Series Path · 8 parts
Inference Optimization
Make models faster at inference — quantization, edge deployment, compilation, efficient attention, KV cache compression, structured generation, and ONNX Runtime.
Curriculum — 8 parts
- 01
Complete Guide to Quantization and Pruning
Model compression techniques are essential for deploying deep learning models in resource-constrained environments. Two of the most effective approaches are…
7 min read - 02
PyTorch Model Deployment on Edge Devices - Complete Code Guide
This guide provides a comprehensive approach to deploying PyTorch models on edge devices. Key takeaways:
9 min read - 03
PyTorch 2.x Compilation Pipeline: From FX to Hardware
PyTorch 2.x introduced a revolutionary compilation stack that transforms high-level Python code into highly optimized machine code. This guide explores the…
11 min read - 04
FlashAttention for Image Models
Attention is the computational core of nearly every modern image model — Vision Transformers (ViT) for classification and representation learning, Diffusion…
21 min read - 05
VL-Cache: A Technical Walkthrough
Sparsity- and Modality-Aware KV Cache Compression for Vision-Language Model Inference Acceleration
33 min read - 06
SGLang: Comprehensive Guide to Structured Generation Language
SGLang (Structured Generation Language) is a revolutionary framework that transforms how developers interact with large language models (LLMs) and…
12 min read - 07
Building Neural Network Architectures Using Only ONNX
ONNX (Open Neural Network Exchange) is most commonly known as an export target — a format you dump a PyTorch or TensorFlow model into for deployment. But ONNX…
34 min read - 08
Training Computer Vision Models and Running Them with ONNX Runtime
Computer vision is one of the most vibrant areas of applied machine learning. Whether you are building an image classifier, a real-time object detector, a…
30 min read