Pytorch-Network-Model-Design

Description

Many tasks of Pytorch for learning Network-Model-Design included.

Some of the tasks can be found in kaggle.com:

• This is just a practice project, and many of the tasks on kaggle.com don't have great results.

• In fact, there is no need to use pytorch for some machine learning tasks (unless you have a high understanding of the underlying principles of machine learning), which can lead to some less efficient work.

• If you can, try to use machine learning to complete tasks on kaggle.

I'm putting together a collection of simple and helpful tasks that I hope will help myself and others who want to learn deep learning. (Or just for a simple deep learning task.)

Consider introducing tasks for Scikit-Learn, TensorFlow, etc. later. (just consider)

Examples / Tasks

• Chinese to English
• Next Frame Prediction
• Sentiment Analysis on Movie Review
• Binary Classification with a Bank Churn Dataset
• Multi-Class Prediction of Obesity Risk
• Regression with a Mohs Hardness Dataset
• ......

Structure

The sample files are in the template folder.

Note:

• Since the general model parameter file is large, it will not be uploaded to GitHub, if the code does not have the logic to create this folder, please create your own checkpoint folder.
• If the dataset file is small ( <5M ), it is uploaded directly to GitHub in the dataset folder. Some cleaned data is also placed / generated in this folder.

dataset/      # folder to store the data set

checkpoint/   # folder to store model parameters

criterion.py  # custom loss function

dataset.py    # custom dataset

model.py      # model

train.py      # train the model

test.py       # test the model

Contributors

• gralliry(Liang Jianye)
• zuozuo

Contact

Email: [email protected]

License

GNU General Public License v3.0

This project has open source tasks, according to the provisions of the open source agreement, this project is open source