OpenArch — PyTorch implementations of modern LLM architectures

OpenArch is a hand-written PyTorch repository implementing modern open-source LLM architectures one model at a time, prioritizing readability and learning. It documents various attention types, normalization choices, and positional encodings; some implementations support forward passes.

• Raschka releases from-scratch reasoning model video course2026-10-26
• Developer releases manual transformer autograd in PyTorch2026-10-28
• Practical notes for GPU kernel dev setup2026-10-27
• Deepity releases C++/CUDA predictive coding library, edges JAX baseline on MNIST2026-10-27
• Developer proposes speculative decoding using 2-bit draft plus RAM residual2026-10-22