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
ab05a74e
Commit
ab05a74e
authored
Oct 29, 2022
by
Muhammad Rizqi Nur
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
Revert "Add cleanup after training"
This reverts commit
3ce2bfdf
.
parent
a27d19de
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
186 additions
and
200 deletions
+186
-200
hypernetwork.py
modules/hypernetworks/hypernetwork.py
+96
-105
textual_inversion.py
modules/textual_inversion/textual_inversion.py
+90
-95
No files found.
modules/hypernetworks/hypernetwork.py
View file @
ab05a74e
...
@@ -398,112 +398,110 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
...
@@ -398,112 +398,110 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
forced_filename
=
"<none>"
forced_filename
=
"<none>"
pbar
=
tqdm
.
tqdm
(
enumerate
(
ds
),
total
=
steps
-
ititial_step
)
pbar
=
tqdm
.
tqdm
(
enumerate
(
ds
),
total
=
steps
-
ititial_step
)
for
i
,
entries
in
pbar
:
try
:
hypernetwork
.
step
=
i
+
ititial_step
for
i
,
entries
in
pbar
:
if
len
(
loss_dict
)
>
0
:
hypernetwork
.
step
=
i
+
ititial_step
previous_mean_losses
=
[
i
[
-
1
]
for
i
in
loss_dict
.
values
()]
if
len
(
loss_dict
)
>
0
:
previous_mean_loss
=
mean
(
previous_mean_losses
)
previous_mean_losses
=
[
i
[
-
1
]
for
i
in
loss_dict
.
values
()]
previous_mean_loss
=
mean
(
previous_mean_losses
)
scheduler
.
apply
(
optimizer
,
hypernetwork
.
step
)
if
scheduler
.
finished
:
break
if
shared
.
state
.
interrupted
:
break
with
torch
.
autocast
(
"cuda"
):
c
=
stack_conds
([
entry
.
cond
for
entry
in
entries
])
.
to
(
devices
.
device
)
# c = torch.vstack([entry.cond for entry in entries]).to(devices.device)
x
=
torch
.
stack
([
entry
.
latent
for
entry
in
entries
])
.
to
(
devices
.
device
)
loss
=
shared
.
sd_model
(
x
,
c
)[
0
]
del
x
del
c
losses
[
hypernetwork
.
step
%
losses
.
shape
[
0
]]
=
loss
.
item
()
for
entry
in
entries
:
loss_dict
[
entry
.
filename
]
.
append
(
loss
.
item
())
optimizer
.
zero_grad
()
weights
[
0
]
.
grad
=
None
loss
.
backward
()
if
weights
[
0
]
.
grad
is
None
:
steps_without_grad
+=
1
else
:
steps_without_grad
=
0
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'
optimizer
.
step
()
steps_done
=
hypernetwork
.
step
+
1
if
torch
.
isnan
(
losses
[
hypernetwork
.
step
%
losses
.
shape
[
0
]]):
raise
RuntimeError
(
"Loss diverged."
)
if
len
(
previous_mean_losses
)
>
1
:
scheduler
.
apply
(
optimizer
,
hypernetwork
.
step
)
std
=
stdev
(
previous_mean_losses
)
if
scheduler
.
finished
:
break
if
shared
.
state
.
interrupted
:
break
with
torch
.
autocast
(
"cuda"
):
c
=
stack_conds
([
entry
.
cond
for
entry
in
entries
])
.
to
(
devices
.
device
)
# c = torch.vstack([entry.cond for entry in entries]).to(devices.device)
x
=
torch
.
stack
([
entry
.
latent
for
entry
in
entries
])
.
to
(
devices
.
device
)
loss
=
shared
.
sd_model
(
x
,
c
)[
0
]
del
x
del
c
losses
[
hypernetwork
.
step
%
losses
.
shape
[
0
]]
=
loss
.
item
()
for
entry
in
entries
:
loss_dict
[
entry
.
filename
]
.
append
(
loss
.
item
())
optimizer
.
zero_grad
()
weights
[
0
]
.
grad
=
None
loss
.
backward
()
if
weights
[
0
]
.
grad
is
None
:
steps_without_grad
+=
1
else
:
else
:
std
=
0
steps_without_grad
=
0
dataset_loss_info
=
f
"dataset loss:{mean(previous_mean_losses):.3f}"
+
u"
\u00B1
"
+
f
"({std / (len(previous_mean_losses) ** 0.5):.3f})"
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'
pbar
.
set_description
(
dataset_loss_info
)
if
hypernetwork_dir
is
not
None
and
steps_done
%
save_hypernetwork_every
==
0
:
# Before saving, change name to match current checkpoint.
hypernetwork
.
name
=
f
'{hypernetwork_name}-{steps_done}'
last_saved_file
=
os
.
path
.
join
(
hypernetwork_dir
,
f
'{hypernetwork.name}.pt'
)
hypernetwork
.
save
(
last_saved_file
)
textual_inversion
.
write_loss
(
log_directory
,
"hypernetwork_loss.csv"
,
hypernetwork
.
step
,
len
(
ds
),
{
"loss"
:
f
"{previous_mean_loss:.7f}"
,
"learn_rate"
:
scheduler
.
learn_rate
})
if
images_dir
is
not
None
and
steps_done
%
create_image_every
==
0
:
forced_filename
=
f
'{hypernetwork_name}-{steps_done}'
last_saved_image
=
os
.
path
.
join
(
images_dir
,
forced_filename
)
optimizer
.
zero_grad
()
shared
.
sd_model
.
cond_stage_model
.
to
(
devices
.
device
)
shared
.
sd_model
.
first_stage_model
.
to
(
devices
.
device
)
p
=
processing
.
StableDiffusionProcessingTxt2Img
(
sd_model
=
shared
.
sd_model
,
do_not_save_grid
=
True
,
do_not_save_samples
=
True
,
)
if
preview_from_txt2img
:
optimizer
.
step
()
p
.
prompt
=
preview_prompt
p
.
negative_prompt
=
preview_negative_prompt
p
.
steps
=
preview_steps
p
.
sampler_index
=
preview_sampler_index
p
.
cfg_scale
=
preview_cfg_scale
p
.
seed
=
preview_seed
p
.
width
=
preview_width
p
.
height
=
preview_height
else
:
p
.
prompt
=
entries
[
0
]
.
cond_text
p
.
steps
=
20
preview_text
=
p
.
prompt
steps_done
=
hypernetwork
.
step
+
1
processed
=
processing
.
process_images
(
p
)
if
torch
.
isnan
(
losses
[
hypernetwork
.
step
%
losses
.
shape
[
0
]]):
image
=
processed
.
images
[
0
]
if
len
(
processed
.
images
)
>
0
else
None
raise
RuntimeError
(
"Loss diverged."
)
if
len
(
previous_mean_losses
)
>
1
:
std
=
stdev
(
previous_mean_losses
)
else
:
std
=
0
dataset_loss_info
=
f
"dataset loss:{mean(previous_mean_losses):.3f}"
+
u"
\u00B1
"
+
f
"({std / (len(previous_mean_losses) ** 0.5):.3f})"
pbar
.
set_description
(
dataset_loss_info
)
if
hypernetwork_dir
is
not
None
and
steps_done
%
save_hypernetwork_every
==
0
:
# Before saving, change name to match current checkpoint.
hypernetwork
.
name
=
f
'{hypernetwork_name}-{steps_done}'
last_saved_file
=
os
.
path
.
join
(
hypernetwork_dir
,
f
'{hypernetwork.name}.pt'
)
hypernetwork
.
save
(
last_saved_file
)
textual_inversion
.
write_loss
(
log_directory
,
"hypernetwork_loss.csv"
,
hypernetwork
.
step
,
len
(
ds
),
{
"loss"
:
f
"{previous_mean_loss:.7f}"
,
"learn_rate"
:
scheduler
.
learn_rate
})
if
images_dir
is
not
None
and
steps_done
%
create_image_every
==
0
:
forced_filename
=
f
'{hypernetwork_name}-{steps_done}'
last_saved_image
=
os
.
path
.
join
(
images_dir
,
forced_filename
)
optimizer
.
zero_grad
()
shared
.
sd_model
.
cond_stage_model
.
to
(
devices
.
device
)
shared
.
sd_model
.
first_stage_model
.
to
(
devices
.
device
)
if
unload
:
p
=
processing
.
StableDiffusionProcessingTxt2Img
(
shared
.
sd_model
.
cond_stage_model
.
to
(
devices
.
cpu
)
sd_model
=
shared
.
sd_model
,
shared
.
sd_model
.
first_stage_model
.
to
(
devices
.
cpu
)
do_not_save_grid
=
True
,
do_not_save_samples
=
True
,
)
if
image
is
not
None
:
if
preview_from_txt2img
:
shared
.
state
.
current_image
=
image
p
.
prompt
=
preview_prompt
last_saved_image
,
last_text_info
=
images
.
save_image
(
image
,
images_dir
,
""
,
p
.
seed
,
p
.
prompt
,
shared
.
opts
.
samples_format
,
processed
.
infotexts
[
0
],
p
=
p
,
forced_filename
=
forced_filename
,
save_to_dirs
=
False
)
p
.
negative_prompt
=
preview_negative_prompt
last_saved_image
+=
f
", prompt: {preview_text}"
p
.
steps
=
preview_steps
p
.
sampler_index
=
preview_sampler_index
p
.
cfg_scale
=
preview_cfg_scale
p
.
seed
=
preview_seed
p
.
width
=
preview_width
p
.
height
=
preview_height
else
:
p
.
prompt
=
entries
[
0
]
.
cond_text
p
.
steps
=
20
preview_text
=
p
.
prompt
processed
=
processing
.
process_images
(
p
)
image
=
processed
.
images
[
0
]
if
len
(
processed
.
images
)
>
0
else
None
if
unload
:
shared
.
sd_model
.
cond_stage_model
.
to
(
devices
.
cpu
)
shared
.
sd_model
.
first_stage_model
.
to
(
devices
.
cpu
)
shared
.
state
.
job_no
=
hypernetwork
.
step
if
image
is
not
None
:
shared
.
state
.
current_image
=
image
last_saved_image
,
last_text_info
=
images
.
save_image
(
image
,
images_dir
,
""
,
p
.
seed
,
p
.
prompt
,
shared
.
opts
.
samples_format
,
processed
.
infotexts
[
0
],
p
=
p
,
forced_filename
=
forced_filename
,
save_to_dirs
=
False
)
last_saved_image
+=
f
", prompt: {preview_text}"
shared
.
state
.
textinfo
=
f
"""
shared
.
state
.
job_no
=
hypernetwork
.
step
shared
.
state
.
textinfo
=
f
"""
<p>
<p>
Loss: {previous_mean_loss:.7f}<br/>
Loss: {previous_mean_loss:.7f}<br/>
Step: {hypernetwork.step}<br/>
Step: {hypernetwork.step}<br/>
...
@@ -512,14 +510,7 @@ Last saved hypernetwork: {html.escape(last_saved_file)}<br/>
...
@@ -512,14 +510,7 @@ Last saved hypernetwork: {html.escape(last_saved_file)}<br/>
Last saved image: {html.escape(last_saved_image)}<br/>
Last saved image: {html.escape(last_saved_image)}<br/>
</p>
</p>
"""
"""
finally
:
if
weights
:
for
weight
in
weights
:
weight
.
requires_grad
=
False
if
unload
:
shared
.
sd_model
.
cond_stage_model
.
to
(
devices
.
device
)
shared
.
sd_model
.
first_stage_model
.
to
(
devices
.
device
)
report_statistics
(
loss_dict
)
report_statistics
(
loss_dict
)
checkpoint
=
sd_models
.
select_checkpoint
()
checkpoint
=
sd_models
.
select_checkpoint
()
...
...
modules/textual_inversion/textual_inversion.py
View file @
ab05a74e
...
@@ -283,113 +283,111 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc
...
@@ -283,113 +283,111 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc
embedding_yet_to_be_embedded
=
False
embedding_yet_to_be_embedded
=
False
pbar
=
tqdm
.
tqdm
(
enumerate
(
ds
),
total
=
steps
-
ititial_step
)
pbar
=
tqdm
.
tqdm
(
enumerate
(
ds
),
total
=
steps
-
ititial_step
)
for
i
,
entries
in
pbar
:
embedding
.
step
=
i
+
ititial_step
try
:
scheduler
.
apply
(
optimizer
,
embedding
.
step
)
for
i
,
entries
in
pbar
:
if
scheduler
.
finished
:
embedding
.
step
=
i
+
ititial_step
break
scheduler
.
apply
(
optimizer
,
embedding
.
step
)
if
shared
.
state
.
interrupted
:
if
scheduler
.
finished
:
break
break
with
torch
.
autocast
(
"cuda"
):
if
shared
.
state
.
interrupted
:
c
=
cond_model
([
entry
.
cond_text
for
entry
in
entries
])
break
x
=
torch
.
stack
([
entry
.
latent
for
entry
in
entries
])
.
to
(
devices
.
device
)
loss
=
shared
.
sd_model
(
x
,
c
)[
0
]
with
torch
.
autocast
(
"cuda"
):
del
x
c
=
cond_model
([
entry
.
cond_text
for
entry
in
entries
])
x
=
torch
.
stack
([
entry
.
latent
for
entry
in
entries
])
.
to
(
devices
.
device
)
losses
[
embedding
.
step
%
losses
.
shape
[
0
]]
=
loss
.
item
()
loss
=
shared
.
sd_model
(
x
,
c
)[
0
]
del
x
optimizer
.
zero_grad
()
loss
.
backward
()
losses
[
embedding
.
step
%
losses
.
shape
[
0
]]
=
loss
.
item
()
optimizer
.
step
()
optimizer
.
zero_grad
()
steps_done
=
embedding
.
step
+
1
loss
.
backward
()
optimizer
.
step
()
epoch_num
=
embedding
.
step
//
len
(
ds
)
epoch_step
=
embedding
.
step
%
len
(
ds
)
steps_done
=
embedding
.
step
+
1
pbar
.
set_description
(
f
"[Epoch {epoch_num}: {epoch_step+1}/{len(ds)}]loss: {losses.mean():.7f}"
)
epoch_num
=
embedding
.
step
//
len
(
ds
)
epoch_step
=
embedding
.
step
%
len
(
ds
)
if
embedding_dir
is
not
None
and
steps_done
%
save_embedding_every
==
0
:
# Before saving, change name to match current checkpoint.
pbar
.
set_description
(
f
"[Epoch {epoch_num}: {epoch_step+1}/{len(ds)}]loss: {losses.mean():.7f}"
)
embedding
.
name
=
f
'{embedding_name}-{steps_done}'
last_saved_file
=
os
.
path
.
join
(
embedding_dir
,
f
'{embedding.name}.pt'
)
if
embedding_dir
is
not
None
and
steps_done
%
save_embedding_every
==
0
:
embedding
.
save
(
last_saved_file
)
# Before saving, change name to match current checkpoint.
embedding_yet_to_be_embedded
=
True
embedding
.
name
=
f
'{embedding_name}-{steps_done}'
last_saved_file
=
os
.
path
.
join
(
embedding_dir
,
f
'{embedding.name}.pt'
)
write_loss
(
log_directory
,
"textual_inversion_loss.csv"
,
embedding
.
step
,
len
(
ds
),
{
embedding
.
save
(
last_saved_file
)
"loss"
:
f
"{losses.mean():.7f}"
,
embedding_yet_to_be_embedded
=
True
"learn_rate"
:
scheduler
.
learn_rate
})
write_loss
(
log_directory
,
"textual_inversion_loss.csv"
,
embedding
.
step
,
len
(
ds
),
{
"loss"
:
f
"{losses.mean():.7f}"
,
if
images_dir
is
not
None
and
steps_done
%
create_image_every
==
0
:
"learn_rate"
:
scheduler
.
learn_rate
forced_filename
=
f
'{embedding_name}-{steps_done}'
})
last_saved_image
=
os
.
path
.
join
(
images_dir
,
forced_filename
)
p
=
processing
.
StableDiffusionProcessingTxt2Img
(
if
images_dir
is
not
None
and
steps_done
%
create_image_every
==
0
:
sd_model
=
shared
.
sd_model
,
forced_filename
=
f
'{embedding_name}-{steps_done}'
do_not_save_grid
=
True
,
last_saved_image
=
os
.
path
.
join
(
images_dir
,
forced_filename
)
do_not_save_samples
=
True
,
p
=
processing
.
StableDiffusionProcessingTxt2Img
(
do_not_reload_embeddings
=
True
,
sd_model
=
shared
.
sd_model
,
)
do_not_save_grid
=
True
,
do_not_save_samples
=
True
,
if
preview_from_txt2img
:
do_not_reload_embeddings
=
True
,
p
.
prompt
=
preview_prompt
)
p
.
negative_prompt
=
preview_negative_prompt
p
.
steps
=
preview_steps
if
preview_from_txt2img
:
p
.
sampler_index
=
preview_sampler_index
p
.
prompt
=
preview_prompt
p
.
cfg_scale
=
preview_cfg_scale
p
.
negative_prompt
=
preview_negative_prompt
p
.
seed
=
preview_seed
p
.
steps
=
preview_steps
p
.
width
=
preview_width
p
.
sampler_index
=
preview_sampler_index
p
.
height
=
preview_height
p
.
cfg_scale
=
preview_cfg_scale
else
:
p
.
seed
=
preview_seed
p
.
prompt
=
entries
[
0
]
.
cond_text
p
.
width
=
preview_width
p
.
steps
=
20
p
.
height
=
preview_height
p
.
width
=
training_width
else
:
p
.
height
=
training_height
p
.
prompt
=
entries
[
0
]
.
cond_text
p
.
steps
=
20
p
.
width
=
training_width
p
.
height
=
training_height
preview_text
=
p
.
prompt
preview_text
=
p
.
prompt
processed
=
processing
.
process_images
(
p
)
processed
=
processing
.
process_images
(
p
)
image
=
processed
.
images
[
0
]
image
=
processed
.
images
[
0
]
shared
.
state
.
current_image
=
image
shared
.
state
.
current_image
=
image
if
save_image_with_stored_embedding
and
os
.
path
.
exists
(
last_saved_file
)
and
embedding_yet_to_be_embedded
:
if
save_image_with_stored_embedding
and
os
.
path
.
exists
(
last_saved_file
)
and
embedding_yet_to_be_embedded
:
last_saved_image_chunks
=
os
.
path
.
join
(
images_embeds_dir
,
f
'{embedding_name}-{steps_done}.png'
)
last_saved_image_chunks
=
os
.
path
.
join
(
images_embeds_dir
,
f
'{embedding_name}-{steps_done}.png'
)
info
=
PngImagePlugin
.
PngInfo
()
info
=
PngImagePlugin
.
PngInfo
()
data
=
torch
.
load
(
last_saved_file
)
data
=
torch
.
load
(
last_saved_file
)
info
.
add_text
(
"sd-ti-embedding"
,
embedding_to_b64
(
data
))
info
.
add_text
(
"sd-ti-embedding"
,
embedding_to_b64
(
data
))
title
=
"<{}>"
.
format
(
data
.
get
(
'name'
,
'???'
))
title
=
"<{}>"
.
format
(
data
.
get
(
'name'
,
'???'
))
try
:
try
:
vectorSize
=
list
(
data
[
'string_to_param'
]
.
values
())[
0
]
.
shape
[
0
]
vectorSize
=
list
(
data
[
'string_to_param'
]
.
values
())[
0
]
.
shape
[
0
]
except
Exception
as
e
:
except
Exception
as
e
:
vectorSize
=
'?'
vectorSize
=
'?'
checkpoint
=
sd_models
.
select_checkpoint
()
checkpoint
=
sd_models
.
select_checkpoint
()
footer_left
=
checkpoint
.
model_name
footer_left
=
checkpoint
.
model_name
footer_mid
=
'[{}]'
.
format
(
checkpoint
.
hash
)
footer_mid
=
'[{}]'
.
format
(
checkpoint
.
hash
)
footer_right
=
'{}v {}s'
.
format
(
vectorSize
,
steps_done
)
footer_right
=
'{}v {}s'
.
format
(
vectorSize
,
steps_done
)
captioned_image
=
caption_image_overlay
(
image
,
title
,
footer_left
,
footer_mid
,
footer_right
)
captioned_image
=
caption_image_overlay
(
image
,
title
,
footer_left
,
footer_mid
,
footer_right
)
captioned_image
=
insert_image_data_embed
(
captioned_image
,
data
)
captioned_image
=
insert_image_data_embed
(
captioned_image
,
data
)
captioned_image
.
save
(
last_saved_image_chunks
,
"PNG"
,
pnginfo
=
info
)
captioned_image
.
save
(
last_saved_image_chunks
,
"PNG"
,
pnginfo
=
info
)
embedding_yet_to_be_embedded
=
False
embedding_yet_to_be_embedded
=
False
last_saved_image
,
last_text_info
=
images
.
save_image
(
image
,
images_dir
,
""
,
p
.
seed
,
p
.
prompt
,
shared
.
opts
.
samples_format
,
processed
.
infotexts
[
0
],
p
=
p
,
forced_filename
=
forced_filename
,
save_to_dirs
=
False
)
last_saved_image
,
last_text_info
=
images
.
save_image
(
image
,
images_dir
,
""
,
p
.
seed
,
p
.
prompt
,
shared
.
opts
.
samples_format
,
processed
.
infotexts
[
0
],
p
=
p
,
forced_filename
=
forced_filename
,
save_to_dirs
=
False
)
last_saved_image
+=
f
", prompt: {preview_text}"
last_saved_image
+=
f
", prompt: {preview_text}"
shared
.
state
.
job_no
=
embedding
.
step
shared
.
state
.
job_no
=
embedding
.
step
shared
.
state
.
textinfo
=
f
"""
shared
.
state
.
textinfo
=
f
"""
<p>
<p>
Loss: {losses.mean():.7f}<br/>
Loss: {losses.mean():.7f}<br/>
Step: {embedding.step}<br/>
Step: {embedding.step}<br/>
...
@@ -398,9 +396,6 @@ Last saved embedding: {html.escape(last_saved_file)}<br/>
...
@@ -398,9 +396,6 @@ Last saved embedding: {html.escape(last_saved_file)}<br/>
Last saved image: {html.escape(last_saved_image)}<br/>
Last saved image: {html.escape(last_saved_image)}<br/>
</p>
</p>
"""
"""
finally
:
if
embedding
and
embedding
.
vec
is
not
None
:
embedding
.
vec
.
requires_grad
=
False
checkpoint
=
sd_models
.
select_checkpoint
()
checkpoint
=
sd_models
.
select_checkpoint
()
...
...
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