[Model] Support Llama4 in vLLM (#16104)

tests/models/test_registry.py

@@ -23,6 +23,11 @@ from .registry import HF_EXAMPLE_MODELS
 
 @pytest.mark.parametrize("model_arch", ModelRegistry.get_supported_archs())
 def test_registry_imports(model_arch):
+
+    # Llama4ForCausalLM does not have a standalone model
+    if model_arch == "Llama4ForCausalLM":
+        return
+
     model_info = HF_EXAMPLE_MODELS.get_hf_info(model_arch)
     model_info.check_transformers_version(on_fail="skip")
 
@@ -91,8 +96,11 @@ def test_registry_is_pp(model_arch, is_pp, init_cuda):
 
 
 def test_hf_registry_coverage():
-    untested_archs = (ModelRegistry.get_supported_archs() -
-                      HF_EXAMPLE_MODELS.get_supported_archs())
+    untested_archs = set(ModelRegistry.get_supported_archs() -
+                         HF_EXAMPLE_MODELS.get_supported_archs())
+
+    # Llama4ForCausalLM does not have a standalone model
+    untested_archs.discard("Llama4ForCausalLM")
 
     assert not untested_archs, (
         "Please add the following architectures to "