fix: llama_grammar_accept_token arg order (abetlen#1649)

tc-wolf · web-flow · commit 5575fedd3314 · 2024-08-04T17:07:54.000-04:00
Old was:
llama_grammar_accept_token(ctx, grammar, token)

Now this is:
llama_grammar_accept_token(grammar, ctx, token)
diff --git a/llama_cpp/_internals.py b/llama_cpp/_internals.py
@@ -511,7 +511,7 @@ def sample_token(self, candidates: "_LlamaTokenDataArray") -> int:
     def grammar_accept_token(self, grammar: LlamaGrammar, token: int):
         assert self.ctx is not None
         assert grammar.grammar is not None
-        llama_cpp.llama_grammar_accept_token(self.ctx, grammar.grammar, token)
+        llama_cpp.llama_grammar_accept_token(grammar.grammar, self.ctx, token)
 
     def reset_timings(self):
         assert self.ctx is not None
