[ { "50.00 percentile latency (ns)": 342384343514, "90.00 percentile latency (ns)": 604195425642, "95.00 percentile latency (ns)": 636922856989, "97.00 percentile latency (ns)": 650002146442, "99.00 percentile latency (ns)": 663090793650, "99.90 percentile latency (ns)": 669748171267, "Max latency (ns)": 671061622442, "Mean latency (ns)": 2988210534, "Min duration satisfied": "Yes", "Min latency (ns)": 14986097900, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 1376920.0, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "24 GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A30", "accelerator_on-chip_memories": "", "accelerators_per_node": 10, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.samples_per_second": 1376920.0, "characteristics.samples_per_second.normalized_per_core": 137692.0, "characteristics.samples_per_second.normalized_per_processor": 137692.0, "ck_system": "A30x10_TRT", "ck_used": false, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "dlrm", "formal_model_accuracy": 99.9, "formal_model_link": "", "framework": "TensorRT 8.0.1, CUDA 11.3", "host_memory_capacity": "2 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 38, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8368", "host_processors_per_node": 2, "host_storage_capacity": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "dlrm-99.9", "input_data_types": "int8", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 10, "normalize_processors": 10, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Nettrix/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Nettrix/results/A30x10_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.1", "other_hardware": "", "other_software_stack": "TensorRT 8.0.1, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, DALI 0.31.0", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 204800, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 924000000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "Nettrix", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Nettrix", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A30x10_TRT", "system_name": "Nettrix X640 G40 (10x A30, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 1400000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 76, "uid": "e2a3e71a855b719f", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 342384343514, "90.00 percentile latency (ns)": 604195425642, "95.00 percentile latency (ns)": 636922856989, "97.00 percentile latency (ns)": 650002146442, "99.00 percentile latency (ns)": 663090793650, "99.90 percentile latency (ns)": 669748171267, "Max latency (ns)": 671061622442, "Mean latency (ns)": 2988210534, "Min duration satisfied": "Yes", "Min latency (ns)": 14986097900, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 1376920.0, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "24 GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A30", "accelerator_on-chip_memories": "", "accelerators_per_node": 10, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.samples_per_second": 1376920.0, "characteristics.samples_per_second.normalized_per_core": 137692.0, "characteristics.samples_per_second.normalized_per_processor": 137692.0, "ck_system": "A30x10_TRT", "ck_used": false, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "dlrm", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "TensorRT 8.0.1, CUDA 11.3", "host_memory_capacity": "2 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 38, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8368", "host_processors_per_node": 2, "host_storage_capacity": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "dlrm-99", "input_data_types": "int8", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 10, "normalize_processors": 10, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Nettrix/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Nettrix/results/A30x10_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.1", "other_hardware": "", "other_software_stack": "TensorRT 8.0.1, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, DALI 0.31.0", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 204800, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 924000000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "Nettrix", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Nettrix", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A30x10_TRT", "system_name": "Nettrix X640 G40 (10x A30, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 1400000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 76, "uid": "9044677b9b6a222b", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 33971106964, "90.00 percentile latency (ns)": 60466696354, "95.00 percentile latency (ns)": 63804248510, "97.00 percentile latency (ns)": 65135348649, "99.00 percentile latency (ns)": 66482768720, "99.90 percentile latency (ns)": 67079004131, "Max latency (ns)": 67140668359, "Mean latency (ns)": 34043368841, "Min duration satisfied": "Yes", "Min latency (ns)": 1286588700, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 534758, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA T4", "accelerator_on-chip_memories": "", "accelerators_per_node": 16, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.samples_per_second": 534758, "characteristics.samples_per_second.normalized_per_core": 33422.375, "characteristics.samples_per_second.normalized_per_processor": 33422.375, "ck_system": "T4x16_TRT", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "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": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC off", "informal_model": "dlrm-99.9", "input_data_types": "int8", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1, "mlperf_version": 0.7, "normalize_cores": 16, "normalize_processors": 16, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/results/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": 35904000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "Nettrix", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Nettrix", "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/T4x16_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 544000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "4a030f9b34709b16", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 33971106964, "90.00 percentile latency (ns)": 60466696354, "95.00 percentile latency (ns)": 63804248510, "97.00 percentile latency (ns)": 65135348649, "99.00 percentile latency (ns)": 66482768720, "99.90 percentile latency (ns)": 67079004131, "Max latency (ns)": 67140668359, "Mean latency (ns)": 34043368841, "Min duration satisfied": "Yes", "Min latency (ns)": 1286588700, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 534758, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA T4", "accelerator_on-chip_memories": "", "accelerators_per_node": 16, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.samples_per_second": 534758, "characteristics.samples_per_second.normalized_per_core": 33422.375, "characteristics.samples_per_second.normalized_per_processor": 33422.375, "ck_system": "T4x16_TRT", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "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": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC off", "informal_model": "dlrm-99", "input_data_types": "int8", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1, "mlperf_version": 0.7, "normalize_cores": 16, "normalize_processors": 16, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/results/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": 35904000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "Nettrix", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Nettrix", "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/T4x16_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 544000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "11bdaefe1cde0841", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 33918223323, "90.00 percentile latency (ns)": 59438624759, "95.00 percentile latency (ns)": 62637442146, "97.00 percentile latency (ns)": 63916185540, "99.00 percentile latency (ns)": 65195558061, "99.90 percentile latency (ns)": 65770291472, "Max latency (ns)": 65834348877, "Mean latency (ns)": 33930860292, "Min duration satisfied": "Yes", "Min latency (ns)": 2015429209, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 921417, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A100-PCIe", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.samples_per_second": 921417, "characteristics.samples_per_second.normalized_per_core": 230354.25, "characteristics.samples_per_second.normalized_per_processor": 230354.25, "ck_system": "A100-PCIex4_TRT", "ck_used": false, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "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": "192 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 56, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8280M CPU @ 2.70GHz", "host_processors_per_node": 2, "host_storage_capacity": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "dlrm-99.9", "input_data_types": "int8", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1, "mlperf_version": 0.7, "normalize_cores": 4, "normalize_processors": 4, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/results/A100-PCIex4_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.3", "other_software_stack": "TensorRT 7.2, CUDA 11.0 Update 1, cuDNN 8.0.2, DALI 0.25.0", "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": 60660864, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "Nettrix", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Nettrix", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A100-PCIex4_TRT", "system_name": "Nettrix X640 G30 (4x A100-PCIe)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 919104, "task": "recommendation", "task2": "recommendation", "total_cores": 112, "uid": "ae6d4bbd8cadb090", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 33918223323, "90.00 percentile latency (ns)": 59438624759, "95.00 percentile latency (ns)": 62637442146, "97.00 percentile latency (ns)": 63916185540, "99.00 percentile latency (ns)": 65195558061, "99.90 percentile latency (ns)": 65770291472, "Max latency (ns)": 65834348877, "Mean latency (ns)": 33930860292, "Min duration satisfied": "Yes", "Min latency (ns)": 2015429209, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 921417, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A100-PCIe", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.samples_per_second": 921417, "characteristics.samples_per_second.normalized_per_core": 230354.25, "characteristics.samples_per_second.normalized_per_processor": 230354.25, "ck_system": "A100-PCIex4_TRT", "ck_used": false, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "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": "192 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 56, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8280M CPU @ 2.70GHz", "host_processors_per_node": 2, "host_storage_capacity": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "dlrm-99", "input_data_types": "int8", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1, "mlperf_version": 0.7, "normalize_cores": 4, "normalize_processors": 4, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/results/A100-PCIex4_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.3", "other_software_stack": "TensorRT 7.2, CUDA 11.0 Update 1, cuDNN 8.0.2, DALI 0.25.0", "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": 60660864, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "Nettrix", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Nettrix", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A100-PCIex4_TRT", "system_name": "Nettrix X640 G30 (4x A100-PCIe)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 919104, "task": "recommendation", "task2": "recommendation", "total_cores": 112, "uid": "1246a38a71d4fbf9", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 35194168885, "90.00 percentile latency (ns)": 62338233855, "95.00 percentile latency (ns)": 65739356577, "97.00 percentile latency (ns)": 67100543465, "99.00 percentile latency (ns)": 68461088955, "99.90 percentile latency (ns)": 69071508797, "Max latency (ns)": 69141539836, "Mean latency (ns)": 35240372213, "Min duration satisfied": "Yes", "Min latency (ns)": 1475564379, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 681558, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA T4", "accelerator_on-chip_memories": "", "accelerators_per_node": 21, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.samples_per_second": 681558, "characteristics.samples_per_second.normalized_per_core": 32455.14285714286, "characteristics.samples_per_second.normalized_per_processor": 32455.14285714286, "ck_system": "T4x21_TRT", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "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": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC off", "informal_model": "dlrm-99.9", "input_data_types": "int8", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1, "mlperf_version": 0.7, "normalize_cores": 21, "normalize_processors": 21, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/results/T4x21_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": 47124000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "Nettrix", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Nettrix", "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/T4x21_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 714000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "38a31d634cd79f1a", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 35194168885, "90.00 percentile latency (ns)": 62338233855, "95.00 percentile latency (ns)": 65739356577, "97.00 percentile latency (ns)": 67100543465, "99.00 percentile latency (ns)": 68461088955, "99.90 percentile latency (ns)": 69071508797, "Max latency (ns)": 69141539836, "Mean latency (ns)": 35240372213, "Min duration satisfied": "Yes", "Min latency (ns)": 1475564379, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 681558, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA T4", "accelerator_on-chip_memories": "", "accelerators_per_node": 21, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.samples_per_second": 681558, "characteristics.samples_per_second.normalized_per_core": 32455.14285714286, "characteristics.samples_per_second.normalized_per_processor": 32455.14285714286, "ck_system": "T4x21_TRT", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "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": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC off", "informal_model": "dlrm-99", "input_data_types": "int8", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1, "mlperf_version": 0.7, "normalize_cores": 21, "normalize_processors": 21, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Nettrix/results/T4x21_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": 47124000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "Nettrix", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Nettrix", "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/T4x21_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 714000, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "882ea4cdeaa25e57", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" } ]