Mirror/stable-diffusion-webui
1
0
Fork 0
mirror of https://github.com/AUTOMATIC1111/stable-diffusion-webui.git synced 2024-06-07 21:20:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

Patch timestep embedding to create tensor on-device

Browse Source
This commit is contained in:
drhead 2024-05-17 12:12:57 -04:00 committed by GitHub
parent ddb28b33a3
commit 53d67088ee
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 27 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
modules/sd_hijack_unet.py
View File

@ -1,5 +1,7 @@
import torch
from packaging import version
from einops import repeat
import math
from modules import devices
from modules.sd_hijack_utils import CondFunc
@ -48,6 +50,30 @@ def apply_model(orig_func, self, x_noisy, t, cond, **kwargs):
return orig_func(self, x_noisy.to(devices.dtype_unet), t.to(devices.dtype_unet), cond, **kwargs).float()
# Monkey patch to create timestep embed tensor on device, avoiding a block.
def timestep_embedding(_, timesteps, dim, max_period=10000, repeat_only=False):
"""
Create sinusoidal timestep embeddings.
:param timesteps: a 1-D Tensor of N indices, one per batch element.
These may be fractional.
:param dim: the dimension of the output.
:param max_period: controls the minimum frequency of the embeddings.
:return: an [N x dim] Tensor of positional embeddings.
"""
if not repeat_only:
half = dim // 2
freqs = torch.exp(
-math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32, device=timesteps.device) / half
)
args = timesteps[:, None].float() * freqs[None]
embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
if dim % 2:
embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
else:
embedding = repeat(timesteps, 'b -> b d', d=dim)
return embedding
class GELUHijack(torch.nn.GELU, torch.nn.Module):
def __init__(self, *args, **kwargs):
torch.nn.GELU.__init__(self, *args, **kwargs)
@ -69,6 +95,7 @@ def hijack_ddpm_edit():
unet_needs_upcast = lambda *args, **kwargs: devices.unet_needs_upcast
CondFunc('ldm.models.diffusion.ddpm.LatentDiffusion.apply_model', apply_model, unet_needs_upcast)
CondFunc('ldm.modules.diffusionmodules.openaimodel.timestep_embedding', timestep_embedding)
CondFunc('ldm.modules.diffusionmodules.openaimodel.timestep_embedding', lambda orig_func, timesteps, *args, **kwargs: orig_func(timesteps, *args, **kwargs).to(torch.float32 if timesteps.dtype == torch.int64 else devices.dtype_unet), unet_needs_upcast)
if version.parse(torch.__version__) <= version.parse("1.13.2") or torch.cuda.is_available():
CondFunc('ldm.modules.diffusionmodules.util.GroupNorm32.forward', lambda orig_func, self, *args, **kwargs: orig_func(self.float(), *args, **kwargs), unet_needs_upcast)