fix(frontend): return HTTP 400 for out-of-vocabulary token_ids_logprob (#28088)

Signed-off-by: Ting Sun <suntcrick@gmail.com>

test/registered/openai_server/validation/test_request_length_validation.py

@@ -1,6 +1,7 @@
 import unittest
 
 import openai
+import requests
 
 from sglang.srt.utils import kill_process_tree
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
@@ -104,6 +105,70 @@ class TestRequestLengthValidation(CustomTestCase):
             str(cm.exception),
         )
 
+    def test_token_ids_logprob_out_of_vocabulary(self):
+        headers = {"Authorization": f"Bearer {self.api_key}"}
+        for token_ids_logprob in ([-1], [2_000_000_000]):
+            response = requests.post(
+                f"{self.base_url}/generate",
+                headers=headers,
+                json={
+                    "text": "hi",
+                    "sampling_params": {"max_new_tokens": 1},
+                    "return_logprob": True,
+                    "token_ids_logprob": token_ids_logprob,
+                },
+            )
+            self.assertEqual(response.status_code, 400)
+            self.assertIn("out-of-vocabulary", response.text)
+
+    def test_token_ids_logprob_rejects_nested_list(self):
+        # Nested lists are a batch-level wire format; a single request must
+        # pass a flat list of ints. A ragged nested list with in-vocab ids
+        # would otherwise crash the scheduler in the sampler gather.
+        headers = {"Authorization": f"Bearer {self.api_key}"}
+        for token_ids_logprob in ([[0]], [[0], [1, 2]]):
+            response = requests.post(
+                f"{self.base_url}/generate",
+                headers=headers,
+                json={
+                    "text": "hi",
+                    "sampling_params": {"max_new_tokens": 1},
+                    "return_logprob": True,
+                    "token_ids_logprob": token_ids_logprob,
+                },
+            )
+            self.assertEqual(response.status_code, 400)
+            self.assertIn("flat list of integers", response.text)
+
+    def test_token_ids_logprob_batch_with_one_oov(self):
+        headers = {"Authorization": f"Bearer {self.api_key}"}
+        response = requests.post(
+            f"{self.base_url}/generate",
+            headers=headers,
+            json={
+                "text": ["hi", "hi"],
+                "sampling_params": {"max_new_tokens": 1},
+                "return_logprob": True,
+                "token_ids_logprob": [[0], [2_000_000_000]],
+            },
+        )
+        self.assertEqual(response.status_code, 400)
+        self.assertIn("out-of-vocabulary", response.text)
+
+    def test_token_ids_logprob_valid(self):
+        headers = {"Authorization": f"Bearer {self.api_key}"}
+        response = requests.post(
+            f"{self.base_url}/generate",
+            headers=headers,
+            json={
+                "text": "hi",
+                "sampling_params": {"max_new_tokens": 1},
+                "return_logprob": True,
+                "token_ids_logprob": [0],
+            },
+        )
+        self.assertEqual(response.status_code, 200)
+
 
 if __name__ == "__main__":
     unittest.main()