Outlines v0.0.9

What's Changed

• Fix ReAct example link path by @lukestanley in #220
• Fix typo in README.md by @eltociear in #235
• Make transformers dependency explicit by @rlouf in #240
• Explicitly list the packages in modules by @rlouf in #242
• Fix some typos and language by @jrysana in #254
• Fix non-matching function definition and call by @krasserm in #252
• Remove unused from typing import List in README.md by @alvarobartt in #256
• Fix some typos/language in regex and sequence by @jrysana in #257
• Allow users to choose devices for models by @BramVanroy in #264
• Escape curly braces in the regexes by @rlouf in #243
• Change default generate behavior from fixed seed to random seed by @mondaychen in #228
• Use FSMs for scanning during grammar-guided generation by @brandonwillard in #178
• Added a fix for mismatched logit and mask tensor shapes by @xaviruvpadhiyar98 in #236
• Misc. Refactoring and Link Updates by @brandonwillard in #279
• Fix missing spaces in Tokenizer.convert_token_to_string by @brandonwillard in #280
• More HF Llama tokenizer space fixes by @brandonwillard in #282
• Fix whitespace and control character handling in JSON guidance by @brandonwillard in #283
• Add py.typed and package data information by @brandonwillard in #294
• Example to generate dating app profiles- combines prompt templating with JSON generation by @veezbo in #287
• Refactor for Python 3.8-3.9 compatibility by @brandonwillard in #295
• Introduce Numba-based FSM utilities by @brandonwillard in #272
• Use device_map value when device is unspecified by @brandonwillard in #300
• Fix typos in model usage examples by @Barahlush in #307
• Modify JSON schema $ref resolution strategy to support resolving values internal to the schema by @mattkindy in #312
• Fix JSON schema generation to accept schemas without 'title' property by @mattkindy in #311
• Update resize padding multiple in test_transformers_logits_vocab_size by @brandonwillard in #313
• Avoid unhashable additional_special_tokens values in special_tokens_map by @brandonwillard in #314
• Remove unnecessary array allocations in generation process and enable caching by @brandonwillard in #308
• Allow configurable sampling steps by @brandonwillard in #318
• Fix some Sequence interface issues by @brandonwillard in #319
• Fix potential REDUCE issues in partial parse states by @brandonwillard in #320
• Add missing sample tests by @brandonwillard in #321
• Support recursive arrays in JSON when an item is an array by @AL-377 in #316
• Fix and simplify walk_fsm by @brandonwillard in #322
• Disable eager Numba FSM info construction step in parser by @brandonwillard in #323
• Add another missing walk_fsm condition by @brandonwillard in #324

New Contributors

• @lukestanley made their first contribution in #220
• @eltociear made their first contribution in #235
• @jrysana made their first contribution in #254
• @krasserm made their first contribution in #252
• @alvarobartt made their first contribution in #256
• @BramVanroy made their first contribution in #264
• @xaviruvpadhiyar98 made their first contribution in #236
• @veezbo made their first contribution in #287
• @Barahlush made their first contribution in #307
• @mattkindy made their first contribution in #312
• @AL-377 made their first contribution in #316

Full Changelog: 0.0.8...0.0.9