actions:
  target_fps: 35
  valid_keys:
  - forward
  - backward
  - strafe_right
  - strafe_left
  - weapon1
  - weapon2
  - weapon3
  - weapon4
  - weapon5
  - weapon6
  - weapon7
  - attack
  - speed
wandb:
  entity: null
  project: mira-codec
  name: doom-codec-260721-1912
  group: null
  mode: online
  remove_old_runs: false
run:
  seed: 28
  steps: 20000
  batch_size: 3
  checkpoint_every: 2000
  checkpoint_keep_recent: 3
  checkpoint_keep_permanent_every: 25000
  log_every: 50
  output_dir: /runs/doom_codec
  compile: true
  viz_every: 1000
  continue_from: null
  finetune_from: null
  latents_ema_decay: 0.99
dataloader:
  num_workers: 12
  shuffle_buffer_size: 100
validation:
  batch_size: null
  val_first: true
  val_n_samples: 512
  val_every: 2%
optim:
  optimizer:
    lr: 0.0001
    betas:
    - 0.9
    - 0.95
    weight_decay: 0.1
  scheduler:
    warmup_steps: 1000
    constant_steps: 0
    decay_steps: 19000
    min_lr: 1.0e-06
  model_ema_decay: 0.9999
model:
  loss:
    _target_: mira.codec.CodecLoss
    _convert_: all
    weights:
      _target_: mira.codec.CodecLossWeights
      loss_mae: 1.0
      loss_lpips_perceptual: 1.0
      lpips_perceptual_frame_frac: 0.25
      loss_dino_latent_consistency: 1.0
      dino_latent_consistency_frame_frac: 0.25
      compile_dino: true
      auto_weight: true
  architecture:
    _target_: mira.codec.VideoCodec
    _convert_: all
    config:
      _target_: mira.codec.VideoCodecConfig
      encoder:
        latent_dim: 32
        rae_model: dinov3_vitl16
        aggregation_layers:
        - 11
        - 13
        - 15
        - 17
        - 19
        - 21
        - 23
        bottleneck:
          stride: 2
          temporal_stride: 2
          noise_tau: 0.0
        compile_dino: true
        video:
          height: 384
          width: 512
          channels: 3
          timesteps: 40
          fps: 35
      decoder:
        latent_dim: 32
        activation_checkpointing: true
        bottleneck:
          stride: 2
        vit_width: 1152
        vit_depth: 28
        vit_num_heads: 16
        mlp_dim_multiplier: 4
        qk_norm: layernorm
        patch_size: 16
        patch_size_t: 2
        video:
          height: 384
          width: 512
          channels: 3
          timesteps: 40
          fps: 35
dataset:
  actions:
    target_fps: 35
    valid_keys:
    - forward
    - backward
    - strafe_right
    - strafe_left
    - weapon1
    - weapon2
    - weapon3
    - weapon4
    - weapon5
    - weapon6
    - weapon7
    - attack
    - speed
  train_index: /data/doom_mira/train
  test_index: /data/doom_mira/test
  n_players: 1
  target_fps: 35
  frame_size:
  - 384
  - 512
  exclude_replays: false