[
  {
    "50.00 percentile latency (ns)": 670092138661,
    "90.00 percentile latency (ns)": 1204090050815,
    "95.00 percentile latency (ns)": 1270732788411,
    "97.00 percentile latency (ns)": 1297391039330,
    "99.00 percentile latency (ns)": 1324043980162,
    "99.90 percentile latency (ns)": 1346442207544,
    "Max latency (ns)": 1349126565012,
    "Mean latency (ns)": 21746158455,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 4820087993,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Samples per second": 273955,
    "Scenario": "offline",
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "24 GB",
    "accelerator_memory_configuration": "HBM2",
    "accelerator_model_name": "NVIDIA A30",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 2,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.samples_per_second": 273955,
    "characteristics.samples_per_second.normalized_per_core": 136977.5,
    "characteristics.samples_per_second.normalized_per_processor": 136977.5,
    "ck_system": "A30x2_TRT",
    "ck_used": false,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.samples_per_second",
    "dim_x_maximize": true,
    "dim_y_default": "characteristics.AUC",
    "dim_y_maximize": true,
    "disk_controllers": "",
    "disk_drives": "",
    "division": "closed",
    "filesystem": "",
    "formal_model": "dlrm",
    "formal_model_accuracy": 99.9,
    "formal_model_link": "",
    "framework": "TensorRT 8.0.1, 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 7302",
    "host_processors_per_node": 2,
    "host_storage_capacity": "1 TB",
    "host_storage_type": " SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99.9",
    "input_data_types": "int8",
    "management_firmware_version": "",
    "max_async_queries": 1,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 600000,
    "min_query_count": 1,
    "mlperf_version": 1.1,
    "network_speed_mbit": "",
    "nics_enabled_connected": "",
    "nics_enabled_firmware": "",
    "nics_enabled_os": "",
    "normalize_cores": 2,
    "normalize_processors": 2,
    "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Fujitsu/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Fujitsu/results/A30x2_TRT",
    "number_of_nodes": 1,
    "number_of_type_nics_installed": "",
    "operating_system": "Ubuntu 20.04.2 LTS",
    "other_hardware": "",
    "other_software_stack": "TensorRT 8.0.1, CUDA 11.3, cuDNN 8.2.1.32, Driver 470.57, DALI 0.31.0",
    "performance_issue_same": 0,
    "performance_issue_same_index": 0,
    "performance_issue_unique": 0,
    "performance_sample_count": 204800,
    "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": 369600000,
    "schedule_rng_seed": 10051496985653635065,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "preview",
    "submitter": "Fujitsu",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Fujitsu",
    "sw_notes": "",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A30x2_TRT",
    "system_name": "PRIMERGY GX2460M1 (2x A30, TensorRT)",
    "system_type": "datacenter",
    "target_latency (ns)": 0,
    "target_qps": 560000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 128,
    "uid": "feaea66944d86a15",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 670092138661,
    "90.00 percentile latency (ns)": 1204090050815,
    "95.00 percentile latency (ns)": 1270732788411,
    "97.00 percentile latency (ns)": 1297391039330,
    "99.00 percentile latency (ns)": 1324043980162,
    "99.90 percentile latency (ns)": 1346442207544,
    "Max latency (ns)": 1349126565012,
    "Mean latency (ns)": 21746158455,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 4820087993,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Samples per second": 273955,
    "Scenario": "offline",
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "24 GB",
    "accelerator_memory_configuration": "HBM2",
    "accelerator_model_name": "NVIDIA A30",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 2,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.samples_per_second": 273955,
    "characteristics.samples_per_second.normalized_per_core": 136977.5,
    "characteristics.samples_per_second.normalized_per_processor": 136977.5,
    "ck_system": "A30x2_TRT",
    "ck_used": false,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.samples_per_second",
    "dim_x_maximize": true,
    "dim_y_default": "characteristics.AUC",
    "dim_y_maximize": true,
    "disk_controllers": "",
    "disk_drives": "",
    "division": "closed",
    "filesystem": "",
    "formal_model": "dlrm",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "",
    "framework": "TensorRT 8.0.1, 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 7302",
    "host_processors_per_node": 2,
    "host_storage_capacity": "1 TB",
    "host_storage_type": " SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99",
    "input_data_types": "int8",
    "management_firmware_version": "",
    "max_async_queries": 1,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 600000,
    "min_query_count": 1,
    "mlperf_version": 1.1,
    "network_speed_mbit": "",
    "nics_enabled_connected": "",
    "nics_enabled_firmware": "",
    "nics_enabled_os": "",
    "normalize_cores": 2,
    "normalize_processors": 2,
    "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Fujitsu/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Fujitsu/results/A30x2_TRT",
    "number_of_nodes": 1,
    "number_of_type_nics_installed": "",
    "operating_system": "Ubuntu 20.04.2 LTS",
    "other_hardware": "",
    "other_software_stack": "TensorRT 8.0.1, CUDA 11.3, cuDNN 8.2.1.32, Driver 470.57, DALI 0.31.0",
    "performance_issue_same": 0,
    "performance_issue_same_index": 0,
    "performance_issue_unique": 0,
    "performance_sample_count": 204800,
    "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": 369600000,
    "schedule_rng_seed": 10051496985653635065,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "preview",
    "submitter": "Fujitsu",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Fujitsu",
    "sw_notes": "",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A30x2_TRT",
    "system_name": "PRIMERGY GX2460M1 (2x A30, TensorRT)",
    "system_type": "datacenter",
    "target_latency (ns)": 0,
    "target_qps": 560000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 128,
    "uid": "b3a15b94b3a80404",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 343804706827,
    "90.00 percentile latency (ns)": 615829190259,
    "95.00 percentile latency (ns)": 649825421067,
    "97.00 percentile latency (ns)": 663424244596,
    "99.00 percentile latency (ns)": 677020715744,
    "99.90 percentile latency (ns)": 684096521976,
    "Max latency (ns)": 685457198723,
    "Mean latency (ns)": 33323784978,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 4975672633,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Samples per second": 519945,
    "Scenario": "offline",
    "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": 2,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.samples_per_second": 519945,
    "characteristics.samples_per_second.normalized_per_core": 259972.5,
    "characteristics.samples_per_second.normalized_per_processor": 259972.5,
    "ck_system": "A100-PCIex2_TRT",
    "ck_used": true,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.samples_per_second",
    "dim_x_maximize": true,
    "dim_y_default": "characteristics.AUC",
    "dim_y_maximize": true,
    "disk_controllers": "",
    "disk_drives": "",
    "division": "closed",
    "filesystem": "",
    "formal_model": "dlrm",
    "formal_model_accuracy": 99.9,
    "formal_model_link": "",
    "framework": "TensorRT 7.2.3, CUDA 11.1",
    "host_memory_capacity": "384 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": "1.92 TB",
    "host_storage_type": "SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99.9",
    "input_data_types": "int8",
    "management_firmware_version": "",
    "max_async_queries": 1,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 600000,
    "min_query_count": 1,
    "mlperf_version": 1.0,
    "network_speed_mbit": "",
    "nics_enabled_connected": "",
    "nics_enabled_firmware": "",
    "nics_enabled_os": "",
    "normalize_cores": 2,
    "normalize_processors": 2,
    "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Fujitsu/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Fujitsu/results/A100-PCIex2_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 455.45.01, 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": 204800,
    "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": 356400000,
    "schedule_rng_seed": 3507442325620259414,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "preview",
    "submitter": "Fujitsu",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Fujitsu",
    "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/A100-PCIex2_TRT",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 0,
    "target_qps": 540000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 240,
    "uid": "2df9f81134e3ea5a",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 343804706827,
    "90.00 percentile latency (ns)": 615829190259,
    "95.00 percentile latency (ns)": 649825421067,
    "97.00 percentile latency (ns)": 663424244596,
    "99.00 percentile latency (ns)": 677020715744,
    "99.90 percentile latency (ns)": 684096521976,
    "Max latency (ns)": 685457198723,
    "Mean latency (ns)": 33323784978,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 4975672633,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Samples per second": 519945,
    "Scenario": "offline",
    "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": 2,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.samples_per_second": 519945,
    "characteristics.samples_per_second.normalized_per_core": 259972.5,
    "characteristics.samples_per_second.normalized_per_processor": 259972.5,
    "ck_system": "A100-PCIex2_TRT",
    "ck_used": true,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.samples_per_second",
    "dim_x_maximize": true,
    "dim_y_default": "characteristics.AUC",
    "dim_y_maximize": true,
    "disk_controllers": "",
    "disk_drives": "",
    "division": "closed",
    "filesystem": "",
    "formal_model": "dlrm",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "",
    "framework": "TensorRT 7.2.3, CUDA 11.1",
    "host_memory_capacity": "384 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": "1.92 TB",
    "host_storage_type": "SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99",
    "input_data_types": "int8",
    "management_firmware_version": "",
    "max_async_queries": 1,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 600000,
    "min_query_count": 1,
    "mlperf_version": 1.0,
    "network_speed_mbit": "",
    "nics_enabled_connected": "",
    "nics_enabled_firmware": "",
    "nics_enabled_os": "",
    "normalize_cores": 2,
    "normalize_processors": 2,
    "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Fujitsu/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Fujitsu/results/A100-PCIex2_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 455.45.01, 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": 204800,
    "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": 356400000,
    "schedule_rng_seed": 3507442325620259414,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "preview",
    "submitter": "Fujitsu",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Fujitsu",
    "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/A100-PCIex2_TRT",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 0,
    "target_qps": 540000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 240,
    "uid": "cc948d29d2e148ee",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 390726534926,
    "90.00 percentile latency (ns)": 704350041344,
    "95.00 percentile latency (ns)": 743511063227,
    "97.00 percentile latency (ns)": 759168033629,
    "99.00 percentile latency (ns)": 774817784224,
    "99.90 percentile latency (ns)": 787698658663,
    "Max latency (ns)": 789270303075,
    "Mean latency (ns)": 38387340569,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 1722883310,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Samples per second": 198685,
    "Scenario": "offline",
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "16 GB",
    "accelerator_memory_configuration": "GDDR6",
    "accelerator_model_name": "NVIDIA A10",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 2,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.samples_per_second": 198685,
    "characteristics.samples_per_second.normalized_per_core": 99342.5,
    "characteristics.samples_per_second.normalized_per_processor": 99342.5,
    "ck_system": "A10x2_TRT",
    "ck_used": false,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.samples_per_second",
    "dim_x_maximize": true,
    "dim_y_default": "characteristics.AUC",
    "dim_y_maximize": true,
    "disk_controllers": "",
    "disk_drives": "",
    "division": "closed",
    "filesystem": "",
    "formal_model": "dlrm",
    "formal_model_accuracy": 99.9,
    "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": 16,
    "host_processor_frequency": "",
    "host_processor_interconnect": "",
    "host_processor_model_name": "AMD EPYC 7302",
    "host_processors_per_node": 2,
    "host_storage_capacity": "1.92 TB",
    "host_storage_type": "SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99.9",
    "input_data_types": "int8",
    "management_firmware_version": "",
    "max_async_queries": 1,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 600000,
    "min_query_count": 1,
    "mlperf_version": 1.0,
    "network_speed_mbit": "",
    "nics_enabled_connected": "",
    "nics_enabled_firmware": "",
    "nics_enabled_os": "",
    "normalize_cores": 2,
    "normalize_processors": 2,
    "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Fujitsu/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Fujitsu/results/A10x2_TRT",
    "number_of_nodes": 1,
    "number_of_type_nics_installed": "",
    "operating_system": "Ubuntu 18.04.5",
    "other_hardware": "",
    "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 455.45.01, DALI 0.30.0",
    "performance_issue_same": 0,
    "performance_issue_same_index": 0,
    "performance_issue_unique": 0,
    "performance_sample_count": 204800,
    "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": 156816000,
    "schedule_rng_seed": 3507442325620259414,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "preview",
    "submitter": "Fujitsu",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Fujitsu",
    "sw_notes": "",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A10x2_TRT",
    "system_name": "Fujitsu PRIMERGY GX2460 m1 (2x A10, TensorRT)",
    "system_type": "datacenter",
    "target_latency (ns)": 0,
    "target_qps": 237600,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 32,
    "uid": "5b2e01445d663581",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 390726534926,
    "90.00 percentile latency (ns)": 704350041344,
    "95.00 percentile latency (ns)": 743511063227,
    "97.00 percentile latency (ns)": 759168033629,
    "99.00 percentile latency (ns)": 774817784224,
    "99.90 percentile latency (ns)": 787698658663,
    "Max latency (ns)": 789270303075,
    "Mean latency (ns)": 38387340569,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 1722883310,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Samples per second": 198685,
    "Scenario": "offline",
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "16 GB",
    "accelerator_memory_configuration": "GDDR6",
    "accelerator_model_name": "NVIDIA A10",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 2,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.samples_per_second": 198685,
    "characteristics.samples_per_second.normalized_per_core": 99342.5,
    "characteristics.samples_per_second.normalized_per_processor": 99342.5,
    "ck_system": "A10x2_TRT",
    "ck_used": false,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.samples_per_second",
    "dim_x_maximize": true,
    "dim_y_default": "characteristics.AUC",
    "dim_y_maximize": true,
    "disk_controllers": "",
    "disk_drives": "",
    "division": "closed",
    "filesystem": "",
    "formal_model": "dlrm",
    "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": 16,
    "host_processor_frequency": "",
    "host_processor_interconnect": "",
    "host_processor_model_name": "AMD EPYC 7302",
    "host_processors_per_node": 2,
    "host_storage_capacity": "1.92 TB",
    "host_storage_type": "SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99",
    "input_data_types": "int8",
    "management_firmware_version": "",
    "max_async_queries": 1,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 600000,
    "min_query_count": 1,
    "mlperf_version": 1.0,
    "network_speed_mbit": "",
    "nics_enabled_connected": "",
    "nics_enabled_firmware": "",
    "nics_enabled_os": "",
    "normalize_cores": 2,
    "normalize_processors": 2,
    "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Fujitsu/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Fujitsu/results/A10x2_TRT",
    "number_of_nodes": 1,
    "number_of_type_nics_installed": "",
    "operating_system": "Ubuntu 18.04.5",
    "other_hardware": "",
    "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 455.45.01, DALI 0.30.0",
    "performance_issue_same": 0,
    "performance_issue_same_index": 0,
    "performance_issue_unique": 0,
    "performance_sample_count": 204800,
    "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": 156816000,
    "schedule_rng_seed": 3507442325620259414,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "preview",
    "submitter": "Fujitsu",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Fujitsu",
    "sw_notes": "",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A10x2_TRT",
    "system_name": "Fujitsu PRIMERGY GX2460 m1 (2x A10, TensorRT)",
    "system_type": "datacenter",
    "target_latency (ns)": 0,
    "target_qps": 237600,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 32,
    "uid": "47d9bab46f8f187f",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  }
]