# Recorded training settings; machine-specific paths replaced with portable examples.
# Validation prompts are preserved verbatim as training provenance.
# These are training settings, not the recommended distilled inference defaults.
acceleration:
  load_text_encoder_in_8bit: false
  mixed_precision_mode: bf16
  offload_optimizer_during_validation: true
  quantization: null
checkpoints:
  interval: 250
  keep_last_n: -1
  no_resume: false
  precision: bfloat16
  save_training_state: minimal
data:
  num_dataloader_workers: 0
  preprocessed_data_root: dataset/.precomputed
flow_matching:
  timestep_sampling_mode: shifted_logit_normal
  timestep_sampling_params: {}
hub:
  hub_model_id: null
  push_to_hub: false
lora:
  alpha: 64
  dropout: 0.0
  rank: 64
  target_modules:
  - attn1.to_k
  - attn1.to_q
  - attn1.to_v
  - attn1.to_out.0
  - attn2.to_k
  - attn2.to_q
  - attn2.to_v
  - attn2.to_out.0
model:
  audio_vae_path: models/vae/ltx-2.5-audio-vae-bf16.safetensors
  load_checkpoint: null
  model_path: models/diffusion_models/ltx-2.5-22b-dev-transformer-bf16.safetensors
  text_encoder_path: models/text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors
  training_mode: lora
  video_vae_path: models/vae/ltx-2.5-video-vae-bf16.safetensors
optimization:
  batch_size: 1
  enable_gradient_checkpointing: true
  gradient_accumulation_steps: 4
  learning_rate: 0.0001
  max_grad_norm: 1.0
  optimizer_type: adamw
  scheduler_params: {}
  scheduler_type: constant
  steps: 4000
output_dir: outputs/xpldv_lora_v5
seed: 42
training_strategy:
  audio: null
  name: flexible
  video:
    conditions:
    - probability: 0.75
      type: first_frame
    is_generated: true
    latents_dir: latents
validation:
  audio_cfg_scale: 4.0
  audio_modality_guidance_scale: 3.0
  audio_stg_scale: 1.0
  frame_rate: 24.0
  generate_audio: false
  generate_video: true
  guidance_rescale: 0.7
  images: null
  include_reference_in_output: false
  inference_steps: 30
  interval: 500
  negative_prompt: worst quality, inconsistent motion, blurry, jittery, distorted
  prompts: []
  reference_downscale_factor: 1
  reference_videos: null
  samples:
  - conditions:
    - image_or_video: validation_images/xpld_house.png
      type: first_frame
    prompt: A two-storey craftsman house separates into an ordered exploded view.
      The roof sections and the chimney move outward first and travel furthest, followed
      by the gables and the exterior walls, then the porch columns, the garage doors
      and the floor decks, all moving outward along controlled paths together. The
      parts come to rest suspended in an organised arrangement. The camera is locked
      off, facing the house head-on. Warm exterior lighting at dusk.
    seed: null
    video_dims: null
  - conditions:
    - image_or_video: validation_images/xpld_watch.png
      type: first_frame
    prompt: A stainless steel automatic wristwatch separates into an ordered exploded
      view. The bracelet links and the case move outward first and travel furthest,
      followed by the bezel and the dial, then the hands and the movement gears, all
      moving outward along controlled paths together. The parts come to rest suspended
      in an organised radial arrangement. The camera is locked off in a three-quarter
      view. A harder key light from one side, with deep fill shadow.
    seed: null
    video_dims: null
  - conditions:
    - image_or_video: validation_images/xpld_suv_2.png
      type: first_frame
    prompt: A black SUV separates into an ordered exploded view. The doors and the
      body panels move outward first and travel furthest, followed by the hood, the
      wheels and the glass, then the engine and the chassis, all moving outward along
      controlled radial paths together. The parts come to rest suspended in an organised
      radial arrangement. The camera arcs slowly to the right around the SUV. Night
      lighting from overhead street lamps.
    seed: null
    video_dims: null
  - conditions:
    - image_or_video: validation_images/xpld_laptop.png
      type: first_frame
    prompt: A silver laptop separates into an ordered exploded view. The display panel
      and the lower case move outward first and travel furthest, followed by the keyboard
      and the trackpad, then the battery, the logic board and the cooling fans, all
      moving apart horizontally along controlled paths together. The parts come to
      rest suspended in an ordered line. The camera arcs slowly to the left around
      the laptop. Soft, even daylight from a window.
    seed: null
    video_dims: null
  seed: 42
  skip_initial_validation: true
  stg_blocks:
  - 28
  video_cfg_scale: 4.0
  video_dims:
  - 960
  - 544
  - 121
  video_modality_guidance_scale: 3.0
  video_stg_scale: 1.0
wandb:
  enabled: false
  entity: null
  log_validation_videos: true
  project: ltxv-trainer
  tags: []