Stable diffusion has revolutionalized text2image models by producing high quality images based on a prompt. Dreambooth is a approach for personalization of text-to-image diffusion models. With images as input subject, we can fine-tune a pretrained text-to-image model
Although the used to finetune the pre-trained model since both the Imagen model and Dreambooth code are closed source, several opensource projects have emerged using stable diffusion.
Dreambooth makes stable-diffusion even more powered with the ability to generate realistic looking pictures of humans, animals or any other object by just training them on 20-30 images.
In this example tutorial, we will be fine-tuning a pretrained stable diffusion using images of a human and generating images of him drinking coffee.
TL;DR
The following command generates the following:
Subject: SBF
Prompt: a photo of SBF without hair
bacalhau docker run \
--gpu 1 \
--timeout 3600 \
--wait-timeout-secs 3600 \
-i ipfs://QmRKnvqvpFzLjEoeeNNGHtc7H8fCn9TvNWHFnbBHkK8Mhy \
jsacex/dreambooth:full \
-- bash finetune.sh /inputs /outputs "a photo of sbf man" "a photo of man" 3000 "/man" "/model"
bacalhau docker run \
--gpu 1 \
-i ipfs://QmUEJPr5pfV6tRzWQuNSSb3wdcN6tRQS5tdk3dYSCJ55Xs:/SBF.ckpt \
jsacex/stable-diffusion-ckpt \
-- conda run --no-capture-output -n ldm python scripts/txt2img.py --prompt "a photo of sbf without hair" --plms --ckpt ../SBF.ckpt --skip_grid --n_samples 1 --skip_grid --outdir ../outputs
Output:
Building this container requires you to have a supported GPU which needs to have 16gb+ of memory, since it can be resource intensive.
We will create a Dockerfile and add the desired configuration to the file. Following commands specify how the image will be built, and what extra requirements will be included:
FROM pytorch/pytorch:1.12.1-cuda11.3-cudnn8-devel
WORKDIR /
# Install requirements
# RUN git clone https://github.com/TheLastBen/diffusers
RUN apt update && apt install wget git unzip -y
RUN wget -q https://gist.githubusercontent.com/js-ts/28684a7e6217214ec944a9224584e9af/raw/d7492bc8f36700b75d51e3346259d1a466b99a40/train_dreambooth.py
RUN wget -q https://github.com/TheLastBen/diffusers/raw/main/scripts/convert_diffusers_to_original_stable_diffusion.py
# RUN cp /content/convert_diffusers_to_original_stable_diffusion.py /content/diffusers/scripts/convert_diffusers_to_original_stable_diffusion.py
RUN pip install -qq git+https://github.com/TheLastBen/diffusers
RUN pip install -q accelerate==0.12.0 transformers ftfy bitsandbytes gradio natsort
# Install xformers
RUN pip install -q https://github.com/metrolobo/xformers_wheels/releases/download/1d31a3ac_various_6/xformers-0.0.14.dev0-cp37-cp37m-linux_x86_64.whl
RUN wget 'https://github.com/TheLastBen/fast-stable-diffusion/raw/main/Dreambooth/Regularization/Women' -O woman.zip
RUN wget 'https://github.com/TheLastBen/fast-stable-diffusion/raw/main/Dreambooth/Regularization/Men' -O man.zip
RUN wget 'https://github.com/TheLastBen/fast-stable-diffusion/raw/main/Dreambooth/Regularization/Mix' -O mix.zip
RUN unzip -j woman.zip -d woman
RUN unzip -j man.zip -d man
RUN unzip -j mix.zip -d mix
This container is using the pytorch/pytorch:1.12.1-cuda11.3-cudnn8-devel image and the working directory is set. Next, we add our custom code and pull the dependent repositories.
The shell script is there to make things much simpler since the command to train the model needs many parameters to pass and later convert the model weights to the checkpoint, you can edit this script and add in your own parameters
To download the models and run a test job in the Docker file, copy the following:
FROM pytorch/pytorch:1.12.1-cuda11.3-cudnn8-devel
WORKDIR /
# Install requirements
# RUN git clone https://github.com/TheLastBen/diffusers
RUN apt update && apt install wget git unzip -y
RUN wget -q https://gist.githubusercontent.com/js-ts/28684a7e6217214ec944a9224584e9af/raw/d7492bc8f36700b75d51e3346259d1a466b99a40/train_dreambooth.py
RUN wget -q https://github.com/TheLastBen/diffusers/raw/main/scripts/convert_diffusers_to_original_stable_diffusion.py
# RUN cp /content/convert_diffusers_to_original_stable_diffusion.py /content/diffusers/scripts/convert_diffusers_to_original_stable_diffusion.py
RUN pip install -qq git+https://github.com/TheLastBen/diffusers
RUN pip install -q accelerate==0.12.0 transformers ftfy bitsandbytes gradio natsort
# Install xformers
RUN pip install -q https://github.com/metrolobo/xformers_wheels/releases/download/1d31a3ac_various_6/xformers-0.0.14.dev0-cp37-cp37m-linux_x86_64.whl
# You need to accept the model license before downloading or using the Stable Diffusion weights. Please, visit the [model card](https://huggingface.co/runwayml/stable-diffusion-v1-5), read the license and tick the checkbox if you agree. You have to be a registered user in 🤗 Hugging Face Hub, and you'll also need to use an access token for the code to work.
# https://huggingface.co/settings/tokens
RUN mkdir -p ~/.huggingface
RUN echo -n "<your-hugging-face-token>" > ~/.huggingface/token
# copy the test dataset from a local file
# COPY jfk /jfk
# Download and extract the test dataset
RUN wget https://github.com/js-ts/test-images/raw/main/jfk.zip
RUN unzip -j jfk.zip -d jfk
RUN mkdir model
RUN wget 'https://github.com/TheLastBen/fast-stable-diffusion/raw/main/Dreambooth/Regularization/Women' -O woman.zip
RUN wget 'https://github.com/TheLastBen/fast-stable-diffusion/raw/main/Dreambooth/Regularization/Men' -O man.zip
RUN wget 'https://github.com/TheLastBen/fast-stable-diffusion/raw/main/Dreambooth/Regularization/Mix' -O mix.zip
RUN unzip -j woman.zip -d woman
RUN unzip -j man.zip -d man
RUN unzip -j mix.zip -d mix
RUN accelerate launch train_dreambooth.py \
--image_captions_filename \
--train_text_encoder \
--save_starting_step=5\
--stop_text_encoder_training=31 \
--class_data_dir=/man \
--save_n_steps=5 \
--pretrained_model_name_or_path="CompVis/stable-diffusion-v1-4" \
--instance_data_dir="/jfk" \
--output_dir="/model" \
--instance_prompt="a photo of jfk man" \
--class_prompt="a photo of man" \
--instance_prompt="" \
--seed=96576 \
--resolution=512 \
--mixed_precision="fp16" \
--train_batch_size=1 \
--gradient_accumulation_steps=1 \
--use_8bit_adam \
--learning_rate=2e-6 \
--lr_scheduler="polynomial" \
--center_crop \
--lr_warmup_steps=0 \
--max_train_steps=30
COPY finetune.sh /finetune.sh
RUN wget -q https://gist.githubusercontent.com/js-ts/624fecc3fff807d4948688cb28993a94/raw/fd69ac084debe26a815485c1f363b8a45566f1ba/clear_mem.py
# Removing your token
RUN rm -rf ~/.huggingface/token
We will run docker build command to build the container:
docker build -t <hub-user>/<repo-name>:<tag> .
Before running the command replace:
repo-name with the name of the container, you can name it anything you want.
tag this is not required but you can use the latest tag
Now you can push this repository to the registry designated by its name or tag.
docker push <hub-user>/<repo-name>:<tag>
The optimal dataset size is between 20-30 images. You can choose the images of the subject in different positions, full body images, half body, pictures of the face etc.
Only the subject should appear in the image so you can crop the image to just fit the subject. Make sure that the images are 512x512 size and are named in the following pattern:
Subject Name.jpg, Subject Name (2).jpg ... Subject Name (n).jpg
After the Subject dataset is created we upload it to IPFS.
To upload your dataset using NFTup just drag and drop your directory it will upload it to IPFS:
After the checkpoint file has been uploaded, copy its CID which will look like this:
Since there are a lot of combinations that you can try, processing of finetuned model can take almost 1hr+ to complete. Here are a few approaches that you can try based on your requirements:
bacalhau docker run: call to bacalhau
The --gpu 1 flag is set to specify hardware requirements, a GPU is needed to run such a job
-i ipfs://bafybeidqbuphwkqwgrobv2vakwsh3l6b4q2mx7xspgh4l7lhulhc3dfa7a Mounts the data from IPFS via its CID
jsacex/dreambooth:latest Name and tag of the docker image we are using
-- bash finetune.sh /inputs /outputs "a photo of David Aronchick man" "a photo of man" 3000 "/man" execute script with following paramters:
/inputs Path to the subject Images
/outputs Path to save the generated outputs
"a photo of < name of the subject > < class >" -> "a photo of David Aronchick man" Subject name along with class
"a photo of < class >" -> "a photo of man" Name of the class
bacalhau docker run \
--gpu 1 \
--timeout 3600 \
--wait-timeout-secs 3600 \
--timeout 3600 \
--wait-timeout-secs 3600 \
-i <CID-OF-THE-SUBJECT> \
jsacex/dreambooth:full \
-- bash finetune.sh /inputs /outputs "a photo of <name-of-the-subject> man" "a photo of man" 3000 "/man" "/model"
The number of iterations is 3000. This number should be no of subject images x 100. So if there are 30 images, it would be 3000. It takes around 32 minutes on a v100 for 3000 iterations, but you can increase/decrease the number based on your requirements.
Here is our command with our parameters replaced:
bacalhau docker run \
--gpu 1 \
--timeout 3600 \
--wait-timeout-secs 3600 \
--timeout 3600 \
--wait-timeout-secs 3600 \
-i ipfs://bafybeidqbuphwkqwgrobv2vakwsh3l6b4q2mx7xspgh4l7lhulhc3dfa7a \
--wait \
--id-only \
jsacex/dreambooth:full \
-- bash finetune.sh /inputs /outputs "a photo of David Aronchick man" "a photo of man" 3000 "/man" "/model"
If your subject fits the above class, but has a different name you just need to replace the input CID and the subject name.
Use the /woman class images
bacalhau docker run \
--gpu 1 \
--timeout 3600 \
--wait-timeout-secs 3600 \
-i <CID-OF-THE-SUBJECT> \
jsacex/dreambooth:full \
-- bash finetune.sh /inputs /outputs "a photo of <name-of-the-subject> woman" "a photo of woman" 3000 "/woman" "/model"
Here you can provide your own regularization images or use the mix class.
Use the /mix class images if the class of the subject is mix
bacalhau docker run \
--gpu 1 \
--timeout 3600 \
--wait-timeout-secs 3600 \
-i <CID-OF-THE-SUBJECT> \
jsacex/dreambooth:full \
-- bash finetune.sh /inputs /outputs "a photo of <name-of-the-subject> mix" "a photo of mix" 3000 "/mix" "/model"
You can upload the model to IPFS and then create a gist, mount the model and script to the lightweight container
bacalhau docker run \
--gpu 1 \
--timeout 3600 \
--wait-timeout-secs 3600 \
-i ipfs://bafybeidqbuphwkqwgrobv2vakwsh3l6b4q2mx7xspgh4l7lhulhc3dfa7a:/aronchick \
-i ipfs://<CID-OF-THE-MODEL>:/model
-i https://gist.githubusercontent.com/js-ts/54b270a36aa3c35fdc270640680b3bd4/raw/7d8e8fa47bc3811ef63772f7fc7f4360aa9d51a8/finetune.sh
--wait \
--id-only \
jsacex/dreambooth:lite \
-- bash /inputs/finetune.sh /aronchick /outputs "a photo of aronchick man" "a photo of man" 3000 "/man" "/model"
When a job is submitted, Bacalhau prints out the related job_id. Use the export JOB_ID=$(bacalhau docker run ...) wrapper to store that in an environment variable so that we can reuse it later on.
name: Stable Diffusion Dreambooth Finetuning
type: batch
count: 1
tasks:
- name: My main task
Engine:
type: docker
params:
Image: "jsacex/dreambooth:full"
Parameters:
- bash finetune.sh /inputs /outputs "a photo of aronchick man" "a photo of man" 3000 "/man" "/model"
InputSources:
- Target: "/inputs/data"
Source:
Type: "ipfs"
Params:
CID: "QmRKnvqvpFzLjEoeeNNGHtc7H8fCn9TvNWHFnbBHkK8Mhy"
Resources:
GPU: "1"
You can check the status of the job using bacalhau job list.
bacalhau job list --id-filter ${JOB_ID}
When it says Completed, that means the job is done, and we can get the results.
You can find out more information about your job by using bacalhau job describe.
bacalhau job describe ${JOB_ID}
You can download your job results directly by using bacalhau job get. Alternatively, you can choose to create a directory to store your results. In the command below, we created a directory and downloaded our job output to be stored in that directory.
After the download has finished you should see the following contents in results directory
Now you can find the file in the results/outputs folder. You can view results by running following commands:
ls results # list the contents of the current directory
In the next steps, we will be doing inference on the finetuned model
Bacalhau currently doesn't support mounting subpaths of the CID, so instead of just mounting the model.ckpt file we need to mount the whole output CID which is 6.4GB, which might result in errors like FAILED TO COPY /inputs. So you have to manually copy the CID of the model.ckpt which is of 2GB.
To get the CID of the model.ckpt file go to https://gateway.ipfs.io/ipfs/< YOUR-OUTPUT-CID >/outputs/. For example:
When a job is sumbitted, Bacalhau prints out the related job_id. We store that in an environment variable so that we can reuse it later on.
We got an image like this as a result:
Inference
Prerequisites
To get started, you need to install the Bacalhau client, see more information
Setting up Docker Container
You can skip this section entirely and directly go to
Downloading the models
Build the Docker container
hub-user with your docker hub username, If you don’t have a docker hub account follow to create a Docker account, and use the username of the account you create.
Create the Subject Dataset
You can view the for reference.
Uploading the Subject Images to IPFS
In this case, we will be using (Recommended Option) to upload files and directories with .
Approaches to run a Bacalhau Job on a Finetuned Model
Case 1: If the subject is of class male
Structure of the command
Case 2 : If the subject is of class female
Case 3: If the subject is of class mix
Case 4: If you want a different tokenizer, model, and a different shell script with custom parameters
Declarative job description
The same job can be presented in the format. In this case, the description will look like this. Change the command in the Parameters section and CID to suit your goals.
Checking the State of your Jobs
Job status
Job information
Job download
Viewing your Job Output
Inference on the Fine-Tuned Model
Refer to our for more details of how to build a SD inference container
If you use the browser, you can use following:
Run the Bacalhau Job on the Fine-Tuned Model
To check the status of your job and download results refer back to the .