[diffusion] fix: fix Flux.2 condition image resize (#14232)

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

python/sglang/multimodal_gen/test/server/perf_baselines.json

@@ -295,6 +295,73 @@
             "expected_avg_denoise_ms": 520.09,
             "expected_median_denoise_ms": 528.0
         },
+        "flux_2_ti2i": {
+            "stages_ms": {
+                "InputValidationStage": 99.82,
+                "TextEncodingStage": 519.88,
+                "ImageVAEEncodingStage": 254.56,
+                "ConditioningStage": 0.01,
+                "LatentPreparationStage": 12.4,
+                "TimestepPreparationStage": 2.71,
+                "DenoisingStage": 54705.41,
+                "DecodingStage": 311.13
+            },
+            "denoise_step_ms": {
+                "0": 1067.03,
+                "1": 271.58,
+                "2": 1073.07,
+                "3": 1071.93,
+                "4": 1100.0,
+                "5": 1102.28,
+                "6": 1088.3,
+                "7": 1089.09,
+                "8": 1086.95,
+                "9": 1089.33,
+                "10": 1089.28,
+                "11": 1096.51,
+                "12": 1098.88,
+                "13": 1080.84,
+                "14": 1098.44,
+                "15": 1100.88,
+                "16": 1086.83,
+                "17": 1090.58,
+                "18": 1096.35,
+                "19": 1086.25,
+                "20": 1082.71,
+                "21": 1097.6,
+                "22": 1098.72,
+                "23": 1100.9,
+                "24": 1099.02,
+                "25": 1101.52,
+                "26": 1098.75,
+                "27": 1101.41,
+                "28": 1091.75,
+                "29": 1087.2,
+                "30": 1101.33,
+                "31": 1098.14,
+                "32": 1100.14,
+                "33": 1098.91,
+                "34": 1100.05,
+                "35": 1099.12,
+                "36": 1100.22,
+                "37": 1103.29,
+                "38": 1092.79,
+                "39": 1086.59,
+                "40": 1094.81,
+                "41": 1105.6,
+                "42": 1100.54,
+                "43": 1099.95,
+                "44": 1096.5,
+                "45": 1086.69,
+                "46": 1095.85,
+                "47": 1092.85,
+                "48": 1086.17,
+                "49": 1099.67
+            },
+            "expected_e2e_ms": 56308.23,
+            "expected_avg_denoise_ms": 1077.26,
+            "expected_median_denoise_ms": 1096.5
+        },
         "flux_image_t2i_2_gpus": {
             "stages_ms": {
                 "InputValidationStage": 0.03,

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

@@ -125,7 +125,7 @@ class DiffusionServerArgs:
 class DiffusionSamplingParams:
     """Configuration for a single model/scenario test case."""
 
-    output_size: str = "1024x1024"  # output image dimensions (or video resolution)
+    output_size: str = ""
 
     # inputs and conditioning
     prompt: str | None = None  # text prompt for generation
@@ -310,6 +310,19 @@ ONE_GPU_CASES_B: list[DiffusionTestCase] = [
     #     warmup_edit=0,
     #     custom_validator="video",
     # ),
+    DiffusionTestCase(
+        "flux_2_ti2i",
+        DiffusionServerArgs(
+            model_path="black-forest-labs/FLUX.2-dev",
+            modality="image",
+            warmup_text=0,
+            warmup_edit=1,
+        ),
+        DiffusionSamplingParams(
+            prompt="Convert 2D style to 3D style",
+            image_path="https://github.com/lm-sys/lm-sys.github.io/releases/download/test/TI2I_Qwen_Image_Edit_Input.jpg",
+        ),
+    ),
     DiffusionTestCase(
         "fast_hunyuan_video",
         DiffusionServerArgs(