Ripple-65B-Preview / modeling_ricursive.py

Commit History

9 days ago

  1. Keep a crop-compatible default cache on Transformers 4⁠

    2f007aa

10 days ago

  1. Link regression tests to GPT-QModel main

    95ebf2a
  2. Link follow-up GPT-QModel unit tests

    fe71170
  3. Fix accumulation, tensor parallel heads, mask exclusions, and tied weights

    02996c3
  4. Point native regression coverage to GPT-QModel follow-up PR

    948fa26
  5. Fix Ripple generation, masks, cache allocation and training contracts

    bad8b01
  6. Fix cached prompt disambiguation and capacity-padded positions

    f8fb7f5

11 days ago

  1. Disambiguate exact-capacity static masks⁠

    Verified signature
    728d0de
  2. Handle capacity-padded masks for cached full prompts⁠

    Verified signature
    75fab16
  3. Fix warm-cache new token blocks in generation preparation⁠

    Verified signature
    d160eae

13 days ago

  1. Fix explicit position_ids generation on Transformers 4 (#8)⁠

    da94cc0
  2. Fix HF v5 inference (#1)⁠

    ebrahim-ricursive +1: Qubitium
    4f872a1

15 days ago

  1. Upload folder using huggingface_hub

    Verified signature
    a1407ea