Skip to content
Projects
Groups
Snippets
Help
Loading...
Help
Support
Keyboard shortcuts
?
Submit feedback
Sign in / Register
Toggle navigation
S
Stable Diffusion Webui
Project overview
Project overview
Details
Activity
Releases
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Locked Files
Issues
0
Issues
0
List
Boards
Labels
Service Desk
Milestones
Merge Requests
0
Merge Requests
0
CI / CD
CI / CD
Pipelines
Jobs
Schedules
Security & Compliance
Security & Compliance
Dependency List
License Compliance
Packages
Packages
List
Container Registry
Analytics
Analytics
CI / CD
Code Review
Insights
Issues
Repository
Value Stream
Wiki
Wiki
Snippets
Snippets
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Create a new issue
Jobs
Commits
Issue Boards
Open sidebar
novelai-storage
Stable Diffusion Webui
Commits
f89829ec
Commit
f89829ec
authored
Oct 21, 2022
by
aria1th
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
Revert "fix bugs and optimizations"
This reverts commit
108be155
.
parent
108be155
Changes
1
Show whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
46 additions
and
59 deletions
+46
-59
modules/hypernetworks/hypernetwork.py
modules/hypernetworks/hypernetwork.py
+46
-59
No files found.
modules/hypernetworks/hypernetwork.py
View file @
f89829ec
...
...
@@ -36,14 +36,14 @@ class HypernetworkModule(torch.nn.Module):
linears
.
append
(
torch
.
nn
.
Linear
(
int
(
dim
*
layer_structure
[
i
]),
int
(
dim
*
layer_structure
[
i
+
1
])))
# if skip_first_layer because first parameters potentially contain negative values
# if i < 1: continue
if
add_layer_norm
:
linears
.
append
(
torch
.
nn
.
LayerNorm
(
int
(
dim
*
layer_structure
[
i
+
1
])))
if
activation_func
in
HypernetworkModule
.
activation_dict
:
linears
.
append
(
HypernetworkModule
.
activation_dict
[
activation_func
]())
else
:
print
(
"Invalid key {} encountered as activation function!"
.
format
(
activation_func
))
# if use_dropout:
# linears.append(torch.nn.Dropout(p=0.3))
if
add_layer_norm
:
linears
.
append
(
torch
.
nn
.
LayerNorm
(
int
(
dim
*
layer_structure
[
i
+
1
])))
self
.
linear
=
torch
.
nn
.
Sequential
(
*
linears
)
...
...
@@ -115,24 +115,11 @@ class Hypernetwork:
for
k
,
layers
in
self
.
layers
.
items
():
for
layer
in
layers
:
layer
.
train
()
res
+=
layer
.
trainables
()
return
res
def
eval
(
self
):
for
k
,
layers
in
self
.
layers
.
items
():
for
layer
in
layers
:
layer
.
eval
()
for
items
in
self
.
weights
():
items
.
requires_grad
=
False
def
train
(
self
):
for
k
,
layers
in
self
.
layers
.
items
():
for
layer
in
layers
:
layer
.
train
()
for
items
in
self
.
weights
():
items
.
requires_grad
=
True
def
save
(
self
,
filename
):
state_dict
=
{}
...
...
@@ -303,6 +290,10 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
shared
.
sd_model
.
first_stage_model
.
to
(
devices
.
cpu
)
hypernetwork
=
shared
.
loaded_hypernetwork
weights
=
hypernetwork
.
weights
()
for
weight
in
weights
:
weight
.
requires_grad
=
True
losses
=
torch
.
zeros
((
32
,))
last_saved_file
=
"<none>"
...
...
@@ -313,10 +304,10 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
return
hypernetwork
,
filename
scheduler
=
LearnRateScheduler
(
learn_rate
,
steps
,
ititial_step
)
optimizer
=
torch
.
optim
.
AdamW
(
hypernetwork
.
weights
(),
lr
=
scheduler
.
learn_rate
)
# if optimizer == "AdamW": or else Adam / AdamW / SGD, etc...
optimizer
=
torch
.
optim
.
AdamW
(
weights
,
lr
=
scheduler
.
learn_rate
)
pbar
=
tqdm
.
tqdm
(
enumerate
(
ds
),
total
=
steps
-
ititial_step
)
hypernetwork
.
train
()
for
i
,
entries
in
pbar
:
hypernetwork
.
step
=
i
+
ititial_step
...
...
@@ -337,9 +328,8 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
losses
[
hypernetwork
.
step
%
losses
.
shape
[
0
]]
=
loss
.
item
()
optimizer
.
zero_grad
(
set_to_none
=
True
)
optimizer
.
zero_grad
()
loss
.
backward
()
del
loss
optimizer
.
step
()
mean_loss
=
losses
.
mean
()
if
torch
.
isnan
(
mean_loss
):
...
...
@@ -356,10 +346,9 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
})
if
hypernetwork
.
step
>
0
and
images_dir
is
not
None
and
hypernetwork
.
step
%
create_image_every
==
0
:
torch
.
cuda
.
empty_cache
()
last_saved_image
=
os
.
path
.
join
(
images_dir
,
f
'{hypernetwork_name}-{hypernetwork.step}.png'
)
with
torch
.
no_grad
():
hypernetwork
.
eval
()
optimizer
.
zero_grad
()
shared
.
sd_model
.
cond_stage_model
.
to
(
devices
.
device
)
shared
.
sd_model
.
first_stage_model
.
to
(
devices
.
device
)
...
...
@@ -396,8 +385,6 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
image
.
save
(
last_saved_image
)
last_saved_image
+=
f
", prompt: {preview_text}"
hypernetwork
.
train
()
shared
.
state
.
job_no
=
hypernetwork
.
step
shared
.
state
.
textinfo
=
f
"""
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment