From ee624d0686d0d333d865cd9a35e30fcd6b6bbaa0 Mon Sep 17 00:00:00 2001
From: amy-why-3459 <wuhaiyan17@huawei.com>
Date: Tue, 10 Feb 2026 16:08:20 +0800
Subject: [PATCH] [BugFix] Fixed the issue where ignore_eos was not working.
 (#1286)

Signed-off-by: amy-why-3459 <wuhaiyan17@huawei.com>
---
 .../test_serving_chat_sampling_params.py      | 22 +++++++------------
 vllm_omni/entrypoints/openai/serving_chat.py  |  4 ++++
 2 files changed, 12 insertions(+), 14 deletions(-)

diff --git a/tests/entrypoints/openai_api/test_serving_chat_sampling_params.py b/tests/entrypoints/openai_api/test_serving_chat_sampling_params.py
index 58d4d253f..a015b9b5f 100644
--- a/tests/entrypoints/openai_api/test_serving_chat_sampling_params.py
+++ b/tests/entrypoints/openai_api/test_serving_chat_sampling_params.py
@@ -86,9 +86,13 @@ def mock_request():
     # OpenAI standard sampling fields
     request.temperature = None
     request.top_p = None
+    request.top_k = None
     request.max_tokens = None
+    request.min_tokens = None
     request.seed = None
+    request.ignore_eos = None
     request.stop = None
+    request.stop_token_ids = None
     request.frequency_penalty = None
     request.presence_penalty = None
     return request
@@ -106,9 +110,13 @@ def test_openai_sampling_fields_contains_expected_fields():
     expected_fields = {
         "temperature",
         "top_p",
+        "top_k",
         "max_tokens",
+        "min_tokens",
         "seed",
+        "ignore_eos",
         "stop",
+        "stop_token_ids",
         "frequency_penalty",
         "presence_penalty",
     }
@@ -236,20 +244,6 @@ def test_multiple_params_override_together(serving_chat, mock_request):
     assert comprehension_params.repetition_penalty == 1.05
 
 
-def test_yaml_custom_params_not_overridden_by_request(serving_chat, mock_request):
-    """Test that YAML custom params (top_k, repetition_penalty) are not affected."""
-    # Even if request has these attributes, they should not override YAML
-    # because they're not in _OPENAI_SAMPLING_FIELDS
-    mock_request.top_k = 100  # Not in allowlist
-    mock_request.repetition_penalty = 2.0  # Not in allowlist
-
-    result = serving_chat._build_sampling_params_list_from_request(mock_request)
-
-    comprehension_params = result[0]
-    assert comprehension_params.top_k == 1  # YAML default preserved
-    assert comprehension_params.repetition_penalty == 1.05  # YAML default preserved
-
-
 # =============================================================================
 # Tests for _apply_request_overrides
 # =============================================================================
diff --git a/vllm_omni/entrypoints/openai/serving_chat.py b/vllm_omni/entrypoints/openai/serving_chat.py
index cfee79157..c025c5f06 100644
--- a/vllm_omni/entrypoints/openai/serving_chat.py
+++ b/vllm_omni/entrypoints/openai/serving_chat.py
@@ -489,9 +489,13 @@ class OmniOpenAIServingChat(OpenAIServingChat, AudioMixin):
     _OPENAI_SAMPLING_FIELDS: set[str] = {
         "temperature",
         "top_p",
+        "top_k",
         "max_tokens",
+        "min_tokens",
         "seed",
+        "ignore_eos",
         "stop",
+        "stop_token_ids",
         "frequency_penalty",
         "presence_penalty",
     }