---
job: "extension"
config:
  name: "p0rtman_lora_v1_copy"
  process:
    - type: "diffusion_trainer"
      training_folder: "/app/ai-toolkit/output"
      sqlite_db_path: "./aitk_db.db"
      device: "cuda"
      trigger_word: "p0rtman"
      performance_log_every: 10
      network:
        type: "lokr"
        linear: 32
        linear_alpha: 32
        conv: 16
        conv_alpha: 16
        lokr_full_rank: true
        lokr_factor: 8
        network_kwargs:
          ignore_if_contains: []
      save:
        dtype: "bf16"
        save_every: 100
        max_step_saves_to_keep: 10
        save_format: "diffusers"
        push_to_hub: false
      datasets:
        - folder_path: "/app/ai-toolkit/datasets/np"
          mask_path: null
          mask_min_value: 0.1
          default_caption: ""
          caption_ext: "txt"
          caption_dropout_rate: 0.05
          cache_latents_to_disk: true
          is_reg: false
          network_weight: 1
          resolution:
            - 512
            - 768
            - 1024
          controls: []
          shrink_video_to_frames: true
          num_frames: 1
          flip_x: false
          flip_y: false
          num_repeats:
            - 16
            - 4
            - 1
          loss_split: "diffusion_depth"
      train:
        weight_noise:
          enabled: true
          mode: "relative"
          sigma: 0.0125
          log_every: 1
        max_grad_norm: 1
        batch_size: 4
        bypass_guidance_embedding: false
        steps: 1200
        gradient_accumulation: 1
        train_unet: true
        train_text_encoder: false
        gradient_checkpointing: true
        noise_scheduler: "flowmatch"
        optimizer: "adamw8bit"
        timestep_type: "linear"
        content_or_style: "balanced"
        optimizer_params:
          weight_decay: 0.0001
        unload_text_encoder: false
        cache_text_embeddings: true
        lr: 0.00005
        ema_config:
          use_ema: false
          ema_decay: 0.99
        skip_first_sample: false
        force_first_sample: true
        disable_sampling: false
        dtype: "bf16"
        diff_output_preservation: false
        diff_output_preservation_multiplier: 1
        diff_output_preservation_class: "person"
        switch_boundary_every: 1
        loss_type: "mse"
        diffusion_loss_weight: 1
        diffusion_loss_max_t: 1
        diffusion_loss_min_t: 0
        custom_timestep_distribution: null
        custom_timestep_curve: null
        max_denoising_steps: 999
        min_denoising_steps: 0
        loss_split: null
      logging:
        log_every: 1
        use_ui_logger: true
      model:
        name_or_path: "black-forest-labs/FLUX.2-klein-base-9B"
        quantize: true
        qtype: "qfloat8"
        quantize_te: true
        qtype_te: "qfloat8"
        arch: "flux2_klein_9b"
        low_vram: false
        model_kwargs:
          match_target_res: false
        layer_offloading: false
        layer_offloading_text_encoder_percent: 1
        layer_offloading_transformer_percent: 1
      sample:
        sampler: "flowmatch"
        sample_every: 100
        width: 640
        height: 960
        samples:
          - prompt: "p0rtman, full body, woman, standing, city street, afternoon light"
          - prompt: "p0rtman, medium shot, woman, casual outfit, outdoor natural light"
          - prompt: "p0rtman, close-up portrait of a woman, soft studio lighting, looking at camera"
        neg: ""
        seed: 42
        walk_seed: false
        guidance_scale: 4
        sample_steps: 25
        num_frames: 1
        fps: 1
      face_id:
        enabled: true
        init_scale: 0.3
        identity_loss_weight: 0.05
        landmark_loss_weight: 0
        body_proportion_loss_weight: 0
        body_proportion_loss_min_t: 0.8
        body_proportion_loss_max_t: 1
        body_shape_loss_weight: 0
        body_shape_loss_max_t: 1
        body_shape_loss_min_t: 0.8
        identity_loss_min_t: 0
        identity_loss_max_t: 0.9
        identity_metrics: true
        identity_loss_use_average: true
        identity_loss_min_cos: 0.4
      depth_consistency:
        loss_weight: 0.001
        input_size: 518
        preview_every: 100
        mask_source: "subject"
        loss_max_t: 1
        model_id: "depth-anything/Depth-Anything-V2-Large-hf"
        loss_min_t: 0
        grad_checkpoint: true
        ssi_weight: 0
        grad_weight: 1
        grad_scales: 6
      subject_mask:
        enabled: true
        background_loss_weight: 0
        clothing_loss_weight: 1
        body_loss_weight: 1
        save_debug_previews: true
        cache_resolution: 768
        segformer_res: 768
meta:
  name: "[name]"
  version: "1.0"