An enhanced list, used for intermidate representation of a prompt.

A LogicNode is a tree primitive. It is either a deduction or a leaf fact. Leaf facts are the ones that we use in story generation (if they a

Intermidate prompt template parser, specifically for language models. Args: meta_template (Dict): The meta template for the model.

Intermidate prompt template parser, specifically for API models. Args: meta_template (Dict): The meta template for the model.

In-context Learning Prompt Template Class This class represents a template that guides the generation of prompts in the retrieval or inference

Planning Evaluation Args: dataset_path(str): File path of evaluation dataset bert_score_model(str): the bert_score model for sente

Args: template(str): Format string with keyword-only arguments. For example '{who} like {what}' pred(Any): Parsed dat

分词器

Main datastructure used when creating a MuSR example. It's basically a standard tree with some parameters controlling the shape.

Model wrapper around OpenAI's models. Args: path (str): The name of OpenAI's model. max_seq_len (int): The maximum allowed sequen

Helper class to wrap a callable into a class with a `tokenize` method as used by rouge-score.

对齐错误句子和正确句子， 使用编辑距离算法抽取编辑操作

Widely used attention mechanism Attributes: qg_weights (nn.Parameter): weight matrices for queries and gates qg_bias (nn.Par

Runs basic tokenization (punctuation splitting, lower casing, etc.).

Generation Inferencer class to directly evaluate by generation. Attributes: model (:obj:`BaseModelWrapper`, optional): The module to infe

A IPython executor that can execute Python scripts in a jupyter manner. Args: description (str): The description of the action. Defaults

Instruct Following Evaluation Args: dataset_path(str): File path of evaluation dataset.

A dataset wrapper around the evaluator inputs, designed for OpenCompass's internal use.

A curses menu that allows the user to select one item from each list. Args: lists (list[list[str]]): A list of lists of strings, where ea

合并编辑操作，从Token-Level转换为Span-Level

Criteria to stop on the specified multi-token sequence.

OpenICL Inference Task. This task is used to run the inference process.

Planning Evaluation Args: dataset_path(str): File path of evaluation dataset name_weight(float): the weight of action_name in bert

Evaluator for retrosynthesis models. It calculates Top-K accuracy and Max-Fragment accuracy based on SMILES string comparisons.

Review Capability Evaluation Args: dataset_path(str): File path of evaluation dataset.

A token bucket for rate limiting. Args: query_per_second (float): The rate of the token bucket.

Zeroshot Retriever. The retriever returns empty list for all queries. Args: dataset (`BaseDataset`): Any BaseDataset instances.

The transformation object that holds translation and rotation

APITemplateParser accepts OpenAI format dialog inputs.

错误类型分类器

This object takes in command and executes it with time out.

In-conext Learning Dataset Reader Class Generate an DatasetReader instance through 'dataset'. Attributes: dataset (:obj:`Dataset` or

Filter the repeated message. Args: name (str): name of the filter.

One iteration of GeoFormer

This is the Invariant Point Attention from Jumper et al. (2021) that performs transformer-like operation on frames

LMTemplateParser accepts OpenAI format dialog inputs.

Math agent evaluator for soft condition. Args: action (str): Action for catching internal prediction. Defaults to `PythonInte

Criteria to stop on the specified multi-token sequence.

Configuration for latex normalization. Each field controls a group of related normalizations: - basic_latex: Basic latex command replacements

The Entire OmegaFold model that comprises a pretrained Protein Language Model, an encoder of the primary sequence, as well as a structure mod

One OmegaPLM Layer This layer baked the pre-layernorm configuration into the model Attributes: gau: the underlying GAU layer contain

OpenICL Inference Task. This task is used to run the inference process.

OpenICL Evaluation Task. This task is used to evaluate the metric between predictions and references.

Perspective api client for toxic evaluation. Args: key (str): Perspective API key. If set to `ENV`, find it in environment va

Compute the relative positional embedding, this is the same algorithm in Jumper et al. (2021) Suppl. Alg. 4 "relpos"