BatchNorm on Text Matrixhttps://txtmix.com/tags/batchnorm/Recent content in BatchNorm on Text MatrixHugozh-cnTue, 21 Jul 2026 20:06:14 +0800从零手写神经网络：Karpathy 的 nn-zero-to-hero 到底在教什么https://txtmix.com/posts/tech/karpathy-nn-zero-to-hero-neural-networks-course/Tue, 02 Jun 2026 12:00:00 +0800https://txtmix.com/posts/tech/karpathy-nn-zero-to-hero-neural-networks-course/<p>Andrej Karpathy 的 <a href="https://github.com/karpathy/nn-zero-to-hero" target="_blank" rel="noopener noreffer ">Neural Networks: Zero to Hero</a>（22,500+ Stars，MIT 许可证）把 PyTorch 的高级 API 扔到一边，让你从 <code>numpy</code> 数组和 Python 原始运算开始，一行一行地把神经网络里真正在发生的事写出来。它不教你调参——它逼你搞清楚 <code>loss.backward()</code> 到底做了什么。</p>