Commit History

  1. FIX

    sparkleman
    1fd8dc6
  2. FIX

    sparkleman
    c9f36eb
  3. UPDATE: g1a4-2.9b-20251118, g1a3-1.5b-20251015

    sparkleman
    22ed854

  1. UPDATE

    sparkleman
    7771b8d

  1. UPDATE: G1a for modelscope

    sparkleman
    f3d9166

  1. UPDATE: G1A 2.9B 1.5B 0.4B

    sparkleman
    efe8cd0

  1. UPDATE

    sparkleman
    40f9931

  1. UPDATE: G1A 0.3B

    sparkleman
    0a81e8c

  1. UPDATE

    sparkleman
    e599bbd
  2. UPDATE

    sparkleman
    2833293
  3. UPDATE

    sparkleman
    e5f4bff

  1. UPDATE: g1 2.9B

    sparkleman
    606a355

  1. UPDATE: uv frozen

    sparkleman
    cfe7b3b
  2. UPDATE

    sparkleman
    6448ad4
  3. UPDATE

    sparkleman
    a650820
  4. UPDATE

    sparkleman
    50e995e
  5. UPDATE

    sparkleman
    353b634
  6. UPDATE: test nvidia/cuda:12.9.0-cudnn-devel-ubuntu24.04

    sparkleman
    ea2288b

  1. UPDATE: G1 2.9 73%

    sparkleman
    11e205d

  1. UPDATE: g1 1.5B

    sparkleman
    6f79f37

  1. UPDATE: G1 2.9B 59%

    sparkleman
    c49d215
  2. UPDATE: [0] prefill

    sparkleman
    efc2316

  1. UPDATE: G1 1.5B 80%

    sparkleman
    2dd12cc

  1. UPDATE: G1 1.5B 62% & G1 2.9B 41%

    sparkleman
    944cd13

  1. UPDATE: G1 50%

    sparkleman
    51ee4dc

  1. UPDATE

    sparkleman
    ed0f7b4
  2. UPDATE: RWKV G1 2.9B 32%

    sparkleman
    684d4c2

  1. UPDATE: g1 0.4B

    sparkleman
    4350c55

  1. UPDATE

    sparkleman
    6b82cc0
  2. TRY max_split_size_mb:128

    sparkleman
    43cae0b
  3. UPDATEL: change modelscope loading order

    sparkleman
    3cb7f89
  4. UPDATE

    sparkleman
    96d9268
  5. UPDATE

    sparkleman
    8dbcd8c
  6. UPDATE

    sparkleman
    d3e033f
  7. UPDATE

    sparkleman
    7063b82
  8. UPDATE

    sparkleman
    1c4c774
  9. UPDATE

    sparkleman
    ade4f4a
  10. UPDATE: PYTORCH_CUDA_ALLOC_CONF=expandable_segments

    sparkleman
    95dc3fc
  11. UPDATE: merge modelscope

    sparkleman
    b76f7cc

  1. UPDATE: Torch memory

    sparkleman
    d055fef
  2. UPDATE: HF_HOME persist

    sparkleman
    9b15eee
  3. UPDATE: test cuda off

    sparkleman
    08af64e
  4. UPDATE: remove rwkv dependence

    sparkleman
    f17e87c
  5. UPDATE

    sparkleman
    3b5e872

  1. UPDATE: Change gpu state display

    sparkleman
    aeaf225
  2. UPDATE: RWKV7-G1-2.9B-16%trained-20250313-ctx4k

    sparkleman
    89db2d9

  1. UPDATE: log context

    sparkleman
    2792ede
  2. FIX: Remove print

    sparkleman
    a14fe00
  3. UPDATE: cache_word_list

    sparkleman
    4ed9fde
  4. UPDATE: Remove <think> tag in content & handle EOS token

    sparkleman
    94c4923