LEARNING IN PUBLIC · ENGINEERING LOGS

Research,
made practical.

Research notes rewritten as practical English articles, plus first-hand investigations from edge AI, VLA models, inference performance, and production systems.

11 field notes Papers → Practice focus

RECOMMENDED READING PATH

Follow the ideas as they evolve.

FoundationQuantizationDeployment

Step 1Orientation

Quantization is a systems decision

A practical field guide to INT8, INT4, FP8, NVFP4, and the measurements that decide whether lower precision is actually better.

Step 2Foundation

Attention, viewed from the edge

The Transformer shape flow, why training parallelizes, and where autoregressive inference becomes a memory and latency problem.

Step 3Foundation

INT8 without folklore

Range mapping, calibration, partial quantization, and a production workflow for models that must remain accurate at the edge.

Step 4Activation quantization

SmoothQuant: moving the hard part offline

Why activation outliers break naive W8A8 quantization, and how equivalent scaling restores an efficient INT8 path.

Step 5Weight quantization

GPTQ: quantizing with second-order information

How Hessian-aware error compensation, lazy updates, and Cholesky reformulation made 3–4 bit LLM quantization practical.

Step 6Salience aware

AWQ: protect what activations reveal

An engineering explanation of salient channels, INT4 weight-only quantization, and why the runtime matters.

Step 7Kernel design

LUT-GEMM: low-bit weights without reconstructing them

How binary-coded weights become lookup-table indices—and when that changes decode latency rather than only checkpoint size.

Step 8Learned rotation

SpinQuant: rotate before you round

Why learned rotations tame activation outliers and make W4A4KV4 much less destructive than naive low-bit quantization.

Step 9Efficient rotation

ParoQuant: pairwise rotations that survive deployment

From learned Givens rotations to an INT4 TensorRT pipeline—and what my reduced-calibration experiment actually showed.

Step 10System design

VLA-Perf: How to reason about VLA inference

A practical map of compute, memory, network, and action-generation bottlenecks in Vision-Language-Action systems.

Step 11Debugging

TensorRT Edge-LLM: four fixes from controlled experiments

How I separated compiler fusion bugs, lossy AWQ repacking, and a silent InternVL3 export failure—and verified each fix on NVIDIA edge hardware.