stophobia / pytorch-study Goto Github PK

By studying a book named '파이토치 첫걸음', I organized what I learned about PyTorch in this repository.

Ch1,Ch2 don't include codes. So there are no files about this chapters.
The rest of the chapters are implemented in jupyter notebook named chapter name.
Plus, a file named practice is a file practiced based on what I learned from the this book.

• Concept of DeepLearning.

• install PyTorch.

• Basic tensor manipulation. (eg. torch.Tensor)
• implement Linear Regrssion using gradient descent. (by using torch.nn, torch.optim.. etc)

• implement MLP, then solve non-Linear data. (by using nn.Sequential, nn.Linear, nn.ReLU.. etc)
• draw Learning curve through loss data.

• implement CNN ,then classify MNIST images.
• implement famous vision models. (VGGNet, GoogLeNet, ResNet)

• implement Simple RNN, LSTM, then test these model through tinyshakespeare data.
• implement embedding.

• Concept of overfitting and underfiting.
• Implement methods to prevent overfittig -> Regularization, Dropout, Data Augmentation, Weight Initialization, Learning rate, normalization, BatchNorm, optimizer

• Concept of Transfer Learning and Style Transfer.
• implement the Style Transfer.

• implement the Auto Encoder.(simple version, convolution version)
• implement UNet for semantic segmentation.

• Concept of GAN.
• implement the basic GAN using MNIST images.
• implement DCGAN using MNIST images.