Install vLLM on DGX Spark
Step‑by‑step guide to download, compile and run vLLM on an NVIDIA DGX‑Spark system (CUDA 13.0, Python 3.12).
Floating point
A concrete overview of IEEE 754 representation: mantissa, exponent, precision, rounding errors. We dismantle the machine and look exactly at how numbers are stored in memory and why `0.1` doesn't really exist in binary.
