NLP pytorch 模型训练过程

通过学习 https://www.bilibili.com/video/BV16S4y1P7PN ，完成笔记记录

# 引入必要的包
import torch
import numpy as np
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
import torch.utils.data as Data

# 在gpu上训练模型
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

一、数据集构建

1.encode

定义函数：输入数字，如果15取余=0返回3，5取余返回2，3取余返回1，否则返回数字本身

def fizzbuzz_encode(number):
    """
    number: int
    """
    if number % 15 == 0:
        return 3 # 'fizzbuzz'
    elif number % 5 == 0:
        return 2 # 'buzz'
    elif number % 3 == 0:
        return 1 # 'fizz'
    return 0 # str(number)
    

2.decode

通过0 1 2 3返回对应的label

def fizzbuzz_decode(number, label):
    """
    number: int
    label: 0 1 2 3
    """
    return [str(number), 'fizz', 'buzz', 'fizzbuzz'][label]

测试

def helper(number):
    print(fizzbuzz_decode(number, fizzbuzz_encode(number)))

for i in range(1, 16)