Skip to content

--upcast-sampling support for CUDA #8782

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
AUTOMATIC1111 merged 7 commits into from
Mar 25, 2023
Merged

--upcast-sampling support for CUDA #8782

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
254d994
Update devices.py
FNSpd Mar 21, 2023
91cfa97
Update sd_hijack_unet.py
FNSpd Mar 21, 2023
c84c9df
Update sd_hijack_optimizations.py
FNSpd Mar 21, 2023
2f01814
Update lora.py
FNSpd Mar 21, 2023
beb7dda
Update sd_hijack_unet.py
FNSpd Mar 24, 2023
280ed8f
Update sd_hijack_optimizations.py
FNSpd Mar 24, 2023
a9eab23
Update devices.py
FNSpd Mar 24, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions extensions-builtin/Lora/lora.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -178,6 +178,7 @@ def load_loras(names, multipliers=None):


def lora_forward(module, input, res):
input = devices.cond_cast_unet(input)
if len(loaded_loras) == 0:
return res

Expand Down
4 changes: 2 additions & 2 deletions modules/sd_hijack_optimizations.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -337,7 +337,7 @@ def xformers_attention_forward(self, x, context=None, mask=None):

dtype = q.dtype
if shared.opts.upcast_attn:
q, k = q.float(), k.float()
q, k, v = q.float(), k.float(), v.float()

out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=get_xformers_flash_attention_op(q, k, v))

Expand Down Expand Up @@ -372,7 +372,7 @@ def scaled_dot_product_attention_forward(self, x, context=None, mask=None):

dtype = q.dtype
if shared.opts.upcast_attn:
q, k = q.float(), k.float()
q, k, v = q.float(), k.float(), v.float()

# the output of sdp = (batch, num_heads, seq_len, head_dim)
hidden_states = torch.nn.functional.scaled_dot_product_attention(
Expand Down
2 changes: 1 addition & 1 deletion modules/sd_hijack_unet.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,7 +67,7 @@ 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', 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("1.13.1"):
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)
CondFunc('ldm.modules.attention.GEGLU.forward', lambda orig_func, self, x: orig_func(self.float(), x.float()).to(devices.dtype_unet), unet_needs_upcast)
CondFunc('open_clip.transformer.ResidualAttentionBlock.__init__', lambda orig_func, *args, **kwargs: kwargs.update({'act_layer': GELUHijack}) and False or orig_func(*args, **kwargs), lambda _, *args, **kwargs: kwargs.get('act_layer') is None or kwargs['act_layer'] == torch.nn.GELU)
Expand Down