26 posts in total

---

2026

02-19

Pallas 101 - multi-backend kernel for JAX

02-07

5D parallelism in LLM training

01-25

Memory usage breakdown during Training

2025

12-22

JAX 101

12-05

Jeff Dean & Gemini team QA at NeurIPS ‘25

11-11

Pytorch Conference & Ray Summit 2025 summary

11-09

Intro to PPO in RL

11-08

Truncated Importance Sampling (TIS) in RL

09-19

speculative decoding 02

06-06

vLLM 05 - vLLM multi-modal support