Lora Weights not Loaded from Checkpoint #6252
Description
Describe the bug
After training with diffusers/examples/text_to_image/train_text_to_image_lora.py script
the lora checkpoint couldnt loaded to pipeline
Reproduction
!accelerate launch /content/diffusers/examples/text_to_image/train_text_to_image_lora.py \
--pretrained_model_name_or_path="stabilityai/stable-diffusion-2-1" \
--dataset_name="dataset_path" \
--resolution=512 --center_crop \
--train_batch_size=2 \
--gradient_accumulation_steps=1 \
--dataloader_num_workers=8 \
--num_train_epochs=200 \
--max_train_steps=10000 \
--learning_rate=1e-05 \
--max_grad_norm=1 \
--lr_scheduler="cosine" \
--lr_warmup_steps=100 \
--checkpointing_steps=500 \
--allow_tf32 \
--seed=42 \
--caption_column="caption" \
--rank=32 \
--output_dir="/content/drive/MyDrive/"import torch
from diffusers import DiffusionPipeline, DDPMScheduler
pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-2-1", use_safetensors=True)
pipe.scheduler = DDPMScheduler.from_config(pipe.scheduler.config)
pipe.load_lora_weights("training_path/checkpoint-1000", weight_name="model.safetensors")
pipe.to("cuda")[/usr/local/lib/python3.10/dist-packages/diffusers/loaders/lora.py](https://localhost:8080/#) in load_lora_weights(self, pretrained_model_name_or_path_or_dict, adapter_name, **kwargs)
108 is_correct_format = all("lora" in key for key in state_dict.keys())
109 if not is_correct_format:
--> 110 raise ValueError("Invalid LoRA checkpoint.")
111
112 low_cpu_mem_usage = kwargs.pop("low_cpu_mem_usage", _LOW_CPU_MEM_USAGE_DEFAULT)
ValueError: Invalid LoRA checkpoint.Logs
No response
System Info
diffusersversion: 0.25.0.dev0- Platform: Linux-6.1.58+-x86_64-with-glibc2.35
- Python version: 3.10.12
- PyTorch version (GPU?): 2.1.0+cu121 (True)
- Huggingface_hub version: 0.19.4
- Transformers version: 4.35.2
- Accelerate version: 0.25.0
- xFormers version: not installed
- Using GPU in script?:
- Using distributed or parallel set-up in script?: