【PyTorch】基于 LSTM 的手写数字识别（MNIST）

最新推荐文章于 2023-10-20 14:11:15 发布
最新推荐文章于 2023-10-20 14:11:15 发布
阅读量2.7k 收藏 14
点赞数 1
分类专栏： pytorch 文章标签： lstm pytorch 深度学习 mnist
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。
本文链接：https://blog.csdn.net/qq_42951560/article/details/122745201
版权

引言

基于 PyTorch 实现的 LSTM 模型在 MNIST 数据集上的手写数字识别。

用法

代码托管于 GitHub：https://github.com/XavierJiezou/pytorch-lstm-mnist

git clone https://github.com/XavierJiezou/pytorch-lstm-mnist.git
cd pytorch-lstm-mnist
pip install -r requirements.txt
python ./code/train_on_mnist.py

注意：在训练之前你需要先解压 mnist.7z 文件。

配置

你也可以修改配置文件。

data:
  data_root: ./data/mnist # Path to data
  train_ratio: 0.8 # Ratio of training set
  val_ratio: 0.1 # Ratio of validation set
  batch_size: 64 # How many samples per batch to load
  visualize_data_save: ./image/training_data_mnist.png

model:
  input_size: 28 # Number of expected features in the input
  hidden_size: 64 # Number of features in the hidden state
  num_layers: 1 # Number of recurrent layers
  output_size: 10 # Number of expected features in the output

train:
  num_epochs: 100 # How many epochs to use for data training
  sequence_length: 28 # Length of the input sequence
  learning_rate: 0.001 # Learning_rate
  device: cuda:0 # On which a `torch.Tensor` is or will be allocated
  save_path: ./checkpoint/mnist.pth # Path to save the trained model

log:
  sink: ./log/mnist.log # Path to save the logging file
  level: INFO # Logging level: DEBUG | INFO | WARNING | ERROR | SUCCESS | CRITICAL
  format: '{message}' # logging output format. Example: '{time:YYYY-MM-DD at HH:mm:ss} {level} {message}'
  visualize_log_save: ./image/training_log_mnist.png # Path to save the visualization result

数据

MNIST: http://yann.lecun.com/exdb/mnist/

mnist

结果

数据集序列长度输入维度准确率
MNIST28280.9892

result

EPOCH: 001/100 LR: 0.0010 TRAIN-LOSS: 0.6444 TRAIN-ACC: 0.7916  VAL-LOSS: 0.2733 VAL-ACC: 0.9208  EPOCH-TIME: 0m10s
EPOCH: 002/100 LR: 0.0010 TRAIN-LOSS: 0.1911 TRAIN-ACC: 0.9446  VAL-LOSS: 0.1564 VAL-ACC: 0.9533  EPOCH-TIME: 0m10s
EPOCH: 003/100 LR: 0.0010 TRAIN-LOSS: 0.1317 TRAIN-ACC: 0.9617  VAL-LOSS: 0.1323 VAL-ACC: 0.9621  EPOCH-TIME: 0m9s
EPOCH: 004/100 LR: 0.0010 TRAIN-LOSS: 0.1077 TRAIN-ACC: 0.9686  VAL-LOSS: 0.1204 VAL-ACC: 0.9643  EPOCH-TIME: 0m9s
EPOCH: 005/100 LR: 0.0010 TRAIN-LOSS: 0.0894 TRAIN-ACC: 0.9735  VAL-LOSS: 0.1078 VAL-ACC: 0.9690  EPOCH-TIME: 0m9s
EPOCH: 006/100 LR: 0.0010 TRAIN-LOSS: 0.0776 TRAIN-ACC: 0.9775  VAL-LOSS: 0.0802 VAL-ACC: 0.9756  EPOCH-TIME: 0m9s
EPOCH: 007/100 LR: 0.0010 TRAIN-LOSS: 0.0695 TRAIN-ACC: 0.9794  VAL-LOSS: 0.0675 VAL-ACC: 0.9791  EPOCH-TIME: 0m9s
EPOCH: 008/100 LR: 0.0010 TRAIN-LOSS: 0.0589 TRAIN-ACC: 0.9827  VAL-LOSS: 0.0769 VAL-ACC: 0.9773  EPOCH-TIME: 0m9s
EPOCH: 009/100 LR: 0.0010 TRAIN-LOSS: 0.0536 TRAIN-ACC: 0.9838  VAL-LOSS: 0.0701 VAL-ACC: 0.9790  EPOCH-TIME: 0m9s
EPOCH: 010/100 LR: 0.0010 TRAIN-LOSS: 0.0488 TRAIN-ACC: 0.9850  VAL-LOSS: 0.0669 VAL-ACC: 0.9808  EPOCH-TIME: 0m10s
EPOCH: 011/100 LR: 0.0010 TRAIN-LOSS: 0.0441 TRAIN-ACC: 0.9862  VAL-LOSS: 0.0646 VAL-ACC: 0.9808  EPOCH-TIME: 0m10s
EPOCH: 012/100 LR: 0.0010 TRAIN-LOSS: 0.0403 TRAIN-ACC: 0.9881  VAL-LOSS: 0.0564 VAL-ACC: 0.9837  EPOCH-TIME: 0m10s
EPOCH: 013/100 LR: 0.0010 TRAIN-LOSS: 0.0371 TRAIN-ACC: 0.9889  VAL-LOSS: 0.0485 VAL-ACC: 0.9853  EPOCH-TIME: 0m9s
EPOCH: 014/100 LR: 0.0010 TRAIN-LOSS: 0.0334 TRAIN-ACC: 0.9900  VAL-LOSS: 0.0562 VAL-ACC: 0.9834  EPOCH-TIME: 0m10s
EPOCH: 015/100 LR: 0.0010 TRAIN-LOSS: 0.0320 TRAIN-ACC: 0.9905  VAL-LOSS: 0.0575 VAL-ACC: 0.9834  EPOCH-TIME: 0m9s
EPOCH: 016/100 LR: 0.0010 TRAIN-LOSS: 0.0318 TRAIN-ACC: 0.9906  VAL-LOSS: 0.0617 VAL-ACC: 0.9821  EPOCH-TIME: 0m9s
EPOCH: 017/100 LR: 0.0010 TRAIN-LOSS: 0.0278 TRAIN-ACC: 0.9913  VAL-LOSS: 0.0573 VAL-ACC: 0.9837  EPOCH-TIME: 0m10s
EPOCH: 018/100 LR: 0.0010 TRAIN-LOSS: 0.0258 TRAIN-ACC: 0.9918  VAL-LOSS: 0.0554 VAL-ACC: 0.9841  EPOCH-TIME: 0m9s
EPOCH: 019/100 LR: 0.0010 TRAIN-LOSS: 0.0217 TRAIN-ACC: 0.9933  VAL-LOSS: 0.0490 VAL-ACC: 0.9853  EPOCH-TIME: 0m10s
EPOCH: 020/100 LR: 0.0010 TRAIN-LOSS: 0.0215 TRAIN-ACC: 0.9933  VAL-LOSS: 0.0460 VAL-ACC: 0.9864  EPOCH-TIME: 0m10s
EPOCH: 021/100 LR: 0.0010 TRAIN-LOSS: 0.0206 TRAIN-ACC: 0.9933  VAL-LOSS: 0.0586 VAL-ACC: 0.9847  EPOCH-TIME: 0m10s
EPOCH: 022/100 LR: 0.0010 TRAIN-LOSS: 0.0191 TRAIN-ACC: 0.9939  VAL-LOSS: 0.0533 VAL-ACC: 0.9863  EPOCH-TIME: 0m10s
EPOCH: 023/100 LR: 0.0010 TRAIN-LOSS: 0.0179 TRAIN-ACC: 0.9945  VAL-LOSS: 0.0470 VAL-ACC: 0.9856  EPOCH-TIME: 0m10s
EPOCH: 024/100 LR: 0.0010 TRAIN-LOSS: 0.0182 TRAIN-ACC: 0.9940  VAL-LOSS: 0.0434 VAL-ACC: 0.9884  EPOCH-TIME: 0m10s
EPOCH: 025/100 LR: 0.0010 TRAIN-LOSS: 0.0140 TRAIN-ACC: 0.9957  VAL-LOSS: 0.0458 VAL-ACC: 0.9877  EPOCH-TIME: 0m10s
EPOCH: 026/100 LR: 0.0010 TRAIN-LOSS: 0.0149 TRAIN-ACC: 0.9954  VAL-LOSS: 0.0599 VAL-ACC: 0.9838  EPOCH-TIME: 0m10s
EPOCH: 027/100 LR: 0.0010 TRAIN-LOSS: 0.0154 TRAIN-ACC: 0.9951  VAL-LOSS: 0.0508 VAL-ACC: 0.9858  EPOCH-TIME: 0m10s
EPOCH: 028/100 LR: 0.0010 TRAIN-LOSS: 0.0160 TRAIN-ACC: 0.9949  VAL-LOSS: 0.0471 VAL-ACC: 0.9878  EPOCH-TIME: 0m10s
EPOCH: 029/100 LR: 0.0010 TRAIN-LOSS: 0.0128 TRAIN-ACC: 0.9957  VAL-LOSS: 0.0477 VAL-ACC: 0.9860  EPOCH-TIME: 0m10s
EPOCH: 030/100 LR: 0.0010 TRAIN-LOSS: 0.0107 TRAIN-ACC: 0.9966  VAL-LOSS: 0.0552 VAL-ACC: 0.9856  EPOCH-TIME: 0m10s
EPOCH: 031/100 LR: 0.0010 TRAIN-LOSS: 0.0107 TRAIN-ACC: 0.9967  VAL-LOSS: 0.0514 VAL-ACC: 0.9867  EPOCH-TIME: 0m9s
EPOCH: 032/100 LR: 0.0010 TRAIN-LOSS: 0.0122 TRAIN-ACC: 0.9958  VAL-LOSS: 0.0491 VAL-ACC: 0.9883  EPOCH-TIME: 0m10s
EPOCH: 033/100 LR: 0.0010 TRAIN-LOSS: 0.0106 TRAIN-ACC: 0.9967  VAL-LOSS: 0.0491 VAL-ACC: 0.9878  EPOCH-TIME: 0m9s
EPOCH: 034/100 LR: 0.0010 TRAIN-LOSS: 0.0081 TRAIN-ACC: 0.9973  VAL-LOSS: 0.0516 VAL-ACC: 0.9873  EPOCH-TIME: 0m10s
EPOCH: 035/100 LR: 0.0010 TRAIN-LOSS: 0.0136 TRAIN-ACC: 0.9958  VAL-LOSS: 0.0547 VAL-ACC: 0.9851  EPOCH-TIME: 0m9s
EPOCH: 036/100 LR: 0.0010 TRAIN-LOSS: 0.0082 TRAIN-ACC: 0.9975  VAL-LOSS: 0.0484 VAL-ACC: 0.9878  EPOCH-TIME: 0m10s
EPOCH: 037/100 LR: 0.0010 TRAIN-LOSS: 0.0104 TRAIN-ACC: 0.9965  VAL-LOSS: 0.0533 VAL-ACC: 0.9870  EPOCH-TIME: 0m9s
EPOCH: 038/100 LR: 0.0010 TRAIN-LOSS: 0.0084 TRAIN-ACC: 0.9975  VAL-LOSS: 0.0653 VAL-ACC: 0.9840  EPOCH-TIME: 0m10s
EPOCH: 039/100 LR: 0.0010 TRAIN-LOSS: 0.0084 TRAIN-ACC: 0.9975  VAL-LOSS: 0.0516 VAL-ACC: 0.9880  EPOCH-TIME: 0m9s
EPOCH: 040/100 LR: 0.0010 TRAIN-LOSS: 0.0081 TRAIN-ACC: 0.9974  VAL-LOSS: 0.0474 VAL-ACC: 0.9886  EPOCH-TIME: 0m9s
EPOCH: 041/100 LR: 0.0010 TRAIN-LOSS: 0.0079 TRAIN-ACC: 0.9976  VAL-LOSS: 0.0574 VAL-ACC: 0.9858  EPOCH-TIME: 0m9s
EPOCH: 042/100 LR: 0.0010 TRAIN-LOSS: 0.0072 TRAIN-ACC: 0.9979  VAL-LOSS: 0.0564 VAL-ACC: 0.9866  EPOCH-TIME: 0m9s
EPOCH: 043/100 LR: 0.0010 TRAIN-LOSS: 0.0095 TRAIN-ACC: 0.9971  VAL-LOSS: 0.0557 VAL-ACC: 0.9868  EPOCH-TIME: 0m10s
EPOCH: 044/100 LR: 0.0010 TRAIN-LOSS: 0.0061 TRAIN-ACC: 0.9980  VAL-LOSS: 0.0469 VAL-ACC: 0.9896  EPOCH-TIME: 0m10s
EPOCH: 045/100 LR: 0.0010 TRAIN-LOSS: 0.0089 TRAIN-ACC: 0.9971  VAL-LOSS: 0.0569 VAL-ACC: 0.9868  EPOCH-TIME: 0m10s
EPOCH: 046/100 LR: 0.0010 TRAIN-LOSS: 0.0068 TRAIN-ACC: 0.9979  VAL-LOSS: 0.0540 VAL-ACC: 0.9867  EPOCH-TIME: 0m9s
EPOCH: 047/100 LR: 0.0010 TRAIN-LOSS: 0.0049 TRAIN-ACC: 0.9985  VAL-LOSS: 0.0569 VAL-ACC: 0.9881  EPOCH-TIME: 0m10s
EPOCH: 048/100 LR: 0.0010 TRAIN-LOSS: 0.0079 TRAIN-ACC: 0.9975  VAL-LOSS: 0.0524 VAL-ACC: 0.9880  EPOCH-TIME: 0m9s
EPOCH: 049/100 LR: 0.0010 TRAIN-LOSS: 0.0037 TRAIN-ACC: 0.9991  VAL-LOSS: 0.0541 VAL-ACC: 0.9876  EPOCH-TIME: 0m9s
EPOCH: 050/100 LR: 0.0010 TRAIN-LOSS: 0.0075 TRAIN-ACC: 0.9976  VAL-LOSS: 0.0564 VAL-ACC: 0.9866  EPOCH-TIME: 0m9s
EPOCH: 051/100 LR: 0.0001 TRAIN-LOSS: 0.0031 TRAIN-ACC: 0.9992  VAL-LOSS: 0.0504 VAL-ACC: 0.9890  EPOCH-TIME: 0m9s
EPOCH: 052/100 LR: 0.0001 TRAIN-LOSS: 0.0013 TRAIN-ACC: 0.9998  VAL-LOSS: 0.0497 VAL-ACC: 0.9893  EPOCH-TIME: 0m10s
EPOCH: 053/100 LR: 0.0001 TRAIN-LOSS: 0.0010 TRAIN-ACC: 0.9999  VAL-LOSS: 0.0491 VAL-ACC: 0.9901  EPOCH-TIME: 0m10s
EPOCH: 054/100 LR: 0.0001 TRAIN-LOSS: 0.0008 TRAIN-ACC: 0.9999  VAL-LOSS: 0.0496 VAL-ACC: 0.9897  EPOCH-TIME: 0m9s
EPOCH: 055/100 LR: 0.0001 TRAIN-LOSS: 0.0007 TRAIN-ACC: 0.9999  VAL-LOSS: 0.0504 VAL-ACC: 0.9897  EPOCH-TIME: 0m9s
EPOCH: 056/100 LR: 0.0001 TRAIN-LOSS: 0.0006 TRAIN-ACC: 0.9999  VAL-LOSS: 0.0503 VAL-ACC: 0.9896  EPOCH-TIME: 0m10s
EPOCH: 057/100 LR: 0.0001 TRAIN-LOSS: 0.0005 TRAIN-ACC: 0.9999  VAL-LOSS: 0.0503 VAL-ACC: 0.9900  EPOCH-TIME: 0m10s
EPOCH: 058/100 LR: 0.0001 TRAIN-LOSS: 0.0005 TRAIN-ACC: 0.9999  VAL-LOSS: 0.0525 VAL-ACC: 0.9896  EPOCH-TIME: 0m10s
EPOCH: 059/100 LR: 0.0001 TRAIN-LOSS: 0.0004 TRAIN-ACC: 0.9999  VAL-LOSS: 0.0532 VAL-ACC: 0.9896  EPOCH-TIME: 0m10s
EPOCH: 060/100 LR: 0.0001 TRAIN-LOSS: 0.0003 TRAIN-ACC: 0.9999  VAL-LOSS: 0.0534 VAL-ACC: 0.9898  EPOCH-TIME: 0m10s
EPOCH: 061/100 LR: 0.0001 TRAIN-LOSS: 0.0003 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0546 VAL-ACC: 0.9891  EPOCH-TIME: 0m10s
EPOCH: 062/100 LR: 0.0001 TRAIN-LOSS: 0.0003 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0555 VAL-ACC: 0.9894  EPOCH-TIME: 0m10s
EPOCH: 063/100 LR: 0.0001 TRAIN-LOSS: 0.0002 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0556 VAL-ACC: 0.9894  EPOCH-TIME: 0m10s
EPOCH: 064/100 LR: 0.0001 TRAIN-LOSS: 0.0002 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0586 VAL-ACC: 0.9898  EPOCH-TIME: 0m9s
EPOCH: 065/100 LR: 0.0001 TRAIN-LOSS: 0.0002 TRAIN-ACC: 0.9999  VAL-LOSS: 0.0584 VAL-ACC: 0.9894  EPOCH-TIME: 0m10s
EPOCH: 066/100 LR: 0.0001 TRAIN-LOSS: 0.0001 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0579 VAL-ACC: 0.9893  EPOCH-TIME: 0m9s
EPOCH: 067/100 LR: 0.0001 TRAIN-LOSS: 0.0001 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0599 VAL-ACC: 0.9887  EPOCH-TIME: 0m9s
EPOCH: 068/100 LR: 0.0001 TRAIN-LOSS: 0.0001 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0611 VAL-ACC: 0.9891  EPOCH-TIME: 0m10s
EPOCH: 069/100 LR: 0.0001 TRAIN-LOSS: 0.0001 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0634 VAL-ACC: 0.9890  EPOCH-TIME: 0m9s
EPOCH: 070/100 LR: 0.0001 TRAIN-LOSS: 0.0001 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0621 VAL-ACC: 0.9888  EPOCH-TIME: 0m9s
EPOCH: 071/100 LR: 0.0001 TRAIN-LOSS: 0.0001 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0620 VAL-ACC: 0.9890  EPOCH-TIME: 0m10s
EPOCH: 072/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0657 VAL-ACC: 0.9888  EPOCH-TIME: 0m10s
EPOCH: 073/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0667 VAL-ACC: 0.9887  EPOCH-TIME: 0m10s
EPOCH: 074/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0665 VAL-ACC: 0.9888  EPOCH-TIME: 0m10s
EPOCH: 075/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0675 VAL-ACC: 0.9888  EPOCH-TIME: 0m10s
EPOCH: 076/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0686 VAL-ACC: 0.9888  EPOCH-TIME: 0m10s
EPOCH: 077/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0692 VAL-ACC: 0.9888  EPOCH-TIME: 0m9s
EPOCH: 078/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0707 VAL-ACC: 0.9888  EPOCH-TIME: 0m10s
EPOCH: 079/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0725 VAL-ACC: 0.9884  EPOCH-TIME: 0m10s
EPOCH: 080/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0749 VAL-ACC: 0.9887  EPOCH-TIME: 0m10s
EPOCH: 081/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0750 VAL-ACC: 0.9886  EPOCH-TIME: 0m10s
EPOCH: 082/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0759 VAL-ACC: 0.9887  EPOCH-TIME: 0m9s
EPOCH: 083/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0784 VAL-ACC: 0.9884  EPOCH-TIME: 0m10s
EPOCH: 084/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0810 VAL-ACC: 0.9884  EPOCH-TIME: 0m10s
EPOCH: 085/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0788 VAL-ACC: 0.9887  EPOCH-TIME: 0m10s
EPOCH: 086/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0815 VAL-ACC: 0.9888  EPOCH-TIME: 0m10s
EPOCH: 087/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0817 VAL-ACC: 0.9886  EPOCH-TIME: 0m10s
EPOCH: 088/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0824 VAL-ACC: 0.9884  EPOCH-TIME: 0m10s
EPOCH: 089/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0850 VAL-ACC: 0.9887  EPOCH-TIME: 0m9s
EPOCH: 090/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0869 VAL-ACC: 0.9886  EPOCH-TIME: 0m10s
EPOCH: 091/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0892 VAL-ACC: 0.9883  EPOCH-TIME: 0m9s
EPOCH: 092/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0903 VAL-ACC: 0.9883  EPOCH-TIME: 0m10s
EPOCH: 093/100 LR: 0.0001 TRAIN-LOSS: 0.0001 TRAIN-ACC: 1.0000  VAL-LOSS: 0.1042 VAL-ACC: 0.9868  EPOCH-TIME: 0m9s
EPOCH: 094/100 LR: 0.0001 TRAIN-LOSS: 0.0003 TRAIN-ACC: 0.9999  VAL-LOSS: 0.0972 VAL-ACC: 0.9874  EPOCH-TIME: 0m10s
EPOCH: 095/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0969 VAL-ACC: 0.9874  EPOCH-TIME: 0m9s
EPOCH: 096/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0966 VAL-ACC: 0.9877  EPOCH-TIME: 0m9s
EPOCH: 097/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0962 VAL-ACC: 0.9877  EPOCH-TIME: 0m9s
EPOCH: 098/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0959 VAL-ACC: 0.9877  EPOCH-TIME: 0m9s
EPOCH: 099/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0956 VAL-ACC: 0.9877  EPOCH-TIME: 0m9s
EPOCH: 100/100 LR: 0.0001 TRAIN-LOSS: 0.0000 TRAIN-ACC: 1.0000  VAL-LOSS: 0.0953 VAL-ACC: 0.9877  EPOCH-TIME: 0m10s

参考

Xavier Jiezou
关注
  • 1
    点赞
  • 踩
  • 14
    收藏
    觉得还不错? 一键收藏
  • 打赏
    打赏
  • 0
    评论
专栏目录
【图像分类】基于PyTorch搭建卷积神经网络实现MNIST手写数字识别(附项目完整代码)
学无止境、积少成多、厚积薄发
02-16 794
基于PyTorch搭建卷积神经网络实现MNIST手写数字识别(附项目完整代码)
pytorch 利用lstm做mnist手写数字识别分类的实例
01-21
代码如下，U我认为对于新手来说最重要的是学会rnn读取数据的格式。 # -*- coding: utf-8 -*- Created on Tue Oct 9 08:53:25 2018 @author: www import sys sys.path.append('..') import torch import datetime from torch.autograd import Variable from torch import nn from torch.utils.data import DataLoader from torchvision import tr
pytorch 入门 （三）案例一：mnist手写数字识别
最新发布
qq_33489955的博客
10-20 1170
本文为内部限免文章。
CNN手写数字识别
qq_54452082的博客
09-03 1094
基于卷积神经网络（CNN）的手写数字识别
深度学习实战—手写数字识别
qq_44113630的博客
04-30 1832
文章目录一、mnist数据集下载二、pytorch pycharm三、直接运行代码四、参考视频和开源代码 提示：以下是本篇文章正文内容，下面案例可供参考 一、mnist数据集下载 http://yann.lecun.com/exdb/mnist/ 二、pytorch pycharm 三、直接运行代码 https://github.com/pytorch/examples/blob/main/mnist/main.py 四、参考视频和开源代码 1.https://www.bilibili.com/vid
基于LSTM实现mnist手写数字识别
qq_29547673的博客
04-27 2438
首先读取数据，数据源是mnist库，可以通过input_data中read_data函数直接读取数据，数据图像为28*28。 #导入库 import tensorflow as tf #下载数据对应的库 import input_data import numpy as np import matplotlib.pyplot as plt print ("Packages imported")...
基于LSTM来实现mnist手写数字识别（内附代码）
flinjin的博客
03-17 831
import tensorflow as tf import input_data import numpy as np import matplotlib.pyplot as plt print ("Packages imported") mnist = input_data.read_data_sets("data/", one_hot=True) trainimgs, trainlabel...
人工智能&深度学习必练案例-PyTorch基于RNN和LSTM进行MNIST手写数字识别（完整代码，含完整注释 + 数据集
05-02
这是一个完整的深度学习案例，适合正在学习深度学习和人工智能的朋友练手使用，该案例所用框架为PyTorch，所用编程语言为Python，所使用的模型，最后各自运用两种模型（一个是RNN，一个是LSTM）完成了该项目，已经在...
pythorch mnist手写数字 CNN MLP LSTM 识别
03-27
在本项目中，我们主要关注的是使用PyTorch框架对MNIST手写数字识别的实现。MNIST数据集是机器学习领域的一个经典基准，它包含了0-9的手写数字图像，用于训练和测试图像分类模型。PyTorch是一个强大的深度学习库，...
【PyTorch 深度学习实战（2）】基于LSTM实现 MNIST手写数字识别 (含源代码+数据集 详解)-2023.5.2
05-02
主题是基于LSTM循环神经网络实现MNIST手写数字识别，所用的框架为目前的主流框架PyTorch，所用编程语言为Python。该项目我已经在博客中详细记录，包含每一步的结果和通俗的解释等，建议结合专栏【Python从入门到人工...
基于四层神经网络的MNIST手写数字识别（GUI-tkinter）
weixin_61279829的博客
09-06 1901
暑假做的神经网络(neural network)手写数字识别项目，基于MNIST数据集和tkinter图形界面编程。
LSTM基本理论及手写数字识别实战应用（pytorch）
gary101818的博客
04-07 4009
LSTM介绍 LSTM的特点（与RNN的区别） 具体实现流程 公式汇总及总结 LSTM实现手写数字识别（pytorch代码） 导入环境 定义超参数 训练和测试数据定义 定义LSTM模型 LSTM模型训练和预测
【图像分类】基于PyTorch搭建LSTM实现MNIST手写数字体识别（单向LSTM，附完整代码和数据集）
学无止境、积少成多、厚积薄发
02-13 2499
基于PyTorch搭建LSTM实现MNIST手写数字体识别（单向LSTM，附完整代码和数据集）
基于Pytorch的MNIST手写数字识别实现（含代码+讲解）
M_ovo的博客
06-22 7572
mnist代码+讲解
【深度学习案例】手写数字项目实现-1.数据集介绍
grape_yan的博客
08-16 2799
该项目通过手写数字识别实现简单的深度学习模型训练实现，充分认识深度学习项目构建过程，提供Matlab以及Python实现版本。 手写数字识别是指给定一系列的手写数字图片以及对应的数字标签，构建模型进行学习，目标是对于一张新的手写数字图片能够自动识别出对应的数字。图像识别是指利用计算机对图像进行处理，通过模型对其分析和理解，得到图片文件中所写的数字。在人工智能领域，手写数字识别被问题转换为自动分类问题。将0~9之内的10个数字分为10类，通过模型训练，实现对数字图片的分类，间接获取数字图片上的手写数字。...
Pytorch入门实战-----LSTM识别手写数据集
rongpeisheng666的博客
06-01 3242
LSTM长短期记忆网络，听说用来识别手写数据集有点大材小做，不过一直对语音、序列模型等等没怎么实践过，以后有空了再玩玩。 跑了一个epoch，正确率为97。 完整代码： import torch import torch.nn as nn import torchvision.transforms as transforms import torchvision.datasets as...
5. LSTM的C++实现
weixin_42216799的博客
11-29 3275
Lstm是RNN网络中最有趣的结构之一，不仅仅使得模型可以从长序列中学习，还创建了长短期记忆模块，模块中所记忆的数值在需要时可以得到更改。遗忘门遗忘单元可以将输入信息和隐藏信息进行信息整合，并进行信息更替，更替步骤如右图公式，其中与乘上权重矩阵后，加上偏置项后，经过激活函数，此时输出值为位于[0,1]之间,并将上一个时间步的与激活函数输出值相乘，更新为输入门当有输入进入时，输入门会结合输入信息与隐藏信息进行整合，并对信息进行更替。
pytorch 使用LSTM进行手写数字识别
qq_40107571的博客
12-19 332
pytorch 使用LSTM进行手写数字识别
基于pytorch的mnist手写数字识别步骤
05-23
以下是基于PyTorch的MNIST手写数字识别的步骤： 1.导入必要的库 ```python import torch import torch.nn as nn import torch.optim as optim from torch.utils.data import DataLoader from torchvision import datasets, transforms ``` 2.定义数据预处理操作 ```python transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,)) ]) ``` 3.加载数据集 ```python train_data = datasets.MNIST(root='data', train=True, download=True, transform=transform) test_data = datasets.MNIST(root='data', train=False, download=True, transform=transform) ``` 4.定义数据加载器 ```python batch_size = 64 train_loader = DataLoader(train_data, batch_size=batch_size, shuffle=True) test_loader = DataLoader(test_data, batch_size=batch_size, shuffle=True) ``` 5.定义模型 ```python class Net(nn.Module): def __init__(self): super(Net, self).__init__() self.conv1 = nn.Conv2d(1, 32, kernel_size=3, padding=1) self.conv2 = nn.Conv2d(32, 64, kernel_size=3, padding=1) self.dropout1 = nn.Dropout2d(0.25) self.dropout2 = nn.Dropout2d(0.5) self.fc1 = nn.Linear(64*7*7, 128) self.fc2 = nn.Linear(128, 10) def forward(self, x): x = nn.functional.relu(self.conv1(x)) x = nn.functional.relu(self.conv2(x)) x = nn.functional.max_pool2d(x, 2) x = self.dropout1(x) x = torch.flatten(x, 1) x = nn.functional.relu(self.fc1(x)) x = self.dropout2(x) x = self.fc2(x) return nn.functional.log_softmax(x, dim=1) model = Net() ``` 6.定义优化器和损失函数 ```python learning_rate = 0.01 optimizer = optim.SGD(model.parameters(), lr=learning_rate, momentum=0.5) criterion = nn.CrossEntropyLoss() ``` 7.训练模型 ```python epochs = 10 for epoch in range(epochs): for batch_idx, (data, target) in enumerate(train_loader): optimizer.zero_grad() output = model(data) loss = criterion(output, target) loss.backward() optimizer.step() if batch_idx % 100 == 0: print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format( epoch, batch_idx*len(data), len(train_loader.dataset), 100.*batch_idx/len(train_loader), loss.item())) ``` 8.测试模型 ```python test_loss = 0 correct = 0 with torch.no_grad(): for data, target in test_loader: output = model(data) test_loss += criterion(output, target).item() pred = output.data.max(1, keepdim=True)[1] correct += pred.eq(target.data.view_as(pred)).sum() test_loss /= len(test_loader.dataset) accuracy = 100. * correct / len(test_loader.dataset) print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format( test_loss, correct, len(test_loader.dataset), accuracy)) ``` 这就是基于PyTorch的MNIST手写数字识别的步骤。

“相关推荐”对你有帮助么？

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
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

打赏作者

Xavier Jiezou

你的鼓励将是我创作的最大动力

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付：¥1
获取中
扫码支付

您的余额不足，请更换扫码支付或充值

打赏作者

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