From d8a32a6a87edba12e574aa88fc6b73ec0202b5a2 Mon Sep 17 00:00:00 2001
From: UsamaKenway <56207634+UsamaKenway@users.noreply.github.com>
Date: Sun, 22 Feb 2026 14:17:09 +0100
Subject: [PATCH] [Misc] Add Qwen2.5-Omni-3B model support to Gradio demo
 (#1382)

Signed-off-by: UsamaKenway <usamakenway@gmail.com>
---
 .../qwen2_5_omni/gradio_demo.py               | 65 ++++++++++---------
 1 file changed, 35 insertions(+), 30 deletions(-)

diff --git a/examples/online_serving/qwen2_5_omni/gradio_demo.py b/examples/online_serving/qwen2_5_omni/gradio_demo.py
index a6d3a67e8..4fdef5248 100644
--- a/examples/online_serving/qwen2_5_omni/gradio_demo.py
+++ b/examples/online_serving/qwen2_5_omni/gradio_demo.py
@@ -15,38 +15,43 @@ from PIL import Image
 
 SEED = 42
 
+DEFAULT_SAMPLING_PARAMS = {
+    "thinker": {
+        "temperature": 0.0,
+        "top_p": 1.0,
+        "top_k": -1,
+        "max_tokens": 2048,
+        "seed": SEED,
+        "detokenize": True,
+        "repetition_penalty": 1.1,
+    },
+    "talker": {
+        "temperature": 0.9,
+        "top_p": 0.8,
+        "top_k": 40,
+        "max_tokens": 2048,
+        "seed": SEED,
+        "detokenize": True,
+        "repetition_penalty": 1.05,
+        "stop_token_ids": [8294],
+    },
+    "code2wav": {
+        "temperature": 0.0,
+        "top_p": 1.0,
+        "top_k": -1,
+        "max_tokens": 2048,
+        "seed": SEED,
+        "detokenize": True,
+        "repetition_penalty": 1.1,
+    },
+}
+
 SUPPORTED_MODELS: dict[str, dict[str, Any]] = {
+    "Qwen/Qwen2.5-Omni-3B": {
+        "sampling_params": DEFAULT_SAMPLING_PARAMS,
+    },
     "Qwen/Qwen2.5-Omni-7B": {
-        "sampling_params": {
-            "thinker": {
-                "temperature": 0.0,
-                "top_p": 1.0,
-                "top_k": -1,
-                "max_tokens": 2048,
-                "seed": SEED,
-                "detokenize": True,
-                "repetition_penalty": 1.1,
-            },
-            "talker": {
-                "temperature": 0.9,
-                "top_p": 0.8,
-                "top_k": 40,
-                "max_tokens": 2048,
-                "seed": SEED,
-                "detokenize": True,
-                "repetition_penalty": 1.05,
-                "stop_token_ids": [8294],
-            },
-            "code2wav": {
-                "temperature": 0.0,
-                "top_p": 1.0,
-                "top_k": -1,
-                "max_tokens": 2048,
-                "seed": SEED,
-                "detokenize": True,
-                "repetition_penalty": 1.1,
-            },
-        },
+        "sampling_params": DEFAULT_SAMPLING_PARAMS,
     },
 }
 # Ensure deterministic behavior across runs.