ai-toolkit/config/examples/train_lora_flux_kontext_24gb.yaml

---
job: extension
config:
  # this name will be the folder and filename name
  name: "my_first_flux_kontext_lora_v1"
  process:
    - type: 'sd_trainer'
      # root folder to save training sessions/samples/weights
      training_folder: "output"
      # uncomment to see performance stats in the terminal every N steps
#      performance_log_every: 1000
      device: cuda:0
      # if a trigger word is specified, it will be added to captions of training data if it does not already exist
      # alternatively, in your captions you can add [trigger] and it will be replaced with the trigger word
#      trigger_word: "p3r5on"
      network:
        type: "lora"
        linear: 16
        linear_alpha: 16
      save:
        dtype: float16 # precision to save
        save_every: 250 # save every this many steps
        max_step_saves_to_keep: 4 # how many intermittent saves to keep
        push_to_hub: false #change this to True to push your trained model to Hugging Face.
        # You can either set up a HF_TOKEN env variable or you'll be prompted to log-in         
#       hf_repo_id: your-username/your-model-slug
#       hf_private: true #whether the repo is private or public
      datasets:
        # datasets are a folder of images. captions need to be txt files with the same name as the image
        # for instance image2.jpg and image2.txt. Only jpg, jpeg, and png are supported currently
        # images will automatically be resized and bucketed into the resolution specified
        # on windows, escape back slashes with another backslash so
        # "C:\\path\\to\\images\\folder"
        - folder_path: "/path/to/images/folder"
          # control path is the input images for kontext for a paired dataset. These are the source images you want to change.  
          # You can comment this out and only use normal images if you don't have a paired dataset. 
          # Control images need to match the filenames on the folder path but in
          # a different folder. These do not need captions.
          control_path: "/path/to/control/folder"
          caption_ext: "txt"
          caption_dropout_rate: 0.05  # will drop out the caption 5% of time
          shuffle_tokens: false  # shuffle caption order, split by commas
          cache_latents_to_disk: true  # leave this true unless you know what you're doing
          # Kontext runs images in at 2x the latent size. It may OOM at 1024 resolution with 24GB vram.
          resolution: [ 512, 768 ]  # flux enjoys multiple resolutions
          # resolution: [ 512, 768, 1024 ]
      train:
        batch_size: 1
        steps: 3000  # total number of steps to train 500 - 4000 is a good range
        gradient_accumulation_steps: 1
        train_unet: true
        train_text_encoder: false  # probably won't work with flux
        gradient_checkpointing: true  # need the on unless you have a ton of vram
        noise_scheduler: "flowmatch" # for training only
        optimizer: "adamw8bit"
        lr: 1e-4
        timestep_type: "weighted" # sigmoid, linear, or weighted.
        # uncomment this to skip the pre training sample
#        skip_first_sample: true
        # uncomment to completely disable sampling
#        disable_sampling: true

        # ema will smooth out learning, but could slow it down.

        # ema_config:
        #   use_ema: true
        #   ema_decay: 0.99

        # will probably need this if gpu supports it for flux, other dtypes may not work correctly
        dtype: bf16
      model:
        # huggingface model name or path. This model is gated.
        # visit https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev to accept the terms and conditions
        # and then you can use this model.
        name_or_path: "black-forest-labs/FLUX.1-Kontext-dev"
        arch: "flux_kontext"
        quantize: true  # run 8bit mixed precision
#        low_vram: true  # uncomment this if the GPU is connected to your monitors. It will use less vram to quantize, but is slower.
      sample:
        sampler: "flowmatch" # must match train.noise_scheduler
        sample_every: 250 # sample every this many steps
        width: 1024
        height: 1024
        prompts:
          # you can add [trigger] to the prompts here and it will be replaced with the trigger word
          # the --ctrl_img path is the one loaded to apply the kontext editing to
#          - "[trigger] holding a sign that says 'I LOVE PROMPTS!'"\
          - "make the person smile  --ctrl_img /path/to/control/folder/person1.jpg"
          - "give the person an afro  --ctrl_img /path/to/control/folder/person1.jpg"
          - "turn this image into a cartoon  --ctrl_img /path/to/control/folder/person1.jpg"
          - "put this person in an action film  --ctrl_img /path/to/control/folder/person1.jpg"
          - "make this person a rapper in a rap music video  --ctrl_img /path/to/control/folder/person1.jpg"
          - "make the person smile  --ctrl_img /path/to/control/folder/person1.jpg"
          - "give the person an afro  --ctrl_img /path/to/control/folder/person1.jpg"
          - "turn this image into a cartoon  --ctrl_img /path/to/control/folder/person1.jpg"
          - "put this person in an action film  --ctrl_img /path/to/control/folder/person1.jpg"
          - "make this person a rapper in a rap music video  --ctrl_img /path/to/control/folder/person1.jpg"
        neg: ""  # not used on flux
        seed: 42
        walk_seed: true
        guidance_scale: 4
        sample_steps: 20
# you can add any additional meta info here. [name] is replaced with config name at top
meta:
  name: "[name]"
  version: '1.0'