Takes a string as input and returns an object that can be given to either cook_refs or cook_test. This is optional: cook_refs and cook_test

Get the policy action and state from an observation (and optional state). Includes sugar-coating to handle different observations (e.

r""" Generates sequences of token ids for models with a language modeling head using **multinomial sampling** and can be used for text

For stable baselines (only return tensor items)

Update policy using the currently gathered rollout buffer (one gradient step over whole data).

Create the networks and the optimizer. :param lr_schedule: Learning rate schedule lr_schedule(1) is the initial learning

Create the policy and value networks. Part of the layers can be shared.

Builds policy and value models and sets self._policy_model and self._value_model

If `inputs` is None and `name` is in both forward function and keyword arguments, then inputs can be retrieved from name

Compute actor gradients for kl div and surrogate objectives. :param kl_div: The KL divergence objective :param policy_object

This class returns a [`LogitsProcessorList`] list object that contains all relevant [`LogitsProcessor`] instances used to modify the

:param callback: Callback(s) called at every step with state of the algorithm. :param eval_freq: How many steps between evaluations;

Computes the length of the LCS between two seqs. The implementation below uses a DP programming algorithm and runs in O(nm) time where n = le

Returns the length of the Longest Common Subsequence between two seqs. Source: http://www.algorithmist.com/index.php/Longest_Common_Subsequence

Yields all ngrams of given order in sequence.

Get the action according to the policy for a given observation. :param observation: :param deterministic: Whether to use sto

Computes max of Rouge-L (https://github.com/allenai/unifiedqa/blob/bad6ef339db6286f0d8bd0661a2daeeb0f800f59/evaluation/evaluate_narrativeqa.p

Initialize different variables needed for training. :param total_timesteps: The total number of samples (env steps) to train on

Returns list of references in both the single and multi-reference setting. Performs lowercasing and white-space tokenization on each line before

:param action_masks: Masks applied to constrain the choice of possible actions.

r""" Generates sequences of token ids for models with a language modeling head using **beam search multinomial sampling** and can be u

r""" Generates sequences of token ids for models with a language modeling head using **beam search decoding** and can be used for text

Compute raw logits for beam search

Compute term frequency for reference data. This will be used to compute idf (inverse document frequency later) The term frequ

Finds an approximate solution to a set of linear equations Ax = b Sources: - https://github.com/ajlangley/trpo-pytorch/blob/master/conj

r""" Generates sequences of token ids for models with a language modeling head using **constrained beam search decoding** and can be u

Takes a list of reference sentences for a single segment and returns an object that encapsulates everything that BLEU needs to know about them

Takes a test sentence and returns an object that encapsulates everything that BLEU needs to know about it. :param test: list of string : hypot

copy the refs.

MD5 hash of a dictionary.

Runs policy for ``n_eval_episodes`` episodes and returns average reward. If a vector env is passed in, this divides the episodes to evaluate