Function maps counts of ngram to vector of tfidf weights. The function returns vec, an array of dictionary that store mapping

Evaluates specified <observation, action> and returns log_probs, values, entropy This is invoked for each mini-batch in roll

Forward pass in all the networks (actor and critic) :param obs: Observation :param deterministic: Whether to sample or use d

Performs a forward pass on the policy and gets log_probs, entropy etc corresponding to specified observation, actions This i

Performs a forward pass on the value network and gets values corresponding to observations This is invoked during rollout generation

r""" Generates sequences of token ids for models with a language modeling head. The method supports the following generation methods

Lists all possible actions indices and its meaning Returns: ActionSpace -- an instance of action space

Extract out the highlighted part of a table.

Extract out the highlighted part of a table.

Extracts the prompt inputs and attention masks which is used as seed for generation

Performs a forward pass on the reference policy and gets log_probs corresponding to specified observation, actions This is i

Returns the observation dim

Returns the type of policy (causal or seq2seq)

Returns list of samples available in the environment Returns: List[Sample]: list of samples in the environment

Convert subtable to PARENT format. Do not include section text.

Convert table to format required by PARENT.

Computes the matrix-vector product with the Fisher information matrix. :param params: list of parameters used to compute the Hessian

Linearize full table with localized headers and return a string.

Linearize the highlighted subtable and return a string of its contents.