Commit History

Today

  1. Prompt as inputs (no baked questions), split Neural Engine model for iOS, card and code

    Verified signature
    7011ac6
  2. d1_ane_prompt_split_fp16: the Neural Engine export in 4 pieces (iOS's Neural Engine compiler can't take 3.1 GB at once), questions as inputs

    Verified signature
    94ea14c
  3. embed_tokens_fp16.npy: d1-2B's input embeddings, to build the prompt inputs

    Verified signature
    bc008ab
  4. d1_ane_prompt_fp16: Neural Engine layout, the questions as inputs

    Verified signature
    caa73de
  5. d1_gpu_prompt_fp16: the questions as inputs (prompt padded to 640 tokens)

    Verified signature
    2634400

Yesterday

  1. Card: new Neural Engine speed, energy and parity; code: MLP split, power.py

    Verified signature
    8a7b200
  2. d1_ane_fp16.aimodel: MLPs in 8 chunks, 1.4x faster on the Neural Engine (236 -> 171 ms)

    Verified signature
    126db3f
  3. code: coreai_backend downloads the models from this repo; add coreai_py.py (used by parity.py)

    Verified signature
    16c436e
  4. d1_ane_fp16.aimodel: Neural Engine, fp16

    Verified signature
    8f63ca0
  5. d1_gpu_fp16.aimodel: GPU, fp16

    Verified signature
    605e496
  6. Card, LFM Open License v1.0, calibration, prompt and code

    Verified signature
    7a102c1
  7. initial commit

    Verified signature
    733988c