules.wrappers.OpenAIWrapper.forward', apply_model)
@ -151,6 +150,5 @@ def timestep_embedding_cast_result(orig_func, timesteps, *args, **kwargs):
return orig_func(timesteps, *args, **kwargs).to(dtype=dtype)
# Always make sure timestep calculation is in correct dtype
CondFunc('ldm.modules.diffusionmodules.openaimodel.timestep_embedding', timestep_embedding_cast_result)
CondFunc('sgm.modules.diffusionmodules.openaimodel.timestep_embedding', timestep_embedding_cast_result)

7
modules/sd_models_config.py
View File

@ -56,19 +56,14 @@ def is_using_v_parameterization_for_sd2(state_dict):
unet.eval()
with torch.no_grad():
unet_dtype = torch.float
original_unet_dtype = devices.dtype_unet
unet_sd = {k.replace("model.diffusion_model.", ""): v for k, v in state_dict.items() if "model.diffusion_model." in k}
unet.load_state_dict(unet_sd, strict=True)
unet.to(device=device, dtype=unet_dtype)
devices.dtype_unet = unet_dtype
unet.to(device=device, dtype=torch.float)
test_cond = torch.ones((1, 2, 1024), device=device) * 0.5
x_test = torch.ones((1, 4, 8, 8), device=device) * 0.5
out = (unet(x_test, torch.asarray([999], device=device), context=test_cond) - x_test).mean().item()
devices.dtype_unet = original_unet_dtype
return out < -1