This repository provides some exercise codes to learn PyTorch. Since this repo doesn't provide the basic tutorial, please see after reading pytorch-exercise or Official Tutorial. NOTE: All the codes work on PyTorch 0.4.0.

• MNIST Classification
• Flower Classification (Custom Dataset)
• Utilities

• Image Super-resolution
• Image Colorization
• Style Transfer
• Conditional DCGAN

• Char-RNN
• Text Classification
• Show and Tell (CUB dataset)
• Show Attend and Tell (COCO dataset)

Make sure you have Python 3.5 or newer version. Installing the requirements are as follow:

pip install -r requirements.txt

For the RNN Applications codes, Some tokenizer pacakages such as SpaCy or NLTK are needed. You need to install these packages and its English model and data.

# install SpaCy
pip install spacy
python -m spacy download en

# install NLTK
pip install nltk
python -m nltk.downloader perluniprops nonbreaking_prefixes

Visdom and torchsummary are used in utilities exercise code. Please install these packages before run it.

pip install visdom torchsummary

Since dataset is downloaded and preprocessed automatically in the runtime (except few exercises) simply running below start training.

python train.py

Input arguments are vary among the codes, so please check the train.py for more details.

1. UnicodeDecodeError: 'ascii' codec can't decode byte 0xc3 when run the text classification.
   : Please refer this issue.

• pytorch-tutorial by yunjey
• practical-torchtext by keitakurita