[diffusion] feat: support lightweight e2e warmup for benchmarking (#16213)

python/sglang/multimodal_gen/test/server/testcase_configs.py

@@ -151,6 +151,8 @@ class DiffusionServerArgs:
     ring_degree: int | None = None
     # LoRA
     lora_path: str | None = None  # LoRA adapter path (HF repo or local path)
+    # misc
+    enable_warmup: bool = False
 
     dit_layerwise_offload: bool = False
     enable_cache_dit: bool = False
@@ -359,6 +361,13 @@ ONE_GPU_CASES_A: list[DiffusionTestCase] = [
         ),
         T2I_sampling_params,
     ),
+    DiffusionTestCase(
+        "zimage_image_t2i_warmup",
+        DiffusionServerArgs(
+            model_path="Tongyi-MAI/Z-Image-Turbo", modality="image", enable_warmup=True
+        ),
+        T2I_sampling_params,
+    ),
     # === Text and Image to Image (TI2I) ===
     DiffusionTestCase(
         "qwen_image_edit_ti2i",