[ { "50.00 percentile latency (ns)": 336659110978, "90.00 percentile latency (ns)": 605775202580, "95.00 percentile latency (ns)": 639450947374, "97.00 percentile latency (ns)": 652892839140, "99.00 percentile latency (ns)": 666331193935, "99.90 percentile latency (ns)": 672415664349, "Max latency (ns)": 673058591304, "Mean latency (ns)": 336661471230, "Min duration satisfied": "Yes", "Min latency (ns)": 192149502, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 4853.96, "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.mAP": 20.117, "characteristics.samples_per_second": 4853.96, "characteristics.samples_per_second.normalized_per_core": 485.396, "characteristics.samples_per_second.normalized_per_processor": 485.396, "ck_system": "A30x10_TRT", "ck_used": false, "cooling": "", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "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": "ssd-resnet34", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "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": 64, "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": 3267000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "resnet34-ssd1200.pytorch", "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": 4950, "task": "object detection", "task2": "object detection", "total_cores": 76, "uid": "4858af541ded5a2d", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 329714783482, "90.00 percentile latency (ns)": 594055560708, "95.00 percentile latency (ns)": 627094910752, "97.00 percentile latency (ns)": 640312213751, "99.00 percentile latency (ns)": 653517281638, "99.90 percentile latency (ns)": 659452196097, "Max latency (ns)": 660027148660, "Mean latency (ns)": 329764743068, "Min duration satisfied": "Yes", "Min latency (ns)": 199591706, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 2499.9, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA A10", "accelerator_on-chip_memories": "", "accelerators_per_node": 8, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.mAP": 20.113, "characteristics.samples_per_second": 2499.9, "characteristics.samples_per_second.normalized_per_core": 312.4875, "characteristics.samples_per_second.normalized_per_processor": 312.4875, "ck_system": "A10x8_TRT", "ck_used": false, "cooling": "", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.1, CUDA 11.3", "host_memory_capacity": "2048 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 40, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8380 CPU @ 2.30GHz", "host_processors_per_node": 2, "host_storage_capacity": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "ssd-resnet34", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "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": 8, "normalize_processors": 8, "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/A10x8_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": 64, "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": 1650000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "resnet34-ssd1200.pytorch", "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/A10x8_TRT", "system_name": "Nettrix X620 G40 (8x A10, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 2500, "task": "object detection", "task2": "object detection", "total_cores": 80, "uid": "e4e1e424a84f11a8", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 34110464572, "90.00 percentile latency (ns)": 61741648993, "95.00 percentile latency (ns)": 65218033272, "97.00 percentile latency (ns)": 66603887039, "99.00 percentile latency (ns)": 67998100612, "99.90 percentile latency (ns)": 68617266729, "Max latency (ns)": 68706438244, "Mean latency (ns)": 34203998552, "Min duration satisfied": "Yes", "Min latency (ns)": 77748837, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 2257.43, "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.mAP": 20.111, "characteristics.samples_per_second": 2257.43, "characteristics.samples_per_second.normalized_per_core": 141.089375, "characteristics.samples_per_second.normalized_per_processor": 141.089375, "ck_system": "T4x16_TRT", "ck_used": true, "cooling": "", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "division": "closed", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "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": "ssd-resnet34", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "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": 64, "print_timestamps": true, "problem": false, "qsl_rng_seed": 12786827339337101903, "retraining": "N", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 155100, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "resnet34-ssd1200.pytorch", "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": 2350, "task": "object detection", "task2": "object detection", "total_cores": 240, "uid": "d85bbabfba41f261", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 35471702151, "90.00 percentile latency (ns)": 63923335560, "95.00 percentile latency (ns)": 67477549183, "97.00 percentile latency (ns)": 68906379126, "99.00 percentile latency (ns)": 70329188874, "99.90 percentile latency (ns)": 70935086417, "Max latency (ns)": 70988694258, "Mean latency (ns)": 35509373080, "Min duration satisfied": "Yes", "Min latency (ns)": 164991059, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 3439.98, "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.mAP": 20.111, "characteristics.samples_per_second": 3439.98, "characteristics.samples_per_second.normalized_per_core": 859.995, "characteristics.samples_per_second.normalized_per_processor": 859.995, "ck_system": "A100-PCIex4_TRT", "ck_used": false, "cooling": "", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "division": "closed", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "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": "ssd-resnet34", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "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": 64, "print_timestamps": true, "problem": false, "qsl_rng_seed": 12786827339337101903, "retraining": "N", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 244200, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "resnet34-ssd1200.pytorch", "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": 3700, "task": "object detection", "task2": "object detection", "total_cores": 112, "uid": "49b788aa5ff9b97e", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 35850532296, "90.00 percentile latency (ns)": 64600920679, "95.00 percentile latency (ns)": 68180319033, "97.00 percentile latency (ns)": 69649259022, "99.00 percentile latency (ns)": 71060212741, "99.90 percentile latency (ns)": 71679402535, "Max latency (ns)": 71720616628, "Mean latency (ns)": 35879311669, "Min duration satisfied": "Yes", "Min latency (ns)": 164545740, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 3404.88, "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.mAP": 20.111, "characteristics.samples_per_second": 3404.88, "characteristics.samples_per_second.normalized_per_core": 851.22, "characteristics.samples_per_second.normalized_per_processor": 851.22, "ck_system": "A100-PCIex4_TRT_AEP", "ck_used": true, "cooling": "", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "division": "closed", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2, CUDA 11.0 Update 1", "host_memory_capacity": "128 GB DDR4 + 512 GB AEP ", "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": "ssd-resnet34", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "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_AEP", "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": 64, "print_timestamps": true, "problem": false, "qsl_rng_seed": 12786827339337101903, "retraining": "N", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 244200, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "resnet34-ssd1200.pytorch", "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/A100-PCIex4_TRT_AEP", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 3700, "task": "object detection", "task2": "object detection", "total_cores": 240, "uid": "064603b2c6304469", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 33492457591, "90.00 percentile latency (ns)": 60410332100, "95.00 percentile latency (ns)": 63784818044, "97.00 percentile latency (ns)": 65142631409, "99.00 percentile latency (ns)": 66489652111, "99.90 percentile latency (ns)": 67057066833, "Max latency (ns)": 67103072733, "Mean latency (ns)": 33549547232, "Min duration satisfied": "Yes", "Min latency (ns)": 159479618, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 2950.68, "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.mAP": 20.111, "characteristics.samples_per_second": 2950.68, "characteristics.samples_per_second.normalized_per_core": 140.50857142857143, "characteristics.samples_per_second.normalized_per_processor": 140.50857142857143, "ck_system": "T4x21_TRT", "ck_used": true, "cooling": "", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "division": "closed", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "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": "ssd-resnet34", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "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": 64, "print_timestamps": true, "problem": false, "qsl_rng_seed": 12786827339337101903, "retraining": "N", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 198000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "resnet34-ssd1200.pytorch", "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": 3000, "task": "object detection", "task2": "object detection", "total_cores": 240, "uid": "dbcdca5626c6ead6", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 34219674633, "90.00 percentile latency (ns)": 61895096921, "95.00 percentile latency (ns)": 65373771403, "97.00 percentile latency (ns)": 66764642503, "99.00 percentile latency (ns)": 68148939018, "99.90 percentile latency (ns)": 68780640099, "Max latency (ns)": 68842795422, "Mean latency (ns)": 34300419651, "Min duration satisfied": "Yes", "Min latency (ns)": 79021923, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 2252.96, "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.mAP": 20.111, "characteristics.samples_per_second": 2252.96, "characteristics.samples_per_second.normalized_per_core": 140.81, "characteristics.samples_per_second.normalized_per_processor": 140.81, "ck_system": "T4x16_TRT_AEP", "ck_used": true, "cooling": "", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "division": "closed", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2, CUDA 11.0 Update 1", "host_memory_capacity": "512 GB DDR4 +512 GB AEP", "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": "ssd-resnet34", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "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_AEP", "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": 64, "print_timestamps": true, "problem": false, "qsl_rng_seed": 12786827339337101903, "retraining": "N", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 155100, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "resnet34-ssd1200.pytorch", "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_AEP", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 2350, "task": "object detection", "task2": "object detection", "total_cores": 240, "uid": "26488403ea941aee", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" } ]