Code

StableLM support#3586

Code

Merged

Galunid merged 33 commits intoggerganov:masterggerganov/llama.cpp:masterfrom Galunid:stablelm-supportGalunid/llama.cpp:stablelm-supportCopy head branch name to clipboardNov 14, 2023

+322-12

Commits

Commits on Oct 11, 2023

• Initial support - model loads, generates random stuff

  Galunid

  committed
  27467a5

  Copy full SHA for 27467a5

• Remove gpt neox references

  Galunid

  committed
  80b2e72

  Copy full SHA for 80b2e72

Commits on Oct 12, 2023

• Fixes suggested by @mmnga

  Galunid

  committed
  605e701

  Copy full SHA for 605e701

Commits on Oct 18, 2023

• Make stablelm conversion script use .safetensors

  Galunid

  committed
  1ee5cc3

  Copy full SHA for 1ee5cc3

Commits on Oct 19, 2023

• Remove random junk print

  Galunid

  committed
  f1dd430

  Copy full SHA for f1dd430

Commits on Oct 21, 2023

• Fix model conversion script

  Galunid

  committed
  4fbce39

  Copy full SHA for 4fbce39

Commits on Oct 22, 2023

• Use ggml_norm not ggml_rms_norm

  Galunid

  committed
  a71041a

  Copy full SHA for a71041a

• Fix rope parameters

  Galunid

  committed
  76b4495

  Copy full SHA for 76b4495

• Fix formatting in gguf.py

  Galunid

  committed
  e167ebc

  Copy full SHA for e167ebc

• Fix formatting in llama.cpp

  Galunid

  committed
  839a183

  Copy full SHA for 839a183

• Merge branch 'master' into stablelm-support

  Galunid

  committed
  db09c02

  Copy full SHA for db09c02

• batch.seq_id[j] -> batch.seq_id[j][0]

  Galunid

  committed
  0153376

  Copy full SHA for 0153376

• Fix added_tokens crashes

  Galunid

  committed
  e399050

  Copy full SHA for e399050

• Merge branch 'master' into stablelm-support

  Galunid

  committed
  cf5eff3

  Copy full SHA for cf5eff3

• Add tests for stablelm tokenizer

  Galunid

  committed
  a92fd2d

  Copy full SHA for a92fd2d

• Update readme with stablelm support

  Galunid

  committed
  d9c0332

  Copy full SHA for d9c0332

Commits on Oct 24, 2023

• Add special token handling to conver script

  Galunid

  committed
  fa2cd7e

  Copy full SHA for fa2cd7e

• Merge branch 'master' into stablelm-support

  Galunid

  committed
  27d0c11

  Copy full SHA for 27d0c11

• Prevent offloading of more than 33 layers

  Galunid

  committed
  51b3b56

  Copy full SHA for 51b3b56

Commits on Oct 26, 2023

• Make convert script with pytorch files

  Galunid

  committed
  a00bb06

  Copy full SHA for a00bb06

Commits on Nov 5, 2023

• Merge branch 'master' into stablelm-support

  Galunid

  committed
  8917767

  Copy full SHA for 8917767

• Update after ggerganov#3382

  Galunid

  committed
  c959376

  Copy full SHA for c959376

Commits on Nov 7, 2023

• Merge branch 'master' into stablelm-support

  Galunid

  committed
  698c945

  Copy full SHA for 698c945

• LLAMA_BACKEND_OFFLOAD* -> llama_backend_offload*

  Galunid

  committed
  4713a40

  Copy full SHA for 4713a40

Commits on Nov 9, 2023

• Merge branch 'master' into stablelm-support

  Galunid

  committed
  2f41552

  Copy full SHA for 2f41552

• Update conversion script to convert-hf-to-gguf.py

  Galunid

  committed
  6be3356

  Copy full SHA for 6be3356

Commits on Nov 10, 2023

• Use ggml_view_3d

  Galunid

  committed
  a371a8b

  Copy full SHA for a371a8b

• Cleanup for review

  Galunid

  committed
  e87d709

  Copy full SHA for e87d709

Commits on Nov 11, 2023

• Add vision model support

  Galunid

  committed
  9e035cd

  Copy full SHA for 9e035cd

Commits on Nov 12, 2023

• Duh - add llava in another place

  Galunid

  committed
  047032d

  Copy full SHA for 047032d

• Make qrot, krot contiguous

  Galunid

  committed
  be2ac38

  Copy full SHA for be2ac38

Commits on Nov 13, 2023

• Merge branch 'master' into stablelm-support

  Galunid

  committed
  beb17a7

  Copy full SHA for beb17a7

• Fix gguf post merge

  Galunid

  committed
  853fe04

  Copy full SHA for 853fe04