[Dreambooth] TypeError: get_scheduler() got an unexpected keyword argument 'num_cycles'

[claudiogreco]

Describe the bug

Hello,

When I run diffusers/examples/dreambooth/train_dreambooth.py, I get the following error:

│   590 │   │   args.max_train_steps = args.num_train_epochs * num_update_step │
│   591 │   │   overrode_max_train_steps = True                                │
│   592 │                                                                      │
│ ❱ 593 │   lr_scheduler = get_scheduler(                                      │
│   594 │   │   args.lr_scheduler,                                             │
│   595 │   │   optimizer=optimizer,                                           │
│   596 │   │   num_warmup_steps=args.lr_warmup_steps * args.gradient_accumula │

TypeError: get_scheduler() got an unexpected keyword argument 'num_cycles'

The command line was:

accelerate launch train_dreambooth.py --train_text_encoder --pretrained_model_name_or_path="stabilityai/stable-diffusion-2-1-base" --instance_data_dir="/images" --output_dir="/models" --instance_prompt="a photo of ABC person" --resolution=512 --train_batch_size=2 --gradient_accumulation_steps=1 --learning_rate=1e-06 --lr_scheduler=constant --lr_warmup_steps=0 --max_train_steps=1000

I get the same issue described above also by using "runwayml/stable-diffusion-v1-5" as model.

How could I solve this issue? Thank you in advance for your attention.

Reproduction

The command line was:

accelerate launch train_dreambooth.py --train_text_encoder --pretrained_model_name_or_path="stabilityai/stable-diffusion-2-1-base" --instance_data_dir="/images" --output_dir="/models" --instance_prompt="a photo of ABC person" --resolution=512 --train_batch_size=2 --gradient_accumulation_steps=1 --learning_rate=1e-06 --lr_scheduler=constant --lr_warmup_steps=0 --max_train_steps=1000

Logs

No response

System Info

• diffusers version: 0.11.1
• Platform: Linux-5.15.0-1021-gcp-x86_64-with-glibc2.31
• Python version: 3.9.15
• PyTorch version (GPU?): 1.12.1 (True)
• Huggingface_hub version: 0.11.1
• Transformers version: 4.25.1
• Using GPU in script?:
• Using distributed or parallel set-up in script?:

[pcuenca]

Hi @claudiogreco!

To run the training scripts, we recommend you install diffusers from main. Please, take a look at these instructions or, if you have already checked out the repo, run something like pip install --upgrade -e . at the root of your copy. Please, let us know if that works!

[claudiogreco]

Hi @pcuenca,

I solved the issue thanks to your suggestions. Thank you!