Code

finetune: Add -ngl parameter#3762

Code

Merged

ggerganov merged 16 commits intoggerganov:masterggerganov/llama.cpp:masterfrom AndrewGodfrey:finetune_enableGpuAndrewGodfrey/llama.cpp:finetune_enableGpuCopy head branch name to clipboardNov 1, 2023

+108-17

Commits

Commits on Oct 24, 2023

• Add '-ngl' support to finetune.cpp

  AndrewGodfrey

  committed
  facb1a3

  Copy full SHA for facb1a3

• Add fprintf in ggml_cuda_op_add

  Show description for 4d452db
  AndrewGodfrey

  committed
  4d452db

  Copy full SHA for 4d452db

• Add 'finetune.sh', which currently fails when using GPU

  Show description for e1ebce0
  AndrewGodfrey

  committed
  e1ebce0

  Copy full SHA for e1ebce0

• tweak finetune.sh

  AndrewGodfrey

  committed
  1758d0a

  Copy full SHA for 1758d0a

• Suppress some warnings in ggml.c

  AndrewGodfrey

  committed
  9ea91ce

  Copy full SHA for 9ea91ce

• Add f16 implementation to ggml_compute_forward_add_f16_f32

  AndrewGodfrey

  committed
  19097c9

  Copy full SHA for 19097c9

• Add an f16 case to ggml_add_cast_impl and llama_build_lora_finetune_graphs

  AndrewGodfrey

  committed
  7cbf5b2

  Copy full SHA for 7cbf5b2

• finetune.sh: Edit comments

  AndrewGodfrey

  committed
  9587ab4

  Copy full SHA for 9587ab4

• Add "add_f16_f32_f32_cuda"

  AndrewGodfrey

  committed
  86ceda4

  Copy full SHA for 86ceda4

• Tweak an error message

  AndrewGodfrey

  committed
  81dabd8

  Copy full SHA for 81dabd8

Commits on Oct 25, 2023

• finetune.sh: Add an optional LLAMA_MODEL_DIR variable

  AndrewGodfrey

  committed
  6359c15

  Copy full SHA for 6359c15

• finetune.sh: Add an optional LLAMA_TRAINING_DIR variable

  AndrewGodfrey

  committed
  fe44ded

  Copy full SHA for fe44ded

Commits on Oct 28, 2023

• train : minor

  ggerganov

  authored
  c5c54d1

  Copy full SHA for c5c54d1

Commits on Oct 30, 2023

• tabs to spaces

  AndrewGodfrey

  committed
  998a548

  Copy full SHA for 998a548

• Merge branch 'master' of https://github.com/AndrewGodfrey/llama.cpp into finetune_enableGpu

  AndrewGodfrey

  committed
  653bc1c

  Copy full SHA for 653bc1c

Commits on Oct 31, 2023

• Merge branch 'master' of https://github.com/ggerganov/llama.cpp into finetune_enableGpu

  cebtenzzre

  committed
  3af7756

  Copy full SHA for 3af7756