fix diffuser save_lora_weights crash

[sywangyi]

fix crash in train_dreambooth_lora_sdxl.py in save_lora_weights
python ../../gaudi_spawn.py --world_size 8 --use_mpi train_dreambooth_lora_sdxl.py --pretrained_model_name_or_path="stabilityai/stable-diffusion-xl-base-1.0" --instance_data_dir="dog" --pretrained_vae_model_name_or_path="madebyollin/sdxl-vae-fp16-fix" --output_dir="lora-trained-xl" --mixed_precision="bf16" --instance_prompt="a photo of sks dog" --resolution=1024 --train_batch_size=1 --gradient_accumulation_steps=4 --learning_rate=1e-4 --lr_scheduler="constant" --lr_warmup_steps=0 --max_train_steps=100 --validation_prompt="A photo of sks dog in a bucket" --validation_epochs=25 --seed=0 --use_hpu_graphs_for_inference --use_hpu_graphs_for_training --gaudi_config_name Habana/stable-diffusion

[sywangyi]

File "/root/optimum/optimum-habana/examples/stable-diffusion/training/train_dreambooth_lora_sdxl.py", line 1768, in
main(args)
File "/root/optimum/optimum-habana/examples/stable-diffusion/training/train_dreambooth_lora_sdxl.py", line 1679, in main
GaudiStableDiffusionXLPipeline.save_lora_weights(
File "/usr/local/lib/python3.10/dist-packages/optimum/habana/diffusers/pipelines/pipeline_utils.py", line 411, in save_lora_weights
return super().save_lora_weights(
File "/usr/local/lib/python3.10/dist-packages/diffusers/loaders/lora_pipeline.py", line 1023, in save_lora_weights
state_dict.update(cls.pack_weights(text_encoder_2_lora_layers, "text_encoder_2"))
File "/usr/local/lib/python3.10/dist-packages/diffusers/loaders/lora_base.py", line 716, in pack_weights
layers_state_dict = {f"{prefix}.{module_name}": param for module_name, param in layers_weights.items()}
AttributeError: 'bool' object has no attribute 'items'

[regisss]

@sywangyi I merged #1626 as the change there is more generic

[sywangyi]

#1626 could not fix the crash in train_dreambooth_lora_sdxl , #1633 could fix both crash "train_dreambooth_lora_sdxl" and ddpo

[regisss]

@sywangyi Please take a look at #1643 and let me know if that works on your side too. It should enable us to keep a generic solution while solving this issue for good.

[sywangyi]

@sywangyi Please take a look at #1643 and let me know if that works on your side too. It should enable us to keep a generic solution while solving this issue for good.

see comment in #1643