Commit History

  1. v4: two-lane decoder (two pieces per infer call, grouped-query attention)

    Verified signature
    b8b7f8c
  2. Declare the base-model relation as quantized

    Verified signature
    c54339a
  3. Model card in the style of the MLX conversion's

    Verified signature
    f34b38d
  4. Confucius4-R2T2 converted for Core ML on the Apple Neural Engine (encoder fp16, decoder LUT8 with 128-row prefill and verify head)

    Verified signature
    d0e38c6
  5. initial commit

    Verified signature
    e1c4fea