Files
sglang/python/sglang/multimodal_gen/test/server/perf_baselines.json
2025-11-23 01:03:36 +08:00

729 lines
23 KiB
JSON

{
"metadata": {
"model": "Diffusion Server",
"hardware": "CI H100 80GB pool",
"description": "Reference numbers captured from the CI diffusion server baseline run"
},
"tolerances": {
"e2e": 0.1,
"denoise_stage": 0.05,
"non_denoise_stage": 0.4,
"denoise_step": 0.2,
"denoise_agg": 0.08
},
"improvement_reporting": {
"threshold": 0.2
},
"sampling": {
"step_fractions": [
0.0,
0.2,
0.4,
0.6,
0.8,
1.0
],
"warmup_requests": {
"text": 1,
"image_edit": 0
}
},
"scenarios": {
"qwen_image_t2i": {
"notes": "Single-image generation using the default prompt",
"expected_e2e_ms": 74500.0,
"expected_avg_denoise_ms": 422.42,
"expected_median_denoise_ms": 410.62,
"stages_ms": {
"InputValidationStage": 0.1,
"TextEncodingStage": 834.2,
"ConditioningStage": 0.1,
"TimestepPreparationStage": 10.6,
"LatentPreparationStage": 11.8,
"DenoisingStage": 21202.6,
"DecodingStage": 751.1
},
"denoise_step_ms": {
"0": 1077.77,
"1": 345.13,
"2": 413.8,
"3": 405.49,
"4": 408.14,
"5": 409.06,
"6": 408.85,
"7": 410.53,
"8": 407.51,
"9": 409.44,
"10": 408.65,
"11": 410.14,
"12": 411.74,
"13": 409.59,
"14": 409.17,
"15": 410.78,
"16": 410.66,
"17": 410.58,
"18": 411.27,
"19": 410.51,
"20": 409.03,
"21": 410.16,
"22": 409.42,
"23": 411.03,
"24": 410.18,
"25": 409.72,
"26": 410.26,
"27": 410.21,
"28": 410.71,
"29": 470.76,
"30": 411.06,
"31": 410.1,
"32": 410.55,
"33": 410.77,
"34": 410.74,
"35": 411.75,
"36": 410.78,
"37": 411.56,
"38": 410.85,
"39": 411.08,
"40": 411.12,
"41": 411.1,
"42": 411.09,
"43": 410.87,
"44": 411.37,
"45": 411.68,
"46": 411.0,
"47": 410.09,
"48": 412.72,
"49": 410.42
}
},
"flux_image_t2i": {
"stages_ms": {
"InputValidationStage": 0.03,
"TextEncodingStage": 81.49,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 2.43,
"LatentPreparationStage": 6.29,
"DenoisingStage": 8381.3,
"DecodingStage": 653.03
},
"denoise_step_ms": {
"0": 165.27,
"1": 58.88,
"2": 166.85,
"3": 166.51,
"4": 166.77,
"5": 167.55,
"6": 172.4,
"7": 167.77,
"8": 167.51,
"9": 167.22,
"10": 168.19,
"11": 167.74,
"12": 168.48,
"13": 168.08,
"14": 168.16,
"15": 167.15,
"16": 167.05,
"17": 169.27,
"18": 167.96,
"19": 167.74,
"20": 168.21,
"21": 167.07,
"22": 167.35,
"23": 167.06,
"24": 169.28,
"25": 169.41,
"26": 168.92,
"27": 167.59,
"28": 167.57,
"29": 170.42,
"30": 166.24,
"31": 168.33,
"32": 168.56,
"33": 168.62,
"34": 167.28,
"35": 167.12,
"36": 168.21,
"37": 168.78,
"38": 168.89,
"39": 167.74,
"40": 168.57,
"41": 167.89,
"42": 168.03,
"43": 167.61,
"44": 167.75,
"45": 168.03,
"46": 168.81,
"47": 168.29,
"48": 168.64,
"49": 168.78
},
"expected_e2e_ms": 9275.51,
"expected_avg_denoise_ms": 165.83,
"expected_median_denoise_ms": 169.33
},
"qwen_image_edit_ti2i": {
"notes": "single uploaded reference image, Qwen/Qwen-Image-Edit",
"expected_e2e_ms": 138500.0,
"expected_avg_denoise_ms": 720.0,
"expected_median_denoise_ms": 718.0,
"stages_ms": {
"InputValidationStage": 23,
"ImageEncodingStage": 1485.0,
"ImageVAEEncodingStage": 350.0,
"ConditioningStage": 0.13,
"TimestepPreparationStage": 13.78,
"LatentPreparationStage": 15.0,
"DenoisingStage": 36000.0,
"DecodingStage": 850.0
},
"denoise_step_ms": {
"0": 720.0,
"1": 720.0,
"2": 720.0,
"3": 720.0,
"4": 720.0,
"5": 720.0,
"6": 720.0,
"7": 720.0,
"8": 720.0,
"9": 720.0,
"10": 720.0,
"11": 720.0,
"12": 720.0,
"13": 720.0,
"14": 720.0,
"15": 720.0,
"16": 720.0,
"17": 720.0,
"18": 720.0,
"19": 720.0,
"20": 720.0,
"21": 720.0,
"22": 720.0,
"23": 720.0,
"24": 720.0,
"25": 720.0,
"26": 720.0,
"27": 720.0,
"28": 720.0,
"29": 720.0,
"30": 720.0,
"31": 720.0,
"32": 720.0,
"33": 720.0,
"34": 720.0,
"35": 720.0,
"36": 720.0,
"37": 720.0,
"38": 720.0,
"39": 720.0,
"40": 720.0,
"41": 720.0,
"42": 720.0,
"43": 720.0,
"44": 720.0,
"45": 720.0,
"46": 720.0,
"47": 720.0,
"48": 720.0,
"49": 720.0
}
},
"wan2_1_t2v_1.3b": {
"stages_ms": {
"InputValidationStage": 0.06,
"TextEncodingStage": 3595.12,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 2.39,
"LatentPreparationStage": 15.27,
"DenoisingStage": 91099.4,
"DecodingStage": 4330.65,
"per_frame_generation": null
},
"denoise_step_ms": {
"0": 2918.67,
"1": 1784.23,
"2": 1797.72,
"3": 1798.8,
"4": 1798.19,
"5": 1799.27,
"6": 1798.54,
"7": 1798.67,
"8": 1798.76,
"9": 1798.34,
"10": 1799.22,
"11": 1798.61,
"12": 1799.4,
"13": 1799.04,
"14": 1797.41,
"15": 1799.05,
"16": 1798.32,
"17": 1799.12,
"18": 1799.56,
"19": 1797.01,
"20": 1798.28,
"21": 1799.06,
"22": 1800.05,
"23": 1797.76,
"24": 1798.16,
"25": 1798.62,
"26": 1798.64,
"27": 1799.44,
"28": 1798.79,
"29": 1798.13,
"30": 1797.47,
"31": 1799.4,
"32": 1798.77,
"33": 1799.47,
"34": 1798.49,
"35": 1796.51,
"36": 1799.68,
"37": 1799.24,
"38": 1798.49,
"39": 1799.66,
"40": 1797.04,
"41": 1799.58,
"42": 1797.35,
"43": 1798.07,
"44": 1798.6,
"45": 1798.95,
"46": 1799.51,
"47": 1798.25,
"48": 1799.04,
"49": 1798.34
},
"expected_e2e_ms": 99083.75,
"expected_avg_denoise_ms": 1820.02,
"expected_median_denoise_ms": 1798.65
},
"wan2_2_ti2v_5b": {
"stages_ms": {
"InputValidationStage": 96.27,
"TextEncodingStage": 2238.81,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 2.39,
"LatentPreparationStage": 27.62,
"DenoisingStage": 134069.79,
"DecodingStage": 13559.79,
"per_frame_generation": null
},
"denoise_step_ms": {
"0": 3181.0,
"1": 2561.67,
"2": 2578.49,
"3": 2582.1,
"4": 2572.24,
"5": 2577.72,
"6": 2581.35,
"7": 2578.79,
"8": 2584.98,
"9": 2588.49,
"10": 2594.37,
"11": 2591.19,
"12": 2591.32,
"13": 2595.35,
"14": 2594.35,
"15": 2595.62,
"16": 2596.35,
"17": 2596.11,
"18": 2597.24,
"19": 2603.13,
"20": 2599.9,
"21": 2601.48,
"22": 2603.58,
"23": 2601.13,
"24": 2600.47,
"25": 2604.13,
"26": 2606.04,
"27": 2605.3,
"28": 2602.02,
"29": 2601.83,
"30": 2603.57,
"31": 2606.63,
"32": 2606.1,
"33": 2602.24,
"34": 2603.29,
"35": 2602.34,
"36": 2602.16,
"37": 2608.14,
"38": 2603.48,
"39": 2601.7,
"40": 2603.96,
"41": 2604.58,
"42": 2606.67,
"43": 2603.52,
"44": 2599.88,
"45": 2598.66,
"46": 2600.74,
"47": 2602.31,
"48": 2608.4,
"49": 2606.02
},
"expected_e2e_ms": 150004.2,
"expected_avg_denoise_ms": 2608.84,
"expected_median_denoise_ms": 2601.59
},
"fastwan2_2_ti2v_5b": {
"stages_ms": {
"InputValidationStage": 88.86,
"TextEncodingStage": 2327.87,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 58.66,
"LatentPreparationStage": 28.55,
"DmdDenoisingStage": 4438.3,
"DecodingStage": 14177.77,
"per_frame_generation": null
},
"denoise_step_ms": {
"0": 2022.21,
"1": 1263.17,
"2": 1149.59
},
"expected_e2e_ms": 21133.36,
"expected_avg_denoise_ms": 1478.32,
"expected_median_denoise_ms": 1263.17
},
"fast_hunyuan_video": {
"stages_ms": {
"InputValidationStage": 0.09,
"TextEncodingStage": 845.64,
"ConditioningStage": 0.04,
"TimestepPreparationStage": 125.22,
"LatentPreparationStage": 29.34,
"DenoisingStage": 3860.64,
"DecodingStage": 2580.55
},
"denoise_step_ms": {
"0": 2063.08,
"1": 164.02,
"2": 406.99,
"3": 407.95,
"4": 407.51,
"5": 404.2
},
"expected_e2e_ms": 7487.87,
"expected_avg_denoise_ms": 642.29,
"expected_median_denoise_ms": 407.25
},
"wan2_2_i2v_a14b_2gpu": {
"stages_ms": {
"InputValidationStage": 59.33,
"TextEncodingStage": 6062.41,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 2.2,
"LatentPreparationStage": 8.93,
"ImageVAEEncodingStage": 2075.47,
"DenoisingStage": 382628.41,
"DecodingStage": 2820.89
},
"denoise_step_ms": {
"0": 31228.27,
"1": 7723.86,
"2": 7769.69,
"3": 7795.93,
"4": 7815.58,
"5": 7829.48,
"6": 7827.34,
"7": 7825.35,
"8": 7828.05,
"9": 7809.53,
"10": 7801.29,
"11": 7790.96,
"12": 7785.88,
"13": 7785.5,
"14": 7780.32,
"15": 55411.1,
"16": 7722.27,
"17": 7761.31,
"18": 7789.46,
"19": 7800.6,
"20": 7814.91,
"21": 7799.62,
"22": 7801.25,
"23": 7798.27,
"24": 7797.67,
"25": 7795.97,
"26": 7781.74,
"27": 7784.16,
"28": 7796.64,
"29": 7789.75,
"30": 7792.13,
"31": 7790.99,
"32": 7778.1,
"33": 7777.78,
"34": 7780.56,
"35": 7778.22,
"36": 7770.88,
"37": 7771.56,
"38": 7767.82,
"39": 7769.23
},
"expected_e2e_ms": 393606.77,
"expected_avg_denoise_ms": 9565.48,
"expected_median_denoise_ms": 7790.98
},
"wan2_1_i2v_14b_480P_2gpu": {
"stages_ms": {
"InputValidationStage": 32.94,
"TextEncodingStage": 2316.5,
"ImageEncodingStage": 3026.2,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 2.69,
"LatentPreparationStage": 9.73,
"ImageVAEEncodingStage": 2290.98,
"DenoisingStage": 385080.09,
"DecodingStage": 2984.69,
"per_frame_generation": null
},
"denoise_step_ms": {
"0": 8785.36,
"1": 7644.16,
"2": 7687.27,
"3": 7703.9,
"4": 7710.61,
"5": 7716.32,
"6": 7714.26,
"7": 7711.27,
"8": 7711.08,
"9": 7706.57,
"10": 7700.78,
"11": 7696.03,
"12": 7704.73,
"13": 7699.99,
"14": 7705.33,
"15": 7701.11,
"16": 7704.04,
"17": 7695.31,
"18": 7693.63,
"19": 7686.34,
"20": 7683.27,
"21": 7689.82,
"22": 7688.74,
"23": 7686.01,
"24": 7675.43,
"25": 7679.86,
"26": 7676.75,
"27": 7671.65,
"28": 7667.0,
"29": 7669.83,
"30": 7660.5,
"31": 7666.82,
"32": 7660.89,
"33": 7668.75,
"34": 7662.27,
"35": 7659.71,
"36": 7661.36,
"37": 7664.87,
"38": 7666.93,
"39": 7661.05,
"40": 7661.88,
"41": 7657.96,
"42": 7660.6,
"43": 7669.82,
"44": 7655.78,
"45": 7654.25,
"46": 7656.56,
"47": 7652.37,
"48": 7657.61,
"49": 7644.6
},
"expected_e2e_ms": 395758.23,
"expected_avg_denoise_ms": 7701.42,
"expected_median_denoise_ms": 7676.09
},
"wan2_1_i2v_14b_720P_2gpu": {
"stages_ms": {
"InputValidationStage": 53.67,
"TextEncodingStage": 2838,
"ImageEncodingStage": 3123.99,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 3.39,
"LatentPreparationStage": 6.68,
"ImageVAEEncodingStage": 2261.05,
"DenoisingStage": 386761.14,
"DecodingStage": 2968.35,
"per_frame_generation": null
},
"denoise_step_ms": {
"0": 10021.98,
"1": 7633.62,
"2": 7676.46,
"3": 7704.68,
"4": 7725.09,
"5": 7732.86,
"6": 7735.42,
"7": 7739.05,
"8": 7740.89,
"9": 7724.35,
"10": 7730.2,
"11": 7713.23,
"12": 7715.93,
"13": 7710.93,
"14": 7699.95,
"15": 7704.72,
"16": 7704.03,
"17": 7700.47,
"18": 7702.0,
"19": 7705.92,
"20": 7704.35,
"21": 7705.11,
"22": 7693.85,
"23": 7696.91,
"24": 7689.6,
"25": 7681.2,
"26": 7675.63,
"27": 7678.95,
"28": 7683.82,
"29": 7681.07,
"30": 7671.07,
"31": 7674.65,
"32": 7679.56,
"33": 7674.59,
"34": 7672.16,
"35": 7679.68,
"36": 7670.81,
"37": 7661.84,
"38": 7668.58,
"39": 7667.1,
"40": 7670.22,
"41": 7664.97,
"42": 7667.3,
"43": 7668.87,
"44": 7663.43,
"45": 7656.34,
"46": 7662.81,
"47": 7662.05,
"48": 7654.13,
"49": 7648.62
},
"expected_e2e_ms": 397541.45,
"expected_avg_denoise_ms": 7735.02,
"expected_median_denoise_ms": 7681.14
},
"wan2_2_t2v_a14b_2gpu": {
"stages_ms": {
"InputValidationStage": 0.09,
"TextEncodingStage": 2322.57,
"ConditioningStage": 0.03,
"TimestepPreparationStage": 2.29,
"LatentPreparationStage": 3.08,
"DenoisingStage": 79913.08,
"DecodingStage": 1339.58
},
"denoise_step_ms": {
"0": 19269.37,
"1": 691.64,
"2": 699.28,
"3": 696.55,
"4": 698.6,
"5": 704.56,
"6": 699.26,
"7": 700.84,
"8": 700.27,
"9": 704.15,
"10": 699.04,
"11": 704.79,
"12": 701.48,
"13": 707.24,
"14": 697.54,
"15": 698.89,
"16": 697.97,
"17": 699.34,
"18": 697.68,
"19": 697.42,
"20": 697.14,
"21": 700.14,
"22": 696.75,
"23": 702.36,
"24": 697.3,
"25": 703.97,
"26": 33676.93,
"27": 700.4,
"28": 703.68,
"29": 691.86,
"30": 706.1,
"31": 704.18,
"32": 700.34,
"33": 698.62,
"34": 698.66,
"35": 699.77,
"36": 700.96,
"37": 701.02,
"38": 703.98,
"39": 702.18
},
"expected_e2e_ms": 83595.94,
"expected_avg_denoise_ms": 1988.81,
"expected_median_denoise_ms": 700.2
},
"wan2_1_t2v_14b_2gpu": {
"stages_ms": {
"InputValidationStage": 0.05,
"TextEncodingStage": 2310.34,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 2.42,
"LatentPreparationStage": 27.7,
"DenoisingStage": 803631.52,
"DecodingStage": 8898.74,
"per_frame_generation": null
},
"denoise_step_ms": {
"0": 17347.88,
"1": 15956.93,
"2": 16027.54,
"3": 16054.15,
"4": 16081.46,
"5": 16062.7,
"6": 16058.56,
"7": 16057.58,
"8": 16061.04,
"9": 16120.97,
"10": 16036.84,
"11": 16019.6,
"12": 16042.29,
"13": 16039.87,
"14": 16063.0,
"15": 16036.16,
"16": 16079.82,
"17": 16019.7,
"18": 16061.5,
"19": 16039.95,
"20": 16009.42,
"21": 16051.01,
"22": 16039.31,
"23": 16048.22,
"24": 16071.41,
"25": 16078.75,
"26": 16061.78,
"27": 16018.39,
"28": 16041.44,
"29": 16039.64,
"30": 16041.89,
"31": 16039.6,
"32": 16038.97,
"33": 15999.48,
"34": 16019.93,
"35": 16040.27,
"36": 16020.3,
"37": 16039.38,
"38": 15999.4,
"39": 16022.15,
"40": 16042.32,
"41": 16016.62,
"42": 15998.92,
"43": 16041.48,
"44": 15999.63,
"45": 16003.21,
"46": 15995.91,
"47": 16023.52,
"48": 16016.64,
"49": 16019.6
},
"expected_e2e_ms": 814884.71,
"expected_avg_denoise_ms": 16062.92,
"expected_median_denoise_ms": 16039.62
}
}
}