pytorch实现generating names with a character-level RNN

最新推荐文章于 2023-08-12 13:53:47 发布
原创 最新推荐文章于 2023-08-12 13:53:47 发布 · 883 阅读 · 0 ·
CC 4.0 BY-SA版权
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
文章标签：

#pytorch #rnn

本文深入探讨了循环神经网络(RNN)和长短期记忆网络(LSTM)的工作原理，通过实例展示了这两种网络如何处理序列数据，同时提供了详细的网络结构配置和训练过程，包括损失函数、优化器选择和早期停止策略。

papers

The Unreasonable Effectiveness of Recurrent Neural Networks
https://karpathy.github.io/2015/05/21/rnn-effectiveness/

Understanding LSTM Networks
https://colah.github.io/posts/2015-08-Understanding-LSTMs/

输出
在这里插入图片描述
在这里插入图片描述

output

RNN(
  (i2h): Linear(in_features=205, out_features=128, bias=True)
  (i2o): Linear(in_features=205, out_features=59, bias=True)
  (o2o): Linear(in_features=187, out_features=59, bias=True)
  (dropout): Dropout(p=0.1)
  (softmax): LogSoftmax()
)
config:
eval_epoch_steps : 10
train_load_check_point_file : True
num_workers : 4
momentum : 0.9
early_stop_epoch_limit : 10
optimizer : SGD
early_stop_epoch : True
train_epoch_steps : 10
dataset : names
steps : 100000
early_stop_step : True
n_hidden : 128
epoch_only : True
learn_rate : 0.0005
all_letters : abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ .,;-
epochs : 100
n_letters : 59
batch_size : 1000
max_epoch_stop : True
max_step_stop : True
device : cpu
print_every : 5
data_path : ./data/names
loss : NLL
early_stop_step_limit : 100
[E: 0/100] [S: 5/100000] [Train: Loss:2.958800  5/10 (50%)] [Val: Loss:2.839336 0/10000 (0%)] [Best epoch:0 Loss:2.839336] [Best step:5 Loss:2.839336] [0.00s 0.0s]
[E: 0/100] [S: 10/100000] [Train: Loss:2.647621  10/10 (100%)] [Val: Loss:2.631263 0/10000 (0%)] [Best epoch:0 Loss:2.631263] [Best step:10 Loss:2.631263] [0.00s 0.0s]
[E: 0/100] [S: 10/100000] [Train: Loss:2.968491  10/10 (100%)] [Val: Loss:2.629430 0/10000 (0%)] [Best epoch:0 Loss:2.629430] [Best step:10 Loss:2.631263] [40.92s 40.9s]
[E: 1/100] [S: 15/100000] [Train: Loss:2.736446  5/10 (50%)] [Val: Loss:2.579854 0/10000 (0%)] [Best epoch:1 Loss:2.579854] [Best step:15 Loss:2.579854] [40.92s 40.9s]
[E: 1/100] [S: 20/100000] [Train: Loss:2.443919  10/10 (100%)] [Val: Loss:2.463921 0/10000 (0%)] [Best epoch:1 Loss:2.463921] [Best step:20 Loss:2.463921] [40.92s 40.9s]
[E: 1/100] [S: 20/100000] [Train: Loss:2.628853  10/10 (100%)] [Val: Loss:2.466182 0/10000 (0%)] [Best epoch:1 Loss:2.463921] [Best step:20 Loss:2.463921] [40.96s 81.9s]
[E: 2/100] [S: 25/100000] [Train: Loss:2.594146  5/10 (50%)] [Val: Loss:2.435748 0/10000 (0%)] [Best epoch:2 Loss:2.435748] [Best step:25 Loss:2.435748] [40.96s 81.9s]
[E: 2/100] [S: 30/100000] [Train: Loss:2.365779  10/10 (100%)] [Val: Loss:2.369891 0/10000 (0%)] [Best epoch:2 Loss:2.369891] [Best step:30 Loss:2.369891] [40.96s 81.9s]
[E: 1/100] [S: 25/100000] [Train: Loss:2.570118  5/10 (50%)] [Val: Loss:2.437161 0/10000 (0%)] [Best epoch:1 Loss:2.437161] [Best step:25 Loss:2.437161] [40.96s 81.9s] Step Status
[E: 1/100] [S: 30/100000] [Train: Loss:2.366355  10/10 (100%)] [Val: Loss:2.368035 0/10000 (0%)] [Best epoch:1 Loss:2.368035] [Best step:30 Loss:2.368035] [40.96s 81.9s] Step Status
[E: 1/100] [S: 30/100000] [Train: Loss:2.366355  10/10 (100%)] [Val: Loss:2.368035 0/10000 (0%)] [Best epoch:1 Loss:2.357966] [Best step:30 Loss:2.368035] [40.80s 122.7s] Epoch Status
[E: 2/100] [S: 35/100000] [Train: Loss:2.515979  5/10 (50%)] [Val: Loss:2.340214 0/10000 (0%)] [Best epoch:2 Loss:2.340214] [Best step:35 Loss:2.340214] [40.80s 122.7s] Step Status
[E: 2/100] [S: 40/100000] [Train: Loss:2.323779  10/10 (100%)] [Val: Loss:2.298239 0/10000 (0%)] [Best epoch:2 Loss:2.298239] [Best step:40 Loss:2.298239] [40.80s 122.7s] Step Status
[E: 2/100] [S: 40/100000] [Train: Loss:2.323779  10/10 (100%)] [Val: Loss:2.298239 0/10000 (0%)] [Best epoch:2 Loss:2.287091] [Best step:40 Loss:2.298239] [41.18s 163.9s] Epoch Status
[E: 3/100] [S: 45/100000] [Train: Loss:2.438657  5/10 (50%)] [Val: Loss:2.293576 0/10000 (0%)] [Best epoch:2 Loss:2.287091] [Best step:45 Loss:2.293576] [41.18s 163.9s] Step Status
[E: 3/100] [S: 50/100000] [Train: Loss:2.246784  10/10 (100%)] [Val: Loss:2.264124 0/10000 (0%)] [Best epoch:3 Loss:2.264124] [Best step:50 Loss:2.264124] [41.18s 163.9s] Step Status
[E: 3/100] [S: 50/100000] [Train: Loss:2.246784  10/10 (100%)] [Val: Loss:2.264124 0/10000 (0%)] [Best epoch:3 Loss:2.260864] [Best step:50 Loss:2.264124] [41.46s 205.3s] Epoch Status
[E: 4/100] [S: 55/100000] [Train: Loss:2.416693  5/10 (50%)] [Val: Loss:2.260192 0/10000 (0%)] [Best epoch:4 Loss:2.260192] [Best step:55 Loss:2.260192] [41.46s 205.3s] Step Status
[E: 4/100] [S: 60/100000] [Train: Loss:2.222714  10/10 (100%)] [Val: Loss:2.227812 0/10000 (0%)] [Best epoch:4 Loss:2.227812] [Best step:60 Loss:2.227812] [41.46s 205.3s] Step Status
[E: 4/100] [S: 60/100000] [Train: Loss:2.222714  10/10 (100%)] [Val: Loss:2.227812 0/10000 (0%)] [Best epoch:4 Loss:2.227812] [Best step:60 Loss:2.227812] [41.39s 246.7s] Epoch Status
[E: 5/100] [S: 65/100000] [Train: Loss:2.408526  5/10 (50%)] [Val: Loss:2.221667 0/10000 (0%)] [Best epoch:5 Loss:2.221667] [Best step:65 Loss:2.221667] [41.39s 246.7s] Step Status
[E: 5/100] [S: 70/100000] [Train: Loss:2.190458  10/10 (100%)] [Val: Loss:2.185826 0/10000 (0%)] [Best epoch:5 Loss:2.185826] [Best step:70 Loss:2.185826] [41.39s 246.7s] Step Status
[E: 5/100] [S: 70/100000] [Train: Loss:2.190458  10/10 (100%)] [Val: Loss:2.185826 0/10000 (0%)] [Best epoch:5 Loss:2.185826] [Best step:70 Loss:2.185826] [41.28s 288.0s] Epoch Status
[E: 6/100] [S: 75/100000] [Train: Loss:2.313870  5/10 (50%)] [Val: Loss:2.186711 0/10000 (0%)] [Best epoch:5 Loss:2.185826] [Best step:70 Loss:2.185826] [41.28s 288.0s] Step Status
[E: 6/100] [S: 80/100000] [Train: Loss:2.149164  10/10 (100%)] [Val: Loss:2.153440 0/10000 (0%)] [Best epoch:6 Loss:2.153440] [Best step:80 Loss:2.153440] [41.28s 288.0s] Step Status
[E: 6/100] [S: 80/100000] [Train: Loss:2.149164  10/10 (100%)] [Val: Loss:2.153440 0/10000 (0%)] [Best epoch:6 Loss:2.150477] [Best step:80 Loss:2.153440] [39.34s 327.3s] Epoch Status
[E: 7/100] [S: 85/100000] [Train: Loss:2.325105  5/10 (50%)] [Val: Loss:2.165497 0/10000 (0%)] [Best epoch:6 Loss:2.150477] [Best step:80 Loss:2.153440] [39.34s 327.3s] Step Status
[E: 7/100] [S: 90/100000] [Train: Loss:2.171192  10/10 (100%)] [Val: Loss:2.125976 0/10000 (0%)] [Best epoch:7 Loss:2.125976] [Best step:90 Loss:2.125976] [39.34s 327.3s] Step Status
[E: 7/100] [S: 90/100000] [Train: Loss:2.171192  10/10 (100%)] [Val: Loss:2.125976 0/10000 (0%)] [Best epoch:7 Loss:2.125976] [Best step:90 Loss:2.125976] [41.46s 368.8s] Epoch Status
[E: 8/100] [S: 95/100000] [Train: Loss:2.307920  5/10 (50%)] [Val: Loss:2.147382 0/10000 (0%)] [Best epoch:7 Loss:2.125976] [Best step:90 Loss:2.125976] [41.46s 368.8s] Step Status
[E: 8/100] [S: 100/100000] [Train: Loss:2.107451  10/10 (100%)] [Val: Loss:2.116395 0/10000 (0%)] [Best epoch:8 Loss:2.116395] [Best step:100 Loss:2.116395] [41.46s 368.8s] Step Status
[E: 8/100] [S: 100/100000] [Train: Loss:2.107451  10/10 (100%)] [Val: Loss:2.116395 0/10000 (0%)] [Best epoch:8 Loss:2.115855] [Best step:100 Loss:2.116395] [41.35s 410.1s] Epoch Status
[E: 9/100] [S: 105/100000] [Train: Loss:2.316520  5/10 (50%)] [Val: Loss:2.133158 0/10000 (0%)] [Best epoch:8 Loss:2.115855] [Best step:100 Loss:2.116395] [41.35s 410.1s] Step Status
[E: 9/100] [S: 110/100000] [Train: Loss:2.101270  10/10 (100%)] [Val: Loss:2.119242 0/10000 (0%)] [Best epoch:8 Loss:2.115855] [Best step:100 Loss:2.116395] [41.35s 410.1s] Step Status
[E: 9/100] [S: 110/100000] [Train: Loss:2.101270  10/10 (100%)] [Val: Loss:2.119242 0/10000 (0%)] [Best epoch:9 Loss:2.107559] [Best step:100 Loss:2.116395] [41.22s 451.4s] Epoch Status
[E: 10/100] [S: 115/100000] [Train: Loss:2.301298  5/10 (50%)] [Val: Loss:2.107354 0/10000 (0%)] [Best epoch:10 Loss:2.107354] [Best step:115 Loss:2.107354] [41.22s 451.4s] Step Status
[E: 10/100] [S: 120/100000] [Train: Loss:2.120560  10/10 (100%)] [Val: Loss:2.098027 0/10000 (0%)] [Best epoch:10 Loss:2.098027] [Best step:120 Loss:2.098027] [41.22s 451.4s] Step Status
[E: 10/100] [S: 120/100000] [Train: Loss:2.120560  10/10 (100%)] [Val: Loss:2.098027 0/10000 (0%)] [Best epoch:10 Loss:2.081124] [Best step:120 Loss:2.098027] [41.28s 492.6s] Epoch Status
[E: 11/100] [S: 125/100000] [Train: Loss:2.274269  5/10 (50%)] [Val: Loss:2.093999 0/10000 (0%)] [Best epoch:10 Loss:2.081124] [Best step:125 Loss:2.093999] [41.28s 492.6s] Step Status
[E: 11/100] [S: 130/100000] [Train: Loss:2.076419  10/10 (100%)] [Val: Loss:2.086819 0/10000 (0%)] [Best epoch:10 Loss:2.081124] [Best step:130 Loss:2.086819] [41.28s 492.6s] Step Status
[E: 11/100] [S: 130/100000] [Train: Loss:2.076419  10/10 (100%)] [Val: Loss:2.086819 0/10000 (0%)] [Best epoch:10 Loss:2.081124] [Best step:130 Loss:2.086819] [41.07s 533.7s] Epoch Status
[E: 12/100] [S: 135/100000] [Train: Loss:2.307883  5/10 (50%)] [Val: Loss:2.092391 0/10000 (0%)] [Best epoch:10 Loss:2.081124] [Best step:130 Loss:2.086819] [41.07s 533.7s] Step Status
[E: 12/100] [S: 140/100000] [Train: Loss:2.105602  10/10 (100%)] [Val: Loss:2.056846 0/10000 (0%)] [Best epoch:12 Loss:2.056846] [Best step:140 Loss:2.056846] [41.07s 533.7s] Step Status
[E: 12/100] [S: 140/100000] [Train: Loss:2.105602  10/10 (100%)] [Val: Loss:2.056846 0/10000 (0%)] [Best epoch:12 Loss:2.056846] [Best step:140 Loss:2.056846] [40.88s 574.6s] Epoch Status
[E: 13/100] [S: 145/100000] [Train: Loss:2.229566  5/10 (50%)] [Val: Loss:2.066366 0/10000 (0%)] [Best epoch:12 Loss:2.056846] [Best step:140 Loss:2.056846] [40.88s 574.6s] Step Status
[E: 13/100] [S: 150/100000] [Train: Loss:2.048885  10/10 (100%)] [Val: Loss:2.070952 0/10000 (0%)] [Best epoch:12 Loss:2.056846] [Best step:140 Loss:2.056846] [40.88s 574.6s] Step Status
[E: 13/100] [S: 150/100000] [Train: Loss:2.048885  10/10 (100%)] [Val: Loss:2.070952 0/10000 (0%)] [Best epoch:13 Loss:2.052732] [Best step:140 Loss:2.056846] [40.91s 615.5s] Epoch Status
[E: 14/100] [S: 155/100000] [Train: Loss:2.234232  5/10 (50%)] [Val: Loss:2.075228 0/10000 (0%)] [Best epoch:13 Loss:2.052732] [Best step:140 Loss:2.056846] [40.91s 615.5s] Step Status
[E: 13/100] [S: 155/100000] [Train: Loss:2.257734  5/10 (50%)] [Val: Loss:2.064717] [Best epoch:13 Loss:2.052732] [Best step:140 Loss:2.056846] [40.91s 615.5s] Step Status
[E: 13/100] [S: 160/100000] [Train: Loss:2.035141  10/10 (100%)] [Val: Loss:2.039078] [Best epoch:13 Loss:2.039078] [Best step:160 Loss:2.039078] [40.91s 615.5s] Step Status
[E: 13/100] [S: 160/100000] [Train: Loss:2.035141  10/10 (100%)] [Val: Loss:2.039078] [Best epoch:13 Loss:2.029834] [Best step:160 Loss:2.039078] [42.38s 657.9s] Epoch Status
[E: 14/100] [S: 165/100000] [Train: Loss:2.241119  5/10 (50%)] [Val: Loss:2.037968] [Best epoch:13 Loss:2.029834] [Best step:165 Loss:2.037968] [42.38s 657.9s] Step Status
[E: 14/100] [S: 170/100000] [Train: Loss:2.031683  10/10 (100%)] [Val: Loss:2.042479] [Best epoch:13 Loss:2.029834] [Best step:165 Loss:2.037968] [42.38s 657.9s] Step Status
[E: 14/100] [S: 170/100000] [Train: Loss:2.031683  10/10 (100%)] [Val: Loss:2.042479] [Best epoch:13 Loss:2.029834] [Best step:165 Loss:2.037968] [40.78s 698.6s] Epoch Status
[E: 15/100] [S: 175/100000] [Train: Loss:2.227793  5/10 (50%)] [Val: Loss:2.041410] [Best epoch:13 Loss:2.029834] [Best step:165 Loss:2.037968] [40.78s 698.6s] Step Status
[E: 15/100] [S: 180/100000] [Train: Loss:2.041359  10/10 (100%)] [Val: Loss:2.021471] [Best epoch:15 Loss:2.021471] [Best step:180 Loss:2.021471] [40.78s 698.6s] Step Status
[E: 15/100] [S: 180/100000] [Train: Loss:2.041359  10/10 (100%)] [Val: Loss:2.021471] [Best epoch:15 Loss:2.016103] [Best step:180 Loss:2.021471] [41.15s 739.8s] Epoch Status
[E: 16/100] [S: 185/100000] [Train: Loss:2.255320  5/10 (50%)] [Val: Loss:2.027396] [Best epoch:15 Loss:2.016103] [Best step:180 Loss:2.021471] [41.15s 739.8s] Step Status
[E: 16/100] [S: 190/100000] [Train: Loss:2.016954  10/10 (100%)] [Val: Loss:2.019632] [Best epoch:15 Loss:2.016103] [Best step:190 Loss:2.019632] [41.15s 739.8s] Step Status
[E: 16/100] [S: 190/100000] [Train: Loss:2.016954  10/10 (100%)] [Val: Loss:2.019632] [Best epoch:15 Loss:2.016103] [Best step:190 Loss:2.019632] [40.85s 780.6s] Epoch Status
[E: 17/100] [S: 195/100000] [Train: Loss:2.210386  5/10 (50%)] [Val: Loss:2.018931] [Best epoch:15 Loss:2.016103] [Best step:195 Loss:2.018931] [40.85s 780.6s] Step Status
[E: 17/100] [S: 200/100000]
最低0.47元/天 解锁文章
采用RNN为小Baby起个英文名字by SixInNight
hongwangdb的博客
06-27 2280
采用RNN为小Baby起个英文名字实验描述准备工作安装PyTorch观察数据集创建字典转换格式张量构造神经网络训练神经网络预测和计算损失打印训练进度和损失、准备绘图作图反应神经网络学习情况预测名字测试及打印结果预测男孩的名字,首字母为L预测女孩的名字,首字母为R完整代码 实验描述 本实验提供了 8000 多个英文名字，试训练一个环神经网络语言模型，进而给定一个开始字母，由语言模型自动生成后续的字母，直到生成一个名字的结束符，与此同时，输出模型为每个时刻预测的前几个最有可能的候选字母。 准备工作 安装Py
PyTorch -- RNN 快速实践
最新发布
哒哒哒哒 哒哒
06-19 1621
步骤一：确定 RNN Layer 相关参数值并基于此创建。步骤二：确定 训练流程。步骤三：测试训练结果。
Generating Names with a Character-Level RNN
walden的博客
12-08 304
目录结构： 1.准备数据 2.创建网络 3.训练 3.1为训练做准备 3.2训练网络 3.3画loss损失率 4.检验网络 3.1为训练做准备 EOS是 3.2训练网络 all_letters = string.ascii_letters + " .,;'-" n_letters = len(all_letters) + 1 # Plus EOS marker cr
GENERATING NAMES WITH A CHARACTER-LEVEL RNN教程解析
lyyiangang的博客
03-29 1487
pytorch RNN 名字生成代码解析
Generating Names and Classifying Names with Character-Level RNN
weixin_30915275的博客
10-23 201
原文地址: Generating Nameswith Character-Level RNN 搬运只为督促自己学习,没有其他目的。 Preparing the Data Download the data from here and extract it to the current directory In short, there are a bunch of plain text...
PyTorch翻译官网教程-NLP FROM SCRATCH: GENERATING NAMES WITH A CHARACTER-LEVEL RNN
poorCoder_的博客
08-12 868
最大的区别在于，我们不是在读取一个名字的所有字母后预测一个类别，而是输入一个类别并每次输出一个字母。为了进行示例，我们给网络一个字母并询问下一个字母是什么，将其作为下一个字母输入，并重复直到EOS令牌。当训练时，我们在每个时间步向网络提供它，这是一个设计选择，它可以作为初始隐藏状态的一部分或其他策略。与其给它一个起始字母，另一种策略是在训练中包含一个“字符串起始”标记，并让网络选择自己的起始字母。我们将创建(“A”，“B”)，(“B”，“C”)，(“C”，“D”)，(“D”，“EOS”)。
精选资源
Dual-path RNN： Pytorch实现的时域单通道语音分离的 高效长序列建模
06-01
Dual-path RNN：Pytorch实现的时域单通道语音分离的高效长序列建模 使用 WSJ0 数据集作为我们的训练、测试和验证集。下面是WSJ0的数据下载链接和混合音频代码。 音频混合样本 WSJ0 数据集 创建数据集
精选资源
pytorch-polygon-rnn:Pytorch实现的Polygon-RNN（http
05-03
pytorch-polygon-rnn Pytorch实现。 注意，我使用另一种方法来处理第一个顶点，而不是像本文中那样训练另一个模型。 与原纸的不同 我使用两个虚拟起始顶点来处理第一个顶点，如图像标题所示。 我需要在ConvLSTM层...
Neural_Speed_Reading_via_Skim-RNN_PyTorch:PyTorch实施“通过Skim-RNN进行神经速度读取”
05-12
这是在ICLR 2018上发布的进行的PyTorch实现。 默认情况下，将使用imdb数据集并将其存储在./data文件夹中。 此外，还使用了经过8400训练的300维GloVe词嵌入。 与目标是离散的或不同，Skim RNN引入了Gumbel-softmax...
精选资源
pytorch-sentiment-analysis-classification:情感分析分类的PyTorch教程（RNN，LSTM，Bi-LSTM，LSTM + Attention，CNN）
05-24
情感分析分类 先决条件 安装依赖项 pip install -r requirements.txt 安装Spacy英语数据 python -m spacy download en 框架 火炬 数据集 Cornell MR（电影评论）数据集...RNN LSTM 双LSTM LSTM +注意 有线电视新闻网
from_scratch：从头开始实现的ML示例
02-16
从头开始 从头开始实施的ML /优化示例。 该代码库旨在在ML和优化常用功能背后建立直觉。 现在，该代码库包含以下示例： KMeans聚类 主成分分析（PCA）：刮擦法使用幂迭代来计算奇异值和特征向量 线性回归：对多个自变量进行线性回归。 通过求解法线方程，梯度下降和随机（小批量）梯度下降确定的参数估计值 使用最速下降法和共轭梯度法求解线性方程组Ax = b 逻辑回归
[翻译Pytorch教程]NLP从零开始：使用字符级RNN进行名字生成
wmq104的博客
02-28 591
翻译自官网手册:NLP From Scratch: Generating Names with a Character-Level RNN Author: Sean Robertson 原文github代码 这是NLP从零开始三个教程的第二个。在第一个教程char_rnn_classification_tutorial中，使用RNN将名字按照其来源的语言进行了分类。本教程将进行名字生成。 >...
RNN 与语言模型
an 橘
11-30 662
RNN 与语言模型 学习笔记，原贴来源： The Unreasonable Effectiveness of Recurrent Neural Networks, May 21, 2015 BY Andrej Karpathy 总结：用 RNN 训练语言模型，并基于此算法介绍 5 个文本生成案例。 文章目录1. RNN 的特别之处：处理序列2. RNN 的隐藏层：“记忆”更进一步：2 层 RNN更...
从零开始 NLP：使用字符级 RNN 生成名字
qq_36646531的博客
05-11 613
从零开始 NLP：使用字符级 RNN 生成名字 pytorch官方例子学习心得 具体任务： 根据语言生成名字 官方项目连接 项目流程： 数据预处理 原始数据 构建模型 训练网络 可视化结果 ...
2020.7.13学习笔记
weixin_43793544的博客
07-19 250
2020.7.13论文笔记TransformerBERTNLP FROM SCRATCH: CLASSIFYING NAMES WITH A CHARACTER-LEVEL RNN Transformer 看了知乎上很详细的解释link 和哈佛的注释link 这里面有些代码陈旧了需要略微修改 稍微理解了原文的思想 BERT NLP FROM SCRATCH: CLASSIFYING NAMES WITH A CHARACTER-LEVEL RNN 按pytorch的文档 link 搭了一个模型，自己写了一些
MNIST数据集的格式以及读取方式
热门推荐
DarrenXf的专栏
12-24 1万+
MNIST 网站 http://yann.lecun.com/exdb/mnist/ 四个文件 train-images-idx3-ubyte.gz: training set images (9912422 bytes) train-labels-idx1-ubyte.gz: training set labels (28881 bytes) t10k-images-idx3-ubyte...
cifar10数据格式以及读取方式
DarrenXf的专栏
12-31 1万+
cifar10 数据网站 http://www.cs.toronto.edu/~kriz/cifar.html 读取下面的文件 CIFAR-10 binary version (suitable for C programs) 162 MB c32a1d4ab5d03f1284b67883e8d87530 下载cifar-10-binary.tar.gz 到./data/文件夹下 cd ./da...
GLUE多任务数据集介绍
DarrenXf的专栏
04-07 1万+
GLUE 是一个自然语言任务集合，包括以下这些数据集 name full name task chinese MNLI Multi-Genre NLI Natural language inference 自然语言推断 QQP Quora Quora Question Pairs Semantic textual similarity/Paraphrase identifica...
pytorch gpu 显存 内存 管理 调试 监控 日志
DarrenXf的专栏
06-02 7206
pytorch gpu 显存调试 如何运行 gpu_memory_log import torch from gpu_memory_log import gpu_memory_log dtype = torch.float N, D_in, H, D_out = 64, 1000, 100, 10 device = torch.device("cuda") x = torch.randn(N,...
评论
成就一亿技术人!
拼手气红包6.0元
还能输入1000个字符
 
红包 添加红包
表情包 插入表情
表情包 代码片
 条评论被折叠 查看
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值