之前我们聊过了 Transformer 架构和 PyTorch 框架的基础,今天来聊聊深度学习里的另一个经典架构 —— CNN (Convolutional Neural Network,卷积神经网络)。如果说 Transformer 是 NLP 领域的王者,那 CNN 就是计算机视觉 (Computer Vision) 的开山鼻祖。虽然现在 Vi…
从 Self-Attention 到 Multi-Head Attention,用 PyTorch 实现 Transformer 核心机制的学习笔记。涵盖 QKV 计算、Position Encoding、Encoder-Decoder 结构等关键概念。
A comprehensive overview of PyTorch — the open-source deep learning framework originally built on Torch with C and Lua, now powering modern AI research with Python-first design and dynamic computation graphs.
A beginner-friendly guide to MNIST handwritten digit recognition using deep learning — covering dataset structure, model architecture, and training pipeline with practical code examples.