CycleDiffusion: Text-to-Image Diffusion Models Are Image-to-Image Editors via Inferring "Random Seed"

Official PyTorch implementation of (Sections 4.1 and 4.2 of) our paper
Unifying Diffusion Models' Latent Space, with Applications to CycleDiffusion and Guidance
Chen Henry Wu, Fernando De la Torre
Carnegie Mellon University
Preprint, Oct 2022

[Paper link] | [Diffusers 🧨 implementation] | [HuggingFace 🤗 demo]

Updates

[Oct 13 2022] Code released.

[Oct 16 2022] Updated customized use for zero-shot image editing.

[Nov 9 2022] CycleDiffusion is now available as a pipeline on HuggingFace 🤗 Diffusers 🧨. Please check the pipeline doc. The CycleDiffusion pipeline supports Cross Attention Control (CAC).

[Nov 10 2022] A demo built with HuggingFace 🤗 Spaces is available at Stable CycleDiffusion. CAC is supported.

Notes

1. Section 4.3 of this paper is open-sourced at Unified Generative Zoo.
2. The code is based on Generative Visual Prompt.
3. Feel free to email me if you think I should cite your work!

Overview

We think the randomness in diffusion models is like magic! Accumulated evidence has shown that fixing the "random seed" helps diffusion models generate images from two image distributions with minimal differences. Our paper is exactly about how to formalize this "random seed" and how to infer it from a given real image.

Our formalization and derivation are purely by definition, while we show that some amazing consequences follow! This repository contains code for CycleDiffusion, an embarrassingly simple method capable of

1. Zero-shot image-to-image translation with text-to-image diffusion models such as Stable Diffusion.
2. Traditional unpaired image-to-image translation with diffusion models trained on two related domains.

Check our results on zero-shot image-to-image translation below! We formulate the task input as a triplet $(\boldsymbol{x}, \boldsymbol{t}, \hat{\boldsymbol{t}})$:

1. $\boldsymbol{x}$ is the source image, displayed with a purple margin.
2. $\boldsymbol{t}$ is the source text, with text spans marked in purple.
3. $\hat{\boldsymbol{t}}$ is the target text, with text spans abbreviated as $[\ldots]$ if overlapped with the source text.

We used Stable Diffusion in our experiments. Notably, all source images $\boldsymbol{x}$ are real images! Yes, you find that some of them are generated by DALL∙E 2, but these images can be seen as real for Stable Diffusion :)

Here are some comparisons with baselines.

Contents

• CycleDiffusion: Text-to-Image Diffusion Models Are Image-to-Image Editors via Inferring "Random Seed"
  • Updates
  • TODOs
  • Notes
  • Overview
  • Contents
  • Dependencies
  • Evaluation data
  • Pre-trained diffusion models
  • Usage
    • Zero-shot image-to-image translation with text-to-image diffusion models
    • Customized use for zero-shot image-to-image translation
    • Unpaired image-to-image translation with diffusion models trained on two domains
  • Citation
  • Potential Societal Impact
  • License
  • Contact

Dependencies

1. Create environment by running

conda env create -f environment.yml
conda activate generative_prompt
pip install git+https://github.com/openai/CLIP.git

2. Install torch and torchvision based on your CUDA version.
3. Install taming-transformers by running

cd ../
git clone git@github.com:CompVis/taming-transformers.git
cd taming-transformers/
pip install -e .
cd ../

4. Set up wandb for logging (registration is required). You should modify the setup_wandb function in main.py to accomodate your wandb credentials. You may want to run something like

wandb login

Evaluation data

1. Most data for zero-shot image-to-image translation are already included in data/. Some images are from the AFHQ validation set, detailed below.
2. Prepare the AFHQ validation set for unpaired image-to-image translation (also for some images used by zero-shot image-to-image translation) by running

git clone git@github.com:clovaai/stargan-v2.git
cd stargan-v2/
bash download.sh afhq-v2-dataset

Pre-trained diffusion models

1. Stable Diffusion

cd ckpts/
mkdir stable_diffusion
cd stable_diffusion/
# Download pre-trained checkpoints for Stable Diffusion here.
# You should download this version: https://huggingface.co/CompVis/stable-diffusion-v-1-4-original
# Due to licence issues, we cannot share the pre-trained checkpoints directly.

2. Latent Diffusion Model

cd ckpts/
wget https://www.dropbox.com/s/9lpdgs83l7tjk6c/ldm_models.zip
unzip ldm_models.zip
cd ldm_models/
mkdir text2img-large
cd text2img-large/
wget https://ommer-lab.com/files/latent-diffusion/nitro/txt2img-f8-large/model.ckpt
wget https://www.dropbox.com/s/7pdttimz78ll0km/txt2img-1p4B-eval.yaml

3. DDPM (AFHQ-Dog and FFHQ are from ILVR; CelebAHQ is from SDEdit; AFHQ-Cat and -Wild are trained by ourselves)

cd ckpts/
mkdir ddpm
cd ddpm/
wget https://image-editing-test-12345.s3-us-west-2.amazonaws.com/checkpoints/celeba_hq.ckpt
wget https://www.dropbox.com/s/g4h8sv07i3hj83d/ffhq_10m.pt
wget https://www.dropbox.com/s/u74w8vaw1f8lc4k/afhq_dog_4m.pt
wget https://www.dropbox.com/s/8i5aznjwdl3b5iq/cat_ema_0.9999_050000.pt
wget https://www.dropbox.com/s/tplximipy8zxaub/wild_ema_0.9999_050000.pt
wget https://www.dropbox.com/s/vqm6bxj0zslrjxv/configs.zip
unzip configs.zip

Usage

Zero-shot image-to-image translation with text-to-image diffusion models

1. Zero-shot image-to-image translation with Stable Diffusion v1-4. We divided the 128 test samples into 8 groups (16 samples in each group), so the averaged metrics are reported.

export CUDA_VISIBLE_DEVICES=0
export RUN_NAME=translate_text2img256_stable_diffusion_stochastic_1
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1405 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 4 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=1
export RUN_NAME=translate_text2img256_stable_diffusion_stochastic_2
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1424 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 4 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=2
export RUN_NAME=translate_text2img256_stable_diffusion_stochastic_3
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1423 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 4 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=3
export RUN_NAME=translate_text2img256_stable_diffusion_stochastic_4
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1422 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 4 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=4
export RUN_NAME=translate_text2img256_stable_diffusion_stochastic_5
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1429 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 4 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=5
export RUN_NAME=translate_text2img256_stable_diffusion_stochastic_6
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1428 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 4 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=6
export RUN_NAME=translate_text2img256_stable_diffusion_stochastic_7
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1427 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 4 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=7
export RUN_NAME=translate_text2img256_stable_diffusion_stochastic_8
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1426 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 4 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

2. Zero-shot image-to-image translation with the LDM text-to-image checkpoint. We divided the 128 test samples into 8 groups, so the averaged metrics are reported.

export CUDA_VISIBLE_DEVICES=0
export RUN_NAME=translate_text2img256_latentdiff_stochastic_1
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1465 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 16 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=1
export RUN_NAME=translate_text2img256_latentdiff_stochastic_2
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1485 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 16 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=2
export RUN_NAME=translate_text2img256_latentdiff_stochastic_3
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1486 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 16 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=3
export RUN_NAME=translate_text2img256_latentdiff_stochastic_4
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1487 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 16 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=4
export RUN_NAME=translate_text2img256_latentdiff_stochastic_5
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1488 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 16 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=5
export RUN_NAME=translate_text2img256_latentdiff_stochastic_6
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1489 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 16 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=6
export RUN_NAME=translate_text2img256_latentdiff_stochastic_7
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1411 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 16 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

export CUDA_VISIBLE_DEVICES=7
export RUN_NAME=translate_text2img256_latentdiff_stochastic_8
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1412 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 16 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

Customized use for zero-shot image-to-image translation

1. Add your own image path and source-target text pairs at the end of this json file. You can add as many as you want.
2. Suggested hyperparameter tuning in this config file

• decoder_unconditional_guidance_scales: larger value means more weight on the target text
• skip_steps: larger value means more similar to the original image
• random seed: different random seed will generate different results

3. Note that each combination of decoder_unconditional_guidance_scales $\times$ skip_steps will be enumerated, and the best one is returned.
4. Run the following command to generate the image. Outputs will be saved in the output folder.

export CUDA_VISIBLE_DEVICES=0
export RUN_NAME=translate_text2img256_stable_diffusion_stochastic_custom
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1426 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 4 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

Unpaired image-to-image translation with diffusion models trained on two domains

1. AFHQ-Cat to AFHQ-Dog with DDIM $\eta=0.1$

export CUDA_VISIBLE_DEVICES=1
export RUN_NAME=translate_afhqcat256_to_afhqdog256_ddim_eta01
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1446 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

2. AFHQ-Wild to AFHQ-Dog with DDIM $\eta=0.1$

export CUDA_VISIBLE_DEVICES=5
export RUN_NAME=translate_afhqwild256_to_afhqdog256_ddim_eta01
export SEED=42
nohup python -m torch.distributed.launch --nproc_per_node 1 --master_port 1498 main.py --seed $SEED --cfg experiments/$RUN_NAME.cfg --run_name $RUN_NAME$SEED --logging_strategy steps --logging_first_step true --logging_steps 4 --evaluation_strategy steps --eval_steps 50 --metric_for_best_model CLIPEnergy --greater_is_better false --save_strategy steps --save_steps 50 --save_total_limit 1 --load_best_model_at_end --gradient_accumulation_steps 4 --num_train_epochs 0 --adafactor false --learning_rate 1e-3 --do_eval --output_dir output/$RUN_NAME$SEED --overwrite_output_dir --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --eval_accumulation_steps 4 --ddp_find_unused_parameters true --verbose true > $RUN_NAME$SEED.log 2>&1 &

Citation

If you find this repository helpful, please cite as

@inproceedings{unifydiffusion2022,
  title={Unifying Diffusion Models' Latent Space, with Applications to {CycleDiffusion} and Guidance},
  author={Chen Henry Wu and Fernando De la Torre},
  booktitle={ArXiv},
  year={2022},
}

Potential Societal Impact

Pre-trained on large-scale text-image pair, text-to-image diffusion models may learn unintended biases, which can be reflected or even amplified by CycleDiffusion. We encourage researchers and practitioners to consider this risk when building and deploying systems upon our work.

License

We use the X11 License. This license is identical to the MIT License, but with an extra sentence that prohibits using the copyright holders' names (Carnegie Mellon University in our case) for advertising or promotional purposes without written permission.

Contact

Issues are welcome if you have any question about the code. If you would like to discuss the method, please contact Chen Henry Wu.