scores: scores that the model assigned to each action at each step counts: count of available actions at each step selections: actions th

Returns: a dict of representations at three different levels: node, dag, and global.

returns a tensor of shape [num_nodes, embed_dim]

returns a tensor of shape [num_dags, embed_dim]

returns a tensor of shape [num_observations, embed_dim]

returns a tensor of shape [num_nodes,]

returns a sparse COO adjacency matrix

returns a batch of edge masks of shape (msg passing depth, num edges), where the i'th mask indicates which edges participate in the i'th root-to-l

converts an env observation into a PyG `Batch` object

samples a new time limit prior to resetting

modifies `truncated` signal when time limit is reached

lazy-init the window