[
  {
    "50.00 percentile latency (ns)": 5463131,
    "90.00 percentile latency (ns)": 5869285,
    "95.00 percentile latency (ns)": 5940906,
    "97.00 percentile latency (ns)": 5991730,
    "99.00 percentile latency (ns)": 6510670,
    "99.90 percentile latency (ns)": 11445524,
    "Completed samples per second": 130020.85,
    "Max latency (ns)": 32363503,
    "Mean latency (ns)": 5492948,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 4843600,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "LWIS_Server",
    "Scenario": "server",
    "Scheduled samples per second": 130021.94,
    "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": 76.046,
    "characteristics.good": 38023,
    "characteristics.scheduled_queries_per_second": 130021.94,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 32505.485,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 32505.485,
    "characteristics.total": 50000,
    "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT",
    "ck_used": false,
    "cooling": "",
    "dataset": "ImageNet 2012",
    "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md",
    "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": "resnet50-v1.5",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package",
    "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": "resnet50",
    "input_data_types": "int8",
    "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": 2048,
    "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": "resnet50_v1.onnx",
    "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)": 15000000,
    "target_qps": 130000,
    "task": "image classification",
    "task2": "image classification",
    "total_cores": 64,
    "uid": "718394db6965417e",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 5915311,
    "90.00 percentile latency (ns)": 6409306,
    "95.00 percentile latency (ns)": 6521198,
    "97.00 percentile latency (ns)": 6608699,
    "99.00 percentile latency (ns)": 7030112,
    "99.90 percentile latency (ns)": 13248061,
    "Completed samples per second": 94855.78,
    "Max latency (ns)": 1004814577,
    "Mean latency (ns)": 5941747,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 5071718,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "Triton_Server",
    "Scenario": "server",
    "Scheduled samples per second": 95014.52,
    "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": 76.044,
    "characteristics.good": 38022,
    "characteristics.scheduled_queries_per_second": 95014.52,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 23753.63,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 23753.63,
    "characteristics.total": 50000,
    "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT_Triton",
    "ck_used": false,
    "cooling": "",
    "dataset": "ImageNet 2012",
    "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md",
    "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": "resnet50-v1.5",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package",
    "framework": "TensorRT 8.0.2, CUDA 11.3",
    "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": "15 TB",
    "host_storage_type": "NVMe SSD",
    "hw_notes": "",
    "informal_model": "resnet50",
    "input_data_types": "int8",
    "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_Triton",
    "number_of_nodes": 1,
    "number_of_type_nics_installed": "",
    "operating_system": "Ubuntu 20.04.4",
    "other_hardware": "",
    "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, DALI 0.31.0, Triton 21.07",
    "performance_issue_same": 0,
    "performance_issue_same_index": 0,
    "performance_issue_unique": 0,
    "performance_sample_count": 2048,
    "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": "resnet50_v1.onnx",
    "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_Triton",
    "system_name": "HPE ProLiant XL645d Gen10 Plus (4x A100-SXM-80GB, TensorRT, Triton)",
    "system_type": "datacenter",
    "target_latency (ns)": 15000000,
    "target_qps": 95000,
    "task": "image classification",
    "task2": "image classification",
    "total_cores": 64,
    "uid": "4a4928320e0880cd",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 3553799,
    "90.00 percentile latency (ns)": 3936188,
    "95.00 percentile latency (ns)": 4045319,
    "97.00 percentile latency (ns)": 4156691,
    "99.00 percentile latency (ns)": 5475035,
    "99.90 percentile latency (ns)": 13218214,
    "Completed samples per second": 79861.25,
    "Max latency (ns)": 1001547165,
    "Mean latency (ns)": 3599314,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 2595700,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "Triton_Server",
    "Scenario": "server",
    "Scheduled samples per second": 79994.56,
    "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": 76.03,
    "characteristics.good": 38015,
    "characteristics.scheduled_queries_per_second": 79994.56,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 19998.64,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 19998.64,
    "characteristics.total": 50000,
    "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT_Triton",
    "ck_used": true,
    "cooling": "",
    "dataset": "ImageNet 2012",
    "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md",
    "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": "resnet50-v1.5",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package",
    "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": "resnet50",
    "input_data_types": "int8",
    "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_Triton",
    "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, Triton 21.02; GCC 7.5.0; Python 3.7.10",
    "performance_issue_same": 0,
    "performance_issue_same_index": 0,
    "performance_issue_unique": 0,
    "performance_sample_count": 2048,
    "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": "resnet50_v1.onnx",
    "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_Triton",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 15000000,
    "target_qps": 80015,
    "task": "image classification",
    "task2": "image classification",
    "total_cores": 120,
    "uid": "9ddf0432d8cbf484",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 4222172,
    "90.00 percentile latency (ns)": 4801215,
    "95.00 percentile latency (ns)": 5017636,
    "97.00 percentile latency (ns)": 5183267,
    "99.00 percentile latency (ns)": 5849071,
    "99.90 percentile latency (ns)": 11760043,
    "Completed samples per second": 122986.19,
    "Max latency (ns)": 20344001,
    "Mean latency (ns)": 4233398,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 2397131,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "LWIS_Server",
    "Scenario": "server",
    "Scheduled samples per second": 122986.96,
    "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": 76.03,
    "characteristics.good": 38015,
    "characteristics.scheduled_queries_per_second": 122986.96,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 30746.74,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 30746.74,
    "characteristics.total": 50000,
    "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT",
    "ck_used": true,
    "cooling": "",
    "dataset": "ImageNet 2012",
    "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md",
    "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": "resnet50-v1.5",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package",
    "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": "resnet50",
    "input_data_types": "int8",
    "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": 2048,
    "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": "resnet50_v1.onnx",
    "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)": 15000000,
    "target_qps": 123000,
    "task": "image classification",
    "task2": "image classification",
    "total_cores": 120,
    "uid": "a9caa83292c43da8",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  }
]