diff --git a/python/sglang/multimodal_gen/test/server/consistency_threshold.json b/python/sglang/multimodal_gen/test/server/consistency_threshold.json index 66512f1c3..007991fe1 100644 --- a/python/sglang/multimodal_gen/test/server/consistency_threshold.json +++ b/python/sglang/multimodal_gen/test/server/consistency_threshold.json @@ -109,6 +109,12 @@ "psnr_threshold": 25.0, "mean_abs_diff_threshold": 8.0 }, + "cosmos3_nano_t2v": { + "clip_threshold": 0.90, + "ssim_threshold": 0.89, + "psnr_threshold": 24.0, + "mean_abs_diff_threshold": 10.0 + }, "ltx_2_two_stage_t2v": { "clip_threshold": 0.90, "ssim_threshold": 0.89, diff --git a/python/sglang/multimodal_gen/test/server/gpu_cases.py b/python/sglang/multimodal_gen/test/server/gpu_cases.py index 90f3f187e..018bcead4 100644 --- a/python/sglang/multimodal_gen/test/server/gpu_cases.py +++ b/python/sglang/multimodal_gen/test/server/gpu_cases.py @@ -239,6 +239,32 @@ ONE_GPU_CASES: list[DiffusionTestCase] = [ model_path=DEFAULT_WAN_2_1_T2V_1_3B_MODEL_NAME_FOR_TEST, ), ), + DiffusionTestCase( + "cosmos3_nano_t2v", + DiffusionServerArgs( + model_path=DEFAULT_COSMOS3_NANO_MODEL_NAME_FOR_TEST, + modality="video", + env_vars={"SGLANG_DISABLE_COSMOS3_GUARDRAILS": "1"}, + ), + DiffusionSamplingParams( + prompt="A blue box slides across a clean warehouse floor.", + output_size="832x480", + seconds=1, + num_frames=9, + extras={ + "num_inference_steps": 4, + "seed": 0, + "max_sequence_length": 128, + "flow_shift": 10.0, + "use_guardrails": False, + "use_duration_template": False, + "use_resolution_template": False, + }, + ), + run_perf_check=False, + run_consistency_check=True, + run_component_accuracy_check=False, + ), # TeaCache acceleration test for Wan video model DiffusionTestCase( "wan2_1_t2v_1.3b_teacache_enabled", diff --git a/python/sglang/multimodal_gen/test/server/perf_baselines.json b/python/sglang/multimodal_gen/test/server/perf_baselines.json index e136acbd8..4c8c85e8f 100644 --- a/python/sglang/multimodal_gen/test/server/perf_baselines.json +++ b/python/sglang/multimodal_gen/test/server/perf_baselines.json @@ -751,7 +751,8 @@ }, "expected_e2e_ms": 961.86, "expected_avg_denoise_ms": 57.95, - "expected_median_denoise_ms": 64.35 + "expected_median_denoise_ms": 64.35, + "estimated_full_test_time_s": 59.9 }, "qwen_image_edit_ti2i": { "stages_ms": { @@ -874,7 +875,8 @@ }, "expected_e2e_ms": 28350.2, "expected_avg_denoise_ms": 672.19, - "expected_median_denoise_ms": 679.16 + "expected_median_denoise_ms": 679.16, + "estimated_full_test_time_s": 117.6 }, "qwen_image_t2i_cache_dit_enabled": { "stages_ms": { @@ -1454,7 +1456,8 @@ }, "expected_e2e_ms": 11816.27, "expected_avg_denoise_ms": 982.57, - "expected_median_denoise_ms": 1118.83 + "expected_median_denoise_ms": 1118.83, + "estimated_full_test_time_s": 77.0 }, "wan2_2_i2v_a14b_2gpu": { "stages_ms": { @@ -2654,6 +2657,206 @@ "expected_avg_denoise_ms": 1325.11, "expected_median_denoise_ms": 1547.31, "estimated_full_test_time_s": 363.2 + }, + "qwen_image_t2i_cache_dit_scm_config_diffusers_1gpu": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 98.3 + }, + "cosmos3_nano_t2i": { + "stages_ms": { + "Cosmos3ImagePreprocessStage": 0.01, + "Cosmos3TokenizationStage": 155.89, + "Cosmos3LatentPreparationStage": 0.15, + "Cosmos3TimestepPreparationStage": 0.64, + "Cosmos3DenoisingStage": 846.85, + "Cosmos3DecodingStage": 12.14 + }, + "denoise_step_ms": { + "0": 58.61, + "1": 21.71, + "2": 44.31, + "3": 22.05, + "4": 21.79, + "5": 21.78, + "6": 21.79, + "7": 22.25, + "8": 22.59, + "9": 22.73, + "10": 22.75, + "11": 22.71, + "12": 24.63, + "13": 22.43, + "14": 21.97, + "15": 22.74, + "16": 22.93, + "17": 22.74, + "18": 22.68, + "19": 22.55, + "20": 22.52, + "21": 22.37, + "22": 22.29, + "23": 22.45, + "24": 22.63, + "25": 22.64, + "26": 22.53, + "27": 22.48, + "28": 22.41, + "29": 22.18, + "30": 22.19, + "31": 22.48, + "32": 22.68, + "33": 22.65, + "34": 22.37 + }, + "expected_e2e_ms": 1019.02, + "expected_avg_denoise_ms": 24.13, + "expected_median_denoise_ms": 22.52, + "estimated_full_test_time_s": 65.0 + }, + "cosmos3_nano_t2v": { + "stages_ms": { + "Cosmos3ImagePreprocessStage": 0.01, + "Cosmos3TokenizationStage": 152.97, + "Cosmos3LatentPreparationStage": 0.17, + "Cosmos3TimestepPreparationStage": 0.64, + "Cosmos3DenoisingStage": 294.92, + "Cosmos3DecodingStage": 47.82, + "per_frame_generation": null + }, + "denoise_step_ms": { + "0": 89.22, + "1": 61.52, + "2": 85.06, + "3": 58.08 + }, + "expected_e2e_ms": 499.70, + "expected_avg_denoise_ms": 73.47, + "expected_median_denoise_ms": 73.29, + "estimated_full_test_time_s": 65.0 + }, + "flux_2_t2i_customized_vae_path": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 574.4 + }, + "wan2_2_t2v_a14b_teacache_2gpu": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 253.9 + }, + "wan2_1_t2v_1_3b_cache_dit_sp_only_2gpu": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 55.4 + }, + "mova_360p_tp2": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 160.2 + }, + "mova_360p_ring1_uly2": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 236.8 + }, + "zimage_image_t2i_2_gpus_non_square": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 67.9 + }, + "flux1_modelopt_fp8_t2i": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 50.3 + }, + "flux2_modelopt_fp8_t2i": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 498.1 + }, + "wan22_modelopt_fp8_t2v": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 89.5 + }, + "hunyuanvideo_modelopt_fp8_t2v": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 64.3 + }, + "qwen_image_modelopt_fp8_t2i": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 65.5 + }, + "qwen_image_edit_modelopt_fp8_ti2i": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 73.7 + }, + "flux1_modelopt_nvfp4_t2i": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 71.2 + }, + "flux2_modelopt_nvfp4_t2i": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 592.3 + }, + "wan22_modelopt_nvfp4_t2v": { + "stages_ms": {}, + "denoise_step_ms": {}, + "expected_e2e_ms": 0.0, + "expected_avg_denoise_ms": 0.0, + "expected_median_denoise_ms": 0.0, + "estimated_full_test_time_s": 181.8 } } }