job: extension
config:
  name: hash_cache_v1
  process:
  - type: diffusion_trainer
    training_folder: /app/ai-toolkit/output
    sqlite_db_path: /app/ai-toolkit/aitk_db.db
    device: cuda
    trigger_word: null
    performance_log_every: 10
    network:
      type: lora
      linear: 32
      linear_alpha: 32
      conv: 16
      conv_alpha: 16
      lokr_full_rank: true
      lokr_factor: -1
      network_kwargs:
        ignore_if_contains: []
    save:
      dtype: bf16
      save_every: 250
      max_step_saves_to_keep: 4
      save_format: diffusers
      push_to_hub: false
    datasets:
    - folder_path: /app/ai-toolkit/datasets/hash_cache
      mask_path: null
      mask_min_value: 0.1
      default_caption: ''
      caption_ext: txt
      caption_dropout_rate: 0.05
      cache_latents_to_disk: false
      is_reg: false
      network_weight: 1
      resolution:
      - 512
      - 768
      - 1024
      controls: []
      shrink_video_to_frames: true
      num_frames: 1
      do_i2v: true
      flip_x: false
      flip_y: false
    train:
      batch_size: 1
      bypass_guidance_embedding: false
      steps: 3000
      gradient_accumulation: 1
      train_unet: true
      train_text_encoder: false
      gradient_checkpointing: true
      noise_scheduler: flowmatch
      optimizer: adamw8bit
      timestep_type: weighted
      content_or_style: balanced
      optimizer_params:
        weight_decay: 0.0001
      unload_text_encoder: false
      cache_text_embeddings: false
      lr: 0.0001
      ema_config:
        use_ema: false
        ema_decay: 0.99
      skip_first_sample: false
      force_first_sample: false
      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
    model:
      name_or_path: Qwen/Qwen-Image
      quantize: true
      qtype: qfloat8
      quantize_te: true
      qtype_te: qfloat8
      arch: qwen_image
      low_vram: true
      model_kwargs: {}
      layer_offloading: false
      layer_offloading_text_encoder_percent: 1
      layer_offloading_transformer_percent: 1
    sample:
      sampler: flowmatch
      sample_every: 250
      width: 1024
      height: 1024
      samples:
      - prompt: "Two adult people sitting side-by-side on a couch in a simple room,\
          \ full bodies fully visible in frame. Neutral lighting, no stylization,\
          \ no dramatic poses. Both figures look forward calmly with relaxed posture,\
          \ hands resting naturally. The background is minimal and uncluttered, with\
          \ a plain wall, soft shadows and a low-table in front. No neon, no props\
          \ \u2014 just clean, natural photographic realism."
      - prompt: 'Two adults slightly slouched on a couch, clearly drunk but not cartoonish:
          relaxed posture, unfocused expressions, leaning into each other casually.
          Full bodies visible. The room is dim but neutral in tone, with soft warm
          lighting, subtle background blur and no dramatic color effects. A few bottles
          or cans on a table in front for context, but environment remains minimal
          and clean.'
      - prompt: "A clean black-and-white line-art drawing of two people sitting on\
          \ a couch, full bodies visible, simple outlines with minimal shading. Smooth,\
          \ consistent linework, no stylization like anime or comic \u2014 neutral\
          \ sketchbook quality. Background lightly suggested with simple lines for\
          \ the couch and wall. No color, no text, no shadows \u2014 pure structural\
          \ line drawing."
      - prompt: "Two human silhouettes sitting on a couch, shown as solid matte shapes\
          \ with no internal details. Full bodies visible, clean side-by-side posture.\
          \ The couch is a simple silhouette shape as well. Background is blank white.\
          \ No shading, no texture \u2014 pure flat graphical silhouette intended\
          \ only for pose and composition reference."
      neg: ''
      seed: 42
      walk_seed: true
      guidance_scale: 4
      sample_steps: 25
      num_frames: 1
      fps: 1
meta:
  name: hash_cache_v1
  version: '1.0'