Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
OpenDAS
diffusers
Commits
24563ca6
Unverified
Commit
24563ca6
authored
Sep 20, 2023
by
Bagheera
Committed by
GitHub
Sep 20, 2023
Browse files
SNR gamma fixes for v_prediction training (#5106)
Co-authored-by:
bghira
<
bghira@users.github.com
>
parent
914586f5
Changes
5
Show whitespace changes
Inline
Side-by-side
Showing
5 changed files
with
15 additions
and
0 deletions
+15
-0
examples/controlnet/train_controlnet_flax.py
examples/controlnet/train_controlnet_flax.py
+3
-0
examples/research_projects/onnxruntime/text_to_image/train_text_to_image.py
...projects/onnxruntime/text_to_image/train_text_to_image.py
+3
-0
examples/text_to_image/train_text_to_image.py
examples/text_to_image/train_text_to_image.py
+3
-0
examples/text_to_image/train_text_to_image_lora.py
examples/text_to_image/train_text_to_image_lora.py
+3
-0
examples/text_to_image/train_text_to_image_lora_sdxl.py
examples/text_to_image/train_text_to_image_lora_sdxl.py
+3
-0
No files found.
examples/controlnet/train_controlnet_flax.py
View file @
24563ca6
...
...
@@ -908,6 +908,9 @@ def main():
if
args
.
snr_gamma
is
not
None
:
snr
=
jnp
.
array
(
compute_snr
(
timesteps
))
snr_loss_weights
=
jnp
.
where
(
snr
<
args
.
snr_gamma
,
snr
,
jnp
.
ones_like
(
snr
)
*
args
.
snr_gamma
)
/
snr
if
noise_scheduler
.
config
.
prediction_type
==
"v_prediction"
:
# velocity objective prediction requires SNR weights to be floored to a min value of 1.
snr_loss_weights
=
snr_loss_weights
+
1
loss
=
loss
*
snr_loss_weights
loss
=
loss
.
mean
()
...
...
examples/research_projects/onnxruntime/text_to_image/train_text_to_image.py
View file @
24563ca6
...
...
@@ -875,6 +875,9 @@ def main():
mse_loss_weights
=
(
torch
.
stack
([
snr
,
args
.
snr_gamma
*
torch
.
ones_like
(
timesteps
)],
dim
=
1
).
min
(
dim
=
1
)[
0
]
/
snr
)
if
noise_scheduler
.
config
.
prediction_type
==
"v_prediction"
:
# velocity objective prediction requires SNR weights to be floored to a min value of 1.
mse_loss_weights
=
mse_loss_weights
+
1
# We first calculate the original loss. Then we mean over the non-batch dimensions and
# rebalance the sample-wise losses with their respective loss weights.
# Finally, we take the mean of the rebalanced loss.
...
...
examples/text_to_image/train_text_to_image.py
View file @
24563ca6
...
...
@@ -955,6 +955,9 @@ def main():
mse_loss_weights
=
(
torch
.
stack
([
snr
,
args
.
snr_gamma
*
torch
.
ones_like
(
timesteps
)],
dim
=
1
).
min
(
dim
=
1
)[
0
]
/
snr
)
if
noise_scheduler
.
config
.
prediction_type
==
"v_prediction"
:
# velocity objective prediction requires SNR weights to be floored to a min value of 1.
mse_loss_weights
=
mse_loss_weights
+
1
# We first calculate the original loss. Then we mean over the non-batch dimensions and
# rebalance the sample-wise losses with their respective loss weights.
# Finally, we take the mean of the rebalanced loss.
...
...
examples/text_to_image/train_text_to_image_lora.py
View file @
24563ca6
...
...
@@ -786,6 +786,9 @@ def main():
mse_loss_weights
=
(
torch
.
stack
([
snr
,
args
.
snr_gamma
*
torch
.
ones_like
(
timesteps
)],
dim
=
1
).
min
(
dim
=
1
)[
0
]
/
snr
)
if
noise_scheduler
.
config
.
prediction_type
==
"v_prediction"
:
# velocity objective prediction requires SNR weights to be floored to a min value of 1.
mse_loss_weights
=
mse_loss_weights
+
1
# We first calculate the original loss. Then we mean over the non-batch dimensions and
# rebalance the sample-wise losses with their respective loss weights.
# Finally, we take the mean of the rebalanced loss.
...
...
examples/text_to_image/train_text_to_image_lora_sdxl.py
View file @
24563ca6
...
...
@@ -1075,6 +1075,9 @@ def main(args):
mse_loss_weights
=
(
torch
.
stack
([
snr
,
args
.
snr_gamma
*
torch
.
ones_like
(
timesteps
)],
dim
=
1
).
min
(
dim
=
1
)[
0
]
/
snr
)
if
noise_scheduler
.
config
.
prediction_type
==
"v_prediction"
:
# velocity objective prediction requires SNR weights to be floored to a min value of 1.
mse_loss_weights
=
mse_loss_weights
+
1
# We first calculate the original loss. Then we mean over the non-batch dimensions and
# rebalance the sample-wise losses with their respective loss weights.
# Finally, we take the mean of the rebalanced loss.
...
...
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