[Multimodal][Speculative Decoding]Eagle3 mm support, enablement on qwen3vl (#29594)

Signed-off-by: Tsai, Louie <louie.tsai@intel.com>
Signed-off-by: EanWang211123 <wangyiheng@sangfor.com.cn>
Co-authored-by: Louie Tsai <louie.tsai@intel.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>

tests/v1/e2e/test_spec_decode.py

@@ -283,6 +283,19 @@ def test_speculators_model_integration(
     ["model_setup", "mm_enabled", "enable_chunked_prefill"],
     [
         (("eagle3", "Qwen/Qwen3-8B", "AngelSlim/Qwen3-8B_eagle3", 1), False, False),
+        pytest.param(
+            (
+                "eagle3",
+                "Qwen/Qwen3-VL-8B-Instruct",
+                "taobao-mnn/Qwen3-VL-8B-Instruct-Eagle3",
+                1,
+            ),
+            False,
+            False,
+            marks=pytest.mark.skip(
+                reason="architecture of its eagle3 is LlamaForCausalLMEagle3"
+            ),
+        ),
         pytest.param(
             (
                 "eagle3",
@@ -352,6 +365,7 @@ def test_speculators_model_integration(
     ],
     ids=[
         "qwen3_eagle3",
+        "qwen3_vl_eagle3",
         "qwen2_5_vl_eagle3",
         "llama3_eagle",
         "llama3_eagle3",