Yi Wang

Yi Wang’s Page

• Compare Deep Learning Toolkits
• Roofline Analysis of Apple Silicon GPUs using MLX
• Activations Are Critial for Deep Learning: Train an MLP Using Your Brain to Address the XOR Classification Challenge
• Why Transformer Models Need Positional Encoding
• Efficient Implementation of Rotary Positional Embedding
• FlashAttention (Part 1): Tiled Matrix Multiplication
• FlashAttention (Part 2): Online Softmax
• FlashAttention (Part 3.)
• Automatic Differentiation (Part 1): The Jacobian for JVP and VJP
• Automatic Differentiation (Part 2): Built a Simple Deep Learning Toolkit with VJP
• Evaluate Hessian-Vector Product Without The Hessian Matrix
• Decipher JAX’s Tracing and JIT Compilation
• A Minimalist HTTP Server Using asyncio
• Unit Test Library Or Framework?
• JAX’s vmap: The Fundamentals
• JAX’s vmap: An Application of Speculative Decoding
• Curried Functions and JAX Type Annotation
• Trio: Simplifying Python Coroutines with Go’s Goroutine Style
• Using Conv2D for Linear Projection on Apple Neural Engine
• JAX’s Device Mesh and Tensor Sharding
• The Right Way to Configure LLM Training