Y Hacker News new | ask | show | jobs

user: gpjt created: 2009-01-12 karma: 1798 https://www.gilesthomas.com/ submissions: Why do OpenAI's GPT-2 weights beat mine? 4 points | 0 comments Benchmarking Qwen 3.6 35B MoE (3B active) on an RTX 3090 16 points | 0 comments Building intuition about LLM parameter counts 2 points | 0 comments Poppy the training box, part 1: the beginnings 3 points | 0 comments From bigrams to GPT-2, one component at a time (in Jax) 1 points | 0 comments Building a Jax training loop for an LLM training run 2 points | 0 comments Thoughts on Role Confusion 3 points | 0 comments 0 points | 0 comments 0 points | 0 comments Flax debugging: making a hash of things 2 points | 0 comments 0 points | 0 comments 0 points | 0 comments 10Gb/s Ethernet: switching to a Broadcom SFP+ module 195 points | 170 comments Jax: Commitment Issues 4 points | 0 comments 0 points | 0 comments Jax Back Ends and Devices 2 points | 0 comments 0 points | 0 comments 0 points | 0 comments Using Safetensors with Flax 2 points | 0 comments First Looking into Jax 3 points | 0 comments 0 points | 0 comments 0 points | 0 comments 0 points | 0 comments 0 points | 0 comments 10Gb/s Ethernet: using mini-heatsinks with a 10GBASE-T SFP+ module 3 points | 0 comments 10Gb/s Ethernet: what I did to get it working in my home 232 points | 177 comments 10Gb Ethernet: what I had to (re)learn 1 points | 1 comments 0 points | 0 comments 0 points | 0 comments