Skip to content
Projects
Groups
Snippets
Help
Loading...
Help
Contribute to GitLab
Sign in / Register
Toggle navigation
S
stable-diffusion-webui
Project
Project
Details
Activity
Cycle Analytics
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Charts
Issues
0
Issues
0
List
Board
Labels
Milestones
Merge Requests
0
Merge Requests
0
CI / CD
CI / CD
Pipelines
Jobs
Schedules
Charts
Wiki
Wiki
Snippets
Snippets
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Charts
Create a new issue
Jobs
Commits
Issue Boards
Open sidebar
Administrator
stable-diffusion-webui
Commits
16451ca5
Commit
16451ca5
authored
Oct 28, 2022
by
Muhammad Rizqi Nur
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
Learning rate sched syntax support for grad clipping
parent
1618df41
Hide whitespace changes
Inline
Side-by-side
Showing
4 changed files
with
30 additions
and
13 deletions
+30
-13
hypernetwork.py
modules/hypernetworks/hypernetwork.py
+10
-3
learn_schedule.py
modules/textual_inversion/learn_schedule.py
+8
-3
textual_inversion.py
modules/textual_inversion/textual_inversion.py
+9
-3
ui.py
modules/ui.py
+3
-4
No files found.
modules/hypernetworks/hypernetwork.py
View file @
16451ca5
...
@@ -383,11 +383,15 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
...
@@ -383,11 +383,15 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
ititial_step
=
hypernetwork
.
step
or
0
ititial_step
=
hypernetwork
.
step
or
0
if
ititial_step
>
steps
:
if
ititial_step
>
steps
:
return
hypernetwork
,
filename
return
hypernetwork
,
filename
clip_grad_mode_value
=
clip_grad_mode
==
"value"
clip_grad_mode_value
=
clip_grad_mode
==
"value"
clip_grad_mode_norm
=
clip_grad_mode
==
"norm"
clip_grad_mode_norm
=
clip_grad_mode
==
"norm"
clip_grad_enabled
=
clip_grad_mode_value
or
clip_grad_mode_norm
if
clip_grad_enabled
:
clip_grad_sched
=
LearnRateScheduler
(
clip_grad_value
,
steps
,
ititial_step
,
verbose
=
False
)
scheduler
=
LearnRateScheduler
(
learn_rate
,
steps
,
ititial_step
)
scheduler
=
LearnRateScheduler
(
learn_rate
,
steps
,
ititial_step
)
# if optimizer == "AdamW": or else Adam / AdamW / SGD, etc...
# if optimizer == "AdamW": or else Adam / AdamW / SGD, etc...
optimizer
=
torch
.
optim
.
AdamW
(
weights
,
lr
=
scheduler
.
learn_rate
)
optimizer
=
torch
.
optim
.
AdamW
(
weights
,
lr
=
scheduler
.
learn_rate
)
...
@@ -407,6 +411,9 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
...
@@ -407,6 +411,9 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
if
shared
.
state
.
interrupted
:
if
shared
.
state
.
interrupted
:
break
break
if
clip_grad_enabled
:
clip_grad_sched
.
step
(
hypernetwork
.
step
)
with
torch
.
autocast
(
"cuda"
):
with
torch
.
autocast
(
"cuda"
):
c
=
stack_conds
([
entry
.
cond
for
entry
in
entries
])
.
to
(
devices
.
device
)
c
=
stack_conds
([
entry
.
cond
for
entry
in
entries
])
.
to
(
devices
.
device
)
# c = torch.vstack([entry.cond for entry in entries]).to(devices.device)
# c = torch.vstack([entry.cond for entry in entries]).to(devices.device)
...
@@ -430,9 +437,9 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
...
@@ -430,9 +437,9 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
assert
steps_without_grad
<
10
,
'no gradient found for the trained weight after backward() for 10 steps in a row; this is a bug; training cannot continue'
assert
steps_without_grad
<
10
,
'no gradient found for the trained weight after backward() for 10 steps in a row; this is a bug; training cannot continue'
if
clip_grad_mode_value
:
if
clip_grad_mode_value
:
torch
.
nn
.
utils
.
clip_grad_value_
(
weights
,
clip_value
=
clip_grad_
valu
e
)
torch
.
nn
.
utils
.
clip_grad_value_
(
weights
,
clip_value
=
clip_grad_
sched
.
learn_rat
e
)
elif
clip_grad_mode_norm
:
elif
clip_grad_mode_norm
:
torch
.
nn
.
utils
.
clip_grad_norm_
(
weights
,
max_norm
=
clip_grad_
valu
e
)
torch
.
nn
.
utils
.
clip_grad_norm_
(
weights
,
max_norm
=
clip_grad_
sched
.
learn_rat
e
)
optimizer
.
step
()
optimizer
.
step
()
...
...
modules/textual_inversion/learn_schedule.py
View file @
16451ca5
...
@@ -51,14 +51,19 @@ class LearnRateScheduler:
...
@@ -51,14 +51,19 @@ class LearnRateScheduler:
self
.
finished
=
False
self
.
finished
=
False
def
apply
(
self
,
optimizer
,
step_number
):
def
step
(
self
,
step_number
):
if
step_number
<=
self
.
end_step
:
if
step_number
<=
self
.
end_step
:
return
return
False
try
:
try
:
(
self
.
learn_rate
,
self
.
end_step
)
=
next
(
self
.
schedules
)
(
self
.
learn_rate
,
self
.
end_step
)
=
next
(
self
.
schedules
)
except
Excep
tion
:
except
StopItera
tion
:
self
.
finished
=
True
self
.
finished
=
True
return
False
return
True
def
apply
(
self
,
optimizer
,
step_number
):
if
not
self
.
step
(
step_number
):
return
return
if
self
.
verbose
:
if
self
.
verbose
:
...
...
modules/textual_inversion/textual_inversion.py
View file @
16451ca5
...
@@ -255,9 +255,12 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc
...
@@ -255,9 +255,12 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc
ititial_step
=
embedding
.
step
or
0
ititial_step
=
embedding
.
step
or
0
if
ititial_step
>
steps
:
if
ititial_step
>
steps
:
return
embedding
,
filename
return
embedding
,
filename
clip_grad_mode_value
=
clip_grad_mode
==
"value"
clip_grad_mode_value
=
clip_grad_mode
==
"value"
clip_grad_mode_norm
=
clip_grad_mode
==
"norm"
clip_grad_mode_norm
=
clip_grad_mode
==
"norm"
clip_grad_enabled
=
clip_grad_mode_value
or
clip_grad_mode_norm
if
clip_grad_enabled
:
clip_grad_sched
=
LearnRateScheduler
(
clip_grad_value
,
steps
,
ititial_step
,
verbose
=
False
)
scheduler
=
LearnRateScheduler
(
learn_rate
,
steps
,
ititial_step
)
scheduler
=
LearnRateScheduler
(
learn_rate
,
steps
,
ititial_step
)
optimizer
=
torch
.
optim
.
AdamW
([
embedding
.
vec
],
lr
=
scheduler
.
learn_rate
)
optimizer
=
torch
.
optim
.
AdamW
([
embedding
.
vec
],
lr
=
scheduler
.
learn_rate
)
...
@@ -273,6 +276,9 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc
...
@@ -273,6 +276,9 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc
if
shared
.
state
.
interrupted
:
if
shared
.
state
.
interrupted
:
break
break
if
clip_grad_enabled
:
clip_grad_sched
.
step
(
embedding
.
step
)
with
torch
.
autocast
(
"cuda"
):
with
torch
.
autocast
(
"cuda"
):
c
=
cond_model
([
entry
.
cond_text
for
entry
in
entries
])
c
=
cond_model
([
entry
.
cond_text
for
entry
in
entries
])
x
=
torch
.
stack
([
entry
.
latent
for
entry
in
entries
])
.
to
(
devices
.
device
)
x
=
torch
.
stack
([
entry
.
latent
for
entry
in
entries
])
.
to
(
devices
.
device
)
...
@@ -285,9 +291,9 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc
...
@@ -285,9 +291,9 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc
loss
.
backward
()
loss
.
backward
()
if
clip_grad_mode_value
:
if
clip_grad_mode_value
:
torch
.
nn
.
utils
.
clip_grad_value_
(
embedding
.
vec
,
clip_value
=
clip_grad_
valu
e
)
torch
.
nn
.
utils
.
clip_grad_value_
(
embedding
.
vec
,
clip_value
=
clip_grad_
sched
.
learn_rat
e
)
elif
clip_grad_mode_norm
:
elif
clip_grad_mode_norm
:
torch
.
nn
.
utils
.
clip_grad_norm_
(
embedding
.
vec
,
max_norm
=
clip_grad_
valu
e
)
torch
.
nn
.
utils
.
clip_grad_norm_
(
embedding
.
vec
,
max_norm
=
clip_grad_
sched
.
learn_rat
e
)
optimizer
.
step
()
optimizer
.
step
()
...
...
modules/ui.py
View file @
16451ca5
...
@@ -1305,7 +1305,9 @@ def create_ui(wrap_gradio_gpu_call):
...
@@ -1305,7 +1305,9 @@ def create_ui(wrap_gradio_gpu_call):
with
gr
.
Row
():
with
gr
.
Row
():
embedding_learn_rate
=
gr
.
Textbox
(
label
=
'Embedding Learning rate'
,
placeholder
=
"Embedding Learning rate"
,
value
=
"0.005"
)
embedding_learn_rate
=
gr
.
Textbox
(
label
=
'Embedding Learning rate'
,
placeholder
=
"Embedding Learning rate"
,
value
=
"0.005"
)
hypernetwork_learn_rate
=
gr
.
Textbox
(
label
=
'Hypernetwork Learning rate'
,
placeholder
=
"Hypernetwork Learning rate"
,
value
=
"0.00001"
)
hypernetwork_learn_rate
=
gr
.
Textbox
(
label
=
'Hypernetwork Learning rate'
,
placeholder
=
"Hypernetwork Learning rate"
,
value
=
"0.00001"
)
with
gr
.
Row
():
clip_grad_mode
=
gr
.
Dropdown
(
value
=
"disabled"
,
label
=
"Gradient Clipping"
,
choices
=
[
"disabled"
,
"value"
,
"norm"
])
clip_grad_value
=
gr
.
Textbox
(
placeholder
=
"Gradient clip value"
,
value
=
"1.0"
,
show_label
=
False
)
batch_size
=
gr
.
Number
(
label
=
'Batch size'
,
value
=
1
,
precision
=
0
)
batch_size
=
gr
.
Number
(
label
=
'Batch size'
,
value
=
1
,
precision
=
0
)
dataset_directory
=
gr
.
Textbox
(
label
=
'Dataset directory'
,
placeholder
=
"Path to directory with input images"
)
dataset_directory
=
gr
.
Textbox
(
label
=
'Dataset directory'
,
placeholder
=
"Path to directory with input images"
)
log_directory
=
gr
.
Textbox
(
label
=
'Log directory'
,
placeholder
=
"Path to directory where to write outputs"
,
value
=
"textual_inversion"
)
log_directory
=
gr
.
Textbox
(
label
=
'Log directory'
,
placeholder
=
"Path to directory where to write outputs"
,
value
=
"textual_inversion"
)
...
@@ -1313,9 +1315,6 @@ def create_ui(wrap_gradio_gpu_call):
...
@@ -1313,9 +1315,6 @@ def create_ui(wrap_gradio_gpu_call):
training_width
=
gr
.
Slider
(
minimum
=
64
,
maximum
=
2048
,
step
=
64
,
label
=
"Width"
,
value
=
512
)
training_width
=
gr
.
Slider
(
minimum
=
64
,
maximum
=
2048
,
step
=
64
,
label
=
"Width"
,
value
=
512
)
training_height
=
gr
.
Slider
(
minimum
=
64
,
maximum
=
2048
,
step
=
64
,
label
=
"Height"
,
value
=
512
)
training_height
=
gr
.
Slider
(
minimum
=
64
,
maximum
=
2048
,
step
=
64
,
label
=
"Height"
,
value
=
512
)
steps
=
gr
.
Number
(
label
=
'Max steps'
,
value
=
100000
,
precision
=
0
)
steps
=
gr
.
Number
(
label
=
'Max steps'
,
value
=
100000
,
precision
=
0
)
with
gr
.
Row
():
clip_grad_mode
=
gr
.
Dropdown
(
value
=
"disabled"
,
label
=
"Gradient Clipping"
,
choices
=
[
"disabled"
,
"value"
,
"norm"
])
clip_grad_value
=
gr
.
Number
(
value
=
1.0
,
show_label
=
False
)
create_image_every
=
gr
.
Number
(
label
=
'Save an image to log directory every N steps, 0 to disable'
,
value
=
500
,
precision
=
0
)
create_image_every
=
gr
.
Number
(
label
=
'Save an image to log directory every N steps, 0 to disable'
,
value
=
500
,
precision
=
0
)
save_embedding_every
=
gr
.
Number
(
label
=
'Save a copy of embedding to log directory every N steps, 0 to disable'
,
value
=
500
,
precision
=
0
)
save_embedding_every
=
gr
.
Number
(
label
=
'Save a copy of embedding to log directory every N steps, 0 to disable'
,
value
=
500
,
precision
=
0
)
save_image_with_stored_embedding
=
gr
.
Checkbox
(
label
=
'Save images with embedding in PNG chunks'
,
value
=
True
)
save_image_with_stored_embedding
=
gr
.
Checkbox
(
label
=
'Save images with embedding in PNG chunks'
,
value
=
True
)
...
...
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