[
  {
    "50.00 percentile latency (ns)": 2584707,
    "90.00 percentile latency (ns)": 4674027,
    "95.00 percentile latency (ns)": 5474032,
    "97.00 percentile latency (ns)": 6056683,
    "99.00 percentile latency (ns)": 19768540,
    "99.90 percentile latency (ns)": 61435117,
    "Completed samples per second": 189000.42,
    "Max latency (ns)": 68536999,
    "Mean latency (ns)": 3145175,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 199400,
    "Min queries satisfied": "Yes",
    "Mode": "Performance",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 189018.81,
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "16 GB",
    "accelerator_memory_configuration": "GDDR6",
    "accelerator_model_name": "NVIDIA T4",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 6,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "characteristics.scheduled_queries_per_second": 189018.81,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 31503.135,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 31503.135,
    "ck_system": "T4x6_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,
    "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": "768 GB",
    "host_memory_configuration": "16 x 64G 2Rx4 PC4-2933Y-R",
    "host_networking": "",
    "host_networking_topology": "",
    "host_processor_caches": "38.5 MB",
    "host_processor_core_count": 120,
    "host_processor_frequency": "2.70 GHz",
    "host_processor_interconnect": "",
    "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor",
    "host_processors_per_node": 2,
    "host_storage_capacity": "800 GB",
    "host_storage_type": "SATA 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": 6,
    "normalize_processors": 6,
    "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Cisco/code",
    "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Cisco/results/T4x6_TRT",
    "number_of_nodes": 1,
    "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)",
    "other_software_stack": "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": "CISCO",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Cisco",
    "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/T4x6_TRT",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 30000000,
    "target_qps": 189000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 240,
    "uid": "31d33850bf64d9ab",
    "use_accelerator": true,
    "weight_data_types": "int8,fp16",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 2584707,
    "90.00 percentile latency (ns)": 4674027,
    "95.00 percentile latency (ns)": 5474032,
    "97.00 percentile latency (ns)": 6056683,
    "99.00 percentile latency (ns)": 19768540,
    "99.90 percentile latency (ns)": 61435117,
    "Completed samples per second": 189000.42,
    "Max latency (ns)": 68536999,
    "Mean latency (ns)": 3145175,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 199400,
    "Min queries satisfied": "Yes",
    "Mode": "Performance",
    "Performance constraints satisfied": "Yes",
    "Result is": "VALID",
    "SUT name": "DLRM SERVER",
    "Scenario": "server",
    "Scheduled samples per second": 189018.81,
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "16 GB",
    "accelerator_memory_configuration": "GDDR6",
    "accelerator_model_name": "NVIDIA T4",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 6,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "characteristics.scheduled_queries_per_second": 189018.81,
    "characteristics.scheduled_queries_per_second.normalized_per_core": 31503.135,
    "characteristics.scheduled_queries_per_second.normalized_per_processor": 31503.135,
    "ck_system": "T4x6_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,
    "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": "768 GB",
    "host_memory_configuration": "16 x 64G 2Rx4 PC4-2933Y-R",
    "host_networking": "",
    "host_networking_topology": "",
    "host_processor_caches": "38.5 MB",
    "host_processor_core_count": 120,
    "host_processor_frequency": "2.70 GHz",
    "host_processor_interconnect": "",
    "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor",
    "host_processors_per_node": 2,
    "host_storage_capacity": "800 GB",
    "host_storage_type": "SATA 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": 6,
    "normalize_processors": 6,
    "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Cisco/code",
    "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Cisco/results/T4x6_TRT",
    "number_of_nodes": 1,
    "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)",
    "other_software_stack": "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": "CISCO",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Cisco",
    "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/T4x6_TRT",
    "system_name": "Microsoft Corporation 7.0 (Virtual Machine)",
    "system_type": "datacenter",
    "target_latency (ns)": 30000000,
    "target_qps": 189000,
    "task": "recommendation",
    "task2": "recommendation",
    "total_cores": 240,
    "uid": "625725348647b08f",
    "use_accelerator": true,
    "weight_data_types": "int8,fp16",
    "weight_transformations": "quantization, affine fusion"
  }
]