[ { "50.00 percentile latency (ns)": 4017351, "90.00 percentile latency (ns)": 7347266, "95.00 percentile latency (ns)": 8381065, "97.00 percentile latency (ns)": 8943585, "99.00 percentile latency (ns)": 9793473, "99.90 percentile latency (ns)": 11452192, "Completed samples per second": 126530.75, "Max latency (ns)": 19683731, "Mean latency (ns)": 4311500, "Min duration satisfied": "Yes", "Min latency (ns)": 239433, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 126531.36, "accelerator_cooling_type": "passive", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA T4", "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": "1.1.3", "characteristics.power": 835.5230000000006, "characteristics.power.normalized_per_core": 208.88075000000015, "characteristics.power.normalized_per_processor": 208.88075000000015, "characteristics.scheduled_queries_per_second": 126531.36, "characteristics.scheduled_queries_per_second.normalized_per_core": 31632.84, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31632.84, "ck_system": "XE2420_datacenter_T4x4_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", "filesystem": "ext3/ext4", "formal_model": "dlrm", "formal_model_accuracy": 99.9, "formal_model_link": "", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_cooling_type": "passive", "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": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC on", "informal_model": "dlrm-99.9", "input_data_types": "int8", "management_firmware_version": "4.40.10.00", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.0, "normalize_cores": 4, "normalize_processors": 4, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/results/XE2420_datacenter_T4x4_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 7.2.3, 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_settings": "closed/DellEMC/measurements/XE2420_datacenter_T4x4_TRT/XE2420_T4x4_power_settings.adoc", "print_timestamps": 0, "problem": false, "psu_details": "2x2000W", "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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "sw_notes": "Powered by CK v2.5.8 (https://github.com/ctuning/ck)", "system_cooling_type": "air", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/XE2420_datacenter_T4x4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 126550, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "24f561fe894008a6", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 3986754, "90.00 percentile latency (ns)": 7313540, "95.00 percentile latency (ns)": 8363555, "97.00 percentile latency (ns)": 8932407, "99.00 percentile latency (ns)": 9782400, "99.90 percentile latency (ns)": 11334965, "Completed samples per second": 126530.71, "Max latency (ns)": 19464765, "Mean latency (ns)": 4286588, "Min duration satisfied": "Yes", "Min latency (ns)": 233929, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 126531.36, "accelerator_cooling_type": "passive", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA T4", "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": "1.1.3", "characteristics.power": 835.9083333333339, "characteristics.power.normalized_per_core": 208.97708333333347, "characteristics.power.normalized_per_processor": 208.97708333333347, "characteristics.scheduled_queries_per_second": 126531.36, "characteristics.scheduled_queries_per_second.normalized_per_core": 31632.84, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31632.84, "ck_system": "XE2420_datacenter_T4x4_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", "filesystem": "ext3/ext4", "formal_model": "dlrm", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_cooling_type": "passive", "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": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC on", "informal_model": "dlrm-99", "input_data_types": "int8", "management_firmware_version": "4.40.10.00", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.0, "normalize_cores": 4, "normalize_processors": 4, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/results/XE2420_datacenter_T4x4_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 7.2.3, 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_settings": "closed/DellEMC/measurements/XE2420_datacenter_T4x4_TRT/XE2420_T4x4_power_settings.adoc", "print_timestamps": 0, "problem": false, "psu_details": "2x2000W", "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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "sw_notes": "Powered by CK v2.5.8 (https://github.com/ctuning/ck)", "system_cooling_type": "air", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/XE2420_datacenter_T4x4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 126550, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "65532728421b45c5", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 1797214, "90.00 percentile latency (ns)": 4463735, "95.00 percentile latency (ns)": 6166069, "97.00 percentile latency (ns)": 7924552, "99.00 percentile latency (ns)": 12896851, "99.90 percentile latency (ns)": 24772282, "Completed samples per second": 620116.87, "Max latency (ns)": 68669645, "Mean latency (ns)": 2480235, "Min duration satisfied": "Yes", "Min latency (ns)": 181476, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 620118.17, "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": 10, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.scheduled_queries_per_second": 620118.17, "characteristics.scheduled_queries_per_second.normalized_per_core": 62011.817, "characteristics.scheduled_queries_per_second.normalized_per_processor": 62011.817, "ck_system": "DSS8440_A40x10_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": "3.84 TB", "host_storage_type": "NVMe 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": 10, "normalize_processors": 10, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/results/DSS8440_A40x10_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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/DSS8440_A40x10_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 620000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "bea8c144de9abc88", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 1797214, "90.00 percentile latency (ns)": 4463735, "95.00 percentile latency (ns)": 6166069, "97.00 percentile latency (ns)": 7924552, "99.00 percentile latency (ns)": 12896851, "99.90 percentile latency (ns)": 24772282, "Completed samples per second": 620116.87, "Max latency (ns)": 68669645, "Mean latency (ns)": 2480235, "Min duration satisfied": "Yes", "Min latency (ns)": 181476, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 620118.17, "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": 10, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.scheduled_queries_per_second": 620118.17, "characteristics.scheduled_queries_per_second.normalized_per_core": 62011.817, "characteristics.scheduled_queries_per_second.normalized_per_processor": 62011.817, "ck_system": "DSS8440_A40x10_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": "3.84 TB", "host_storage_type": "NVMe 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": 10, "normalize_processors": 10, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/results/DSS8440_A40x10_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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/DSS8440_A40x10_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 620000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "58cfa468c94d0deb", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 1707509, "90.00 percentile latency (ns)": 2896942, "95.00 percentile latency (ns)": 3338644, "97.00 percentile latency (ns)": 3652513, "99.00 percentile latency (ns)": 4772308, "99.90 percentile latency (ns)": 285875300, "Completed samples per second": 400020.59, "Max latency (ns)": 372703976, "Mean latency (ns)": 2658032, "Min duration satisfied": "Yes", "Min latency (ns)": 132351, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 400021.0, "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": 3, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 400021.0, "characteristics.scheduled_queries_per_second.normalized_per_core": 133340.33333333334, "characteristics.scheduled_queries_per_second.normalized_per_processor": 133340.33333333334, "ck_system": "R740_A100-PCIe-40GBx3_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 7.2.3, 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": "1.8 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC on", "informal_model": "dlrm-99.9", "input_data_types": "int8", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.0, "normalize_cores": 3, "normalize_processors": 3, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/results/R740_A100-PCIe-40GBx3_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 7.2.3, 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, "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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R740_A100-PCIe-40GBx3_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 400000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "b96292d69975b099", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 1707509, "90.00 percentile latency (ns)": 2896942, "95.00 percentile latency (ns)": 3338644, "97.00 percentile latency (ns)": 3652513, "99.00 percentile latency (ns)": 4772308, "99.90 percentile latency (ns)": 285875300, "Completed samples per second": 400020.59, "Max latency (ns)": 372703976, "Mean latency (ns)": 2658032, "Min duration satisfied": "Yes", "Min latency (ns)": 132351, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 400021.0, "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": 3, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 400021.0, "characteristics.scheduled_queries_per_second.normalized_per_core": 133340.33333333334, "characteristics.scheduled_queries_per_second.normalized_per_processor": 133340.33333333334, "ck_system": "R740_A100-PCIe-40GBx3_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 7.2.3, 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": "1.8 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC on", "informal_model": "dlrm-99", "input_data_types": "int8", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.0, "normalize_cores": 3, "normalize_processors": 3, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/results/R740_A100-PCIe-40GBx3_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 7.2.3, 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, "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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R740_A100-PCIe-40GBx3_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 400000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "fdd567ff42c44d86", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 5389936, "90.00 percentile latency (ns)": 12019440, "95.00 percentile latency (ns)": 14553985, "97.00 percentile latency (ns)": 16000205, "99.00 percentile latency (ns)": 18346214, "99.90 percentile latency (ns)": 26713562, "Completed samples per second": 223029.98, "Max latency (ns)": 48682446, "Mean latency (ns)": 6306270, "Min duration satisfied": "Yes", "Min latency (ns)": 198999, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 223032.29, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "48 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA Quadro RTX 8000", "accelerator_on-chip_memories": "", "accelerators_per_node": 3, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 223032.29, "characteristics.scheduled_queries_per_second.normalized_per_core": 74344.09666666666, "characteristics.scheduled_queries_per_second.normalized_per_processor": 74344.09666666666, "ck_system": "R7525_QuadroRTX8000x3_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 7.2.3, 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": 2, "host_storage_capacity": "1.8 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC on", "informal_model": "dlrm-99.9", "input_data_types": "int8", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.0, "normalize_cores": 3, "normalize_processors": 3, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/results/R7525_QuadroRTX8000x3_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 7.2.3, 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, "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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R7525_QuadroRTX8000x3_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 223050, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "00998025edc4e919", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 5389936, "90.00 percentile latency (ns)": 12019440, "95.00 percentile latency (ns)": 14553985, "97.00 percentile latency (ns)": 16000205, "99.00 percentile latency (ns)": 18346214, "99.90 percentile latency (ns)": 26713562, "Completed samples per second": 223029.98, "Max latency (ns)": 48682446, "Mean latency (ns)": 6306270, "Min duration satisfied": "Yes", "Min latency (ns)": 198999, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 223032.29, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "48 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA Quadro RTX 8000", "accelerator_on-chip_memories": "", "accelerators_per_node": 3, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 223032.29, "characteristics.scheduled_queries_per_second.normalized_per_core": 74344.09666666666, "characteristics.scheduled_queries_per_second.normalized_per_processor": 74344.09666666666, "ck_system": "R7525_QuadroRTX8000x3_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 7.2.3, 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": 2, "host_storage_capacity": "1.8 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC on", "informal_model": "dlrm-99", "input_data_types": "int8", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.0, "normalize_cores": 3, "normalize_processors": 3, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/results/R7525_QuadroRTX8000x3_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 7.2.3, 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, "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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R7525_QuadroRTX8000x3_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 223050, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "c2d81448492d146d", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 4240920, "90.00 percentile latency (ns)": 7972083, "95.00 percentile latency (ns)": 8883645, "97.00 percentile latency (ns)": 10116231, "99.00 percentile latency (ns)": 28920661, "99.90 percentile latency (ns)": 66610984, "Completed samples per second": 260050.68, "Max latency (ns)": 73151205, "Mean latency (ns)": 5003988, "Min duration satisfied": "Yes", "Min latency (ns)": 226151, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 260062.59, "accelerator_frequency": "1590MHz", "accelerator_host_interconnect": "PCI Express 3.0", "accelerator_interconnect": "PCI Express 3.0", "accelerator_interconnect_topology": "4 Accelerators per CPU", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA Tesla T4", "accelerator_on-chip_memories": "", "accelerators_per_node": 8, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 260062.59, "characteristics.scheduled_queries_per_second.normalized_per_core": 32507.82375, "characteristics.scheduled_queries_per_second.normalized_per_processor": 32507.82375, "ck_system": "R7525_T4x8_TRT", "ck_used": true, "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, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "1 TB", "host_memory_configuration": "8x64GB DDR4-3200 HMAA8GR7AJR4N-XN RDIMM ECC", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "2MB+16MB+128MB", "host_processor_core_count": 120, "host_processor_frequency": "2.35GHz", "host_processor_interconnect": "Infinity Fabric", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "3 TB (5x800GB WUSTR6480ASS200 in RAID5)", "host_storage_type": "3D-TLC Solid State with 12Gbps SAS", "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": 8, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/R7525_T4x8_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": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R7525_T4x8_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 260000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "b38925f9e28c4eda", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 4240920, "90.00 percentile latency (ns)": 7972083, "95.00 percentile latency (ns)": 8883645, "97.00 percentile latency (ns)": 10116231, "99.00 percentile latency (ns)": 28920661, "99.90 percentile latency (ns)": 66610984, "Completed samples per second": 260050.68, "Max latency (ns)": 73151205, "Mean latency (ns)": 5003988, "Min duration satisfied": "Yes", "Min latency (ns)": 226151, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 260062.59, "accelerator_frequency": "1590MHz", "accelerator_host_interconnect": "PCI Express 3.0", "accelerator_interconnect": "PCI Express 3.0", "accelerator_interconnect_topology": "4 Accelerators per CPU", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA Tesla T4", "accelerator_on-chip_memories": "", "accelerators_per_node": 8, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 260062.59, "characteristics.scheduled_queries_per_second.normalized_per_core": 32507.82375, "characteristics.scheduled_queries_per_second.normalized_per_processor": 32507.82375, "ck_system": "R7525_T4x8_TRT", "ck_used": true, "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, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "1 TB", "host_memory_configuration": "8x64GB DDR4-3200 HMAA8GR7AJR4N-XN RDIMM ECC", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "2MB+16MB+128MB", "host_processor_core_count": 120, "host_processor_frequency": "2.35GHz", "host_processor_interconnect": "Infinity Fabric", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "3 TB (5x800GB WUSTR6480ASS200 in RAID5)", "host_storage_type": "3D-TLC Solid State with 12Gbps SAS", "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": 8, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/R7525_T4x8_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": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R7525_T4x8_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 260000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "293a442caaa0f795", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 2814572, "90.00 percentile latency (ns)": 5197274, "95.00 percentile latency (ns)": 6178533, "97.00 percentile latency (ns)": 7093669, "99.00 percentile latency (ns)": 16479284, "99.90 percentile latency (ns)": 52008458, "Completed samples per second": 126507.64, "Max latency (ns)": 62099322, "Mean latency (ns)": 3405626, "Min duration satisfied": "Yes", "Min latency (ns)": 256791, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 126513.98, "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": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 126513.98, "characteristics.scheduled_queries_per_second.normalized_per_core": 31628.495, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31628.495, "ck_system": "XE2420_T4x4_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": "384 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "2.10GHz", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1 TB", "host_storage_type": "NVMe 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/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/XE2420_T4x4_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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/XE2420_T4x4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 126500, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "c2a5bbdf8da6b04a", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 2814572, "90.00 percentile latency (ns)": 5197274, "95.00 percentile latency (ns)": 6178533, "97.00 percentile latency (ns)": 7093669, "99.00 percentile latency (ns)": 16479284, "99.90 percentile latency (ns)": 52008458, "Completed samples per second": 126507.64, "Max latency (ns)": 62099322, "Mean latency (ns)": 3405626, "Min duration satisfied": "Yes", "Min latency (ns)": 256791, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 126513.98, "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": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 126513.98, "characteristics.scheduled_queries_per_second.normalized_per_core": 31628.495, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31628.495, "ck_system": "XE2420_T4x4_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": "384 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "2.10GHz", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1 TB", "host_storage_type": "NVMe 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/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/XE2420_T4x4_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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/XE2420_T4x4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 126500, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "f4592318e9013b58", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 9551549, "90.00 percentile latency (ns)": 12614111, "95.00 percentile latency (ns)": 13580069, "97.00 percentile latency (ns)": 14265206, "99.00 percentile latency (ns)": 16631074, "99.90 percentile latency (ns)": 94473376, "Completed samples per second": 385035.2, "Max latency (ns)": 119675324, "Mean latency (ns)": 9917343, "Min duration satisfied": "Yes", "Min latency (ns)": 200576, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 385085.3, "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, "characteristics.scheduled_queries_per_second": 385085.3, "characteristics.scheduled_queries_per_second.normalized_per_core": 192542.65, "characteristics.scheduled_queries_per_second.normalized_per_processor": 192542.65, "ck_system": "R7525_A100x2_TRT", "ck_used": true, "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, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "2.50GHz", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1.84 TB", "host_storage_type": "NVMe", "hw_notes": "", "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": 2, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/R7525_A100x2_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": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R7525_A100x2_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 385000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "7911a71a701cd29b", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 9551549, "90.00 percentile latency (ns)": 12614111, "95.00 percentile latency (ns)": 13580069, "97.00 percentile latency (ns)": 14265206, "99.00 percentile latency (ns)": 16631074, "99.90 percentile latency (ns)": 94473376, "Completed samples per second": 385035.2, "Max latency (ns)": 119675324, "Mean latency (ns)": 9917343, "Min duration satisfied": "Yes", "Min latency (ns)": 200576, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 385085.3, "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, "characteristics.scheduled_queries_per_second": 385085.3, "characteristics.scheduled_queries_per_second.normalized_per_core": 192542.65, "characteristics.scheduled_queries_per_second.normalized_per_processor": 192542.65, "ck_system": "R7525_A100x2_TRT", "ck_used": true, "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, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "2.50GHz", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1.84 TB", "host_storage_type": "NVMe", "hw_notes": "", "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": 2, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/R7525_A100x2_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": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R7525_A100x2_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 385000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "f446fc98027e3952", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 5846737, "90.00 percentile latency (ns)": 15135023, "95.00 percentile latency (ns)": 18574822, "97.00 percentile latency (ns)": 20089548, "99.00 percentile latency (ns)": 22222354, "99.90 percentile latency (ns)": 25561646, "Completed samples per second": 777727.18, "Max latency (ns)": 38585467, "Mean latency (ns)": 7272711, "Min duration satisfied": "Yes", "Min latency (ns)": 232263, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 777955.77, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "48 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA Quadro RTX 8000", "accelerator_on-chip_memories": "", "accelerators_per_node": 10, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 777955.77, "characteristics.scheduled_queries_per_second.normalized_per_core": 77795.577, "characteristics.scheduled_queries_per_second.normalized_per_processor": 77795.577, "ck_system": "DSS8440_QuadroRTX8000x10_TRT", "ck_used": true, "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, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "768 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "2.10GHz", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1.84 TB", "host_storage_type": "NVMe", "hw_notes": "", "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": 10, "normalize_processors": 10, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/DSS8440_QuadroRTX8000x10_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": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/DSS8440_QuadroRTX8000x10_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 777500, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "9569341809883054", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 5846737, "90.00 percentile latency (ns)": 15135023, "95.00 percentile latency (ns)": 18574822, "97.00 percentile latency (ns)": 20089548, "99.00 percentile latency (ns)": 22222354, "99.90 percentile latency (ns)": 25561646, "Completed samples per second": 777727.18, "Max latency (ns)": 38585467, "Mean latency (ns)": 7272711, "Min duration satisfied": "Yes", "Min latency (ns)": 232263, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 777955.77, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "48 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA Quadro RTX 8000", "accelerator_on-chip_memories": "", "accelerators_per_node": 10, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 777955.77, "characteristics.scheduled_queries_per_second.normalized_per_core": 77795.577, "characteristics.scheduled_queries_per_second.normalized_per_processor": 77795.577, "ck_system": "DSS8440_QuadroRTX8000x10_TRT", "ck_used": true, "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, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "768 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "2.10GHz", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1.84 TB", "host_storage_type": "NVMe", "hw_notes": "", "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": 10, "normalize_processors": 10, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/DSS8440_QuadroRTX8000x10_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": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/DSS8440_QuadroRTX8000x10_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 777500, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "cbe10681684bbb8c", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 5668100, "90.00 percentile latency (ns)": 8099003, "95.00 percentile latency (ns)": 10456538, "97.00 percentile latency (ns)": 12384978, "99.00 percentile latency (ns)": 26580794, "99.90 percentile latency (ns)": 63160889, "Completed samples per second": 503144.5, "Max latency (ns)": 72154623, "Mean latency (ns)": 5614811, "Min duration satisfied": "Yes", "Min latency (ns)": 218992, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 503200.26, "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": 16, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 503200.26, "characteristics.scheduled_queries_per_second.normalized_per_core": 31450.01625, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31450.01625, "ck_system": "DSS8440_T4x16_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": "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": "1.84 TB", "host_storage_type": "NVMe", "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": 16, "normalize_processors": 16, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/DSS8440_T4x16_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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/DSS8440_T4x16_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 503000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "c0d4fc78d5998bd6", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 5668100, "90.00 percentile latency (ns)": 8099003, "95.00 percentile latency (ns)": 10456538, "97.00 percentile latency (ns)": 12384978, "99.00 percentile latency (ns)": 26580794, "99.90 percentile latency (ns)": 63160889, "Completed samples per second": 503144.5, "Max latency (ns)": 72154623, "Mean latency (ns)": 5614811, "Min duration satisfied": "Yes", "Min latency (ns)": 218992, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 503200.26, "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": 16, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 503200.26, "characteristics.scheduled_queries_per_second.normalized_per_core": 31450.01625, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31450.01625, "ck_system": "DSS8440_T4x16_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": "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": "1.84 TB", "host_storage_type": "NVMe", "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": 16, "normalize_processors": 16, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/DSS8440_T4x16_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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/DSS8440_T4x16_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 503000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "d9f4f2b9f4ac791a", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 3163131, "90.00 percentile latency (ns)": 6876396, "95.00 percentile latency (ns)": 9403464, "97.00 percentile latency (ns)": 12380003, "99.00 percentile latency (ns)": 25680651, "99.90 percentile latency (ns)": 59762371, "Completed samples per second": 126485.42, "Max latency (ns)": 67917198, "Mean latency (ns)": 4244183, "Min duration satisfied": "Yes", "Min latency (ns)": 303135, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 126513.98, "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": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 126513.98, "characteristics.scheduled_queries_per_second.normalized_per_core": 31628.495, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31628.495, "ck_system": "R7515_T4x4_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 7.2.0.14, CUDA 11.0.207", "host_memory_capacity": "256 GB", "host_memory_configuration": "DDR-4", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "2.0GHz", "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": "NVMe SSD", "hw_notes": "ECC on", "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/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/R7515_T4x4_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 7.2.0.14, CUDA 11.0.27, 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R7515_T4x4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 126500, "task": "recommendation", "task2": "recommendation", "total_cores": 120, "uid": "a9dfd06495ab53c2", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 3163131, "90.00 percentile latency (ns)": 6876396, "95.00 percentile latency (ns)": 9403464, "97.00 percentile latency (ns)": 12380003, "99.00 percentile latency (ns)": 25680651, "99.90 percentile latency (ns)": 59762371, "Completed samples per second": 126485.42, "Max latency (ns)": 67917198, "Mean latency (ns)": 4244183, "Min duration satisfied": "Yes", "Min latency (ns)": 303135, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 126513.98, "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": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 126513.98, "characteristics.scheduled_queries_per_second.normalized_per_core": 31628.495, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31628.495, "ck_system": "R7515_T4x4_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 7.2.0.14, CUDA 11.0.207", "host_memory_capacity": "256 GB", "host_memory_configuration": "DDR-4", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "2.0GHz", "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": "NVMe SSD", "hw_notes": "ECC on", "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/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/R7515_T4x4_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 7.2.0.14, CUDA 11.0.27, 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R7515_T4x4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 126500, "task": "recommendation", "task2": "recommendation", "total_cores": 120, "uid": "3e42da98b51dd5aa", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 5712177, "90.00 percentile latency (ns)": 8589191, "95.00 percentile latency (ns)": 10499747, "97.00 percentile latency (ns)": 12197523, "99.00 percentile latency (ns)": 25316038, "99.90 percentile latency (ns)": 68447230, "Completed samples per second": 380074.17, "Max latency (ns)": 78639940, "Mean latency (ns)": 6000388, "Min duration satisfied": "Yes", "Min latency (ns)": 253713, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 380082.59, "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": 12, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 380082.59, "characteristics.scheduled_queries_per_second.normalized_per_core": 31673.549166666668, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31673.549166666668, "ck_system": "DSS8440_T4x12_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": "", "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 TB", "host_storage_type": "NVMe 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": 12, "normalize_processors": 12, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/DSS8440_T4x12_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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/DSS8440_T4x12_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 380000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "b6cd73460cb538e5", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 5712177, "90.00 percentile latency (ns)": 8589191, "95.00 percentile latency (ns)": 10499747, "97.00 percentile latency (ns)": 12197523, "99.00 percentile latency (ns)": 25316038, "99.90 percentile latency (ns)": 68447230, "Completed samples per second": 380074.17, "Max latency (ns)": 78639940, "Mean latency (ns)": 6000388, "Min duration satisfied": "Yes", "Min latency (ns)": 253713, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 380082.59, "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": 12, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 380082.59, "characteristics.scheduled_queries_per_second.normalized_per_core": 31673.549166666668, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31673.549166666668, "ck_system": "DSS8440_T4x12_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": "", "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 TB", "host_storage_type": "NVMe 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": 12, "normalize_processors": 12, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/DSS8440_T4x12_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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/DSS8440_T4x12_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 380000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "b10d50148699e80e", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 10950629, "90.00 percentile latency (ns)": 15720559, "95.00 percentile latency (ns)": 17058870, "97.00 percentile latency (ns)": 18825926, "99.00 percentile latency (ns)": 26909301, "99.90 percentile latency (ns)": 39947352, "Completed samples per second": 279344.3, "Max latency (ns)": 84518064, "Mean latency (ns)": 10510213, "Min duration satisfied": "Yes", "Min latency (ns)": 237744, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 279363.86, "accelerator_frequency": "1770MHz", "accelerator_host_interconnect": "PCI Express 3.0", "accelerator_interconnect": "PCI Express 3.0", "accelerator_interconnect_topology": "2 Accelerators per CPU", "accelerator_memory_capacity": "24 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA Quadro RTX 6000", "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": 279363.86, "characteristics.scheduled_queries_per_second.normalized_per_core": 69840.965, "characteristics.scheduled_queries_per_second.normalized_per_processor": 69840.965, "ck_system": "C4140_QuadroRTX6000x4_TRT", "ck_used": true, "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, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "384 GB", "host_memory_configuration": "6x16GB DDR4-2666 HMA82GR7AFR8N-VK RDIMM ECC", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "1.25MB+20MB+27.5MB", "host_processor_core_count": 120, "host_processor_frequency": "2.40GHz", "host_processor_interconnect": "Ultra Path Interconnect", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1.6 TB (1x1.6TB Dell Express Flash PM1725a 1.6TB AIC)", "host_storage_type": "3D-TLC Solid State with PCIe NVME x8 Interface", "hw_notes": "ECC off. RTX6000 is available as a special config thru Dell DSS or OEM for PowerEdge C4140", "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/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/C4140_QuadroRTX6000x4_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": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/C4140_QuadroRTX6000x4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 279300, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "ff24708917a538b5", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 10950629, "90.00 percentile latency (ns)": 15720559, "95.00 percentile latency (ns)": 17058870, "97.00 percentile latency (ns)": 18825926, "99.00 percentile latency (ns)": 26909301, "99.90 percentile latency (ns)": 39947352, "Completed samples per second": 279344.3, "Max latency (ns)": 84518064, "Mean latency (ns)": 10510213, "Min duration satisfied": "Yes", "Min latency (ns)": 237744, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 279363.86, "accelerator_frequency": "1770MHz", "accelerator_host_interconnect": "PCI Express 3.0", "accelerator_interconnect": "PCI Express 3.0", "accelerator_interconnect_topology": "2 Accelerators per CPU", "accelerator_memory_capacity": "24 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA Quadro RTX 6000", "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": 279363.86, "characteristics.scheduled_queries_per_second.normalized_per_core": 69840.965, "characteristics.scheduled_queries_per_second.normalized_per_processor": 69840.965, "ck_system": "C4140_QuadroRTX6000x4_TRT", "ck_used": true, "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, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "384 GB", "host_memory_configuration": "6x16GB DDR4-2666 HMA82GR7AFR8N-VK RDIMM ECC", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "1.25MB+20MB+27.5MB", "host_processor_core_count": 120, "host_processor_frequency": "2.40GHz", "host_processor_interconnect": "Ultra Path Interconnect", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1.6 TB (1x1.6TB Dell Express Flash PM1725a 1.6TB AIC)", "host_storage_type": "3D-TLC Solid State with PCIe NVME x8 Interface", "hw_notes": "ECC off. RTX6000 is available as a special config thru Dell DSS or OEM for PowerEdge C4140", "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/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/C4140_QuadroRTX6000x4_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": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/C4140_QuadroRTX6000x4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 279300, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "9499a917c27cef95", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 9595613, "90.00 percentile latency (ns)": 16201583, "95.00 percentile latency (ns)": 17373119, "97.00 percentile latency (ns)": 18098219, "99.00 percentile latency (ns)": 20427349, "99.90 percentile latency (ns)": 31042959, "Completed samples per second": 233058.29, "Max latency (ns)": 49330232, "Mean latency (ns)": 9869348, "Min duration satisfied": "Yes", "Min latency (ns)": 178184, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 233102.78, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "48 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA Quadro RTX 8000", "accelerator_on-chip_memories": "", "accelerators_per_node": 3, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 233102.78, "characteristics.scheduled_queries_per_second.normalized_per_core": 77700.92666666667, "characteristics.scheduled_queries_per_second.normalized_per_processor": 77700.92666666667, "ck_system": "R7525_QuadroRTX8000x3_TRT", "ck_used": false, "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, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 32, "host_processor_frequency": "2.50GHz", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7502", "host_processors_per_node": 2, "host_storage_capacity": "1.84 TB", "host_storage_type": "NVMe", "hw_notes": "", "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": 3, "normalize_processors": 3, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/R7525_QuadroRTX8000x3_TRT", "number_of_nodes": 1, "operating_system": "CentOS Linux release 8.1.1911", "other_software_stack": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1", "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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/R7525_QuadroRTX8000x3_TRT", "system_name": "Dell EMC PowerEdge R7525 (3x Quadro RTX 8000)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 233050, "task": "recommendation", "task2": "recommendation", "total_cores": 64, "uid": "396f92996c0f85d5", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 9595613, "90.00 percentile latency (ns)": 16201583, "95.00 percentile latency (ns)": 17373119, "97.00 percentile latency (ns)": 18098219, "99.00 percentile latency (ns)": 20427349, "99.90 percentile latency (ns)": 31042959, "Completed samples per second": 233058.29, "Max latency (ns)": 49330232, "Mean latency (ns)": 9869348, "Min duration satisfied": "Yes", "Min latency (ns)": 178184, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 233102.78, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "48 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA Quadro RTX 8000", "accelerator_on-chip_memories": "", "accelerators_per_node": 3, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 233102.78, "characteristics.scheduled_queries_per_second.normalized_per_core": 77700.92666666667, "characteristics.scheduled_queries_per_second.normalized_per_processor": 77700.92666666667, "ck_system": "R7525_QuadroRTX8000x3_TRT", "ck_used": false, "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, cuDNN 8.0.2, cuBLAS 11.2.0, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 32, "host_processor_frequency": "2.50GHz", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7502", "host_processors_per_node": 2, "host_storage_capacity": "1.84 TB", "host_storage_type": "NVMe", "hw_notes": "", "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": 3, "normalize_processors": 3, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/R7525_QuadroRTX8000x3_TRT", "number_of_nodes": 1, "operating_system": "CentOS Linux release 8.1.1911", "other_software_stack": "docker 19.03.12, python 3.6.8, gcc 5.5.0, onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1", "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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/R7525_QuadroRTX8000x3_TRT", "system_name": "Dell EMC PowerEdge R7525 (3x Quadro RTX 8000)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 233050, "task": "recommendation", "task2": "recommendation", "total_cores": 64, "uid": "9764e1389bd95955", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 3137321, "90.00 percentile latency (ns)": 6215910, "95.00 percentile latency (ns)": 7602037, "97.00 percentile latency (ns)": 8701981, "99.00 percentile latency (ns)": 20204875, "99.90 percentile latency (ns)": 56842193, "Completed samples per second": 126010.92, "Max latency (ns)": 65263329, "Mean latency (ns)": 3888135, "Min duration satisfied": "Yes", "Min latency (ns)": 260940, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 126014.85, "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": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 126014.85, "characteristics.scheduled_queries_per_second.normalized_per_core": 31503.7125, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31503.7125, "ck_system": "R740_T4x4_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 7.2.0.14, CUDA 11.0.207", "host_memory_capacity": "384 GB", "host_memory_configuration": "DDR-4", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "3.0GHz", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "3.84 TB", "host_storage_type": "SSD", "hw_notes": "ECC on", "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/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/R740_T4x4_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 7.2.0.14, CUDA 11.0.207, 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R740_T4x4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 126000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "4056805dcf1f61c8", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 3137321, "90.00 percentile latency (ns)": 6215910, "95.00 percentile latency (ns)": 7602037, "97.00 percentile latency (ns)": 8701981, "99.00 percentile latency (ns)": 20204875, "99.90 percentile latency (ns)": 56842193, "Completed samples per second": 126010.92, "Max latency (ns)": 65263329, "Mean latency (ns)": 3888135, "Min duration satisfied": "Yes", "Min latency (ns)": 260940, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "DLRM SERVER", "Scenario": "server", "Scheduled samples per second": 126014.85, "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": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.scheduled_queries_per_second": 126014.85, "characteristics.scheduled_queries_per_second.normalized_per_core": 31503.7125, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31503.7125, "ck_system": "R740_T4x4_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 7.2.0.14, CUDA 11.0.207", "host_memory_capacity": "384 GB", "host_memory_configuration": "DDR-4", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "3.0GHz", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "3.84 TB", "host_storage_type": "SSD", "hw_notes": "ECC on", "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/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/DellEMC/results/R740_T4x4_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 7.2.0.14, CUDA 11.0.207, 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": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "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/R740_T4x4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 126000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "bfcd9573832434b8", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" } ]