from typing import Dict
from ml_collections import FrozenConfigDict

# The tasks we in MetaWorld.
METAWORLDTASKS = frozenset([
    "assembly", 
    "basketball",
    "button-press",
    "button-press-topdown",
    "door-close", 
    "door-open", 
    "drawer-open", 
    "faucet-close", 
    "faucet-open", 
    "hammer", 
    "handle-press", 
    "push", 
    "shelf-place",
])

# The tasks in robosuite
ROBOSUITETASKS = frozenset([
    "NutAssemblySquare",
    "Stack",
    "Lift",
])

# A mapping from MetaWorld task to Gym environment name.
METAWORLD_TASK_TO_ENV_NAME: Dict[str, str] = {
    k: f"{k}-v2-goal-observable"
    for k in METAWORLDTASKS
}

# All available encoders.
ALGORITHMS = frozenset([
    "xirl",
    "tcn",
    "lifs",
    "goal_classifier",
])

# A mapping from x-MAGICAL embodiment to RL training iterations.
MetaWorldTrainingIterations = FrozenConfigDict({
    "assembly": 2_000_000, 
    "basketball": 3_000_000,
    "button-press": 10_000_000,
    "button-press-topdown": 10_000_000,
    "door-close": 2_400_000,
    "door-open": 1_000_000,
    "drawer-open": 1_000_000,
    "faucet-close": 10_000_000,
    "faucet-open": 1_000_000,
    "hammer": 1_500_000,
    "handle-press": 10_000_000,
    "push": 10_000_000,
    "shelf-place": 10_000_000,
})

RobosuiteTrainingIterations = FrozenConfigDict({
    "NutAssemblySquare": 2_000_000,
    "Stack": 2_000_000,
    "Lift": 2_000_000,
})