Code

add PLaMo model#3557

Code

Merged

ggerganov merged 19 commits intoggerganov:masterggerganov/llama.cpp:masterfrom okdshin:add_pfnet_plamo_13bokdshin/llama.cpp:add_pfnet_plamo_13bCopy head branch name to clipboardDec 24, 2023

+307-15

Commits

Commits on Dec 18, 2023

• add plamo mock

  okdshin

  committed
  feb0966

  Copy full SHA for feb0966

• add tensor loading

  okdshin

  committed
  4c585b4

  Copy full SHA for 4c585b4

• plamo convert

  okdshin

  committed
  b2330f5

  Copy full SHA for b2330f5

• update norm

  okdshin

  committed
  9d49236

  Copy full SHA for 9d49236

• able to compile

  okdshin

  committed
  4a3ef4f

  Copy full SHA for 4a3ef4f

• fix norm_rms_eps hparam

  okdshin

  committed
  a22040a

  Copy full SHA for a22040a

• runnable

  okdshin

  committed
  86d5348

  Copy full SHA for 86d5348

• use inp_pos

  okdshin

  committed
  f76fd39

  Copy full SHA for f76fd39

• seems ok

  okdshin

  committed
  ca8f698

  Copy full SHA for ca8f698

• update kqv code

  okdshin

  committed
  febc635

  Copy full SHA for febc635

• remove develop code

  okdshin

  committed
  907b921

  Copy full SHA for 907b921

• update README

  okdshin

  committed
  9339ffc

  Copy full SHA for 9339ffc

Commits on Dec 24, 2023

• shuffle attn_q.weight and attn_output.weight for broadcasting

  okdshin

  committed
  db1b18d

  Copy full SHA for db1b18d

• remove plamo_llm_build_kqv and use llm_build_kqv

  okdshin

  committed
  26340a1

  Copy full SHA for 26340a1

• fix style

  okdshin

  committed
  700f7c6

  Copy full SHA for 700f7c6

• Merge branch 'master' into add_pfnet_plamo_13b_2

  okdshin

  committed
  602cc71

  Copy full SHA for 602cc71

• update

  okdshin

  committed
  307481f

  Copy full SHA for 307481f

• llama : remove obsolete KQ_scale

  ggerganov

  committed
  eedd434

  Copy full SHA for eedd434

• plamo : fix tensor names for correct GPU offload

  ggerganov

  committed
  1949c95

  Copy full SHA for 1949c95