Align Operators

Align Rust eLLM operators with HuggingFace FP32 operator outputs at the operator/unit-test level. Use when comparing mathematical correctness for operators such as Linear, RMSNorm, SiLU, SwiGLU, Softmax, or Attention; generating synthetic inputs; checking FP32 determinism; or debugging shape, layout, transpose, reshape, or epsilon mismatches between Rust and Python implementations.

lucienhuangfu Updated

File contents

lucienhuangfu/ellm commit 9cb8582e0e

Frequently asked questions

npx get-skill add lucienhuangfu/align-operators