[ { "50.00 percentile latency (ns)": 295565, "90.00 percentile latency (ns)": 299260, "90th percentile latency (ns)": 299260, "95.00 percentile latency (ns)": 300971, "97.00 percentile latency (ns)": 302121, "99.00 percentile latency (ns)": 304227, "99.90 percentile latency (ns)": 308081, "Max latency (ns)": 58123410, "Mean latency (ns)": 295760, "Min duration satisfied": "Yes", "Min latency (ns)": 268059, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 3331.68, "QPS w/o loadgen overhead": 3381.12, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "22GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A10", "accelerator_on-chip_memories": "", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 0.29926, "characteristics.90th_percentile_latency_ns": 299260.0, "characteristics.90th_percentile_latency_s": 0.00029926, "characteristics.90th_percentile_latency_us": 299.26, "characteristics.mAP": 22.914, "ck_system": "A10x1_TRT", "ck_used": false, "cooling": "", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "division": "closed", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.1.6, CUDA 11.3", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 26, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8269CY CPU @ 2.50GHz", "host_processors_per_node": 2, "host_storage_capacity": "170 GB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "ssd-mobilenet", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "mlperf_version": 1.1, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Alibaba/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Alibaba/results/A10x1_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 20.04.4", "other_software_stack": "TensorRT 8.0.1.6, CUDA 11.3, cuDNN 8.2.1.32, Driver 470.57.02", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "ssd_mobilenet_v1_coco_2018_01_28/frozen_inference_graph.pb", "status": "available", "submitter": "Alibaba", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Alibaba", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A10x1_TRT", "system_name": "Alibaba Cloud Sinian Platform A10 (1x A10)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 2680.97, "task": "object detection", "task2": "object detection", "total_cores": 52, "uid": "1851124c9aa357c3", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 498953, "90.00 percentile latency (ns)": 508841, "90th percentile latency (ns)": 508841, "95.00 percentile latency (ns)": 511747, "97.00 percentile latency (ns)": 513870, "99.00 percentile latency (ns)": 519541, "99.90 percentile latency (ns)": 582962, "Max latency (ns)": 8369036, "Mean latency (ns)": 498012, "Min duration satisfied": "Yes", "Min latency (ns)": 455323, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 1970.88, "QPS w/o loadgen overhead": 2007.98, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "Nvidia Tesla T4", "accelerator_on-chip_memories": "", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 0.508841, "characteristics.90th_percentile_latency_ns": 508841.0, "characteristics.90th_percentile_latency_s": 0.000508841, "characteristics.90th_percentile_latency_us": 508.841, "characteristics.mAP": 22.914, "ck_system": "T4x1_TRT", "ck_used": false, "cooling": "", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "division": "closed", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.1.6, CUDA 11.3", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 24, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8163 CPU @ 2.50GHz", "host_processors_per_node": 2, "host_storage_capacity": "3.5 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "ssd-mobilenet", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "mlperf_version": 1.1, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Alibaba/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Alibaba/results/T4x1_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 20.04.4", "other_software_stack": "TensorRT 8.0.1.6, CUDA 11.3, cuDNN 8.2.1.32, Driver 470.57.02", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "ssd_mobilenet_v1_coco_2018_01_28/frozen_inference_graph.pb", "status": "available", "submitter": "Alibaba", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Alibaba", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/T4x1_TRT", "system_name": "Alibaba Cloud Sinian Platform T4 (1x T4)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1327.22, "task": "object detection", "task2": "object detection", "total_cores": 48, "uid": "62898046a8a1304f", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 281997, "90.00 percentile latency (ns)": 286313, "90th percentile latency (ns)": 286313, "95.00 percentile latency (ns)": 289491, "97.00 percentile latency (ns)": 292555, "99.00 percentile latency (ns)": 295999, "99.90 percentile latency (ns)": 327998, "Max latency (ns)": 22215887, "Mean latency (ns)": 283179, "Min duration satisfied": "Yes", "Min latency (ns)": 243268, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 3477.76, "QPS w/o loadgen overhead": 3531.34, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80 GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-SXM-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.90th_percentile_latency_ms": 0.286313, "characteristics.90th_percentile_latency_ns": 286313.0, "characteristics.90th_percentile_latency_s": 0.000286313, "characteristics.90th_percentile_latency_us": 286.313, "characteristics.mAP": 22.914, "ck_system": "A100-SXM-80GBx1_TRT", "ck_used": false, "cooling": "", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.1.6, CUDA 11.3", "host_memory_capacity": "768 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 32, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8369B CPU @ 2.90GHz", "host_processors_per_node": 2, "host_storage_capacity": "170 GB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "ssd-mobilenet", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Alibaba/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Alibaba/results/A100-SXM-80GBx1_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.4", "other_software_stack": "TensorRT 8.0.1.6, CUDA 11.3, cuDNN 8.2.1.32, Driver 470.57.02", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "ssd_mobilenet_v1_coco_2018_01_28/frozen_inference_graph.pb", "status": "available", "submitter": "Alibaba", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Alibaba", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A100-SXM-80GBx1_TRT", "system_name": "Alibaba Cloud Sinian Platform A100 (1x A100-SXM-80GB)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 2173.91, "task": "object detection", "task2": "object detection", "total_cores": 64, "uid": "7935944030ced941", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" } ]