diff --git a/python/sglang/multimodal_gen/runtime/entrypoints/openai/protocol.py b/python/sglang/multimodal_gen/runtime/entrypoints/openai/protocol.py index 67c16c727..8a22f6e57 100644 --- a/python/sglang/multimodal_gen/runtime/entrypoints/openai/protocol.py +++ b/python/sglang/multimodal_gen/runtime/entrypoints/openai/protocol.py @@ -80,7 +80,7 @@ class VideoGenerationsRequest(BaseModel): generator_device: Optional[str] = "cuda" # SGLang extensions num_inference_steps: Optional[int] = None - guidance_scale: Optional[float] = 1.0 + guidance_scale: Optional[float] = None guidance_scale_2: Optional[float] = None true_cfg_scale: Optional[float] = ( None # for CFG vs guidance distillation (e.g., QwenImage) diff --git a/python/sglang/multimodal_gen/test/server/perf_baselines.json b/python/sglang/multimodal_gen/test/server/perf_baselines.json index 3f58e3a5b..5de7aadbe 100644 --- a/python/sglang/multimodal_gen/test/server/perf_baselines.json +++ b/python/sglang/multimodal_gen/test/server/perf_baselines.json @@ -828,136 +828,136 @@ "wan2_1_t2v_1.3b_text_encoder_cpu_offload": { "stages_ms": { "InputValidationStage": 0.09, - "TextEncodingStage": 4005.3, - "ConditioningStage": 0.1, - "TimestepPreparationStage": 4.99, - "LatentPreparationStage": 5.49, - "DenoisingStage": 7932.25, - "DecodingStage": 902.53, + "TextEncodingStage": 2480.54, + "ConditioningStage": 0.07, + "TimestepPreparationStage": 3.73, + "LatentPreparationStage": 1.34, + "DenoisingStage": 12514.88, + "DecodingStage": 1147.6, "per_frame_generation": null }, "denoise_step_ms": { - "0": 964.7, - "1": 147.48, - "2": 191.94, - "3": 138.0, - "4": 149.09, - "5": 186.5, - "6": 126.5, - "7": 135.06, - "8": 167.13, - "9": 148.28, - "10": 130.8, - "11": 148.97, - "12": 164.98, - "13": 126.23, - "14": 140.33, - "15": 138.65, - "16": 136.43, - "17": 149.42, - "18": 157.08, - "19": 179.23, - "20": 154.54, - "21": 139.79, - "22": 153.94, - "23": 129.32, - "24": 122.91, - "25": 134.7, - "26": 133.6, - "27": 131.25, - "28": 122.71, - "29": 133.91, - "30": 134.5, - "31": 122.44, - "32": 123.12, - "33": 122.11, - "34": 151.54, - "35": 138.77, - "36": 149.28, - "37": 163.16, - "38": 126.12, - "39": 127.19, - "40": 145.81, - "41": 148.05, - "42": 153.0, - "43": 132.07, - "44": 148.57, - "45": 129.11, - "46": 129.98, - "47": 126.94, - "48": 130.1, - "49": 129.64 + "0": 487.21, + "1": 243.47, + "2": 244.28, + "3": 244.06, + "4": 244.77, + "5": 245.86, + "6": 245.38, + "7": 246.74, + "8": 246.28, + "9": 245.58, + "10": 245.6, + "11": 245.21, + "12": 245.08, + "13": 245.03, + "14": 245.53, + "15": 245.36, + "16": 246.17, + "17": 245.32, + "18": 244.37, + "19": 246.83, + "20": 245.87, + "21": 244.93, + "22": 245.11, + "23": 245.23, + "24": 245.76, + "25": 245.44, + "26": 246.47, + "27": 244.56, + "28": 244.76, + "29": 244.79, + "30": 244.76, + "31": 244.8, + "32": 245.11, + "33": 245.27, + "34": 245.37, + "35": 245.3, + "36": 244.84, + "37": 245.26, + "38": 245.38, + "39": 245.31, + "40": 244.7, + "41": 245.84, + "42": 245.66, + "43": 246.68, + "44": 245.38, + "45": 245.98, + "46": 246.02, + "47": 245.96, + "48": 245.31, + "49": 244.99 }, - "expected_e2e_ms": 12875.42, - "expected_avg_denoise_ms": 158.3, - "expected_median_denoise_ms": 138.32 + "expected_e2e_ms": 16161.11, + "expected_avg_denoise_ms": 250.18, + "expected_median_denoise_ms": 245.32 }, "wan2_1_t2v_1.3b_cfg_parallel": { "stages_ms": { - "InputValidationStage": 0.09, - "TextEncodingStage": 1834.21, - "ConditioningStage": 0.04, - "TimestepPreparationStage": 3.18, - "LatentPreparationStage": 1.27, - "DenoisingStage": 6807.91, - "DecodingStage": 891.98, + "InputValidationStage": 0.08, + "TextEncodingStage": 2700.44, + "ConditioningStage": 0.02, + "TimestepPreparationStage": 2.82, + "LatentPreparationStage": 2.0, + "DenoisingStage": 11640.75, + "DecodingStage": 890.63, "per_frame_generation": null }, "denoise_step_ms": { - "0": 847.76, - "1": 120.56, - "2": 173.14, - "3": 147.17, - "4": 118.31, - "5": 107.02, - "6": 111.75, - "7": 107.35, - "8": 108.29, - "9": 106.22, - "10": 108.6, - "11": 105.37, - "12": 111.32, - "13": 104.72, - "14": 104.12, - "15": 106.77, - "16": 104.02, - "17": 107.16, - "18": 106.08, - "19": 105.54, - "20": 105.53, - "21": 106.69, - "22": 106.55, - "23": 106.72, - "24": 106.66, - "25": 107.57, - "26": 106.68, - "27": 110.17, - "28": 108.17, - "29": 107.83, - "30": 106.11, - "31": 109.55, - "32": 134.08, - "33": 132.01, - "34": 107.83, - "35": 106.15, - "36": 106.1, - "37": 106.32, - "38": 106.08, - "39": 106.71, - "40": 109.31, - "41": 110.77, - "42": 107.72, - "43": 106.32, - "44": 106.3, - "45": 108.05, - "46": 106.09, - "47": 107.59, - "48": 106.98, - "49": 105.91 + "0": 1069.91, + "1": 211.32, + "2": 206.59, + "3": 208.12, + "4": 210.68, + "5": 210.28, + "6": 213.92, + "7": 211.25, + "8": 212.89, + "9": 205.35, + "10": 205.92, + "11": 208.99, + "12": 207.1, + "13": 208.1, + "14": 206.52, + "15": 205.5, + "16": 205.24, + "17": 204.93, + "18": 207.05, + "19": 203.78, + "20": 205.23, + "21": 203.87, + "22": 204.28, + "23": 203.8, + "24": 206.02, + "25": 207.2, + "26": 209.53, + "27": 207.46, + "28": 206.77, + "29": 208.14, + "30": 208.05, + "31": 208.78, + "32": 209.23, + "33": 209.72, + "34": 208.26, + "35": 208.55, + "36": 205.24, + "37": 204.96, + "38": 203.77, + "39": 210.2, + "40": 202.57, + "41": 204.77, + "42": 204.96, + "43": 203.8, + "44": 203.9, + "45": 204.49, + "46": 207.75, + "47": 209.09, + "48": 207.51, + "49": 207.38 }, - "expected_e2e_ms": 9549.87, - "expected_avg_denoise_ms": 125.6, - "expected_median_denoise_ms": 107.0 + "expected_e2e_ms": 15245.6, + "expected_avg_denoise_ms": 224.37, + "expected_median_denoise_ms": 207.15 }, "turbo_wan2_1_t2v_1.3b": { "stages_ms": {