robodojo.yaml

type: robodojo

# Formal layout only:
#   <dataset_dir>/<task>/<embodiment>/data/episode_*.hdf5
dataset_dir: /path/to/robodojo_data
embodiment: arx_x5               # real: arx_x5, piper, piper_x; sim: arx_x5
variant: sim                     # sim: env/world -> arm base; real: native pose

# ── Action space ─────────────────────────────────────────────────────────────
# EEF is the only supported mode; the raw representation is EEF20. Raw EEF20
# is normalized first, then scattered into the shared 80-D layout:
#   left [xyz3, rot6d6, grip1] -> 0:10; right -> 34:44.
action_mode: eef
unify_action: true
unify_action_map: ["0-9", "34-43"]
unify_state_map: null  # null = state reuses unify_action_map

# ── Frame sampling ───────────────────────────────────────────────────────────
# num_frames = state/action window length; action horizon = num_frames - 1.
# video_stride subsamples video only; state/action stay at the raw rate.
# window_stride = sliding window step (1 = every frame starts a window).
num_frames: 33
video_stride: 4
window_stride: 1
split: train

# ── Rendering (3-slot L-shape canvas) ────────────────────────────────────────
# camera_layout = [head, left/bottom-left, right/bottom-right]; null slots are
# rendered as black tiles. At 384x320: top 256Hx320W, bottom halves 128Hx160W,
# each tile BILINEAR-stretched to its slot. target_camera drives single-view
# mode and is ignored when multiview=true.
height: 384
width: 320
multiview: true
camera_layout: [cam_head, cam_left_wrist, cam_right_wrist]
target_camera: cam_head

# ── Normalization ────────────────────────────────────────────────────────────
# Pooled stats are discovered/auto-built under <dataset_dir>/meta/ for this
# variant and embodiment. An explicit normalization_stats_path remains an API
# override.
normalize_mode: min-max          # min-max, z-score, quantile, or null

# ── Video augmentation (optional) ────────────────────────────────────────────
# Load-time color jitter applied by the reader to each clip with the SAME
# random factors across all frames (temporal consistency), TRAIN split only.
# enabled: false keeps video byte-identical while preserving the knobs.
color_jitter:
  enabled: true
  brightness: 0.2
  contrast: 0.2
  saturation: 0.2
  hue: 0.0