infer main

func

init pretraining params

interface

check cost time from lst Returns: TODO Raises: NULL

log

Define the input data layers for the transformer model.

Define the input data layers for the transformer model.

Define the input data layers for the transformer model.

Define the input data layers for the transformer model.

make layer Args: in_planes: 输入特征通道数 places: 输出特征通道数 block: block个数 stride： 步长

Desc: 两个cluster进行合并 Args: self : self cl : new cluster Return: None Raise:

This function calculates and returns the precision, recall and f1-score Args: metric_fn: metric function pointer which calculates sco

Multi-Head Attention. Note that attn_bias is added to the logit before computing softmax activiation to mask certain selected positions so th

Multi-Head Attention. Note that attn_bias is added to the logit before computing softmax activiation to mask certain selected positions so th

Normalize strings to space joined chars. Args: s: a list of strings. Returns: A list of normalized strings.

Lower text and remove punctuation, articles and extra whitespace.

Lower text and remove punctuation, articles and extra whitespace.

:param qid_info: :return:

Pad the instances to the max sequence length in batch, and generate the corresponding position data and attention bias.

Pad the instances to the max sequence length in batch, and generate the corresponding position data and input mask.

padding ids

parse

:returns next idx, (agg_op id, col_id)

:returns next idx, (agg_op id, col_id)

Position-wise Feed-Forward Networks. This module consists of two linear transformations with a ReLU activation in between, which is appli

Position-wise Feed-Forward Networks. This module consists of two linear transformations with a ReLU activation in between, which is appli

result: one batch predict narray format result

result: one batch predict narray format result

Put all padded data needed by training into a dict.

Put all padded data needed by beam search decoder into a dict.

1. Prepare ALL (unique) labels for classification from trainval-set. 2. For each split, generate sample list for level1 & level2 classifi

Preprocess utterance and table value.

Print arguments

print log

Args: Returns:

Args: Returns:

Args: Returns:

Args: Returns:

remove_punctuation

implementation of Dataset.reset

Desc: link反转, 获取当前link的一个反link, 即start, end对调 Args: self : self Return: a new link Raise:

rnn encoder

save model

set pages from 'start' to 'end' with new same status 'status'

set match relation for question Args: q_match_dict (TYPE): [in/out] q_start (TYPE): NULL q_match_len (TYP

implementation of Dataset.size

Return a view of the list of tokens from [i, j).

str to num

tokenzie

Tokenizes a piece of text.

tokenize

topic generalization for text

do inference with given inference_program

inputs: (batch_size, seq_len, emb_dim)

warp an image/tensor (im2) back to im1, according to the optical flow x: [B, C, H, W] (im2) flo: [B, 2, H, W] flow

Desc: 计算这个聚类的x宽度 Args: self : self Return: float, x方向上的宽度 如果类为空, 返回None Raise:

Desc: 计算这个聚类的y宽度 Args: self : self Return: float, y方向上的宽度 如果类为空, 返回None Raise:

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