* Install wsl under windows 11 and ubuntu

* Update ubuntu

sudo apt update
sudo apt upgrade

* Close and reopen terminal window

* Make a place for downloads
mkdir Downloads
cd Downloads

* Download & Install Anaconda

wget https://repo.anaconda.com/archive/Anaconda3-2022.05-Linux-x86_64.sh
chmod +x ./Anaconda3-2022.05-Linux-x86_64.sh
./Anaconda3-2022.05-Linux-x86_64.sh

* Accept license and defaults until

Do you wish the installer to initialize Anaconda3
by running conda init? [yes|no]
[no] >>> yes

* close and reopen terminal window


* Nvidia CUDA toolkit for Ubuntu WSL - https://developer.nvidia.com/cuda-downloads
wget https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-wsl-ubuntu.pin
sudo mv cuda-wsl-ubuntu.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/11.3.1/local_installers/cuda-repo-wsl-ubuntu-11-3-local_11.3.1-1_amd64.deb
sudo dpkg -i cuda-repo-wsl-ubuntu-11-3-local_11.3.1-1_amd64.deb
sudo apt-key add /var/cuda-repo-wsl-ubuntu-11-3-local/7fa2af80.pub
sudo apt-get update
sudo apt-get -y install cuda


* Install g++
sudo apt install build-essential

* New virtual Python environment
conda update -n base -c defaults conda
conda create --name diffusers python=3.9
conda activate diffusers

* Make a directory for all your github downloads, then download diffusers
mkdir ~/github
cd ~/github
git clone https://github.com/Ttl/diffusers.git
cd diffusers
git checkout dreambooth_deepspeed
git pull

* Install required packages
pip install .
cd examples/dreambooth
conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
pip install -r requirements.txt
pip install -U --pre triton
pip install ninja bitsandbytes
pip install git+https://github.com/facebookresearch/xformers@1d31a3a#egg=xformers
pip install deepspeed
pip install diffusers


* Configure / login

accelerate config

* Use following options

In which compute environment are you running? ([0] This machine, [1] AWS (Amazon SageMaker)): 0
Which type of machine are you using? ([0] No distributed training, [1] multi-CPU, [2] multi-GPU, [3] TPU [4] MPS): 0
Do you want to run your training on CPU only (even if a GPU is available)? [yes/NO]:
Do you want to use DeepSpeed? [yes/NO]: yes
Do you want to specify a json file to a DeepSpeed config? [yes/NO]:
What should be your DeepSpeed's ZeRO optimization stage (0, 1, 2, 3)? [2]:
Where to offload optimizer states? [none/cpu/nvme]: cpu
Where to offload parameters? [none/cpu/nvme]: cpu
How many gradient accumulation steps you're passing in your script? [1]:
Do you want to use gradient clipping? [yes/NO]:
Do you want to enable `deepspeed.zero.Init` when using ZeRO Stage-3 for constructing massive models? [yes/NO]:
How many GPU(s) should be used for distributed training? [1]:
Do you wish to use FP16 or BF16 (mixed precision)? [NO/fp16/bf16]: fp16

huggingface-cli login

* And paste token

mkdir -p training
mkdir -p classes
mkdir -p model
export LD_LIBRARY_PATH=/usr/lib/wsl/lib:$LD_LIBRARY_PATH


explorer.exe .

* Copy your class images(person?) to the classes folder. Generate with whatever SD gui you're using
  Generate the amount you'll use in --num_class_images (200?) or you'll run out of vram when the script tries to generate them
* Copy your training images(you?) to the training folder

* Copy your training file from your Windows desktop (or wherever you saved it):
cp /mnt/c/Users/YOUR_USERNAME/Desktop/my_training.sh ./
chmod +x ./my_training.sh
./my_training.sh

*Example of contents of my_training.sh

export MODEL_NAME="CompVis/stable-diffusion-v1-4"
export INSTANCE_DIR="training"
export CLASS_DIR="classes"
export OUTPUT_DIR="model"

accelerate launch train_dreambooth.py \
  --pretrained_model_name_or_path=$MODEL_NAME \
  --instance_data_dir=$INSTANCE_DIR \
  --class_data_dir=$CLASS_DIR \
  --output_dir=$OUTPUT_DIR \
  --with_prior_preservation --prior_loss_weight=1.0 \
  --instance_prompt="a photo of sks person" \
  --class_prompt="a photo of person" \
  --resolution=512 \
  --train_batch_size=1 \
  --gradient_accumulation_steps=1 --gradient_checkpointing \
  --learning_rate=5e-6 \
  --lr_scheduler="constant" \
  --lr_warmup_steps=0 \
  --num_class_images=200 \
  --max_train_steps=800 \
  --mixed_precision=fp16


* Convert to .ckpt

wget https://gist.githubusercontent.com/jachiam/8a5c0b607e38fcc585168b90c686eb05/raw/cd2f93b1a0487006b0eb4f430b00e89aa84228ae/convert_diffusers_to_sd.py

python convert_diffusers_to_sd.py --model_path model --checkpoint_path model.ckpt


* NB: If you get an error like this -
403 Client Error: Forbidden for url: https://huggingface.co/CompVis/stable-diffusion-v1-4/resolve/main/model_index.json

Then:
Make sure you've visited https://huggingface.co/CompVis/stable-diffusion-v1-4 and accepted the licence!
NB. If this is the very first time you've used diffusers, it's about a 6GB download

To train again next time, simply:

* Start Ubuntu, then:
conda activate diffusers
cd ~/github/diffusers/examples/dreambooth
./my_training.sh