Code

support for llguidance grammars#10224

Code

Merged

ggerganov merged 34 commits intoggml-org:masterggml-org/llama.cpp:masterfrom mmoskal:llgmmoskal/llama.cpp:llgCopy head branch name to clipboardFeb 2, 2025

+1,555-9

Commits

Commits on Jan 25, 2025

• initial porting of previous LLG patch

  mmoskal

  committed
  76290d9

  Copy full SHA for 76290d9

• update for new APIs

  mmoskal

  committed
  f19655c

  Copy full SHA for f19655c

• build: integrate llguidance as an external project

  mmoskal

  committed
  f4dc4b8

  Copy full SHA for f4dc4b8

Commits on Jan 26, 2025

• use '%llguidance' as marker to enable llg lark syntax

  mmoskal

  committed
  afb6cac

  Copy full SHA for afb6cac

• add some docs

  mmoskal

  committed
  b5399d4

  Copy full SHA for b5399d4

• clarify docs

  mmoskal

  committed
  adc4aed

  Copy full SHA for adc4aed

• code style fixes

  mmoskal

  committed
  2a92bfb

  Copy full SHA for 2a92bfb

• remove llguidance.h from .gitignore

  mmoskal

  committed
  8cb12d4

  Copy full SHA for 8cb12d4

• fix tests when llg is enabled

  mmoskal

  committed
  de269a1

  Copy full SHA for de269a1

• pass vocab not model to llama_sampler_init_llg()

  mmoskal

  committed
  a7be666

  Copy full SHA for a7be666

• copy test-grammar-integration.cpp to test-llguidance.cpp

  mmoskal

  committed
  3675050

  Copy full SHA for 3675050

• clang fmt

  mmoskal

  committed
  58006dd

  Copy full SHA for 58006dd

• fix ref-count bug

  mmoskal

  committed
  036b91f

  Copy full SHA for 036b91f

• build and run test

  mmoskal

  committed
  f245ca2

  Copy full SHA for f245ca2

• gbnf -> lark syntax

  mmoskal

  committed
  16a5484

  Copy full SHA for 16a5484

• conditionally include llguidance test based on LLAMA_LLGUIDANCE flag

  mmoskal

  committed
  2937537

  Copy full SHA for 2937537

• rename llguidance test file to test-grammar-llguidance.cpp

  mmoskal

  committed
  c7ebf57

  Copy full SHA for c7ebf57

• add gh action for llg test

  mmoskal

  committed
  0a211fc

  Copy full SHA for 0a211fc

• align tests with LLG grammar syntax and JSON Schema spec

  mmoskal

  committed
  8e027f8

  Copy full SHA for 8e027f8

• llama_tokenizer() in fact requires valid utf8

  mmoskal

  committed
  ca88ce7

  Copy full SHA for ca88ce7

• update llg

  mmoskal

  committed
  44e1973

  Copy full SHA for 44e1973

• format file

  mmoskal

  committed
  c9e9853

  Copy full SHA for c9e9853

• add $LLGUIDANCE_LOG_LEVEL support

  mmoskal

  committed
  efc36c9

  Copy full SHA for efc36c9

• fix whitespace

  mmoskal

  committed
  08fefd1

  Copy full SHA for 08fefd1

• fix warning

  mmoskal

  committed
  1afc53a

  Copy full SHA for 1afc53a

• include <cmath> for INFINITY

  mmoskal

  committed
  00fcd98

  Copy full SHA for 00fcd98

• add final newline

  mmoskal

  committed
  437ff31

  Copy full SHA for 437ff31

Commits on Jan 29, 2025

• fail llama_sampler_init_llg() at runtime

  mmoskal

  committed
  5475357

  Copy full SHA for 5475357

• Link gbnf_to_lark.py script; fix links; refer to llg docs for lexemes

  mmoskal

  committed
  d06448a

  Copy full SHA for d06448a

Commits on Jan 30, 2025

• simplify #includes

  mmoskal

  committed
  59da969

  Copy full SHA for 59da969

• improve doc string for LLAMA_LLGUIDANCE

  mmoskal

  committed
  d59d939

  Copy full SHA for d59d939

Commits on Jan 31, 2025

• Merge branch 'master' into llg

  mmoskal

  committed
  6b2de55

  Copy full SHA for 6b2de55

• typo in merge

  mmoskal

  committed
  a049afb

  Copy full SHA for a049afb

• bump llguidance to 0.6.12

  mmoskal

  committed
  7057589

  Copy full SHA for 7057589