Compute embeddings for given task specifications and save them to file. Args: task_spec_list: List of task specifications.

Construct a task-specific prompt. Args: reward_to: Reward given to the agent upon successful completion of the task.

Compute the embeddings of tasks. Args: prompts: List of tasks defined as prompts. Returns: A list of embedd

Extract source code and dependent functions of the evaluation function. Args: eval_fn: Function for task evaluation.

Called before observations are returned from the environment Use this to define custom featurizers. Changing the space itself requires you to

Called before observations are returned from the environment Use this to define custom featurizers. Changing the space itself requires you to

Update stats + info and save replays.

Called before actions are passed from the model to the environment

Process the event log and extract performed actions and achievements.

Called on reward, terminated, truncated, and info before they are returned from the environment Use this to define custom reward shaping.

The progress, the max of which is 1, should * increase small for each eating * increase big for the 1st and 3rd eating * reach 1 with 10 e