extract condition and/or

format

encode images and texts independently by Vit and RoBERTa, get the optimal grounded tokens

symmetric KL-divergence (See SMART, Sec 3.1) q_logits: logits p_logits: delta_logits

Read DocRED json file into examples

normalize the input unicode string

normalize the input unicode string

_pad_batch_records

Reads a tab separated value file.

new a tensor

update state

Make layer define support indicating block, by which we can add layers to other blocks within current block. This will make it easy to define

Desc: 计算region的面积 Args: self : self Return: float, 面积 Raise: None

build data feed

build data feed

构建数据迭代器 Args: mode: 数据迭代器的类型 可选值 ['Train', 'Gallery', 'Query', 'Eval'] Returns: data_loader: 返回对应类型的数据迭代器

build examples, data: ``List[Dict]``

build examples, data: ``List[Dict]``

Desc: 计算两个相邻grid之间的friend关系 Args: self : self grid1 : 网格号(grid_x, grid_y) grid2 : 网格号(grid_x, gri

calculate max sequence length

Desc: 是否能够merge Args: self : self cl : cl Return: None Raise: None

Log error and exit when set use_gpu=true in paddlepaddle cpu version.

conv net

conv net

conv net

Converts a sequence of [tokens|ids] using the vocab.

Convert a set of `InputExample`s to a list of `InputFeatures`.

Converts `text` to Unicode (if it's not already), assuming utf-8 input.

Converts `text` to Unicode (if it's not already), assuming utf-8 input.

Converts `text` to Unicode (if it's not already), assuming utf-8 input.

Converts `text` to Unicode (if it's not already), assuming utf-8 input.

Converts `text` to Unicode (if it's not already), assuming utf-8 input.

convert tokens to vocab ids

define fine-tuning model

func

func

data_process

down layer Args: in_planes: 输入特征通道数 places: 输出特征通道数 stride： 步长 kernel_size： 卷积大小 padding： padding大小 Re

build inputs for model Args: inputs (TYPE): NULL Returns: TODO Raises: NULL

The encoder is composed of a stack of identical layers returned by calling encoder_layer.

evaluate

evaluate

Evaluation interface @param : data_iter @type : DataLoader @param : need_save @type : bool

Obtain model loss or decoding output

evaluation

Position-wise Feed-Forward Network input is LoDTensor

Position-wise Feed-Forward Network input is LoDTensor

Position-wise Feed-Forward Network input is LoDTensor

file_reader

file_reader

外部调用接口

Merges a list of lists into a single list.

:param hidden_emb: [batch_size, seq_size, hidden_size] :param sequence_mask: [batch_size, seq_size, 1] :return:

Get features for the dataset

Get features for the dataset

Gets the list of labels for this data set.

Prepare answers for task 'main'. Args: qid: question_id. pred_result: A dict include all question_id's result information re

Generate a vocabulary from the datasets in sources.

Recall n at m

Get the first feature of each sequence for classification

get time

Gets progress for training phase.

Gets progress for training phase.

Project hidden states to a specific head with a 4D-shape.