[
  {
    "50.00 percentile latency (ns)": 366265971,
    "90.00 percentile latency (ns)": 501790988,
    "95.00 percentile latency (ns)": 539371959,
    "97.00 percentile latency (ns)": 564063154,
    "99.00 percentile latency (ns)": 601025707,
    "99.90 percentile latency (ns)": 666051795,
    "Completed samples per second": 51986.98,
    "Max latency (ns)": 768066586,
    "Mean latency (ns)": 378968754,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 21946734,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "RNNT SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 52004.78,
    "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": 4,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.accuracy": 92.55225596969493,
    "characteristics.scheduled_queries_per_second": 52004.78,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 13001.195,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 13001.195,
    "characteristics.word error rate": 7.4477440303050795,
    "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT",
    "ck_used": false,
    "cooling": "",
    "dataset": "LibriSpeech",
    "dataset_link": "",
    "dim_x_default": "characteristics.scheduled_queries_per_second",
    "dim_y_default": "characteristics.accuracy",
    "dim_y_maximize": true,
    "disk_controllers": "",
    "disk_drives": "",
    "division": "closed",
    "filesystem": "",
    "formal_model": "rnn-t",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "",
    "framework": "TensorRT 8.0.2, CUDA 11.4",
    "host_memory_capacity": "512 GB",
    "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 7742",
    "host_processors_per_node": 1,
    "host_storage_capacity": "5 TB",
    "host_storage_type": "NVMe SSD",
    "hw_notes": "",
    "informal_model": "rnnt",
    "input_data_types": "fp16",
    "key.accuracy": "characteristics.accuracy",
    "management_firmware_version": "",
    "max_async_queries": 0,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 600000,
    "min_query_count": 270336,
    "mlperf_version": 1.1,
    "network_speed_mbit": "",
    "nics_enabled_connected": "",
    "nics_enabled_firmware": "",
    "nics_enabled_os": "",
    "normalize_cores": 4,
    "normalize_processors": 4,
    "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/HPE/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/HPE/results/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT",
    "number_of_nodes": 1,
    "number_of_type_nics_installed": "",
    "operating_system": "Ubuntu 20.04.2",
    "other_hardware": "",
    "other_software_stack": "TensorRT 8.0.2, CUDA 11.4, cuDNN 8.2.1, Driver 470.42.01, DALI 0.31.0",
    "performance_issue_same": 0,
    "performance_issue_same_index": 0,
    "performance_issue_unique": 0,
    "performance_sample_count": 2513,
    "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": "DistributedDataParallel_1576581068.9962234-epoch-100.pt",
    "status": "available",
    "submitter": "HPE",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/HPE",
    "sw_notes": "",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT",
    "system_name": "HPE ProLiant XL645d Gen10 Plus (4x A100-SXM-80GB, TensorRT)",
    "system_type": "datacenter",
    "target_latency (ns)": 1000000000,
    "target_qps": 52000,
    "task": "speech recognition",
    "task2": "speech recognition",
    "total_cores": 64,
    "uid": "5c1884b278440fca",
    "use_accelerator": true,
    "weight_data_types": "fp16",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 299413070,
    "90.00 percentile latency (ns)": 430574207,
    "95.00 percentile latency (ns)": 471466343,
    "97.00 percentile latency (ns)": 494693576,
    "99.00 percentile latency (ns)": 542624469,
    "99.90 percentile latency (ns)": 623958590,
    "Completed samples per second": 44975.76,
    "Max latency (ns)": 760784350,
    "Mean latency (ns)": 307604076,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 60176177,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "RNNT SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 44988.42,
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "40 GB",
    "accelerator_memory_configuration": "HBM2e",
    "accelerator_model_name": "NVIDIA A100-SXM4-40GB",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 4,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.accuracy": 92.5567656542425,
    "characteristics.scheduled_queries_per_second": 44988.42,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 11247.105,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 11247.105,
    "characteristics.word error rate": 7.4432343457575145,
    "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT",
    "ck_used": true,
    "cooling": "",
    "dataset": "LibriSpeech",
    "dataset_link": "",
    "dim_x_default": "characteristics.scheduled_queries_per_second",
    "dim_y_default": "characteristics.accuracy",
    "dim_y_maximize": true,
    "disk_controllers": "",
    "disk_drives": "",
    "division": "closed",
    "filesystem": "",
    "formal_model": "rnn-t",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "",
    "framework": "TensorRT 7.2.3, CUDA 11.1",
    "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": 1,
    "host_storage_capacity": "3.2 TB",
    "host_storage_type": "SSD",
    "hw_notes": "",
    "informal_model": "rnnt",
    "input_data_types": "fp16",
    "key.accuracy": "characteristics.accuracy",
    "management_firmware_version": "",
    "max_async_queries": 0,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 600000,
    "min_query_count": 270336,
    "mlperf_version": 1.0,
    "network_speed_mbit": "",
    "nics_enabled_connected": "",
    "nics_enabled_firmware": "",
    "nics_enabled_os": "",
    "normalize_cores": 4,
    "normalize_processors": 4,
    "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/HPE/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/HPE/results/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT",
    "number_of_nodes": 1,
    "number_of_type_nics_installed": "",
    "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)",
    "other_hardware": "",
    "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 460.27.04, DALI 0.30.0; GCC 7.5.0; Python 3.7.10",
    "performance_issue_same": 0,
    "performance_issue_same_index": 0,
    "performance_issue_unique": 0,
    "performance_sample_count": 2513,
    "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": "DistributedDataParallel_1576581068.9962234-epoch-100.pt",
    "status": "available",
    "submitter": "HPE",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/HPE",
    "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/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 1000000000,
    "target_qps": 45000,
    "task": "speech recognition",
    "task2": "speech recognition",
    "total_cores": 120,
    "uid": "0cd84f4ba2e2cacd",
    "use_accelerator": true,
    "weight_data_types": "fp16",
    "weight_transformations": "quantization, affine fusion"
  }
]