Develop Notes

Contents:

• 1. deepseek模型学习笔记
  • 1.1. Multi-Head Latent Attention (MLA)
  • 1.2. 矩阵吸收absorb
• 2. floating point number
  • 2.1. Microscaling Formats
  • 2.2. IEEE 浮点数
• 3. cuda programming
  • 3.1. MMA (Matrix Multiply-Accumulate Instructions)
  • 3.2. old
  • 3.3. cuda environment
  • 3.4. whitepaper
• 4. multi-platform development
  • 4.1. How to detect cpu architecture
• 5. deep learning operators
  • 5.1. flash attention
  • 5.2. transformer decoder k/v cache
  • 5.3. LSTM
  • 5.4. 方差计算
• 6. Calculation of elementary math functions
• 7. onnnxruntime
  • 7.1. Pre-build packages
  • 7.2. MLAS
• 8. reStructuredText
• 9. Develop Environment
  • 9.1. Centos7
  • 9.2. CMAKE
  • 9.3. Conda
  • 9.4. Docker
  • 9.5. manylinux
  • 9.6. git
  • 9.7. Homebrew
• 10. ML system
  • 10.1. ML system 入门资料
• 11. deep learning miscellanies
  • 11.1. Backpropagation的推导
  • 11.2. convolution arithmetic
  • 11.3. NLP

Indices and tables

• 索引

• 模块索引

• 搜索页面