[Bugfix] Add fix for JSON whitespace (#4189)

Co-authored-by: Ubuntu <ubuntu@ip-172-31-13-147.ec2.internal>

tests/entrypoints/test_openai_server.py

@@ -754,19 +754,20 @@ async def test_guided_choice_chat_logprobs(server, client: openai.AsyncOpenAI,
 
 
 async def test_response_format_json_object(server, client: openai.AsyncOpenAI):
-    resp = await client.chat.completions.create(
-        model=MODEL_NAME,
-        messages=[{
-            "role":
-            "user",
-            "content": ('what is 1+1? please respond with a JSON object, '
-                        'the format is {"result": 2}')
-        }],
-        response_format={"type": "json_object"})
-
-    content = resp.choices[0].message.content
-    loaded = json.loads(content)
-    assert loaded == {"result": 2}, loaded
+    for _ in range(2):
+        resp = await client.chat.completions.create(
+            model=MODEL_NAME,
+            messages=[{
+                "role":
+                "user",
+                "content": ('what is 1+1? please respond with a JSON object, '
+                            'the format is {"result": 2}')
+            }],
+            response_format={"type": "json_object"})
+
+        content = resp.choices[0].message.content
+        loaded = json.loads(content)
+        assert loaded == {"result": 2}, loaded
 
 
 async def test_guided_grammar(server, client: openai.AsyncOpenAI):

vllm/model_executor/guided_decoding/outlines_logits_processors.py

@@ -131,6 +131,11 @@ class CFGLogitsProcessor(BaseLogitsProcessor):
         fsm = CFGFSM(cfg, tokenizer)
         self.fsm = fsm
 
+    def init_state(self):
+        """Initialize state with a CFGFSM copy."""
+        super().init_state()
+        self.fsm = self.fsm.copy()
+
 
 @lru_cache
 def _adapt_tokenizer(tokenizer: PreTrainedTokenizerBase):