Research

Lab notes & recipes

Notes from inside the work. Reproducibility recipes, infrastructure write-ups, and the small things we needed to pin down to make the artifacts ship.

• Apr 27, 2026 · v37

  Deterministic Fine-Tuning on Dual MI100s

  What it actually takes to make a Qwen3.6-MoE LoRA train reproducibly on AMD MI100s. Three patches, one well-known invariant, and why the same inputs really do produce the same weights.

  #reproducibility #training #rocm #qwen #moe