Releases: epfl-dlab/transformers-CFG
Releases · epfl-dlab/transformers-CFG
v0.2.6
What's Changed
- Add missing init.py to transformers_cfg/cli to ensure proper module inclusion by @Saibo-creator in #90
- Fix cli module not found error by @Saibo-creator in #91
Full Changelog: v0.2.5...v0.2.6
v0.2.5
What's Changed
- Resolve Errors Preventing CLI from Running Correctly by @gfesatidis in #88
New Contributors
- @gfesatidis made their first contribution in #88
Full Changelog: v0.2.4...v0.2.5
v0.2.4
What's Changed
- Support for local instances of llama 3 and changes to support reusing trie and homomorphism by @MatthewChang in #74
- Support masking when embedding size is different from vocab size by @x0wllaar in #83
- test: Added new test file to validate case where model'embedding has mismatch with tokenizer's vocab by @Saibo-creator in #85
- Unified cli by @Saibo-creator in #86
New Contributors
Full Changelog: v0.2.3...v0.2.4
v0.2.3
What's Changed
- Refactor: Move can_stop and must_stop to AcceptState by @Saibo-creator in #79
- Switch to extracting version from init.py in GitHub workflow by @Saibo-creator in #80
- Use version from init.py in setup.py by @Saibo-creator in #81
- Remove profiling decorator from GrammarConstrainedLogitsProcessor by @Saibo-creator in #82
Full Changelog: v0.2.2...v0.2.3
v0.2.2
What's Changed
- remove misc files and style by @Saibo-creator in #19
- Hotfix v0.2.0 by @Saibo-creator in #20
- raise error when unicode mode used with model other than GPT2 by @Saibo-creator in #22
- Smiles chemical by @Saibo-creator in #23
- Dev by @Saibo-creator in #25
- Cal flow by @arinaruck in #26
- Adding pddl grammars by @arinaruck in #28
- Update logits_process.py by returning a modified copy of a logit tensor by @Saibo-creator in #27
- add accept.py in example, which are used to check if a sentence is accepted by a grammar by @Saibo-creator in #33
- Fix bug in parsing hex char escapes by @RadixSeven in #36
- add metrics to measure probability gain from constrained generation by @Saibo-creator in #38
- Fix unicode error by @Saibo-creator in #42
- Unicode escaping by @Saibo-creator in #39
- Benchmarking by @arinaruck in #41
- Update vocab_struct.py to fix llama3 tokenizer by @Saibo-creator in #44
- Update logging_config.py by @Saibo-creator in #51
- Custom json grammars by @arinaruck in #45
- Pipeline example by @arinaruck in #43
- fix: update func api in debugging_custom_grammar.md by @Saibo-creator in #56
- Refactor tokenization interface by @Saibo-creator in #65
- Add support phi,bart, t5, mistral by @Saibo-creator in #66
- Check for presence of pretokenizers key when checking model type by @MatthewChang in #61
- Add support phi bart t5 by @Saibo-creator in #68
- Add Meta-Llama-3 tokenizer support, enhance token handling and refactor tests structure by @Saibo-creator in #70
- Update latest news in README by @Saibo-creator in #71
- Add support for DeepSeek by @Saibo-creator in #73
- Add Support to Gemma2 by @fillassuncao in #75
- Update README and requirements: by @Saibo-creator in #77
New Contributors
- @arinaruck made their first contribution in #26
- @RadixSeven made their first contribution in #36
- @MatthewChang made their first contribution in #61
- @fillassuncao made their first contribution in #75
Full Changelog: v0.2.0...v0.2.1
What's Changed
- add GitHub Actions workflow for PyPI releases by @Saibo-creator in #78
Full Changelog: v0.2.1...v0.2.2
code snapshot for SketchGCD paper
v0.2.1-SketchGCD use server demo in README.md
v0.2.0
Delete .github/workflows/python-test.yml
v0.1.3
replace gpt2 by Mistral update version to 0.1.3