[
  {
    "50.00 percentile latency (ns)": 1986788,
    "90.00 percentile latency (ns)": 2206063,
    "90th percentile latency (ns)": 2206063,
    "95.00 percentile latency (ns)": 2289796,
    "97.00 percentile latency (ns)": 2369620,
    "99.00 percentile latency (ns)": 2383631,
    "99.90 percentile latency (ns)": 2410162,
    "Max latency (ns)": 5195651,
    "Mean latency (ns)": 2026029,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 1832242,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "QPS w/ loadgen overhead": 488.82,
    "QPS w/o loadgen overhead": 493.58,
    "Result is": "VALID",
    "SUT name": "BERT SERVER",
    "Scenario": "singlestream",
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "40 GB",
    "accelerator_memory_configuration": "HBM2",
    "accelerator_model_name": "NVIDIA A100-PCIe-40GB",
    "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": 2.206063,
    "characteristics.90th_percentile_latency_ns": 2206063.0,
    "characteristics.90th_percentile_latency_s": 0.002206063,
    "characteristics.90th_percentile_latency_us": 2206.063,
    "ck_system": "A100-PCIex1_TRT",
    "ck_used": false,
    "cooling": "",
    "dataset": "SQuAD v1.1",
    "dataset_link": "",
    "dim_x_default": "seq_number",
    "dim_y_default": "characteristics.90th_percentile_latency_ms",
    "dim_y_maximize": false,
    "disk_controllers": "",
    "disk_drives": "",
    "division": "closed",
    "filesystem": "",
    "formal_model": "bert",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "",
    "framework": "TensorRT 7.2.3, CUDA 11.1",
    "host_memory_capacity": "1 TB",
    "host_memory_configuration": "",
    "host_networking": "",
    "host_networking_topology": "",
    "host_processor_caches": "",
    "host_processor_core_count": 64,
    "host_processor_frequency": "",
    "host_processor_interconnect": "",
    "host_processor_model_name": "AMD EPYC 7713",
    "host_processors_per_node": 1,
    "host_storage_capacity": "3.5 TB",
    "host_storage_type": "NVMe SSD",
    "hw_notes": "",
    "informal_model": "bert-99",
    "input_data_types": "int32",
    "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": 1.0,
    "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_v1.0/tree/master/closed/Supermicro/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Supermicro/results/A100-PCIex1_TRT",
    "number_of_nodes": 1,
    "number_of_type_nics_installed": "",
    "operating_system": "Ubuntu 18.04.4",
    "other_hardware": "",
    "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 460.32.03, DALI 0.30.0",
    "performance_issue_same": 0,
    "performance_issue_same_index": 0,
    "performance_issue_unique": 0,
    "performance_sample_count": 10833,
    "power_management": "",
    "power_supply_details": "",
    "power_supply_quantity_and_rating_watts": "",
    "print_timestamps": 0,
    "problem": false,
    "qsl_rng_seed": 7322528924094909334,
    "retraining": "N",
    "sample_index_rng_seed": 1570999273408051088,
    "samples_per_query": 1,
    "schedule_rng_seed": 3507442325620259414,
    "starting_weights_filename": "bert_large_v1_1_fake_quant.onnx",
    "status": "available",
    "submitter": "Supermicro",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Supermicro",
    "sw_notes": "",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A100-PCIex1_TRT",
    "system_name": "Supermicro SBA-4119SG-X (1x A100-PCIe)",
    "system_type": "edge",
    "target_latency (ns)": 0,
    "target_qps": 588.235,
    "task": "NLP",
    "task2": "nlp",
    "total_cores": 64,
    "uid": "56bd503576b423b8",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  }
]