[Model] Support Step1 Model (#32511)

Signed-off-by: xieli <xieli@stepfun.com>

tests/models/registry.py

@@ -472,6 +472,9 @@ _TEXT_GENERATION_EXAMPLE_MODELS = {
         "ByteDance-Seed/Seed-OSS-36B-Instruct",
         trust_remote_code=True,
     ),
+    "Step1ForCausalLM": _HfExamplesInfo(
+        "stepfun-ai/Step-Audio-EditX", trust_remote_code=True
+    ),
     "SmolLM3ForCausalLM": _HfExamplesInfo("HuggingFaceTB/SmolLM3-3B"),
     "StableLMEpochForCausalLM": _HfExamplesInfo("stabilityai/stablelm-zephyr-3b"),
     "StableLmForCausalLM": _HfExamplesInfo("stabilityai/stablelm-3b-4e1t"),

tests/models/test_initialization.py

@@ -115,8 +115,11 @@ def can_initialize(
         # FIXME: A hack to bypass FA3 assertion because our CI's L4 GPU
         # has cc==8.9 which hasn't supported FA3 yet. Remove this hack when
         # L4 supports FA3.
+        # Step1ForCausalLM requires TRITON_ATTN for use_alibi_sqrt support.
         attention_config = (
-            {"backend": "TRITON_ATTN"} if model_arch == "GptOssForCausalLM" else None
+            {"backend": "TRITON_ATTN"}
+            if model_arch in ("GptOssForCausalLM", "Step1ForCausalLM")
+            else None
         )
         if model_arch == "WhisperForConditionalGeneration":
             m.setenv("VLLM_WORKER_MULTIPROC_METHOD", "spawn")