Skip to content

Add rank argument to train_dreambooth_lora_sdxl.py #4343

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
merged 2 commits into from
Aug 3, 2023
+15 −3
Merged

Add rank argument to train_dreambooth_lora_sdxl.py #4343

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
77b9f04
Add rank argument to train_dreambooth_lora_sdxl.py
levi Jul 28, 2023
8997fc6
Update train_dreambooth_lora_sdxl.py
levi Jul 28, 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
18 changes: 15 additions & 3 deletions examples/dreambooth/train_dreambooth_lora_sdxl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -402,6 +402,12 @@ def parse_args(input_args=None):
parser.add_argument(
"--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
)
parser.add_argument(
"--rank",
type=int,
default=4,
help=("The dimension of the LoRA update matrices."),
)

if input_args is not None:
args = parser.parse_args(input_args)
Expand Down Expand Up @@ -767,7 +773,9 @@ def main(args):
lora_attn_processor_class = (
LoRAAttnProcessor2_0 if hasattr(F, "scaled_dot_product_attention") else LoRAAttnProcessor
)
module = lora_attn_processor_class(hidden_size=hidden_size, cross_attention_dim=cross_attention_dim)
module = lora_attn_processor_class(
hidden_size=hidden_size, cross_attention_dim=cross_attention_dim, rank=args.rank
)
unet_lora_attn_procs[name] = module
unet_lora_parameters.extend(module.parameters())

Expand All @@ -777,8 +785,12 @@ def main(args):
# So, instead, we monkey-patch the forward calls of its attention-blocks.
if args.train_text_encoder:
# ensure that dtype is float32, even if rest of the model that isn't trained is loaded in fp16
text_lora_parameters_one = LoraLoaderMixin._modify_text_encoder(text_encoder_one, dtype=torch.float32)
text_lora_parameters_two = LoraLoaderMixin._modify_text_encoder(text_encoder_two, dtype=torch.float32)
text_lora_parameters_one = LoraLoaderMixin._modify_text_encoder(
text_encoder_one, dtype=torch.float32, rank=args.rank
)
text_lora_parameters_two = LoraLoaderMixin._modify_text_encoder(
text_encoder_two, dtype=torch.float32, rank=args.rank
)

# create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
def save_model_hook(models, weights, output_dir):
Expand Down