EdgeRock / runtime /optimizations /b1_bf16_selected.py

Commit History

3 days ago

  1. Use BF16 state at every CUDA batch and stabilize physical-ring attention (#1)⁠

    rockmatrix-ai +1: hkust-aaron
    49ab3cd

5 days ago

  1. Use BF16 inference and align figures and results with the technical report

    Verified signature
    799fa1f