What Are Tensors?
TL;DR · AI Summary
Tensors are fundamental structures in machine learning that organize numbers, enabling models to process real-world data like text, images, and audio by converting them into numerical forms for the complete input-to-output data flow.
Key Takeaways
- Machine learning models process data through tensors, where scalars are 0D tenso
- Data conversion flow: Text becomes token IDs, images become pixel tensors, audio
- Tools like Transformers.js convert real-world inputs into tensors, allowing mode
Outline
Jump quickly between sections.
Machine learning models process data through tensors, which are numerical structures organized by shape, forming the basis for understanding real-world data.
Scalars are 0D tensors, vectors are 1D tensors, matrices are 2D tensors, with higher dimensions extending this concept.
Real-world inputs (text, images, audio) are converted to tensors for model processing, then output tensors are transformed back to useful results.
Mindmap
See how the topics connect at a glance.
查看大纲文本（无障碍 / 无 JS 友好）
- 张量
- 定义
- 按形状组织的数字
- 类型
- 标量(0D)
- 向量(1D)
- 矩阵(2D)
- 应用
- 数据转换
- 模型推理
- 结果输出
Highlights
Key sentences worth saving and sharing.
Machine learning models do not understand text, images or audio directly. They understand numbers.
A tensor is just numbers organized by shape. A single number is a scalar or a 0D tensor. A list of numbers is a vector or a 1D tensor.
Text becomes tokens and token IDs. Images become pixel tensors. Audio becomes numerical samples or features.