100 Papers with Code

About this website

This website aims to collect and showcase 100 important AI papers and their code implementations. Each paper is accompanied by a link for readers to explore further.

Why build this website?

On the one head, during my study, reading, and practice in the field of AI, I found that many important papers and code implementations were scattered in various places. To facilitate access for myself and others, I decided to build this website to consolidate these important papers and my implementations in one place. On the other hand, there is a gap between topic in papers and practical application. By providing code implementations, I hope to help bridge this gap and make it easier for practitioners to apply the latest research results.

There are some awesome resource about the paper and implementation such as:

But why I still build this website?

  • The Annotated Research Paper focuses more on each component of the paper, it didn’t provide the whole modeling and training process.
  • The Papers with Code provides the paper and code, but the code is sometime hard to understand, and it didn’t provide the explanation of the code.

This website is built to fill these gaps by providing clear explanations and easy-to-understand code implementations for each paper.Each page(paper) will come with a self-contained Jupyter Notebook that can be run directly, making it easier for readers to understand and apply the concepts presented in the papers.

How to use this website?

This website is designed to be user-friendly and accessible for anyone interested in understanding and implementing the latest research in AI and machine learning. Each paper has its own dedicated page that includes

  • The explanation of the paper.

  • Code implementation.

  • Q & A part (you can use as flash cards).

  • The further direction might be interesting to explore (According to my understanding).

    Below is the list of papers that have been implemented or are planned to be documented. You can click on the paper title to view detailed content.

Number Paper Name Description Code Category
01 Attention is All You Need (Transformer ) 引入了 Transformer 架构，该架构完全依赖 自注意力机制（Self-Attention） 进行序列建模，使得模型可以 并行计算，并在 自然语言处理任务 上大幅提升了性能。 NLP / Transformer
02 An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale ( Vision Transformer ) Vision Transformer (ViT) 是一种将图像划分为小块（patches），并将其作为 token 输入标准 Transformer 模型进行图像分类的架构，首次实现了纯注意力机制在视觉任务中的成功应用。 Computer Vision / Transformer
03 Swin Transformer: Hierarchical Vision Transformer using Shifted Windows ( Swin Transformer ) Swin Transformer 是一种使用层次化结构和滑动窗口自注意力机制的视觉 Transformer 模型，既保留了局部建模的高效性，又通过窗口偏移实现跨区域信息交互，适用于图像分类、目标检测和语义分割等多种视觉任务。 Computer Vision / Transformer
04 Learning Transferable Visual Models From Natural Language Supervision ( CLIP ) CLIP 是一种利用大规模图文对比学习，将图像与自然语言映射到同一语义空间，从而实现零样本图像识别与跨模态检索的多模态基础模型 Computer Vision / Transformer
05 FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness ( Flash Attention ) FlashAttention 是一种优化的注意力机制实现，通过减少内存访问和提升计算效率，实现更快、更节省资源的 Transformer 推理与训练。 Transformer / AI Engine
Back to top