mostlyai-engine 2.7.0__tar.gz → 2.7.1__tar.gz

{mostlyai_engine-2.7.0 → mostlyai_engine-2.7.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: mostlyai-engine
-Version: 2.7.0
+Version: 2.7.1
 Summary: Synthetic Data Engine
 Project-URL: homepage, https://github.com/mostly-ai/mostlyai-engine
 Project-URL: repository, https://github.com/mostly-ai/mostlyai-engine

{mostlyai_engine-2.7.0 → mostlyai_engine-2.7.1}/mostlyai/engine/__init__.py

@@ -34,7 +34,7 @@ __all__ = [
     "TabularARGN",
     "LanguageModel",
 ]
-__version__ = "2.7.0"
+__version__ = "2.7.1"
 
 # suppress specific warning related to os.fork() in multi-threaded processes
 warnings.filterwarnings("ignore", category=DeprecationWarning, message=".*multi-threaded.*fork.*")

{mostlyai_engine-2.7.0 → mostlyai_engine-2.7.1}/mostlyai/engine/_language/xgrammar_utils.py

@@ -32,14 +32,17 @@ JSON_NULL = "null"
 
 
 def prepend_grammar_root_with_space(grammar: str) -> str:
-    # XGrammar always starts with "{" when enforcing JSON Schema
-    # training is done on strings like ` {...} {...}`
-    # later, generation is done with prompt like ` {...}`
-    # so the first output token must be a space
-    start_of_grammar = 'root ::= "{"'
-    start_of_grammar_with_space = 'root ::= " {"'
-    assert start_of_grammar in grammar
-    return grammar.replace(start_of_grammar, start_of_grammar_with_space)
+    # JSON-schema grammars start at "{". Training strings look like ` {...} {...}`
+    # and generation prompts look like ` {...}`, so the first output token must be a space.
+    # xgrammar <=0.2.3 emits `root ::= "{"`. xgrammar 0.2.7 emits `root ::= (("{"`.
+    replacements = (
+        ('root ::= "{"', 'root ::= " {"'),
+        ('root ::= (("{"', 'root ::= ((" {"'),
+    )
+    for start_of_grammar, start_of_grammar_with_space in replacements:
+        if start_of_grammar in grammar:
+            return grammar.replace(start_of_grammar, start_of_grammar_with_space)
+    raise AssertionError("unrecognized xgrammar root rule")
 
 
 def ensure_seed_can_be_tokenized(seed_data: pd.DataFrame, tokenizer: PreTrainedTokenizerBase) -> pd.DataFrame:

{mostlyai_engine-2.7.0 → mostlyai_engine-2.7.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "mostlyai-engine"
-version = "2.7.0"
+version = "2.7.1"
 description = "Synthetic Data Engine"
 requires-python = ">=3.11,<3.14"
 readme = "README.md"