From d3589a7251e4df6710e14ac55071585e80ae62c7 Mon Sep 17 00:00:00 2001 From: Mick Date: Sun, 16 Aug 2026 20:54:50 +0800 Subject: [PATCH] [diffusion] CI: tighten NVIDIA perf baselines (#35016) --- .../test/server/perf_baselines/5090.json | 58 +- .../test/server/perf_baselines/h100.json | 2715 +++++++++-------- .../test/server/test_server_common.py | 16 - 3 files changed, 1388 insertions(+), 1401 deletions(-) diff --git a/python/sglang/multimodal_gen/test/server/perf_baselines/5090.json b/python/sglang/multimodal_gen/test/server/perf_baselines/5090.json index 6979a9cbd..75f1e0ac2 100644 --- a/python/sglang/multimodal_gen/test/server/perf_baselines/5090.json +++ b/python/sglang/multimodal_gen/test/server/perf_baselines/5090.json @@ -2,8 +2,8 @@ "metadata": { "model": "Diffusion Server", "hardware": "CI RTX 5090 pool", - "description": "Reference numbers captured from real 5090 CI runner history (PR 29791 run 28492141274, job 84450974283).", - "last_updated": "2026-07-01" + "description": "Reference numbers tightened from RTX 5090 multimodal_gen CI run 31935214180 at source revision 2ee0d38a8553d186faad9c50893480e3c1d79270.", + "last_updated": "2026-08-16" }, "tolerances": { "long_term": { @@ -38,68 +38,68 @@ "flux_2_klein_base_image_t2i": { "stages_ms": { "DecodingStage": 18.19, - "DenoisingStage": 18213.48, + "DenoisingStage": 17612.9, "ImageVAEEncodingStage": 0.01, "InputValidationStage": 0.08, - "LatentPreparationStage": 4.87, + "LatentPreparationStage": 1.67, "TextEncodingStage": 104.46, "TimestepPreparationStage": 142.18 }, "denoise_step_ms": { - "0": 219.58, - "10": 359.79, - "20": 361.55, - "29": 361.99, - "39": 362.57, - "49": 362.89 + "0": 95.07, + "10": 344.0, + "20": 345.85, + "29": 346.2, + "39": 346.9, + "49": 347.31 }, - "expected_e2e_ms": 18641.09, - "expected_avg_denoise_ms": 357.93, - "expected_median_denoise_ms": 361.8, + "expected_e2e_ms": 18255.39, + "expected_avg_denoise_ms": 338.78, + "expected_median_denoise_ms": 345.95, "estimated_full_test_time_s": 94.0 }, "wan2_1_t2v_1.3b": { "stages_ms": { - "DecodingStage": 1202.43, + "DecodingStage": 495.01, "DenoisingStage": 21452.59, "InputValidationStage": 0.09, "LatentPreparationStage": 0.23, - "TextEncodingStage": 1210.22, + "TextEncodingStage": 1025.11, "TimestepPreparationStage": 3.91 }, "denoise_step_ms": { - "0": 480.39, + "0": 169.02, "10": 426.11, "20": 427.32, "29": 428.09, "39": 428.04, "49": 422.31 }, - "expected_e2e_ms": 23877.73, - "expected_avg_denoise_ms": 428.74, + "expected_e2e_ms": 23418.73, + "expected_avg_denoise_ms": 426.59, "expected_median_denoise_ms": 427.94, "estimated_full_test_time_s": 160.9 }, "zimage_image_t2i": { "stages_ms": { "DecodingStage": 7.11, - "DenoisingStage": 2229.48, + "DenoisingStage": 2178.8, "InputValidationStage": 0.04, "LatentPreparationStage": 0.17, - "TextEncodingStage": 252.73, - "TimestepPreparationStage": 38.57 + "TextEncodingStage": 252.2, + "TimestepPreparationStage": 36.04 }, "denoise_step_ms": { "0": 39.34, - "2": 277.35, - "3": 267.94, - "5": 279.01, - "6": 273.01, - "8": 275.12 + "2": 270.3, + "3": 263.6, + "5": 264.28, + "6": 265.1, + "8": 265.56 }, - "expected_e2e_ms": 2533.95, - "expected_avg_denoise_ms": 246.97, - "expected_median_denoise_ms": 273.01, + "expected_e2e_ms": 2492.09, + "expected_avg_denoise_ms": 241.53, + "expected_median_denoise_ms": 265.1, "estimated_full_test_time_s": 329.8 }, "lingbot_video_moe_t2v": { diff --git a/python/sglang/multimodal_gen/test/server/perf_baselines/h100.json b/python/sglang/multimodal_gen/test/server/perf_baselines/h100.json index 46b3fffcf..52845ec8f 100644 --- a/python/sglang/multimodal_gen/test/server/perf_baselines/h100.json +++ b/python/sglang/multimodal_gen/test/server/perf_baselines/h100.json @@ -2,8 +2,8 @@ "metadata": { "model": "Diffusion Server", "hardware": "CI H100 80GB pool", - "description": "Reference numbers captured from CI diffusion runs (latest source: PR 28127 run 27487958425, tightened only from prior baselines).", - "last_updated": "2026-06-14" + "description": "Reference numbers tightened from NVIDIA multimodal_gen CI at source revision 2ee0d38a8553d186faad9c50893480e3c1d79270.", + "last_updated": "2026-08-16" }, "tolerances": { "long_term": { @@ -38,65 +38,65 @@ "qwen_image_t2i": { "stages_ms": { "TextEncodingStage": 232.03, - "DenoisingStage": 12404.23, - "InputValidationStage": 0.05, - "LatentPreparationStage": 0.22, + "DenoisingStage": 12402.12, + "InputValidationStage": 0.04, + "LatentPreparationStage": 0.18, "TimestepPreparationStage": 14.99, "DecodingStage": 17.7 }, "denoise_step_ms": { - "0": 196.21, - "1": 242.25, + "0": 196.19, + "1": 242.18, "2": 243.58, - "3": 256.87, + "3": 256.73, "4": 249.73, - "5": 244.48, + "5": 244.33, "6": 244.27, "7": 253.29, "8": 245.72, - "9": 246.67, - "10": 245.49, + "9": 245.51, + "10": 244.71, "11": 247.89, "12": 248.23, - "13": 247.24, - "14": 247.57, + "13": 246.45, + "14": 245.96, "15": 248.01, "16": 246.63, "17": 248.05, - "18": 248.94, + "18": 247.4, "19": 247.0, "20": 248.84, "21": 248.0, - "22": 249.18, + "22": 246.75, "23": 247.89, - "24": 251.11, + "24": 250.98, "25": 247.79, "26": 247.83, - "27": 250.09, - "28": 249.67, + "27": 249.35, + "28": 249.43, "29": 248.21, - "30": 250.75, + "30": 248.37, "31": 249.95, "32": 248.65, "33": 247.7, - "34": 248.77, - "35": 250.16, + "34": 247.77, + "35": 249.55, "36": 247.89, "37": 248.77, - "38": 249.16, - "39": 250.0, + "38": 248.62, + "39": 248.54, "40": 247.02, "41": 247.97, - "42": 250.37, + "42": 249.28, "43": 247.83, "44": 247.66, "45": 248.01, - "46": 249.32, + "46": 248.85, "47": 247.37, "48": 248.95, "49": 248.9 }, - "expected_e2e_ms": 12799.17, + "expected_e2e_ms": 12770.15, "expected_avg_denoise_ms": 247.95, "expected_median_denoise_ms": 249.01, "estimated_full_test_time_s": 133.1 @@ -105,65 +105,65 @@ "stages_ms": { "InputValidationStage": 0.04, "TextEncodingStage": 457.66, - "LatentPreparationStage": 0.21, + "LatentPreparationStage": 0.15, "TimestepPreparationStage": 20.44, - "DenoisingStage": 10125.16, + "DenoisingStage": 9648.39, "DecodingStage": 13.71 }, "denoise_step_ms": { - "0": 208.21, - "1": 231.64, - "2": 232.46, - "3": 228.42, - "4": 227.95, - "5": 229.55, - "6": 235.94, - "7": 222.3, - "8": 224.02, - "9": 228.53, - "10": 197.15, + "0": 206.32, + "1": 201.5, + "2": 199.16, + "3": 200.37, + "4": 197.53, + "5": 217.91, + "6": 209.45, + "7": 198.87, + "8": 188.8, + "9": 191.42, + "10": 191.39, "11": 183.96, "12": 184.16, "13": 187.16, "14": 189.72, - "15": 190.24, + "15": 189.28, "16": 190.09, "17": 189.98, - "18": 190.28, + "18": 189.54, "19": 189.91, - "20": 199.31, + "20": 188.74, "21": 190.14, "22": 190.16, "23": 189.96, - "24": 189.65, - "25": 189.93, + "24": 187.38, + "25": 188.84, "26": 189.16, - "27": 190.09, + "27": 189.2, "28": 189.99, - "29": 203.0, + "29": 187.87, "30": 190.11, - "31": 189.97, + "31": 189.34, "32": 189.98, "33": 188.8, - "34": 189.1, + "34": 187.93, "35": 190.04, "36": 189.37, "37": 187.14, - "38": 188.47, + "38": 187.5, "39": 188.48, "40": 188.16, "41": 187.64, "42": 188.23, "43": 188.18, - "44": 189.71, + "44": 189.56, "45": 188.61, "46": 189.33, - "47": 190.56, + "47": 190.31, "48": 190.14, - "49": 189.98 + "49": 189.25 }, - "expected_e2e_ms": 10681.51, - "expected_avg_denoise_ms": 202.35, + "expected_e2e_ms": 10192.48, + "expected_avg_denoise_ms": 192.9, "expected_median_denoise_ms": 190.02, "estimated_full_test_time_s": 133.2 }, @@ -233,228 +233,228 @@ "flux_image_t2i": { "stages_ms": { "TimestepPreparationStage": 32.58, - "DenoisingStage": 6876.88, + "DenoisingStage": 6447.42, "DecodingStage": 10.31, "LatentPreparationStage": 0.19, - "TextEncodingStage": 22.38, - "InputValidationStage": 0.05 + "TextEncodingStage": 20.86, + "InputValidationStage": 0.04 }, "denoise_step_ms": { - "0": 45.19, + "0": 30.2, "1": 53.93, - "2": 135.69, - "3": 136.11, - "4": 135.51, - "5": 136.88, - "6": 138.7, - "7": 145.0, - "8": 139.34, - "9": 135.89, - "10": 137.33, - "11": 136.49, - "12": 136.15, - "13": 139.22, - "14": 139.91, - "15": 139.9, - "16": 138.58, - "17": 136.39, - "18": 137.38, - "19": 137.25, - "20": 139.79, - "21": 139.97, - "22": 139.4, - "23": 138.29, - "24": 138.59, - "25": 137.13, - "26": 136.88, - "27": 140.51, - "28": 139.88, - "29": 139.07, - "30": 139.07, - "31": 138.89, - "32": 138.84, - "33": 137.37, - "34": 139.55, - "35": 140.57, - "36": 139.26, - "37": 139.38, - "38": 138.18, - "39": 137.97, - "40": 137.53, - "41": 138.01, - "42": 138.87, - "43": 140.05, - "44": 139.59, - "45": 139.68, - "46": 139.42, - "47": 137.81, - "48": 138.16, - "49": 139.34 + "2": 127.34, + "3": 128.02, + "4": 128.91, + "5": 128.0, + "6": 128.32, + "7": 138.16, + "8": 136.44, + "9": 130.22, + "10": 129.23, + "11": 128.63, + "12": 128.02, + "13": 128.72, + "14": 132.61, + "15": 139.33, + "16": 133.27, + "17": 130.77, + "18": 128.99, + "19": 128.9, + "20": 128.22, + "21": 130.32, + "22": 135.04, + "23": 136.12, + "24": 132.18, + "25": 130.76, + "26": 128.97, + "27": 129.91, + "28": 128.91, + "29": 131.18, + "30": 134.12, + "31": 133.24, + "32": 131.29, + "33": 129.89, + "34": 130.16, + "35": 129.0, + "36": 130.22, + "37": 132.56, + "38": 132.41, + "39": 132.75, + "40": 130.44, + "41": 129.67, + "42": 129.95, + "43": 130.37, + "44": 131.32, + "45": 132.66, + "46": 131.96, + "47": 132.01, + "48": 130.08, + "49": 128.76 }, - "expected_e2e_ms": 7152.96, - "expected_avg_denoise_ms": 137.39, - "expected_median_denoise_ms": 138.77, + "expected_e2e_ms": 6693.01, + "expected_avg_denoise_ms": 128.85, + "expected_median_denoise_ms": 130.22, "estimated_full_test_time_s": 127.4 }, "flux_2_image_t2i": { "stages_ms": { - "TimestepPreparationStage": 20.47, - "DenoisingStage": 23650.96, + "TimestepPreparationStage": 15.77, + "DenoisingStage": 22276.29, "DecodingStage": 7.1, - "LatentPreparationStage": 1.17, - "TextEncodingStage": 375.31, + "LatentPreparationStage": 0.51, + "TextEncodingStage": 370.09, "InputValidationStage": 0.04, "ImageVAEEncodingStage": 0.01 }, "denoise_step_ms": { "0": 65.28, "1": 129.11, - "2": 488.44, - "3": 465.17, - "4": 478.82, - "5": 472.84, - "6": 475.45, - "7": 477.05, - "8": 469.78, - "9": 480.78, - "10": 466.21, - "11": 482.63, - "12": 468.39, - "13": 481.72, - "14": 468.31, - "15": 480.25, - "16": 468.87, - "17": 480.89, - "18": 471.87, - "19": 477.75, - "20": 470.42, - "21": 476.72, - "22": 475.24, - "23": 475.18, - "24": 477.48, - "25": 474.97, - "26": 477.79, - "27": 474.04, - "28": 479.14, - "29": 473.3, - "30": 478.18, - "31": 473.57, - "32": 479.55, - "33": 474.66, - "34": 476.27, - "35": 474.9, - "36": 476.92, - "37": 475.12, - "38": 477.88, - "39": 474.87, - "40": 476.03, - "41": 477.65, - "42": 475.38, - "43": 477.2, - "44": 473.63, - "45": 477.37, - "46": 472.12, - "47": 477.77, - "48": 476.58, - "49": 476.52 + "2": 440.87, + "3": 457.02, + "4": 436.07, + "5": 464.75, + "6": 433.23, + "7": 458.23, + "8": 434.45, + "9": 451.8, + "10": 440.74, + "11": 443.64, + "12": 446.63, + "13": 438.65, + "14": 451.32, + "15": 439.22, + "16": 451.34, + "17": 440.49, + "18": 448.93, + "19": 441.4, + "20": 447.31, + "21": 445.59, + "22": 443.46, + "23": 451.02, + "24": 439.56, + "25": 449.29, + "26": 440.48, + "27": 449.7, + "28": 440.8, + "29": 449.01, + "30": 445.41, + "31": 445.79, + "32": 446.72, + "33": 445.36, + "34": 446.45, + "35": 443.98, + "36": 448.91, + "37": 443.29, + "38": 448.99, + "39": 445.39, + "40": 447.16, + "41": 446.46, + "42": 447.81, + "43": 446.82, + "44": 445.7, + "45": 448.14, + "46": 445.92, + "47": 447.33, + "48": 444.8, + "49": 447.19 }, - "expected_e2e_ms": 24561.42, - "expected_avg_denoise_ms": 464.89, - "expected_median_denoise_ms": 472.74, + "expected_e2e_ms": 23139.56, + "expected_avg_denoise_ms": 436.76, + "expected_median_denoise_ms": 445.86, "estimated_full_test_time_s": 145.2 }, "flux_2_klein_image_t2i": { "stages_ms": { "DecodingStage": 6.41, "InputValidationStage": 0.03, - "DenoisingStage": 266.12, - "TextEncodingStage": 53.59, - "LatentPreparationStage": 0.5, - "TimestepPreparationStage": 19.97, + "DenoisingStage": 236.89, + "TextEncodingStage": 53.18, + "LatentPreparationStage": 0.3, + "TimestepPreparationStage": 18.21, "ImageVAEEncodingStage": 0.01 }, "denoise_step_ms": { - "0": 28.54, + "0": 25.99, "1": 12.61, - "2": 57.24, - "3": 64.49 + "2": 22.29, + "3": 57.75 }, - "expected_e2e_ms": 434.57, - "expected_avg_denoise_ms": 39.98, - "expected_median_denoise_ms": 39.47, + "expected_e2e_ms": 399.62, + "expected_avg_denoise_ms": 30.0, + "expected_median_denoise_ms": 24.14, "estimated_full_test_time_s": 120.5 }, "flux_2_klein_base_image_t2i": { "stages_ms": { "InputValidationStage": 0.03, - "TextEncodingStage": 53.32, + "TextEncodingStage": 53.19, "ImageVAEEncodingStage": 0.01, - "LatentPreparationStage": 1.14, - "TimestepPreparationStage": 20.37, - "DenoisingStage": 6704.85, + "LatentPreparationStage": 0.3, + "TimestepPreparationStage": 19.9, + "DenoisingStage": 5948.03, "DecodingStage": 7.23 }, "denoise_step_ms": { - "0": 72.0, - "1": 86.92, - "2": 130.18, - "3": 130.1, - "4": 131.2, - "5": 130.92, - "6": 130.31, - "7": 144.42, - "8": 136.01, - "9": 132.71, - "10": 130.96, - "11": 130.37, - "12": 130.77, - "13": 132.92, - "14": 138.92, - "15": 138.26, - "16": 132.61, - "17": 133.87, - "18": 133.08, - "19": 131.36, - "20": 132.41, - "21": 136.52, - "22": 136.33, - "23": 135.24, - "24": 134.33, - "25": 132.95, - "26": 132.93, - "27": 130.62, - "28": 133.91, - "29": 135.79, - "30": 135.71, - "31": 134.15, - "32": 133.72, - "33": 133.32, - "34": 132.37, - "35": 135.04, - "36": 135.56, - "37": 136.58, - "38": 135.37, - "39": 134.37, - "40": 135.29, - "41": 132.97, - "42": 132.97, - "43": 134.56, - "44": 136.54, - "45": 135.2, - "46": 135.27, - "47": 133.13, - "48": 131.98, - "49": 133.24 + "0": 40.52, + "1": 78.16, + "2": 114.67, + "3": 115.28, + "4": 114.79, + "5": 115.62, + "6": 114.56, + "7": 115.46, + "8": 130.25, + "9": 125.42, + "10": 117.8, + "11": 116.19, + "12": 115.4, + "13": 115.62, + "14": 116.24, + "15": 116.12, + "16": 128.02, + "17": 125.93, + "18": 121.26, + "19": 117.48, + "20": 115.7, + "21": 115.62, + "22": 115.84, + "23": 116.64, + "24": 122.84, + "25": 125.77, + "26": 122.1, + "27": 119.19, + "28": 117.26, + "29": 116.71, + "30": 117.05, + "31": 116.16, + "32": 119.32, + "33": 125.41, + "34": 121.84, + "35": 118.37, + "36": 118.44, + "37": 117.05, + "38": 116.31, + "39": 116.66, + "40": 118.87, + "41": 123.73, + "42": 122.76, + "43": 119.21, + "44": 118.9, + "45": 117.97, + "46": 115.87, + "47": 117.31, + "48": 119.12, + "49": 121.88 }, - "expected_e2e_ms": 6870.9, - "expected_avg_denoise_ms": 131.92, - "expected_median_denoise_ms": 133.4, + "expected_e2e_ms": 6112.68, + "expected_avg_denoise_ms": 116.49, + "expected_median_denoise_ms": 117.28, "estimated_full_test_time_s": 124.4 }, "flux_2_ti2i": { "stages_ms": { - "TextEncodingStage": 367.56, - "DenoisingStage": 46838.26, + "TextEncodingStage": 364.98, + "DenoisingStage": 44465.25, "InputValidationStage": 43.89, "LatentPreparationStage": 0.27, "TimestepPreparationStage": 10.54, @@ -462,218 +462,218 @@ "ImageVAEEncodingStage": 69.59 }, "denoise_step_ms": { - "0": 910.0, - "1": 148.53, - "2": 932.49, - "3": 930.25, - "4": 928.98, - "5": 928.58, - "6": 933.5, - "7": 934.46, - "8": 934.42, - "9": 935.13, - "10": 934.08, - "11": 934.95, - "12": 934.47, - "13": 935.96, - "14": 935.65, - "15": 937.07, - "16": 936.13, - "17": 932.69, - "18": 938.8, - "19": 937.48, - "20": 939.61, - "21": 938.2, - "22": 936.83, - "23": 939.99, - "24": 935.59, - "25": 934.87, - "26": 937.1, - "27": 939.72, - "28": 938.59, - "29": 938.43, - "30": 939.54, - "31": 938.4, - "32": 936.73, - "33": 936.87, - "34": 936.71, - "35": 937.35, - "36": 937.94, - "37": 936.72, - "38": 936.04, - "39": 937.37, - "40": 937.72, - "41": 939.13, - "42": 938.47, - "43": 939.74, - "44": 939.68, - "45": 939.95, - "46": 940.96, - "47": 939.01, - "48": 940.73, - "49": 940.47 + "0": 860.71, + "1": 33.46, + "2": 875.07, + "3": 888.59, + "4": 887.02, + "5": 889.13, + "6": 887.37, + "7": 884.63, + "8": 881.32, + "9": 885.66, + "10": 883.83, + "11": 887.14, + "12": 887.58, + "13": 890.45, + "14": 889.23, + "15": 889.8, + "16": 888.58, + "17": 887.73, + "18": 888.73, + "19": 890.37, + "20": 894.24, + "21": 893.01, + "22": 894.3, + "23": 894.57, + "24": 889.55, + "25": 889.1, + "26": 889.64, + "27": 889.56, + "28": 889.7, + "29": 889.88, + "30": 890.64, + "31": 890.0, + "32": 891.23, + "33": 890.55, + "34": 889.98, + "35": 891.36, + "36": 890.76, + "37": 889.85, + "38": 890.75, + "39": 892.18, + "40": 890.76, + "41": 891.59, + "42": 891.87, + "43": 890.92, + "44": 892.14, + "45": 891.43, + "46": 891.44, + "47": 891.68, + "48": 890.74, + "49": 890.5 }, - "expected_e2e_ms": 47799.39, - "expected_avg_denoise_ms": 920.57, - "expected_median_denoise_ms": 936.97, + "expected_e2e_ms": 45445.95, + "expected_avg_denoise_ms": 871.81, + "expected_median_denoise_ms": 889.93, "estimated_full_test_time_s": 168.9 }, "flux_2_ti2i_multi_image_cache_dit": { "stages_ms": { - "ImageVAEEncodingStage": 282.83, - "DenoisingStage": 26936.93, - "DecodingStage": 129.33, - "TextEncodingStage": 737.01, - "LatentPreparationStage": 0.84, - "TimestepPreparationStage": 20.57, + "ImageVAEEncodingStage": 156.01, + "DenoisingStage": 23101.61, + "DecodingStage": 21.86, + "TextEncodingStage": 374.3, + "LatentPreparationStage": 0.3, + "TimestepPreparationStage": 13.54, "InputValidationStage": 84.29 }, "denoise_step_ms": { - "0": 1846.55, - "1": 232.6, - "2": 1621.05, - "3": 1606.1, - "4": 1441.78, - "5": 59.2, - "6": 60.32, - "7": 235.36, - "8": 1443.21, - "9": 59.71, - "10": 59.79, - "11": 236.12, - "12": 1439.11, - "13": 59.97, - "14": 60.46, - "15": 239.29, - "16": 1441.32, - "17": 60.76, - "18": 60.68, - "19": 240.16, - "20": 1442.46, - "21": 60.14, - "22": 61.1, - "23": 239.26, - "24": 1443.28, - "25": 59.41, - "26": 60.74, - "27": 238.02, - "28": 1444.38, - "29": 59.09, - "30": 59.12, - "31": 241.69, - "32": 1441.99, - "33": 60.44, - "34": 61.93, - "35": 241.0, - "36": 1443.56, - "37": 60.55, - "38": 61.07, - "39": 238.11, - "40": 1443.33, - "41": 59.08, - "42": 60.43, - "43": 239.23, - "44": 1444.53, - "45": 61.77, - "46": 61.65, - "47": 239.84, - "48": 1443.89, - "49": 59.92 + "0": 1397.91, + "1": 45.33, + "2": 1413.73, + "3": 1417.16, + "4": 1411.26, + "5": 55.42, + "6": 54.37, + "7": 48.08, + "8": 1425.69, + "9": 52.9, + "10": 52.68, + "11": 45.65, + "12": 1419.88, + "13": 54.02, + "14": 53.46, + "15": 47.62, + "16": 1419.75, + "17": 53.05, + "18": 53.55, + "19": 47.92, + "20": 1422.82, + "21": 53.72, + "22": 53.98, + "23": 47.33, + "24": 1421.76, + "25": 53.62, + "26": 54.47, + "27": 47.17, + "28": 1423.94, + "29": 53.67, + "30": 54.58, + "31": 46.8, + "32": 1423.79, + "33": 53.51, + "34": 53.94, + "35": 47.69, + "36": 1425.71, + "37": 53.22, + "38": 54.28, + "39": 47.44, + "40": 1422.02, + "41": 53.09, + "42": 53.87, + "43": 46.69, + "44": 1421.83, + "45": 53.46, + "46": 52.79, + "47": 47.26, + "48": 1418.37, + "49": 52.67 }, - "expected_e2e_ms": 28591.93, - "expected_avg_denoise_ms": 533.42, - "expected_median_denoise_ms": 235.74, + "expected_e2e_ms": 24192.8, + "expected_avg_denoise_ms": 461.7, + "expected_median_denoise_ms": 53.69, "estimated_full_test_time_s": 148.6 }, "flux_image_t2i_2_gpus": { "stages_ms": { - "InputValidationStage": 0.04, + "InputValidationStage": 0.03, "TextEncodingStage": 28.24, "TimestepPreparationStage": 25.74, - "LatentPreparationStage": 0.25, - "DenoisingStage": 4994.21, + "LatentPreparationStage": 0.2, + "DenoisingStage": 4004.08, "DecodingStage": 11.98 }, "denoise_step_ms": { - "0": 73.63, - "1": 81.34, - "2": 99.66, - "3": 99.12, - "4": 99.9, - "5": 100.06, - "6": 100.02, - "7": 99.58, - "8": 99.56, - "9": 100.84, - "10": 100.78, - "11": 100.72, - "12": 100.77, - "13": 100.35, - "14": 100.09, - "15": 100.51, - "16": 99.86, - "17": 100.29, - "18": 100.72, - "19": 99.24, - "20": 100.42, - "21": 100.98, - "22": 100.47, - "23": 101.11, - "24": 100.19, - "25": 101.35, - "26": 100.53, - "27": 100.2, - "28": 99.65, - "29": 98.85, - "30": 99.6, - "31": 100.72, - "32": 100.29, - "33": 100.43, - "34": 101.03, - "35": 100.54, - "36": 99.26, - "37": 100.48, - "38": 100.88, - "39": 100.52, - "40": 100.48, - "41": 100.76, - "42": 100.87, - "43": 100.34, - "44": 100.68, - "45": 100.73, - "46": 99.13, - "47": 100.54, - "48": 100.57, - "49": 100.09 + "0": 42.52, + "1": 57.39, + "2": 80.17, + "3": 80.34, + "4": 80.43, + "5": 80.58, + "6": 80.46, + "7": 80.43, + "8": 80.73, + "9": 80.53, + "10": 80.45, + "11": 81.78, + "12": 83.73, + "13": 81.89, + "14": 81.46, + "15": 80.96, + "16": 80.88, + "17": 80.92, + "18": 80.73, + "19": 80.74, + "20": 80.74, + "21": 80.77, + "22": 81.64, + "23": 81.71, + "24": 83.15, + "25": 82.31, + "26": 82.06, + "27": 81.49, + "28": 80.85, + "29": 81.11, + "30": 81.28, + "31": 80.47, + "32": 80.64, + "33": 80.64, + "34": 80.81, + "35": 81.48, + "36": 81.86, + "37": 82.27, + "38": 82.13, + "39": 81.67, + "40": 81.7, + "41": 81.39, + "42": 81.22, + "43": 81.06, + "44": 80.77, + "45": 80.77, + "46": 81.37, + "47": 81.1, + "48": 81.91, + "49": 81.52 }, - "expected_e2e_ms": 5215.39, - "expected_avg_denoise_ms": 99.72, - "expected_median_denoise_ms": 100.52, + "expected_e2e_ms": 4244.27, + "expected_avg_denoise_ms": 79.98, + "expected_median_denoise_ms": 81.01, "estimated_full_test_time_s": 125.5 }, "zimage_image_t2i": { "stages_ms": { - "DecodingStage": 134.1, - "InputValidationStage": 0.04, - "DenoisingStage": 771.7, - "TextEncodingStage": 130.68, - "LatentPreparationStage": 0.16, - "TimestepPreparationStage": 27.67 + "DecodingStage": 6.23, + "InputValidationStage": 0.03, + "DenoisingStage": 745.43, + "TextEncodingStage": 129.49, + "LatentPreparationStage": 0.09, + "TimestepPreparationStage": 18.32 }, "denoise_step_ms": { - "0": 95.97, + "0": 25.35, "1": 29.8, - "2": 91.06, - "3": 91.41, - "4": 93.26, - "5": 90.11, - "6": 90.88, - "7": 91.93, - "8": 93.16 + "2": 89.97, + "3": 90.15, + "4": 89.95, + "5": 90.0, + "6": 90.31, + "7": 90.2, + "8": 91.01 }, - "expected_e2e_ms": 1068.66, - "expected_avg_denoise_ms": 85.29, - "expected_median_denoise_ms": 91.41, + "expected_e2e_ms": 906.14, + "expected_avg_denoise_ms": 82.61, + "expected_median_denoise_ms": 90.0, "estimated_full_test_time_s": 116.3 }, "zimage_image_t2i_fp8": { @@ -682,73 +682,73 @@ "TextEncodingStage": 131.15, "LatentPreparationStage": 0.18, "TimestepPreparationStage": 26.29, - "DenoisingStage": 959.66, - "DecodingStage": 125.62 + "DenoisingStage": 931.6, + "DecodingStage": 20.72 }, "denoise_step_ms": { - "0": 115.4, + "0": 53.93, "1": 56.32, - "2": 111.77, - "3": 110.02, - "4": 111.43, - "5": 111.32, - "6": 110.77, - "7": 110.53, - "8": 118.39 + "2": 110.1, + "3": 109.53, + "4": 108.29, + "5": 109.82, + "6": 108.71, + "7": 108.73, + "8": 112.84 }, - "expected_e2e_ms": 1247.73, - "expected_avg_denoise_ms": 106.22, - "expected_median_denoise_ms": 111.32, + "expected_e2e_ms": 1131.68, + "expected_avg_denoise_ms": 103.24, + "expected_median_denoise_ms": 108.73, "estimated_full_test_time_s": 123.7 }, "zimage_image_t2i_multi_lora": { "stages_ms": { - "InputValidationStage": 0.04, - "TextEncodingStage": 130.25, + "InputValidationStage": 0.03, + "TextEncodingStage": 129.63, "LatentPreparationStage": 0.12, - "TimestepPreparationStage": 26.5, - "DenoisingStage": 841.93, - "DecodingStage": 142.78 + "TimestepPreparationStage": 18.04, + "DenoisingStage": 749.82, + "DecodingStage": 6.6 }, "denoise_step_ms": { - "0": 102.18, + "0": 29.97, "1": 40.06, - "2": 99.39, - "3": 100.33, - "4": 103.95, - "5": 93.48, - "6": 99.6, - "7": 99.38, - "8": 99.77 + "2": 90.61, + "3": 92.34, + "4": 88.98, + "5": 90.13, + "6": 89.3, + "7": 89.93, + "8": 91.58 }, - "expected_e2e_ms": 1145.53, - "expected_avg_denoise_ms": 93.12, - "expected_median_denoise_ms": 99.6, + "expected_e2e_ms": 910.85, + "expected_avg_denoise_ms": 83.08, + "expected_median_denoise_ms": 89.93, "estimated_full_test_time_s": 162.1 }, "zimage_image_t2i_2_gpus": { "stages_ms": { "InputValidationStage": 0.04, - "TextEncodingStage": 256.67, - "LatentPreparationStage": 0.14, - "TimestepPreparationStage": 36.44, - "DenoisingStage": 525.39, - "DecodingStage": 10.18 + "TextEncodingStage": 255.84, + "LatentPreparationStage": 0.1, + "TimestepPreparationStage": 18.29, + "DenoisingStage": 462.13, + "DecodingStage": 6.05 }, "denoise_step_ms": { - "0": 40.09, + "0": 29.64, "1": 38.17, - "2": 54.71, - "3": 64.29, - "4": 65.0, - "5": 63.7, - "6": 64.61, - "7": 64.42, - "8": 64.35 + "2": 51.86, + "3": 54.69, + "4": 53.9, + "5": 55.3, + "6": 54.3, + "7": 54.42, + "8": 54.55 }, - "expected_e2e_ms": 1425.0, - "expected_avg_denoise_ms": 57.95, - "expected_median_denoise_ms": 64.35, + "expected_e2e_ms": 752.84, + "expected_avg_denoise_ms": 51.13, + "expected_median_denoise_ms": 54.3, "estimated_full_test_time_s": 75.0 }, "qwen_image_edit_ti2i": { @@ -756,7 +756,7 @@ "InputValidationStage": 24.78, "ImageEncodingStage": 714.55, "ImageVAEEncodingStage": 51.33, - "LatentPreparationStage": 0.16, + "LatentPreparationStage": 0.14, "TimestepPreparationStage": 11.8, "DenoisingStage": 31661.25, "DecodingStage": 18.88 @@ -795,36 +795,36 @@ "30": 638.16, "31": 634.96, "32": 636.42, - "33": 640.92, + "33": 638.75, "34": 638.18, "35": 637.46, "36": 641.18, "37": 634.96, "38": 640.21, - "39": 641.08, + "39": 639.08, "40": 636.65, - "41": 640.47, + "41": 639.68, "42": 638.16, "43": 636.12, "44": 637.03, "45": 640.74, "46": 638.69, - "47": 643.17, + "47": 639.58, "48": 636.95, "49": 632.72 }, - "expected_e2e_ms": 33010.17, + "expected_e2e_ms": 32832.9, "expected_avg_denoise_ms": 633.05, "expected_median_denoise_ms": 636.16, "estimated_full_test_time_s": 153.6 }, "joyai_image_edit_ti2i": { "stages_ms": { - "InputValidationStage": 21.56, - "ImageEncodingStage": 612.88, + "InputValidationStage": 16.43, + "ImageEncodingStage": 602.8, "ImageVAEEncodingStage": 67.95, "LatentPreparationStage": 0.17, - "TimestepPreparationStage": 15.19, + "TimestepPreparationStage": 9.53, "DenoisingStage": 26686.38, "DecodingStage": 14.27 }, @@ -877,9 +877,9 @@ }, "qwen_image_t2i_cache_dit_enabled": { "stages_ms": { - "InputValidationStage": 0.04, - "TextEncodingStage": 231.2, - "LatentPreparationStage": 0.16, + "InputValidationStage": 0.03, + "TextEncodingStage": 231.07, + "LatentPreparationStage": 0.14, "TimestepPreparationStage": 14.34, "DenoisingStage": 4216.22, "DecodingStage": 8.11 @@ -889,51 +889,51 @@ "1": 242.42, "2": 243.19, "3": 255.67, - "4": 59.95, - "5": 6.59, + "4": 59.42, + "5": 6.33, "6": 198.65, "7": 30.85, "8": 6.38, "9": 193.27, "10": 48.7, - "11": 6.4, - "12": 6.64, + "11": 6.29, + "12": 6.1, "13": 194.41, "14": 51.4, - "15": 6.56, - "16": 6.36, + "15": 6.45, + "16": 6.28, "17": 205.12, "18": 58.3, - "19": 6.55, - "20": 6.13, + "19": 6.48, + "20": 6.09, "21": 195.49, "22": 56.24, "23": 6.36, "24": 6.17, "25": 194.51, - "26": 56.45, - "27": 6.37, - "28": 6.24, + "26": 55.49, + "27": 6.18, + "28": 6.16, "29": 196.33, "30": 57.49, - "31": 6.6, - "32": 6.24, + "31": 6.4, + "32": 6.2, "33": 201.13, "34": 57.04, - "35": 6.58, - "36": 6.45, + "35": 6.32, + "36": 6.2, "37": 196.17, "38": 55.55, - "39": 6.48, - "40": 7.16, + "39": 6.17, + "40": 6.2, "41": 193.75, "42": 56.32, - "43": 6.43, - "44": 6.48, + "43": 6.27, + "44": 6.22, "45": 196.52, "46": 58.45, - "47": 6.55, - "48": 201.62, + "47": 6.34, + "48": 200.43, "49": 45.11 }, "expected_e2e_ms": 4531.7, @@ -943,82 +943,82 @@ }, "wan2_1_t2v_1.3b_teacache_enabled": { "stages_ms": { - "TextEncodingStage": 618.71, - "DenoisingStage": 3926.63, - "InputValidationStage": 0.05, - "LatentPreparationStage": 0.14, - "TimestepPreparationStage": 2.31, - "DecodingStage": 544.28 + "TextEncodingStage": 534.94, + "DenoisingStage": 3853.51, + "InputValidationStage": 0.04, + "LatentPreparationStage": 0.12, + "TimestepPreparationStage": 2.17, + "DecodingStage": 335.2 }, "denoise_step_ms": { "0": 80.49, - "1": 125.15, - "2": 144.75, - "3": 144.41, - "4": 143.82, - "5": 137.91, - "6": 33.18, + "1": 123.07, + "2": 143.29, + "3": 141.76, + "4": 141.29, + "5": 135.04, + "6": 27.55, "7": 102.12, - "8": 32.52, + "8": 27.15, "9": 104.07, - "10": 37.91, + "10": 29.37, "11": 105.12, - "12": 2.52, - "13": 31.09, + "12": 2.51, + "13": 26.99, "14": 106.91, - "15": 2.79, - "16": 30.23, + "15": 2.44, + "16": 27.16, "17": 95.25, - "18": 2.67, - "19": 29.76, + "18": 2.41, + "19": 26.7, "20": 110.46, - "21": 2.66, - "22": 29.96, + "21": 2.45, + "22": 26.9, "23": 108.86, "24": 2.7, - "25": 32.5, + "25": 28.71, "26": 108.62, - "27": 2.6, - "28": 29.72, + "27": 2.53, + "28": 27.12, "29": 107.12, - "30": 2.53, - "31": 30.15, + "30": 2.42, + "31": 26.77, "32": 112.05, "33": 2.5, - "34": 104.27, + "34": 96.98, "35": 38.24, - "36": 102.6, + "36": 100.31, "37": 37.63, - "38": 104.91, + "38": 100.09, "39": 32.24, - "40": 106.69, + "40": 98.56, "41": 38.64, - "42": 106.66, + "42": 99.47, "43": 38.09, - "44": 109.63, - "45": 145.34, - "46": 148.45, - "47": 143.63, - "48": 148.49, + "44": 97.99, + "45": 144.91, + "46": 144.79, + "47": 142.12, + "48": 143.1, "49": 150.48 }, - "expected_e2e_ms": 5161.84, - "expected_avg_denoise_ms": 78.45, - "expected_median_denoise_ms": 102.49, + "expected_e2e_ms": 4739.06, + "expected_avg_denoise_ms": 76.99, + "expected_median_denoise_ms": 98.28, "estimated_full_test_time_s": 126.0 }, "wan2_1_t2v_1.3b": { "stages_ms": { - "DecodingStage": 503.17, + "DecodingStage": 361.6, "InputValidationStage": 0.03, - "DenoisingStage": 7244.34, - "TextEncodingStage": 624.67, + "DenoisingStage": 7195.52, + "TextEncodingStage": 534.89, "LatentPreparationStage": 0.09, - "TimestepPreparationStage": 2.28 + "TimestepPreparationStage": 2.07 }, "denoise_step_ms": { "0": 92.19, - "1": 121.33, + "1": 121.24, "2": 136.86, "3": 137.63, "4": 138.47, @@ -1068,109 +1068,109 @@ "48": 137.42, "49": 137.45 }, - "expected_e2e_ms": 8629.02, - "expected_avg_denoise_ms": 144.79, - "expected_median_denoise_ms": 145.57, + "expected_e2e_ms": 8106.87, + "expected_avg_denoise_ms": 143.82, + "expected_median_denoise_ms": 144.57, "estimated_full_test_time_s": 129.3 }, "wan2_1_t2v_1.3b_cfg_parallel": { "stages_ms": { - "LatentPreparationStage": 0.29, + "LatentPreparationStage": 0.1, "InputValidationStage": 0.05, - "DenoisingStage": 3735.53, + "DenoisingStage": 3678.22, "DecodingStage": 404.36, - "TimestepPreparationStage": 3.66, - "TextEncodingStage": 652.39 + "TimestepPreparationStage": 1.99, + "TextEncodingStage": 326.35 }, "denoise_step_ms": { "0": 58.84, - "1": 64.87, - "2": 66.46, + "1": 26.67, + "2": 53.47, "3": 63.27, "4": 67.23, "5": 62.74, "6": 60.87, - "7": 74.0, - "8": 75.26, - "9": 76.17, - "10": 73.24, - "11": 76.09, - "12": 87.22, + "7": 72.76, + "8": 72.71, + "9": 73.83, + "10": 72.95, + "11": 73.1, + "12": 80.08, "13": 78.91, - "14": 76.16, - "15": 78.78, - "16": 75.38, - "17": 75.62, - "18": 75.29, - "19": 74.83, - "20": 74.68, + "14": 73.72, + "15": 77.3, + "16": 75.06, + "17": 73.17, + "18": 72.72, + "19": 74.27, + "20": 74.22, "21": 72.13, - "22": 73.38, - "23": 77.21, - "24": 77.8, - "25": 85.28, + "22": 73.34, + "23": 73.81, + "24": 74.0, + "25": 79.87, "26": 74.18, - "27": 78.97, - "28": 77.76, - "29": 76.05, + "27": 77.05, + "28": 76.2, + "29": 75.89, "30": 73.97, - "31": 75.86, - "32": 75.52, - "33": 75.77, - "34": 74.81, - "35": 76.4, - "36": 76.53, - "37": 77.94, - "38": 78.81, - "39": 78.26, - "40": 78.24, - "41": 77.43, - "42": 76.87, - "43": 76.96, - "44": 76.15, - "45": 76.11, - "46": 75.26, - "47": 76.75, - "48": 79.17, + "31": 72.66, + "32": 74.06, + "33": 74.21, + "34": 73.25, + "35": 73.97, + "36": 73.34, + "37": 76.6, + "38": 77.01, + "39": 77.91, + "40": 76.23, + "41": 76.24, + "42": 75.02, + "43": 75.3, + "44": 74.38, + "45": 74.1, + "46": 73.69, + "47": 74.09, + "48": 74.04, "49": 72.62 }, - "expected_e2e_ms": 4956.47, - "expected_avg_denoise_ms": 74.54, - "expected_median_denoise_ms": 75.03, + "expected_e2e_ms": 4435.63, + "expected_avg_denoise_ms": 73.51, + "expected_median_denoise_ms": 74.05, "estimated_full_test_time_s": 127.6 }, "turbo_wan2_1_t2v_1.3b": { "stages_ms": { - "InputValidationStage": 0.06, - "TextEncodingStage": 2508.95, - "TimestepPreparationStage": 73.51, - "LatentPreparationStage": 1.34, - "DmdDenoisingStage": 1285.25, - "DecodingStage": 805.04, + "InputValidationStage": 0.03, + "TextEncodingStage": 364.61, + "TimestepPreparationStage": 34.97, + "LatentPreparationStage": 0.11, + "DmdDenoisingStage": 294.63, + "DecodingStage": 260.15, "per_frame_generation": null }, "denoise_step_ms": { - "0": 897.62, - "1": 126.04, - "2": 126.52, - "3": 128.26 + "0": 103.03, + "1": 44.95, + "2": 72.14, + "3": 73.41 }, - "expected_e2e_ms": 4686.66, - "expected_avg_denoise_ms": 319.61, - "expected_median_denoise_ms": 127.39, + "expected_e2e_ms": 959.14, + "expected_avg_denoise_ms": 73.38, + "expected_median_denoise_ms": 72.78, "estimated_full_test_time_s": 124.7 }, "ltx_2_two_stage_t2v": { "stages_ms": { - "InputValidationStage": 0.04, - "TextEncodingStage": 775.78, - "LTX2TextConnectorStage": 10.72, + "InputValidationStage": 0.03, + "TextEncodingStage": 403.5, + "LTX2TextConnectorStage": 10.29, "LTX2HalveResolutionStage": 0.05, - "LTX2LoRASwitchStage": 2.8, - "LTX2SigmaPreparationStage": 0.17, - "TimestepPreparationStage": 11.39, + "LTX2LoRASwitchStage": 2.62, + "LTX2SigmaPreparationStage": 0.09, + "TimestepPreparationStage": 7.2, "LTX2AVLatentPreparationStage": 0.18, - "LTX2ImageEncodingStage": 0.02, + "LTX2ImageEncodingStage": 0.01, "LTX2AVDenoisingStage": 4521.8, "LTX2UpsampleStage": 2.52, "LTX2RefinementStage": 544.95, @@ -1179,12 +1179,12 @@ }, "denoise_step_ms": { "0": 116.82, - "1": 166.1, + "1": 160.19, "2": 113.57, "3": 116.56, "4": 115.93, - "5": 116.1, - "6": 116.65, + "5": 115.1, + "6": 116.31, "7": 115.37, "8": 109.93, "9": 116.34, @@ -1193,8 +1193,8 @@ "12": 114.94, "13": 120.27, "14": 117.81, - "15": 118.43, - "16": 118.83, + "15": 116.67, + "16": 114.69, "17": 107.31, "18": 115.91, "19": 114.51, @@ -1206,10 +1206,10 @@ "25": 106.07, "26": 114.65, "27": 115.08, - "28": 115.54, + "28": 114.2, "29": 114.29, "30": 115.5, - "31": 123.66, + "31": 114.95, "32": 104.07, "33": 103.24, "34": 104.41, @@ -1230,73 +1230,73 @@ "wan2_2_ti2v_5b": { "stages_ms": { "InputValidationStage": 706.07, - "TextEncodingStage": 328.44, - "LatentPreparationStage": 0.13, - "TimestepPreparationStage": 2.32, - "DenoisingStage": 15171.58, - "DecodingStage": 1100.28 + "TextEncodingStage": 327.65, + "LatentPreparationStage": 0.11, + "TimestepPreparationStage": 2.13, + "DenoisingStage": 12704.94, + "DecodingStage": 1098.67 }, "denoise_step_ms": { - "0": 183.83, - "1": 301.44, - "2": 300.81, - "3": 311.19, - "4": 305.2, - "5": 302.05, - "6": 307.57, - "7": 305.22, - "8": 299.57, - "9": 302.19, - "10": 306.41, - "11": 303.95, - "12": 304.63, - "13": 305.57, - "14": 304.51, - "15": 305.02, - "16": 305.67, - "17": 304.17, - "18": 305.43, - "19": 304.32, - "20": 305.91, - "21": 304.7, - "22": 302.28, - "23": 303.57, - "24": 305.08, - "25": 304.67, - "26": 304.03, - "27": 305.48, - "28": 306.68, - "29": 305.07, - "30": 304.83, - "31": 305.76, - "32": 304.22, - "33": 305.49, - "34": 306.05, - "35": 305.28, - "36": 306.09, - "37": 307.76, - "38": 305.25, - "39": 305.21, - "40": 306.17, - "41": 304.79, - "42": 305.64, - "43": 305.06, - "44": 306.74, - "45": 305.89, - "46": 304.28, - "47": 306.01, - "48": 304.94, - "49": 302.57 + "0": 152.87, + "1": 235.22, + "2": 252.15, + "3": 258.08, + "4": 260.61, + "5": 247.07, + "6": 249.12, + "7": 258.27, + "8": 262.45, + "9": 252.75, + "10": 255.25, + "11": 258.51, + "12": 256.13, + "13": 254.4, + "14": 250.52, + "15": 254.05, + "16": 258.13, + "17": 253.54, + "18": 253.11, + "19": 257.78, + "20": 257.34, + "21": 253.75, + "22": 251.11, + "23": 258.34, + "24": 259.23, + "25": 255.5, + "26": 254.41, + "27": 253.37, + "28": 258.43, + "29": 252.74, + "30": 255.85, + "31": 256.13, + "32": 256.95, + "33": 255.23, + "34": 256.28, + "35": 256.91, + "36": 257.58, + "37": 255.14, + "38": 256.66, + "39": 258.09, + "40": 256.6, + "41": 257.13, + "42": 257.36, + "43": 254.11, + "44": 258.23, + "45": 256.86, + "46": 256.56, + "47": 256.47, + "48": 257.43, + "49": 254.6 }, - "expected_e2e_ms": 17597.35, - "expected_avg_denoise_ms": 302.53, - "expected_median_denoise_ms": 305.07, + "expected_e2e_ms": 14853.56, + "expected_avg_denoise_ms": 253.29, + "expected_median_denoise_ms": 256.21, "estimated_full_test_time_s": 141.7 }, "qwen_image_edit_2509_ti2i": { "stages_ms": { - "ImageEncodingStage": 590.9, - "DenoisingStage": 38451.75, + "ImageEncodingStage": 587.89, + "DenoisingStage": 37539.33, "InputValidationStage": 98.87, "LatentPreparationStage": 0.15, "TimestepPreparationStage": 12.19, @@ -1304,55 +1304,55 @@ "ImageVAEEncodingStage": 91.27 }, "denoise_step_ms": { - "0": 717.34, - "1": 959.86, - "2": 964.82, - "3": 964.14, - "4": 964.39, - "5": 962.79, - "6": 959.88, - "7": 962.3, - "8": 959.08, - "9": 966.39, - "10": 968.95, - "11": 966.02, - "12": 963.19, - "13": 965.61, - "14": 964.04, - "15": 964.3, - "16": 964.12, - "17": 966.04, - "18": 967.82, - "19": 968.01, - "20": 968.46, - "21": 968.58, - "22": 967.91, - "23": 967.14, - "24": 966.89, - "25": 967.93, - "26": 964.2, - "27": 968.56, - "28": 966.23, - "29": 970.81, - "30": 968.49, - "31": 970.28, - "32": 968.94, - "33": 970.14, - "34": 968.76, - "35": 970.11, - "36": 968.54, - "37": 968.7, - "38": 971.43, - "39": 969.25 + "0": 699.03, + "1": 936.47, + "2": 939.29, + "3": 936.7, + "4": 937.41, + "5": 939.63, + "6": 938.12, + "7": 939.71, + "8": 939.43, + "9": 938.08, + "10": 940.05, + "11": 941.22, + "12": 940.93, + "13": 943.6, + "14": 942.94, + "15": 943.23, + "16": 943.51, + "17": 943.98, + "18": 944.47, + "19": 945.8, + "20": 947.21, + "21": 947.75, + "22": 946.88, + "23": 946.35, + "24": 950.18, + "25": 946.9, + "26": 946.87, + "27": 947.91, + "28": 948.72, + "29": 947.01, + "30": 946.16, + "31": 947.72, + "32": 950.85, + "33": 950.25, + "34": 949.38, + "35": 951.84, + "36": 949.58, + "37": 953.33, + "38": 946.98, + "39": 939.83 }, - "expected_e2e_ms": 39483.04, - "expected_avg_denoise_ms": 961.13, - "expected_median_denoise_ms": 967.48, + "expected_e2e_ms": 38630.0, + "expected_avg_denoise_ms": 938.38, + "expected_median_denoise_ms": 945.14, "estimated_full_test_time_s": 160.2 }, "qwen_image_layered_i2i": { "stages_ms": { - "QwenImageLayeredBeforeDenoisingStage": 800, + "QwenImageLayeredBeforeDenoisingStage": 144.26, "TimestepPreparationStage": 0.0, "DenoisingStage": 33064.04, "DecodingStage": 142.89 @@ -1362,7 +1362,7 @@ "1": 671.77, "2": 658.22, "3": 666.15, - "4": 669.94, + "4": 669.86, "5": 663.29, "6": 663.62, "7": 665.89, @@ -1383,30 +1383,30 @@ "22": 665.95, "23": 666.9, "24": 668.24, - "25": 669.13, - "26": 667.71, - "27": 672.43, - "28": 669.31, + "25": 666.29, + "26": 666.38, + "27": 665.91, + "28": 667.71, "29": 660.48, "30": 670.89, "31": 666.7, "32": 669.39, "33": 666.78, - "34": 670.27, + "34": 669.29, "35": 671.19, "36": 668.11, "37": 670.61, "38": 670.22, "39": 660.73, - "40": 674.31, - "41": 672.18, + "40": 674.04, + "41": 667.9, "42": 670.49, - "43": 669.57, - "44": 674.7, + "43": 667.54, + "44": 673.75, "45": 673.28, - "46": 672.62, + "46": 671.01, "47": 672.52, - "48": 673.85, + "48": 670.0, "49": 666.93 }, "expected_e2e_ms": 33451.98, @@ -1417,60 +1417,60 @@ "fastwan2_2_ti2v_5b": { "stages_ms": { "InputValidationStage": 400.0, - "TextEncodingStage": 329.33, + "TextEncodingStage": 327.82, "TimestepPreparationStage": 30.08, - "LatentPreparationStage": 0.12, - "DmdDenoisingStage": 270.39, - "DecodingStage": 1462.06 + "LatentPreparationStage": 0.1, + "DmdDenoisingStage": 252.46, + "DecodingStage": 1073.1 }, "denoise_step_ms": { "0": 35.0, - "1": 30.01, - "2": 114.63 + "1": 25.92, + "2": 112.67 }, - "expected_e2e_ms": 2680.2, - "expected_avg_denoise_ms": 88.05, - "expected_median_denoise_ms": 114.63, + "expected_e2e_ms": 2444.17, + "expected_avg_denoise_ms": 83.86, + "expected_median_denoise_ms": 112.67, "estimated_full_test_time_s": 125.2 }, "fast_hunyuan_video": { "stages_ms": { "InputValidationStage": 0.04, - "TextEncodingStage": 254.74, + "TextEncodingStage": 252.98, "TimestepPreparationStage": 28.98, "LatentPreparationStage": 0.13, - "DenoisingStage": 5812.04, + "DenoisingStage": 5229.75, "DecodingStage": 2888.31, "per_frame_generation": null }, "denoise_step_ms": { "0": 286.89, - "1": 284.01, - "2": 1090.2, - "3": 1094.58, - "4": 1087.51, - "5": 1090.66 + "1": 167.02, + "2": 1010.47, + "3": 1009.41, + "4": 1011.33, + "5": 1011.9 }, - "expected_e2e_ms": 9761.46, - "expected_avg_denoise_ms": 967.68, - "expected_median_denoise_ms": 1090.43, + "expected_e2e_ms": 9123.68, + "expected_avg_denoise_ms": 871.39, + "expected_median_denoise_ms": 1010.9, "estimated_full_test_time_s": 77.0 }, "wan2_2_i2v_a14b_2gpu": { "stages_ms": { - "InputValidationStage": 18.1, - "ImageVAEEncodingStage": 1630.35, + "InputValidationStage": 15.59, + "ImageVAEEncodingStage": 1484.2, "DecodingStage": 1114.63, - "LatentPreparationStage": 0.31, - "TextEncodingStage": 626.49, - "TimestepPreparationStage": 4.42, + "LatentPreparationStage": 0.21, + "TextEncodingStage": 327.94, + "TimestepPreparationStage": 2.47, "DenoisingStage": 68736.39 }, "denoise_step_ms": { "0": 519.5, - "1": 612.04, + "1": 588.09, "2": 1737.52, - "3": 1705.6, + "3": 1699.03, "4": 1711.89, "5": 1714.28, "6": 1711.38, @@ -1482,7 +1482,7 @@ "12": 1716.05, "13": 1718.95, "14": 1720.84, - "15": 1743.12, + "15": 1738.95, "16": 1715.18, "17": 1718.87, "18": 1712.45, @@ -1516,145 +1516,145 @@ "wan2_1_i2v_14b_480P_2gpu": { "stages_ms": { "InputValidationStage": 7.88, - "LatentPreparationStage": 0.19, - "ImageEncodingStage": 30.99, - "DenoisingStage": 33259.34, - "DecodingStage": 570.09, - "TimestepPreparationStage": 2.14, - "TextEncodingStage": 400.75, - "ImageVAEEncodingStage": 451.69 + "LatentPreparationStage": 0.1, + "ImageEncodingStage": 21.87, + "DenoisingStage": 32698.4, + "DecodingStage": 564.54, + "TimestepPreparationStage": 1.69, + "TextEncodingStage": 325.75, + "ImageVAEEncodingStage": 426.58 }, "denoise_step_ms": { "0": 511.75, - "1": 684.53, + "1": 654.08, "2": 655.82, - "3": 667.35, - "4": 669.71, - "5": 664.06, - "6": 667.2, - "7": 669.88, - "8": 662.88, - "9": 665.68, - "10": 665.83, - "11": 665.85, - "12": 663.76, - "13": 666.65, - "14": 663.7, - "15": 666.51, - "16": 667.55, - "17": 666.8, - "18": 670.57, - "19": 670.01, - "20": 667.19, - "21": 667.71, - "22": 666.87, - "23": 668.3, - "24": 668.23, - "25": 668.41, - "26": 668.06, - "27": 665.42, - "28": 671.21, - "29": 669.95, - "30": 667.59, - "31": 670.01, - "32": 667.18, - "33": 670.25, - "34": 669.73, - "35": 670.18, - "36": 671.23, - "37": 671.36, - "38": 670.11, - "39": 668.64, - "40": 668.9, - "41": 670.84, - "42": 667.72, - "43": 670.59, - "44": 670.31, - "45": 667.49, - "46": 668.46, - "47": 670.64, - "48": 666.37, - "49": 662.19 + "3": 657.74, + "4": 654.51, + "5": 659.63, + "6": 655.73, + "7": 659.26, + "8": 654.83, + "9": 657.98, + "10": 651.05, + "11": 656.71, + "12": 658.81, + "13": 661.11, + "14": 659.3, + "15": 660.24, + "16": 664.77, + "17": 661.9, + "18": 658.99, + "19": 658.88, + "20": 660.7, + "21": 662.66, + "22": 660.66, + "23": 656.54, + "24": 651.75, + "25": 656.65, + "26": 661.84, + "27": 660.45, + "28": 651.25, + "29": 649.15, + "30": 649.27, + "31": 657.23, + "32": 655.16, + "33": 656.84, + "34": 652.69, + "35": 656.92, + "36": 655.05, + "37": 653.9, + "38": 656.83, + "39": 656.98, + "40": 650.36, + "41": 649.44, + "42": 658.21, + "43": 653.31, + "44": 652.45, + "45": 649.64, + "46": 649.43, + "47": 650.64, + "48": 655.48, + "49": 651.17 }, - "expected_e2e_ms": 34741.5, - "expected_avg_denoise_ms": 665.03, - "expected_median_denoise_ms": 667.89, + "expected_e2e_ms": 34055.55, + "expected_avg_denoise_ms": 653.89, + "expected_median_denoise_ms": 656.68, "estimated_full_test_time_s": 243.0 }, "wan2_1_i2v_14b_720P_2gpu": { "stages_ms": { "InputValidationStage": 11.58, - "TextEncodingStage": 390.5, + "TextEncodingStage": 327.33, "DecodingStage": 1357.99, - "TimestepPreparationStage": 2.27, - "ImageEncodingStage": 22.34, - "LatentPreparationStage": 0.17, - "DenoisingStage": 98693.53, - "ImageVAEEncodingStage": 938.33 + "TimestepPreparationStage": 1.32, + "ImageEncodingStage": 22.1, + "LatentPreparationStage": 0.1, + "DenoisingStage": 96179.65, + "ImageVAEEncodingStage": 910.6 }, "denoise_step_ms": { - "0": 1520.16, - "1": 2022.96, - "2": 1995.39, - "3": 1991.55, - "4": 1991.27, - "5": 1992.33, - "6": 1995.09, - "7": 1991.43, - "8": 1996.42, - "9": 1995.96, - "10": 1979.55, - "11": 1996.81, - "12": 1995.44, - "13": 1997.31, - "14": 1998.39, - "15": 2001.51, - "16": 1996.36, - "17": 2004.52, - "18": 1998.34, - "19": 1996.5, - "20": 1989.38, - "21": 1999.1, - "22": 2001.81, - "23": 2003.79, - "24": 2000.16, - "25": 2001.36, - "26": 1997.9, - "27": 1998.86, - "28": 2001.61, - "29": 1984.19, - "30": 2001.64, - "31": 2001.72, - "32": 1999.12, - "33": 1999.24, - "34": 1999.12, - "35": 2000.89, - "36": 1997.08, - "37": 1997.03, - "38": 1998.96, - "39": 1986.55, - "40": 2000.45, - "41": 2000.86, - "42": 1997.49, - "43": 2000.21, - "44": 1995.28, - "45": 1998.78, - "46": 1997.67, - "47": 1997.41, - "48": 1990.79, - "49": 1970.68 + "0": 1443.24, + "1": 1953.87, + "2": 1936.47, + "3": 1928.05, + "4": 1921.58, + "5": 1921.31, + "6": 1923.37, + "7": 1927.16, + "8": 1925.09, + "9": 1931.73, + "10": 1928.81, + "11": 1925.07, + "12": 1929.74, + "13": 1929.83, + "14": 1929.4, + "15": 1929.98, + "16": 1931.9, + "17": 1930.28, + "18": 1932.82, + "19": 1930.82, + "20": 1936.68, + "21": 1933.39, + "22": 1935.74, + "23": 1933.21, + "24": 1935.07, + "25": 1934.7, + "26": 1935.59, + "27": 1931.25, + "28": 1933.49, + "29": 1933.91, + "30": 1935.7, + "31": 1938.71, + "32": 1937.04, + "33": 1936.53, + "34": 1938.01, + "35": 1936.17, + "36": 1938.22, + "37": 1935.6, + "38": 1936.37, + "39": 1937.19, + "40": 1935.86, + "41": 1941.13, + "42": 1940.62, + "43": 1936.95, + "44": 1936.68, + "45": 1931.72, + "46": 1936.47, + "47": 1936.04, + "48": 1932.97, + "49": 1924.05 }, - "expected_e2e_ms": 101519.43, - "expected_avg_denoise_ms": 1973.69, - "expected_median_denoise_ms": 1982.25, + "expected_e2e_ms": 98831.21, + "expected_avg_denoise_ms": 1923.51, + "expected_median_denoise_ms": 1933.7, "estimated_full_test_time_s": 248.9 }, "wan2_2_t2v_a14b_2gpu": { "stages_ms": { - "InputValidationStage": 0.05, + "InputValidationStage": 0.03, "TextEncodingStage": 314.11, - "LatentPreparationStage": 0.21, - "TimestepPreparationStage": 1.89, + "LatentPreparationStage": 0.14, + "TimestepPreparationStage": 1.36, "DenoisingStage": 81747.14, "DecodingStage": 784.99 }, @@ -1665,13 +1665,13 @@ "3": 2012.96, "4": 2020.39, "5": 2019.44, - "6": 2056.54, - "7": 2065.04, + "6": 2020.69, + "7": 2020.43, "8": 2020.14, - "9": 2052.19, - "10": 2020.63, - "11": 2075.95, - "12": 2020.56, + "9": 2020.23, + "10": 2020.08, + "11": 2020.65, + "12": 2020.28, "13": 2020.28, "14": 2019.65, "15": 2019.53, @@ -1688,95 +1688,95 @@ "26": 2084.71, "27": 2022.36, "28": 2031.39, - "29": 2048.32, + "29": 2042.74, "30": 2020.36, - "31": 2033.97, - "32": 2036.4, - "33": 2039.62, + "31": 2020.09, + "32": 2023.93, + "33": 2035.01, "34": 2045.06, "35": 2026.33, - "36": 2022.31, - "37": 2026.32, + "36": 2021.0, + "37": 2020.37, "38": 2020.56, "39": 2014.28 }, "expected_e2e_ms": 83297.78, "expected_avg_denoise_ms": 2029.88, - "expected_median_denoise_ms": 2026.87, + "expected_median_denoise_ms": 2020.71, "estimated_full_test_time_s": 204.3 }, "wan2_1_t2v_14b_2gpu": { "stages_ms": { - "TextEncodingStage": 648.05, + "TextEncodingStage": 325.79, "DecodingStage": 637.43, - "TimestepPreparationStage": 3.4, - "InputValidationStage": 0.05, - "LatentPreparationStage": 0.27, - "DenoisingStage": 23262.32 + "TimestepPreparationStage": 1.81, + "InputValidationStage": 0.04, + "LatentPreparationStage": 0.13, + "DenoisingStage": 23061.89 }, "denoise_step_ms": { "0": 307.85, - "1": 190.41, + "1": 143.3, "2": 490.26, - "3": 468.77, - "4": 479.98, - "5": 470.02, - "6": 481.03, - "7": 474.6, - "8": 480.52, - "9": 475.21, - "10": 466.27, - "11": 475.87, - "12": 476.53, - "13": 477.88, - "14": 476.46, - "15": 478.46, - "16": 475.68, - "17": 480.34, - "18": 473.5, - "19": 481.12, - "20": 469.81, - "21": 480.89, - "22": 473.67, - "23": 478.55, - "24": 475.17, - "25": 478.69, - "26": 476.86, - "27": 476.79, - "28": 478.15, + "3": 451.8, + "4": 477.94, + "5": 457.57, + "6": 471.22, + "7": 465.52, + "8": 466.47, + "9": 466.56, + "10": 463.7, + "11": 472.7, + "12": 461.24, + "13": 474.59, + "14": 458.77, + "15": 472.84, + "16": 463.12, + "17": 475.1, + "18": 461.02, + "19": 473.99, + "20": 463.62, + "21": 471.32, + "22": 464.93, + "23": 467.86, + "24": 467.44, + "25": 468.78, + "26": 468.15, + "27": 468.36, + "28": 466.14, "29": 469.87, - "30": 478.06, - "31": 478.14, - "32": 477.48, - "33": 478.13, - "34": 476.76, - "35": 476.79, - "36": 478.36, - "37": 476.93, - "38": 477.99, - "39": 469.25, - "40": 477.28, - "41": 478.29, - "42": 476.71, - "43": 477.91, - "44": 479.14, - "45": 477.02, - "46": 477.17, - "47": 478.45, - "48": 476.97, - "49": 464.05 + "30": 469.61, + "31": 467.89, + "32": 468.39, + "33": 467.53, + "34": 469.45, + "35": 463.33, + "36": 472.97, + "37": 465.35, + "38": 472.17, + "39": 465.02, + "40": 467.49, + "41": 468.55, + "42": 469.94, + "43": 469.35, + "44": 466.62, + "45": 468.37, + "46": 465.78, + "47": 466.77, + "48": 470.86, + "49": 459.09 }, - "expected_e2e_ms": 24749.27, - "expected_avg_denoise_ms": 464.73, - "expected_median_denoise_ms": 470.88, + "expected_e2e_ms": 24072.24, + "expected_avg_denoise_ms": 461.16, + "expected_median_denoise_ms": 467.69, "estimated_full_test_time_s": 173.2 }, "wan2_2_t2v_a14b_lora_2gpu": { "stages_ms": { "InputValidationStage": 0.03, - "TextEncodingStage": 616.22, - "LatentPreparationStage": 0.13, - "TimestepPreparationStage": 2.72, + "TextEncodingStage": 325.03, + "LatentPreparationStage": 0.12, + "TimestepPreparationStage": 1.16, "DenoisingStage": 57638.76, "DecodingStage": 823.49 }, @@ -1807,7 +1807,7 @@ "23": 1419.4, "24": 1417.73, "25": 1417.12, - "26": 1619.69, + "26": 1597.59, "27": 1598.62, "28": 1426.79, "29": 1423.25, @@ -1830,11 +1830,11 @@ "wan2_1_t2v_1_3b_lora_1gpu": { "stages_ms": { "InputValidationStage": 0.05, - "TextEncodingStage": 340.58, + "TextEncodingStage": 329.6, "LatentPreparationStage": 0.16, - "TimestepPreparationStage": 3.26, + "TimestepPreparationStage": 3.19, "DenoisingStage": 7328.48, - "DecodingStage": 563.07 + "DecodingStage": 364.86 }, "denoise_step_ms": { "0": 90.0, @@ -1848,7 +1848,7 @@ "8": 144.97, "9": 145.9, "10": 144.53, - "11": 145.21, + "11": 144.78, "12": 146.25, "13": 152.87, "14": 148.99, @@ -1856,7 +1856,7 @@ "16": 144.81, "17": 144.76, "18": 147.29, - "19": 151.36, + "19": 151.21, "20": 151.05, "21": 146.67, "22": 147.05, @@ -1888,7 +1888,7 @@ "48": 146.76, "49": 145.83 }, - "expected_e2e_ms": 8324.93, + "expected_e2e_ms": 8138.0, "expected_avg_denoise_ms": 146.48, "expected_median_denoise_ms": 147.68, "estimated_full_test_time_s": 129.6 @@ -1896,85 +1896,85 @@ "wan2_1_i2v_14b_lora_2gpu": { "stages_ms": { "InputValidationStage": 12.06, - "TextEncodingStage": 341.83, + "TextEncodingStage": 325.79, "DecodingStage": 1329.75, - "TimestepPreparationStage": 2.14, - "ImageEncodingStage": 33.35, - "LatentPreparationStage": 0.17, - "DenoisingStage": 98821.24, - "ImageVAEEncodingStage": 936.91 + "TimestepPreparationStage": 1.33, + "ImageEncodingStage": 22.9, + "LatentPreparationStage": 0.1, + "DenoisingStage": 95661.38, + "ImageVAEEncodingStage": 910.47 }, "denoise_step_ms": { - "0": 1536.32, - "1": 2013.2, - "2": 1986.65, - "3": 1975.67, - "4": 1976.62, - "5": 1978.03, - "6": 1977.88, - "7": 1979.89, - "8": 1981.97, - "9": 1978.5, - "10": 1982.51, - "11": 1982.54, - "12": 1984.15, - "13": 1987.31, - "14": 1983.76, - "15": 1988.0, - "16": 1988.23, - "17": 1982.14, - "18": 1983.82, - "19": 1986.63, - "20": 1987.46, - "21": 1984.61, - "22": 1985.56, - "23": 1987.15, - "24": 1986.5, - "25": 1986.78, - "26": 1989.74, - "27": 1987.21, - "28": 1986.19, - "29": 1988.72, - "30": 1991.59, - "31": 1985.31, - "32": 1985.79, - "33": 1986.29, - "34": 1989.01, - "35": 1988.83, - "36": 1986.41, - "37": 1990.63, - "38": 1989.91, - "39": 1987.18, - "40": 1986.08, - "41": 1991.05, - "42": 1983.31, - "43": 1984.62, - "44": 1984.96, - "45": 1984.37, - "46": 1981.33, - "47": 1980.99, - "48": 1982.94, - "49": 1967.75 + "0": 1438.56, + "1": 1942.94, + "2": 1931.24, + "3": 1922.22, + "4": 1923.67, + "5": 1921.44, + "6": 1919.48, + "7": 1922.9, + "8": 1923.55, + "9": 1929.76, + "10": 1926.19, + "11": 1922.45, + "12": 1927.88, + "13": 1927.85, + "14": 1922.85, + "15": 1927.86, + "16": 1926.21, + "17": 1927.53, + "18": 1927.63, + "19": 1927.91, + "20": 1922.41, + "21": 1927.57, + "22": 1923.68, + "23": 1922.88, + "24": 1921.76, + "25": 1922.2, + "26": 1923.58, + "27": 1920.5, + "28": 1924.61, + "29": 1923.23, + "30": 1924.05, + "31": 1921.47, + "32": 1920.61, + "33": 1921.39, + "34": 1919.62, + "35": 1920.77, + "36": 1918.86, + "37": 1919.54, + "38": 1919.35, + "39": 1920.17, + "40": 1921.95, + "41": 1918.87, + "42": 1918.74, + "43": 1917.54, + "44": 1920.34, + "45": 1920.99, + "46": 1917.23, + "47": 1918.53, + "48": 1918.12, + "49": 1907.0 }, - "expected_e2e_ms": 101924.55, - "expected_avg_denoise_ms": 1976.24, - "expected_median_denoise_ms": 1985.67, + "expected_e2e_ms": 98305.51, + "expected_avg_denoise_ms": 1913.15, + "expected_median_denoise_ms": 1922.21, "estimated_full_test_time_s": 248.2 }, "flux_2_image_t2i_2_gpus": { "stages_ms": { - "InputValidationStage": 0.05, - "TextEncodingStage": 747.5, + "InputValidationStage": 0.04, + "TextEncodingStage": 372.55, "ImageVAEEncodingStage": 0.01, - "LatentPreparationStage": 0.83, - "TimestepPreparationStage": 30.12, - "DenoisingStage": 13360.47, - "DecodingStage": 19.32 + "LatentPreparationStage": 0.35, + "TimestepPreparationStage": 19.51, + "DenoisingStage": 12372.22, + "DecodingStage": 7.37 }, "denoise_step_ms": { "0": 66.76, - "1": 71.94, - "2": 226.39, + "1": 66.48, + "2": 189.28, "3": 226.37, "4": 226.46, "5": 226.56, @@ -2023,171 +2023,174 @@ "48": 227.69, "49": 227.17 }, - "expected_e2e_ms": 15050.94, - "expected_avg_denoise_ms": 262.66, - "expected_median_denoise_ms": 266.23, + "expected_e2e_ms": 13216.13, + "expected_avg_denoise_ms": 242.72, + "expected_median_denoise_ms": 246.92, "estimated_full_test_time_s": 135.3 }, "qwen_image_edit_2511_ti2i": { "stages_ms": { "DecodingStage": 18.69, "InputValidationStage": 48.98, - "DenoisingStage": 22129.32, + "DenoisingStage": 22080.78, "ImageEncodingStage": 562.02, "LatentPreparationStage": 0.13, "TimestepPreparationStage": 12.36, "ImageVAEEncodingStage": 47.89 }, "denoise_step_ms": { - "0": 415.55, - "1": 558.04, + "0": 412.22, + "1": 555.43, "2": 548.65, - "3": 560.17, - "4": 550.15, + "3": 558.45, + "4": 549.6, "5": 554.87, "6": 548.51, "7": 554.21, - "8": 551.19, + "8": 550.62, "9": 557.15, - "10": 553.7, - "11": 556.73, - "12": 554.35, - "13": 555.94, + "10": 552.9, + "11": 556.55, + "12": 554.32, + "13": 555.01, "14": 556.44, "15": 554.15, "16": 554.86, "17": 555.14, - "18": 558.16, - "19": 554.71, - "20": 557.93, - "21": 556.35, - "22": 555.88, - "23": 558.05, - "24": 556.66, - "25": 557.84, + "18": 555.79, + "19": 554.34, + "20": 556.88, + "21": 555.62, + "22": 555.5, + "23": 556.25, + "24": 555.78, + "25": 557.11, "26": 556.77, - "27": 558.22, - "28": 555.72, - "29": 558.04, + "27": 557.99, + "28": 555.39, + "29": 556.36, "30": 555.66, - "31": 559.13, + "31": 557.42, "32": 555.84, - "33": 559.4, - "34": 557.41, - "35": 559.05, - "36": 558.7, - "37": 556.78, - "38": 556.8, - "39": 554.2 + "33": 557.42, + "34": 557.38, + "35": 558.4, + "36": 554.4, + "37": 553.53, + "38": 550.95, + "39": 551.35 }, - "expected_e2e_ms": 23143.72, - "expected_avg_denoise_ms": 552.98, - "expected_median_denoise_ms": 556.77, + "expected_e2e_ms": 22989.11, + "expected_avg_denoise_ms": 551.91, + "expected_median_denoise_ms": 555.79, "estimated_full_test_time_s": 143.7 }, "fsdp-inference": { "stages_ms": { - "InputValidationStage": 0.04, - "LatentPreparationStage": 0.16, - "TextEncodingStage": 257.5, - "TimestepPreparationStage": 38.44, - "DecodingStage": 16.88, - "DenoisingStage": 2374.14 + "InputValidationStage": 0.03, + "LatentPreparationStage": 0.11, + "TextEncodingStage": 255.96, + "TimestepPreparationStage": 27.44, + "DecodingStage": 6.55, + "DenoisingStage": 742.38 }, "denoise_step_ms": { - "0": 259.26, - "1": 164.46, - "2": 263.36, - "3": 650.0, - "4": 271.46, - "5": 266.71, - "6": 650.0, - "7": 270.08, - "8": 259.96 + "0": 83.2, + "1": 81.75, + "2": 81.7, + "3": 81.23, + "4": 80.34, + "5": 85.95, + "6": 81.3, + "7": 82.17, + "8": 82.66 }, - "expected_e2e_ms": 2850.0, - "expected_avg_denoise_ms": 263.19, - "expected_median_denoise_ms": 261.85, + "expected_e2e_ms": 1042.43, + "expected_avg_denoise_ms": 82.26, + "expected_median_denoise_ms": 81.75, "estimated_full_test_time_s": 122.7 }, "hunyuan3d_shape_gen": { "stages_ms": { - "Hunyuan3DShapeBeforeDenoisingStage": 544.59, - "Hunyuan3DShapeDenoisingStage": 3306.16, - "Hunyuan3DShapeExportStage": 8488.42, - "Hunyuan3DShapeSaveStage": 859.23 + "Hunyuan3DShapeBeforeDenoisingStage": 54.28, + "Hunyuan3DShapeDenoisingStage": 1698.03, + "Hunyuan3DShapeExportStage": 6011.97, + "Hunyuan3DShapeSaveStage": 684.0, + "Hunyuan3DPaintPreprocessStage": 218114.22, + "Hunyuan3DPaintTexGenStage": 3086.62, + "Hunyuan3DPaintPostprocessStage": 11500.0 }, "denoise_step_ms": { - "0": 137.54, - "1": 31.13, - "2": 65.96, - "3": 65.31, - "4": 65.03, - "5": 65.08, - "6": 65.1, - "7": 65.48, - "8": 64.99, - "9": 65.53, - "10": 64.91, - "11": 65.44, - "12": 64.94, - "13": 65.42, - "14": 65.1, - "15": 65.55, - "16": 65.03, - "17": 65.46, - "18": 64.97, - "19": 65.39, - "20": 65.08, - "21": 65.47, - "22": 65.89, - "23": 64.66, - "24": 65.16, - "25": 65.59, - "26": 64.95, - "27": 65.65, - "28": 64.94, - "29": 65.49, - "30": 65.37, - "31": 65.58, - "32": 64.85, - "33": 65.53, - "34": 65.18, - "35": 65.53, - "36": 64.89, - "37": 65.54, - "38": 65.29, - "39": 65.41, - "40": 65.01, - "41": 65.54, - "42": 65.17, - "43": 65.49, - "44": 64.98, - "45": 65.36, - "46": 65.25, - "47": 65.38, - "48": 65.36, - "49": 66.03 + "0": 131.29, + "1": 30.03, + "2": 30.33, + "3": 30.53, + "4": 30.26, + "5": 30.34, + "6": 30.12, + "7": 30.47, + "8": 30.41, + "9": 30.38, + "10": 30.16, + "11": 30.06, + "12": 30.36, + "13": 30.04, + "14": 30.11, + "15": 30.25, + "16": 30.33, + "17": 30.08, + "18": 30.33, + "19": 30.23, + "20": 30.45, + "21": 30.46, + "22": 30.05, + "23": 30.09, + "24": 30.33, + "25": 30.12, + "26": 29.98, + "27": 30.3, + "28": 30.13, + "29": 30.33, + "30": 30.18, + "31": 29.98, + "32": 30.36, + "33": 30.02, + "34": 30.32, + "35": 29.96, + "36": 30.22, + "37": 30.3, + "38": 29.97, + "39": 30.05, + "40": 30.02, + "41": 30.14, + "42": 30.15, + "43": 30.12, + "44": 30.22, + "45": 29.95, + "46": 30.05, + "47": 30.12, + "48": 30.26, + "49": 30.3 }, - "expected_e2e_ms": 300141.39, - "expected_avg_denoise_ms": 66.06, - "expected_median_denoise_ms": 65.36, + "expected_e2e_ms": 237254.68, + "expected_avg_denoise_ms": 32.68, + "expected_median_denoise_ms": 30.2, "estimated_full_test_time_s": 420.1 }, "wan2_1_t2v_1.3b_frame_interp_2x": { "stages_ms": { "InputValidationStage": 0.05, - "TextEncodingStage": 612.62, + "TextEncodingStage": 534.56, "LatentPreparationStage": 0.2, "TimestepPreparationStage": 2.72, "DenoisingStage": 7336.03, - "DecodingStage": 564.05 + "DecodingStage": 361.17 }, "denoise_step_ms": { "0": 88.11, - "1": 122.29, + "1": 108.23, "2": 146.56, - "3": 145.57, + "3": 143.75, "4": 145.12, "5": 144.35, "6": 158.28, @@ -2200,119 +2203,119 @@ "13": 153.84, "14": 147.59, "15": 146.91, - "16": 147.58, + "16": 146.07, "17": 144.84, - "18": 146.74, + "18": 145.05, "19": 150.18, "20": 150.08, "21": 146.22, "22": 146.88, "23": 145.99, - "24": 145.55, + "24": 144.43, "25": 148.05, "26": 149.89, "27": 148.3, - "28": 148.18, + "28": 146.68, "29": 146.65, - "30": 147.5, + "30": 146.33, "31": 145.22, - "32": 149.73, - "33": 150.95, + "32": 147.46, + "33": 149.25, "34": 146.49, "35": 147.33, "36": 146.74, "37": 145.63, - "38": 147.07, + "38": 146.59, "39": 148.72, - "40": 150.98, + "40": 149.12, "41": 146.97, "42": 148.43, - "43": 148.38, - "44": 145.78, - "45": 149.13, + "43": 146.87, + "44": 144.96, + "45": 147.59, "46": 149.84, - "47": 150.0, + "47": 148.06, "48": 145.6, "49": 146.37 }, - "expected_e2e_ms": 8543.5, + "expected_e2e_ms": 8273.83, "expected_avg_denoise_ms": 146.61, "expected_median_denoise_ms": 147.3, "estimated_full_test_time_s": 129.3 }, "flux_2_image_t2i_upscaling_4x": { "stages_ms": { - "TextEncodingStage": 374.99, - "DenoisingStage": 23676.19, - "InputValidationStage": 0.06, - "LatentPreparationStage": 1.2, - "TimestepPreparationStage": 18.66, - "DecodingStage": 9.62, + "TextEncodingStage": 373.42, + "DenoisingStage": 21888.49, + "InputValidationStage": 0.04, + "LatentPreparationStage": 0.53, + "TimestepPreparationStage": 16.91, + "DecodingStage": 7.68, "ImageVAEEncodingStage": 0.01 }, "denoise_step_ms": { "0": 66.32, - "1": 395.03, - "2": 488.73, - "3": 460.96, - "4": 474.42, - "5": 471.59, - "6": 469.96, - "7": 472.41, - "8": 468.92, - "9": 475.33, - "10": 465.3, - "11": 477.98, - "12": 466.21, - "13": 478.19, - "14": 467.91, - "15": 477.69, - "16": 468.21, - "17": 477.27, - "18": 470.51, - "19": 473.63, - "20": 472.38, - "21": 472.47, - "22": 475.89, - "23": 471.61, - "24": 474.66, - "25": 471.33, - "26": 476.31, - "27": 473.64, - "28": 474.4, - "29": 472.35, - "30": 474.65, - "31": 471.61, - "32": 476.36, - "33": 468.74, - "34": 473.22, - "35": 471.23, - "36": 476.13, - "37": 472.25, - "38": 473.52, - "39": 473.61, - "40": 474.92, - "41": 474.98, - "42": 473.82, - "43": 473.32, - "44": 474.41, - "45": 473.61, - "46": 473.71, - "47": 475.03, - "48": 473.97, - "49": 473.92 + "1": 341.96, + "2": 419.84, + "3": 460.86, + "4": 418.25, + "5": 458.92, + "6": 423.03, + "7": 450.03, + "8": 432.51, + "9": 437.41, + "10": 441.7, + "11": 429.47, + "12": 442.33, + "13": 429.25, + "14": 444.51, + "15": 432.19, + "16": 439.45, + "17": 436.03, + "18": 436.47, + "19": 438.65, + "20": 436.16, + "21": 440.88, + "22": 435.62, + "23": 440.49, + "24": 434.96, + "25": 439.19, + "26": 436.86, + "27": 438.94, + "28": 438.28, + "29": 437.45, + "30": 438.63, + "31": 436.51, + "32": 440.12, + "33": 438.91, + "34": 437.07, + "35": 438.81, + "36": 438.74, + "37": 440.95, + "38": 438.64, + "39": 440.95, + "40": 439.37, + "41": 439.5, + "42": 439.12, + "43": 437.93, + "44": 439.23, + "45": 438.25, + "46": 439.57, + "47": 438.03, + "48": 439.43, + "49": 438.64 }, - "expected_e2e_ms": 24551.25, - "expected_avg_denoise_ms": 465.17, - "expected_median_denoise_ms": 473.63, + "expected_e2e_ms": 22732.26, + "expected_avg_denoise_ms": 429.1, + "expected_median_denoise_ms": 438.64, "estimated_full_test_time_s": 145.1 }, "wan2_1_t2v_1.3b_upscaling_4x": { "stages_ms": { - "DecodingStage": 566.48, + "DecodingStage": 362.62, "InputValidationStage": 0.04, "DenoisingStage": 7238.94, - "TextEncodingStage": 623.45, + "TextEncodingStage": 534.42, "LatentPreparationStage": 0.14, "TimestepPreparationStage": 2.35 }, @@ -2321,14 +2324,14 @@ "1": 104.79, "2": 143.85, "3": 144.68, - "4": 145.57, + "4": 144.32, "5": 144.43, "6": 145.13, "7": 144.79, "8": 144.33, "9": 145.01, "10": 144.58, - "11": 146.38, + "11": 145.24, "12": 141.99, "13": 148.34, "14": 144.98, @@ -2368,7 +2371,7 @@ "48": 145.25, "49": 144.24 }, - "expected_e2e_ms": 8612.26, + "expected_e2e_ms": 8333.66, "expected_avg_denoise_ms": 144.68, "expected_median_denoise_ms": 144.74, "estimated_full_test_time_s": 129.3 @@ -2376,76 +2379,76 @@ "wan2_1_t2v_1.3b_frame_interp_2x_upscaling_4x": { "stages_ms": { "InputValidationStage": 0.04, - "TextEncodingStage": 559.97, + "TextEncodingStage": 532.52, "LatentPreparationStage": 0.14, "TimestepPreparationStage": 2.34, - "DenoisingStage": 7419.14, - "DecodingStage": 564.64 + "DenoisingStage": 7189.61, + "DecodingStage": 360.73 }, "denoise_step_ms": { "0": 110.01, - "1": 125.73, - "2": 147.19, - "3": 145.54, - "4": 145.31, - "5": 146.31, - "6": 158.16, - "7": 148.97, - "8": 146.24, - "9": 148.82, - "10": 145.96, - "11": 146.19, - "12": 151.66, - "13": 154.95, - "14": 150.19, - "15": 146.77, - "16": 149.06, - "17": 146.95, - "18": 147.98, - "19": 153.21, - "20": 151.47, - "21": 147.68, - "22": 146.38, - "23": 148.52, - "24": 147.07, - "25": 151.59, - "26": 150.39, - "27": 150.41, - "28": 151.1, - "29": 148.02, - "30": 147.85, - "31": 148.52, - "32": 152.62, - "33": 150.97, - "34": 149.8, - "35": 150.53, - "36": 148.13, - "37": 148.54, - "38": 148.74, - "39": 149.68, - "40": 151.1, - "41": 150.29, - "42": 146.85, - "43": 146.56, - "44": 149.6, - "45": 148.65, - "46": 150.93, - "47": 150.19, - "48": 151.48, - "49": 147.91 + "1": 111.05, + "2": 143.36, + "3": 140.74, + "4": 141.91, + "5": 141.72, + "6": 157.69, + "7": 148.67, + "8": 142.39, + "9": 142.18, + "10": 141.97, + "11": 141.58, + "12": 146.84, + "13": 152.72, + "14": 146.44, + "15": 144.73, + "16": 142.22, + "17": 141.52, + "18": 142.6, + "19": 146.0, + "20": 149.21, + "21": 146.28, + "22": 143.82, + "23": 142.23, + "24": 142.02, + "25": 144.91, + "26": 145.6, + "27": 145.78, + "28": 145.23, + "29": 144.61, + "30": 143.35, + "31": 142.45, + "32": 145.0, + "33": 146.21, + "34": 147.97, + "35": 144.98, + "36": 143.61, + "37": 142.96, + "38": 143.75, + "39": 144.67, + "40": 147.72, + "41": 147.63, + "42": 144.8, + "43": 143.42, + "44": 144.01, + "45": 144.66, + "46": 145.02, + "47": 145.81, + "48": 146.56, + "49": 146.13 }, - "expected_e2e_ms": 8630.42, - "expected_avg_denoise_ms": 148.25, - "expected_median_denoise_ms": 148.78, + "expected_e2e_ms": 8099.03, + "expected_avg_denoise_ms": 143.71, + "expected_median_denoise_ms": 144.67, "estimated_full_test_time_s": 129.4 }, "ltx_2.3_one_stage_ti2v": { "stages_ms": { - "InputValidationStage": 3.6, - "TextEncodingStage": 773.87, + "InputValidationStage": 3.49, + "TextEncodingStage": 397.91, "LTX2TextConnectorStage": 27.35, - "LTX2SigmaPreparationStage": 0.19, - "TimestepPreparationStage": 11.85, + "LTX2SigmaPreparationStage": 0.17, + "TimestepPreparationStage": 10.5, "LTX2AVLatentPreparationStage": 0.11, "LTX2ImageEncodingStage": 29.61, "LTX2AVDenoisingStage": 15025.95, @@ -2490,71 +2493,71 @@ }, "ltx_2.3_two_stage_t2v_2gpus": { "stages_ms": { - "InputValidationStage": 0.05, - "TextEncodingStage": 775.72, - "LTX2TextConnectorStage": 29.83, - "LTX2HalveResolutionStage": 0.06, + "InputValidationStage": 0.04, + "TextEncodingStage": 399.08, + "LTX2TextConnectorStage": 27.15, + "LTX2HalveResolutionStage": 0.05, "LTX2LoRASwitchStage": 0.01, - "LTX2SigmaPreparationStage": 0.61, - "TimestepPreparationStage": 20.77, + "LTX2SigmaPreparationStage": 0.23, + "TimestepPreparationStage": 16.85, "LTX2AVLatentPreparationStage": 0.12, "LTX2ImageEncodingStage": 0.02, - "LTX2AVDenoisingStage": 5460.05, + "LTX2AVDenoisingStage": 4491.78, "LTX2UpsampleStage": 3.31, "LTX2RefinementStage": 589.32, "LTX2AVDecodingStage": 240.26, "per_frame_generation": null }, "denoise_step_ms": { - "0": 225.43, - "1": 283.15, - "2": 193.3, - "3": 166.82, - "4": 200.16, - "5": 196.51, - "6": 178.52, + "0": 199.31, + "1": 171.07, + "2": 146.64, + "3": 146.1, + "4": 145.66, + "5": 147.1, + "6": 170.71, "7": 168.41, - "8": 182.67, - "9": 166.79, - "10": 163.28, - "11": 170.97, - "12": 176.57, + "8": 143.8, + "9": 147.53, + "10": 142.88, + "11": 144.39, + "12": 145.42, "13": 136.28, "14": 138.62, - "15": 172.68, - "16": 172.41, - "17": 184.39, - "18": 161.55, - "19": 176.63, - "20": 200.48, - "21": 182.02, - "22": 161.98, + "15": 143.97, + "16": 144.55, + "17": 142.06, + "18": 144.82, + "19": 143.2, + "20": 147.84, + "21": 142.94, + "22": 144.65, "23": 139.58, "24": 137.21, - "25": 182.74, - "26": 149.38, + "25": 143.94, + "26": 142.25, "27": 144.21, - "28": 171.22, - "29": 183.21, + "28": 154.11, + "29": 144.28, "30": 142.41, "31": 134.02, "32": 138.32 }, - "expected_e2e_ms": 12925.1, - "expected_avg_denoise_ms": 209.57, - "expected_median_denoise_ms": 179.26, + "expected_e2e_ms": 6349.58, + "expected_avg_denoise_ms": 158.2, + "expected_median_denoise_ms": 145.66, "estimated_full_test_time_s": 160.0 }, "ltx_2_3_two_stage_ti2v_2gpus": { "stages_ms": { "InputValidationStage": 3.21, - "TextEncodingStage": 747.28, + "TextEncodingStage": 411.15, "LTX2TextConnectorStage": 27.37, - "LTX2HalveResolutionStage": 0.05, - "LTX2LoRASwitchStage": 3.01, - "LTX2SigmaPreparationStage": 0.22, - "TimestepPreparationStage": 19.94, - "LTX2AVLatentPreparationStage": 0.14, + "LTX2HalveResolutionStage": 0.04, + "LTX2LoRASwitchStage": 2.95, + "LTX2SigmaPreparationStage": 0.17, + "TimestepPreparationStage": 13.18, + "LTX2AVLatentPreparationStage": 0.13, "LTX2ImageEncodingStage": 27.58, "LTX2AVDenoisingStage": 7506.38, "LTX2UpsampleStage": 2.31, @@ -2604,34 +2607,34 @@ }, "longlive2_t2v": { "stages_ms": { - "InputValidationStage": 0.05, - "LongLive2TextEncodingStage": 328.32, + "InputValidationStage": 0.04, + "LongLive2TextEncodingStage": 328.05, "LongLive2ImageVAEEncodingStage": 0.0, - "LongLive2LatentPreparationStage": 0.17, - "LongLive2CausalDenoisingStage": 4879.98, - "DecodingStage": 1397.62, + "LongLive2LatentPreparationStage": 0.16, + "LongLive2CausalDenoisingStage": 3959.31, + "DecodingStage": 1340.81, "per_frame_generation": null }, "denoise_step_ms": {}, - "expected_e2e_ms": 6610.71, - "expected_avg_denoise_ms": 650.0, - "expected_median_denoise_ms": 650.0, + "expected_e2e_ms": 5648.49, + "expected_avg_denoise_ms": 477.49, + "expected_median_denoise_ms": 56.74, "estimated_full_test_time_s": 153.1 }, "longlive2_i2v": { "stages_ms": { "InputValidationStage": 23.02, "LongLive2TextEncodingStage": 327.98, - "LongLive2ImageVAEEncodingStage": 1048.81, + "LongLive2ImageVAEEncodingStage": 920.37, "LongLive2LatentPreparationStage": 0.12, - "LongLive2CausalDenoisingStage": 4975.28, - "DecodingStage": 3051.96, + "LongLive2CausalDenoisingStage": 4326.16, + "DecodingStage": 2548.02, "per_frame_generation": null }, "denoise_step_ms": {}, - "expected_e2e_ms": 9431.65, - "expected_avg_denoise_ms": 800.0, - "expected_median_denoise_ms": 800.0, + "expected_e2e_ms": 9086.81, + "expected_avg_denoise_ms": 492.02, + "expected_median_denoise_ms": 151.9, "estimated_full_test_time_s": 149.4 }, "lingbot_video_moe_t2v": { @@ -2654,19 +2657,19 @@ "stages_ms": { "InputValidationStage": 4.63, "TextEncodingStage": 401.75, - "LTX2TextConnectorStage": 27.48, + "LTX2TextConnectorStage": 27.26, "LTX2HalveResolutionStage": 0.04, - "LTX2LoRASwitchStage": 6.12, - "LTX2SigmaPreparationStage": 0.26, + "LTX2LoRASwitchStage": 3.91, + "LTX2SigmaPreparationStage": 0.23, "TimestepPreparationStage": 14.45, "LTX2AVLatentPreparationStage": 0.13, - "LTX2ImageEncodingStage": 57.62, - "LTX2AVDenoisingStage": 12162.51, - "LTX2UpsampleStage": 11.04, + "LTX2ImageEncodingStage": 55.34, + "LTX2AVDenoisingStage": 11243.85, + "LTX2UpsampleStage": 10.48, "ltx2_lora_switch_stage2": 5.68, - "ltx2_image_encoding_stage2": 64.54, - "LTX2RefinementStage": 3484.86, - "LTX2AVDecodingStage": 1054.91, + "ltx2_image_encoding_stage2": 57.86, + "LTX2RefinementStage": 3198.06, + "LTX2AVDecodingStage": 885.02, "per_frame_generation": null }, "denoise_step_ms": { @@ -2685,12 +2688,12 @@ "12": 760.17, "13": 740.69, "14": 369.53, - "15": 1193.58, - "16": 1144.87, - "17": 1143.19 + "15": 1098.34, + "16": 1047.82, + "17": 1050.03 }, - "expected_e2e_ms": 15981.39, - "expected_avg_denoise_ms": 868.06, + "expected_e2e_ms": 15974.21, + "expected_avg_denoise_ms": 802.06, "expected_median_denoise_ms": 747.62, "estimated_full_test_time_s": 363.2 }, @@ -2707,7 +2710,7 @@ "Cosmos3ImagePreprocessStage": 0.01, "Cosmos3TokenizationStage": 155.89, "Cosmos3LatentPreparationStage": 0.15, - "Cosmos3TimestepPreparationStage": 0.64, + "Cosmos3TimestepPreparationStage": 0.49, "Cosmos3DenoisingStage": 846.85, "Cosmos3DecodingStage": 12.14 }, @@ -2756,10 +2759,10 @@ "cosmos3_nano_t2v": { "stages_ms": { "Cosmos3ImagePreprocessStage": 0.01, - "Cosmos3TokenizationStage": 152.97, + "Cosmos3TokenizationStage": 143.82, "Cosmos3LatentPreparationStage": 0.14, "Cosmos3TimestepPreparationStage": 0.56, - "Cosmos3DenoisingStage": 261.27, + "Cosmos3DenoisingStage": 202.12, "Cosmos3DecodingStage": 47.82, "per_frame_generation": null }, @@ -2802,10 +2805,10 @@ "stages_ms": { "InputValidationStage": 0.05, "MiniMaxH3PartitionAdmissionStage": 0.03, - "MiniMaxH3TextEncodingStage": 620.13, - "MiniMaxH3VisualEncodingStage": 0.11, + "MiniMaxH3TextEncodingStage": 598.33, + "MiniMaxH3VisualEncodingStage": 0.06, "MiniMaxH3AudioEncodingStage": 0.03, - "MiniMaxH3LatentPreparationStage": 18.72, + "MiniMaxH3LatentPreparationStage": 17.5, "MiniMaxH3TimestepPreparationStage": 0.34, "MiniMaxH3DenoisingStage": 16839.0, "MiniMaxH3DecodingStage": 2200.0, @@ -2813,15 +2816,15 @@ }, "denoise_step_ms": { "0": 211.7, - "1": 2427.55, - "2": 2344.0, - "3": 2332.99, - "4": 2332.69, - "5": 2343.56, + "1": 2422.79, + "2": 2338.59, + "3": 2332.68, + "4": 2328.93, + "5": 2342.07, "6": 2326.79 }, "expected_e2e_ms": 19700.0, - "expected_avg_denoise_ms": 2045.61, + "expected_avg_denoise_ms": 2045.39, "expected_median_denoise_ms": 2332.99, "estimated_full_test_time_s": 208.0 }, diff --git a/python/sglang/multimodal_gen/test/server/test_server_common.py b/python/sglang/multimodal_gen/test/server/test_server_common.py index 3c4d7dee1..d2b1ad478 100644 --- a/python/sglang/multimodal_gen/test/server/test_server_common.py +++ b/python/sglang/multimodal_gen/test/server/test_server_common.py @@ -158,22 +158,6 @@ def diffusion_server(case: DiffusionTestCase) -> ServerContext: # picking another one (which causes the test client to connect to the wrong server). extra_args += " --strict-ports" - # Shape-only mesh cases (e.g. hunyuan3d_shape_gen) validate geometry via - # mesh-correctness and must NOT run the paint/texture stages, whose - # verification checks texture artifacts (paint_mesh/normal_maps/renderer) - # that the shape-only path never produces. Inject a pipeline-config override - # disabling paint for these cases. - if server_args.custom_validator == "mesh": - import json as _json - import tempfile as _tempfile - - _paint_off_cfg = os.path.join( - _tempfile.gettempdir(), f"{case.id}_paint_off.json" - ) - with open(_paint_off_cfg, "w") as _f: - _json.dump({"paint_enable": False}, _f) - extra_args += f" --config {_paint_off_cfg}" - for arg in server_args.extras: extra_args += f" {arg}"