Morning Fold
en

Transformers

  1. AI & LLMs

    180B Parameters, 32GB of RAM: Pushing MoE Inference to the Limit with Transformers

    The author compares how llama.cpp, vLLM and Transformers place and move MoE experts when a model doesn't fit in memory. Streaming experts from SSD via Strata gets within 6% of Strata's token generation speed.

  2. AI & LLMs

    Transformers Model Integration Contracts

    A July 2026 change let UMT5's decoder attend to future tokens during training, and the bug stayed in main for almost two months. The post proposes transformers-mic, which pairs shrunk fixture models with recorded baselines to catch such bugs on PRs.

Add Morning Fold to your home screen

  1. Tap the Share button <svg viewBox="0 0 24 24" aria-hidden="true"><path d="M12 15V4M7.5 8.5 12 4l4.5 4.5M5 13v5.5A1.5 1.5 0 0 0 6.5 20h11a1.5 1.5 0 0 0 1.5-1.5V13"/></svg> at the bottom of Safari.
  2. Scroll down and tap Add to Home Screen.
  3. Tap Add. Morning Fold opens from its icon, full screen.

Create a free account

Morning Fold is free. Already have an account? Log in