[
  {
    "50.00 percentile latency (ns)": 6266888,
    "90.00 percentile latency (ns)": 9223940,
    "95.00 percentile latency (ns)": 13398391,
    "97.00 percentile latency (ns)": 16403887,
    "99.00 percentile latency (ns)": 21932393,
    "99.90 percentile latency (ns)": 30249570,
    "Completed samples per second": 270048.3,
    "Max latency (ns)": 51663492,
    "Mean latency (ns)": 6929420,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 411800,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 270049.67,
    "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": 8,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.scheduled_queries_per_second": 270049.67,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 33756.20875,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 33756.20875,
    "ck_system": "A100-PCIex8_TRT",
    "ck_used": false,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.scheduled_queries_per_second",
    "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 8.0.2, CUDA 11.3",
    "host_memory_capacity": "768 GB",
    "host_memory_configuration": "",
    "host_networking": "",
    "host_networking_topology": "",
    "host_processor_caches": "",
    "host_processor_core_count": 28,
    "host_processor_frequency": "",
    "host_processor_interconnect": "",
    "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8280 CPU @ 2.60GHz",
    "host_processors_per_node": 2,
    "host_storage_capacity": "2 TB",
    "host_storage_type": "SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99.9",
    "input_data_types": "int8",
    "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": 8,
    "normalize_processors": 8,
    "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Lenovo/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Lenovo/results/A100-PCIex8_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 11.3, cuDNN 8.2.1, Driver 470.57.02",
    "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": 1,
    "schedule_rng_seed": 10051496985653635065,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "available",
    "submitter": "Lenovo",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Lenovo",
    "sw_notes": "",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A100-PCIex8_TRT",
    "system_name": "Lenovo ThinkSystem SR670-v2 Server",
    "system_type": "datacenter",
    "target_latency (ns)": 30000000,
    "target_qps": 270000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 56,
    "uid": "42e8089639c522a9",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 6266888,
    "90.00 percentile latency (ns)": 9223940,
    "95.00 percentile latency (ns)": 13398391,
    "97.00 percentile latency (ns)": 16403887,
    "99.00 percentile latency (ns)": 21932393,
    "99.90 percentile latency (ns)": 30249570,
    "Completed samples per second": 270048.3,
    "Max latency (ns)": 51663492,
    "Mean latency (ns)": 6929420,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 411800,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 270049.67,
    "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": 8,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.scheduled_queries_per_second": 270049.67,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 33756.20875,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 33756.20875,
    "ck_system": "A100-PCIex8_TRT",
    "ck_used": false,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.scheduled_queries_per_second",
    "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 8.0.2, CUDA 11.3",
    "host_memory_capacity": "768 GB",
    "host_memory_configuration": "",
    "host_networking": "",
    "host_networking_topology": "",
    "host_processor_caches": "",
    "host_processor_core_count": 28,
    "host_processor_frequency": "",
    "host_processor_interconnect": "",
    "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8280 CPU @ 2.60GHz",
    "host_processors_per_node": 2,
    "host_storage_capacity": "2 TB",
    "host_storage_type": "SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99",
    "input_data_types": "int8",
    "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": 8,
    "normalize_processors": 8,
    "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Lenovo/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Lenovo/results/A100-PCIex8_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 11.3, cuDNN 8.2.1, Driver 470.57.02",
    "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": 1,
    "schedule_rng_seed": 10051496985653635065,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "available",
    "submitter": "Lenovo",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Lenovo",
    "sw_notes": "",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A100-PCIex8_TRT",
    "system_name": "Lenovo ThinkSystem SR670-v2 Server",
    "system_type": "datacenter",
    "target_latency (ns)": 30000000,
    "target_qps": 270000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 56,
    "uid": "bca28ae448628cff",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 1936911,
    "90.00 percentile latency (ns)": 4810592,
    "95.00 percentile latency (ns)": 6990297,
    "97.00 percentile latency (ns)": 9313119,
    "99.00 percentile latency (ns)": 15589038,
    "99.90 percentile latency (ns)": 33576046,
    "Completed samples per second": 315995.66,
    "Max latency (ns)": 54632555,
    "Mean latency (ns)": 2656588,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 167913,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 315997.24,
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "48 GB",
    "accelerator_memory_configuration": "GDDR6",
    "accelerator_model_name": "NVIDIA A40",
    "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.scheduled_queries_per_second": 315997.24,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 78999.31,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 78999.31,
    "ck_system": "A40x4_TRT",
    "ck_used": true,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.scheduled_queries_per_second",
    "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": "704 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": "2 TB",
    "host_storage_type": "SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99.9",
    "input_data_types": "int8",
    "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/Lenovo/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Lenovo/results/A40x4_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 460.39; 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": 1,
    "schedule_rng_seed": 3507442325620259414,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "available",
    "submitter": "Lenovo",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Lenovo",
    "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/A40x4_TRT",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 30000000,
    "target_qps": 316000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 240,
    "uid": "0b4a5b8aabb51ed7",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 1936911,
    "90.00 percentile latency (ns)": 4810592,
    "95.00 percentile latency (ns)": 6990297,
    "97.00 percentile latency (ns)": 9313119,
    "99.00 percentile latency (ns)": 15589038,
    "99.90 percentile latency (ns)": 33576046,
    "Completed samples per second": 315995.66,
    "Max latency (ns)": 54632555,
    "Mean latency (ns)": 2656588,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 167913,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 315997.24,
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "48 GB",
    "accelerator_memory_configuration": "GDDR6",
    "accelerator_model_name": "NVIDIA A40",
    "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.scheduled_queries_per_second": 315997.24,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 78999.31,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 78999.31,
    "ck_system": "A40x4_TRT",
    "ck_used": true,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.scheduled_queries_per_second",
    "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": "704 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": "2 TB",
    "host_storage_type": "SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99",
    "input_data_types": "int8",
    "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/Lenovo/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Lenovo/results/A40x4_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 460.39; 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": 1,
    "schedule_rng_seed": 3507442325620259414,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "available",
    "submitter": "Lenovo",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Lenovo",
    "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/A40x4_TRT",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 30000000,
    "target_qps": 316000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 240,
    "uid": "d48c8addd6937f63",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 9407518,
    "90.00 percentile latency (ns)": 14064239,
    "95.00 percentile latency (ns)": 15112533,
    "97.00 percentile latency (ns)": 15804277,
    "99.00 percentile latency (ns)": 17446145,
    "99.90 percentile latency (ns)": 28216405,
    "Completed samples per second": 402260.22,
    "Max latency (ns)": 51050043,
    "Mean latency (ns)": 7023136,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 132198,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 402260.47,
    "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": 4,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.scheduled_queries_per_second": 402260.47,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 100565.1175,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 100565.1175,
    "ck_system": "A100-PCIex4_TRT",
    "ck_used": true,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.scheduled_queries_per_second",
    "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": "768 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": "2 TB",
    "host_storage_type": "SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99.9",
    "input_data_types": "int8",
    "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/Lenovo/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Lenovo/results/A100-PCIex4_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 460.32.03, 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": 1,
    "schedule_rng_seed": 3507442325620259414,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "available",
    "submitter": "Lenovo",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Lenovo",
    "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-PCIex4_TRT",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 30000000,
    "target_qps": 402241,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 240,
    "uid": "2813f94ab44fbb80",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 9407518,
    "90.00 percentile latency (ns)": 14064239,
    "95.00 percentile latency (ns)": 15112533,
    "97.00 percentile latency (ns)": 15804277,
    "99.00 percentile latency (ns)": 17446145,
    "99.90 percentile latency (ns)": 28216405,
    "Completed samples per second": 402260.22,
    "Max latency (ns)": 51050043,
    "Mean latency (ns)": 7023136,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 132198,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 402260.47,
    "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": 4,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "",
    "characteristics.scheduled_queries_per_second": 402260.47,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 100565.1175,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 100565.1175,
    "ck_system": "A100-PCIex4_TRT",
    "ck_used": true,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.scheduled_queries_per_second",
    "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": "768 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": "2 TB",
    "host_storage_type": "SSD",
    "hw_notes": "",
    "informal_model": "dlrm-99",
    "input_data_types": "int8",
    "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/Lenovo/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Lenovo/results/A100-PCIex4_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 460.32.03, 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": 1,
    "schedule_rng_seed": 3507442325620259414,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "available",
    "submitter": "Lenovo",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Lenovo",
    "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-PCIex4_TRT",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 30000000,
    "target_qps": 402241,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 240,
    "uid": "20d5420b895af631",
    "use_accelerator": true,
    "weight_data_types": "int8",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 8418462,
    "90.00 percentile latency (ns)": 10951431,
    "95.00 percentile latency (ns)": 11751886,
    "97.00 percentile latency (ns)": 12306486,
    "99.00 percentile latency (ns)": 18326159,
    "99.90 percentile latency (ns)": 28027725,
    "Completed samples per second": 638196.68,
    "Max latency (ns)": 36723624,
    "Mean latency (ns)": 8307908,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 184192,
    "Min queries satisfied": "Yes",
    "Mode": "Performance",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 638262.86,
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "40GB",
    "accelerator_memory_configuration": "HBM2",
    "accelerator_model_name": "NVIDIA A100-PCIe",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 4,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "characteristics.scheduled_queries_per_second": 638262.86,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 159565.715,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 159565.715,
    "ck_system": "A100-PCIex4_TRT72",
    "ck_used": true,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.scheduled_queries_per_second",
    "dim_y_default": "characteristics.AUC",
    "dim_y_maximize": true,
    "division": "closed",
    "formal_model": "dlrm",
    "formal_model_accuracy": 99.9,
    "formal_model_link": "",
    "framework": "TensorRT 7.2, CUDA 11.0 Update 1",
    "host_memory_capacity": "770 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": "25.6 TB",
    "host_storage_type": "SAS 2.5 SSD",
    "hw_notes": "ECC off",
    "informal_model": "dlrm-99.9",
    "input_data_types": "int8",
    "max_async_queries": 0,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 60000,
    "min_query_count": 270336,
    "mlperf_version": 0.7,
    "normalize_cores": 4,
    "normalize_processors": 4,
    "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Lenovo/code",
    "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Lenovo/results/A100-PCIex4_TRT72",
    "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": "TensorRT 7.2, CUDA 11.0 Update 1, cuDNN 8.0.2, DALI 0.25.0; GCC 7.5.0; Python 3.7.10",
    "performance_issue_same": true,
    "performance_issue_same_index": 0,
    "performance_issue_unique": true,
    "performance_sample_count": 204800,
    "print_timestamps": true,
    "problem": false,
    "qsl_rng_seed": 12786827339337101903,
    "retraining": "N",
    "sample_index_rng_seed": 12640797754436136668,
    "samples_per_query": 1,
    "schedule_rng_seed": 3135815929913719677,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "available",
    "submitter": "Lenovo",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Lenovo",
    "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-PCIex4_TRT72",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 30000000,
    "target_qps": 638000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 240,
    "uid": "d3f31479f2466fbd",
    "use_accelerator": true,
    "weight_data_types": "int8,fp16",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 8418462,
    "90.00 percentile latency (ns)": 10951431,
    "95.00 percentile latency (ns)": 11751886,
    "97.00 percentile latency (ns)": 12306486,
    "99.00 percentile latency (ns)": 18326159,
    "99.90 percentile latency (ns)": 28027725,
    "Completed samples per second": 638196.68,
    "Max latency (ns)": 36723624,
    "Mean latency (ns)": 8307908,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 184192,
    "Min queries satisfied": "Yes",
    "Mode": "Performance",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 638262.86,
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "40GB",
    "accelerator_memory_configuration": "HBM2",
    "accelerator_model_name": "NVIDIA A100-PCIe",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 4,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "characteristics.scheduled_queries_per_second": 638262.86,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 159565.715,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 159565.715,
    "ck_system": "A100-PCIex4_TRT72",
    "ck_used": true,
    "cooling": "",
    "dataset": "1TB Click Logs",
    "dataset_link": "",
    "dim_x_default": "characteristics.scheduled_queries_per_second",
    "dim_y_default": "characteristics.AUC",
    "dim_y_maximize": true,
    "division": "closed",
    "formal_model": "dlrm",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "",
    "framework": "TensorRT 7.2, CUDA 11.0 Update 1",
    "host_memory_capacity": "770 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": "25.6 TB",
    "host_storage_type": "SAS 2.5 SSD",
    "hw_notes": "ECC off",
    "informal_model": "dlrm-99",
    "input_data_types": "int8",
    "max_async_queries": 0,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 60000,
    "min_query_count": 270336,
    "mlperf_version": 0.7,
    "normalize_cores": 4,
    "normalize_processors": 4,
    "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Lenovo/code",
    "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Lenovo/results/A100-PCIex4_TRT72",
    "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": "TensorRT 7.2, CUDA 11.0 Update 1, cuDNN 8.0.2, DALI 0.25.0; GCC 7.5.0; Python 3.7.10",
    "performance_issue_same": true,
    "performance_issue_same_index": 0,
    "performance_issue_unique": true,
    "performance_sample_count": 204800,
    "print_timestamps": true,
    "problem": false,
    "qsl_rng_seed": 12786827339337101903,
    "retraining": "N",
    "sample_index_rng_seed": 12640797754436136668,
    "samples_per_query": 1,
    "schedule_rng_seed": 3135815929913719677,
    "starting_weights_filename": "tb00_40M.pt",
    "status": "available",
    "submitter": "Lenovo",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Lenovo",
    "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-PCIex4_TRT72",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 30000000,
    "target_qps": 638000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 240,
    "uid": "c41abc9a48bf8732",
    "use_accelerator": true,
    "weight_data_types": "int8,fp16",
    "weight_transformations": "quantization, affine fusion"
  }
]