[
  {
    "50.00 percentile latency (ns)": 43046624108,
    "90.00 percentile latency (ns)": 63556287118,
    "95.00 percentile latency (ns)": 65371818807,
    "97.00 percentile latency (ns)": 65995373759,
    "99.00 percentile latency (ns)": 66490666450,
    "99.90 percentile latency (ns)": 66658538764,
    "Max latency (ns)": 66663374934,
    "Mean latency (ns)": 40486200265,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 486199002,
    "Min queries satisfied": "Yes",
    "Mode": "Performance",
    "Result is": "VALID",
    "SUT name": "BERT SERVER",
    "Samples per second": 5643.28,
    "Scenario": "offline",
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "40 GB",
    "accelerator_memory_configuration": "HBM2",
    "accelerator_model_name": "A100-PCIE-40GB",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 2,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "characteristics.samples_per_second": 5643.28,
    "characteristics.samples_per_second.normalized_per_core": 2821.64,
    "characteristics.samples_per_second.normalized_per_processor": 2821.64,
    "ck_system": "QCT_D43KQ_A100-PCIex2_TRT",
    "ck_used": true,
    "cooling": "",
    "dataset": "SQuAD v1.1",
    "dataset_link": "",
    "dim_x_default": "seq_number",
    "dim_x_maximize": true,
    "dim_y_default": "characteristics.samples_per_second",
    "dim_y_maximize": false,
    "division": "closed",
    "formal_model": "bert",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "",
    "framework": "TensorRT 7.2, CUDA 11.0, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch",
    "host_memory_capacity": "512 GB",
    "host_memory_configuration": "",
    "host_networking": "",
    "host_networking_topology": "",
    "host_processor_caches": "",
    "host_processor_core_count": 120,
    "host_processor_frequency": "",
    "host_processor_interconnect": "",
    "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor",
    "host_processors_per_node": 2,
    "host_storage_capacity": "NVMe SSD",
    "host_storage_type": "2.0 TB",
    "hw_notes": "",
    "informal_model": "bert-99",
    "input_data_types": "int32",
    "max_async_queries": 1,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 60000,
    "min_query_count": 1,
    "mlperf_version": 0.7,
    "normalize_cores": 2,
    "normalize_processors": 2,
    "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/QCT/code",
    "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/QCT/results/QCT_D43KQ_A100-PCIex2_TRT",
    "number_of_nodes": 1,
    "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)",
    "other_software_stack": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; GCC 7.5.0; Python 3.7.10",
    "performance_issue_same": true,
    "performance_issue_same_index": 0,
    "performance_issue_unique": true,
    "performance_sample_count": 10833,
    "print_timestamps": true,
    "problem": false,
    "qsl_rng_seed": 12786827339337101903,
    "retraining": "N",
    "sample_index_rng_seed": 12640797754436136668,
    "samples_per_query": 376200,
    "schedule_rng_seed": 3135815929913719677,
    "starting_weights_filename": "bert_large_v1_1_fake_quant.onnx",
    "status": "available",
    "submitter": "QCT",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/QCT",
    "sw_notes": "Powered by CK v2.5.8 (https://github.com/ctuning/ck)",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/QCT_D43KQ_A100-PCIex2_TRT",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 0,
    "target_qps": 5700,
    "task": "NLP",
    "task2": "nlp",
    "total_cores": 240,
    "uid": "49b4a01b6c54e0d0",
    "use_accelerator": true,
    "weight_data_types": "int8,fp16",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 43296852768,
    "90.00 percentile latency (ns)": 63641548998,
    "95.00 percentile latency (ns)": 65408325335,
    "97.00 percentile latency (ns)": 66046479739,
    "99.00 percentile latency (ns)": 66513294307,
    "99.90 percentile latency (ns)": 66634546488,
    "Max latency (ns)": 66645207311,
    "Mean latency (ns)": 40591744021,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 831505070,
    "Min queries satisfied": "Yes",
    "Mode": "Performance",
    "Result is": "VALID",
    "SUT name": "BERT SERVER",
    "Samples per second": 2812.51,
    "Scenario": "offline",
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "40 GB",
    "accelerator_memory_configuration": "HBM2",
    "accelerator_model_name": "A100-PCIE-40GB",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 2,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "characteristics.samples_per_second": 2812.51,
    "characteristics.samples_per_second.normalized_per_core": 1406.255,
    "characteristics.samples_per_second.normalized_per_processor": 1406.255,
    "ck_system": "QCT_D43KQ_A100-PCIex2_TRT",
    "ck_used": true,
    "cooling": "",
    "dataset": "SQuAD v1.1",
    "dataset_link": "",
    "dim_x_default": "seq_number",
    "dim_x_maximize": true,
    "dim_y_default": "characteristics.samples_per_second",
    "dim_y_maximize": false,
    "division": "closed",
    "formal_model": "bert",
    "formal_model_accuracy": 99.9,
    "formal_model_link": "",
    "framework": "TensorRT 7.2, CUDA 11.0, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch",
    "host_memory_capacity": "512 GB",
    "host_memory_configuration": "",
    "host_networking": "",
    "host_networking_topology": "",
    "host_processor_caches": "",
    "host_processor_core_count": 120,
    "host_processor_frequency": "",
    "host_processor_interconnect": "",
    "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor",
    "host_processors_per_node": 2,
    "host_storage_capacity": "NVMe SSD",
    "host_storage_type": "2.0 TB",
    "hw_notes": "",
    "informal_model": "bert-99.9",
    "input_data_types": "int32",
    "max_async_queries": 1,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 60000,
    "min_query_count": 1,
    "mlperf_version": 0.7,
    "normalize_cores": 2,
    "normalize_processors": 2,
    "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/QCT/code",
    "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/QCT/results/QCT_D43KQ_A100-PCIex2_TRT",
    "number_of_nodes": 1,
    "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)",
    "other_software_stack": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; GCC 7.5.0; Python 3.7.10",
    "performance_issue_same": true,
    "performance_issue_same_index": 0,
    "performance_issue_unique": true,
    "performance_sample_count": 10833,
    "print_timestamps": true,
    "problem": false,
    "qsl_rng_seed": 12786827339337101903,
    "retraining": "N",
    "sample_index_rng_seed": 12640797754436136668,
    "samples_per_query": 187440,
    "schedule_rng_seed": 3135815929913719677,
    "starting_weights_filename": "bert_large_v1_1_fake_quant.onnx",
    "status": "available",
    "submitter": "QCT",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/QCT",
    "sw_notes": "Powered by CK v2.5.8 (https://github.com/ctuning/ck)",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/QCT_D43KQ_A100-PCIex2_TRT",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 0,
    "target_qps": 2840,
    "task": "NLP",
    "task2": "nlp",
    "total_cores": 240,
    "uid": "ad189938e26896db",
    "use_accelerator": true,
    "weight_data_types": "int8,fp16",
    "weight_transformations": "quantization, affine fusion"
  }
]