[ { "50.00 percentile": 1, "50.00 percentile latency (ns)": 28158862, "90.00 percentile": 1, "90.00 percentile latency (ns)": 45015400, "95.00 percentile": 1, "95.00 percentile latency (ns)": 45298478, "97.00 percentile": 1, "97.00 percentile latency (ns)": 45451018, "99.00 percentile": 1, "99.00 percentile latency (ns)": 45665932, "99.90 percentile": 1, "99.90 percentile latency (ns)": 45943192, "Max latency (ns)": 94789928, "Mean latency (ns)": 28081551, "Min duration satisfied": "Yes", "Min latency (ns)": 10252833, "Min queries satisfied": "Yes", "Mode": "Performance", "Per-sample latency": "", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per query": 2560, "Scenario": "multistream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "24 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA TITAN RTX", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "characteristics.accuracy": 76.034, "characteristics.good": 38017, "characteristics.samples_per_query": 2560, "characteristics.samples_per_query.normalized_per_core": 640.0, "characteristics.samples_per_query.normalized_per_processor": 640.0, "characteristics.total": 50000, "ck_system": "TitanRTXx4", "ck_used": false, "cooling": "watercooled", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_query", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 6.0, CUDA 10.1, cuDNN 7.6.3, 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": 24, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) 8268", "host_processors_per_node": 2, "host_storage_capacity": "3.84 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "mlperf_version": 0.5, "normalize_cores": 4, "normalize_processors": 4, "note_code": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/code", "note_details": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/results/TitanRTXx4", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.3", "other_software_stack": "docker 18.09.2, 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": 1024, "print_timestamps": true, "problem": false, "qsl_rng_seed": 3133965575612453542, "retraining": "N", "sample_index_rng_seed": 665484352860916858, "samples_per_query": 2560, "schedule_rng_seed": 3622009729038561421, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "NVIDIA", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/NVIDIA", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/TitanRTXx4", "system_name": "SCAN 3XS DBP T496X2 Fluid", "target_latency (ns)": 50000000, "target_qps": 20, "task": "image classification", "task2": "image classification", "total_cores": 48, "uid": "9202c944d090de16", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile": 1, "50.00 percentile latency (ns)": 25029176, "90.00 percentile": 1, "90.00 percentile latency (ns)": 42555508, "95.00 percentile": 1, "95.00 percentile latency (ns)": 45011511, "97.00 percentile": 1, "97.00 percentile latency (ns)": 46012982, "99.00 percentile": 1, "99.00 percentile latency (ns)": 46546546, "99.90 percentile": 2, "99.90 percentile latency (ns)": 48176881, "Max latency (ns)": 92420637, "Mean latency (ns)": 24773667, "Min duration satisfied": "Yes", "Min latency (ns)": 2562045, "Min queries satisfied": "Yes", "Mode": "Performance", "Per-sample latency": "", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per query": 8704, "Scenario": "multistream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "24 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA TITAN RTX", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "characteristics.accuracy": 70.814, "characteristics.good": 35407, "characteristics.samples_per_query": 8704, "characteristics.samples_per_query.normalized_per_core": 2176.0, "characteristics.samples_per_query.normalized_per_processor": 2176.0, "characteristics.total": 50000, "ck_system": "TitanRTXx4", "ck_used": false, "cooling": "watercooled", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_query", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 6.0, CUDA 10.1, cuDNN 7.6.3, 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": 24, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) 8268", "host_processors_per_node": 2, "host_storage_capacity": "3.84 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "mobilenet", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "mlperf_version": 0.5, "normalize_cores": 4, "normalize_processors": 4, "note_code": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/code", "note_details": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/results/TitanRTXx4", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.3", "other_software_stack": "docker 18.09.2, 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": 1024, "print_timestamps": true, "problem": false, "qsl_rng_seed": 3133965575612453542, "retraining": "N", "sample_index_rng_seed": 665484352860916858, "samples_per_query": 8704, "schedule_rng_seed": 3622009729038561421, "starting_weights_filename": "mobilenet_sym_no_bn.onnx", "status": "available", "submitter": "NVIDIA", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/NVIDIA", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/TitanRTXx4", "system_name": "SCAN 3XS DBP T496X2 Fluid", "target_latency (ns)": 50000000, "target_qps": 20, "task": "image classification", "task2": "image classification", "total_cores": 48, "uid": "d0bea453ea62c347", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile": 1, "50.00 percentile latency (ns)": 47224652, "90.00 percentile": 1, "90.00 percentile latency (ns)": 47321477, "95.00 percentile": 1, "95.00 percentile latency (ns)": 47341869, "97.00 percentile": 1, "97.00 percentile latency (ns)": 47354456, "99.00 percentile": 1, "99.00 percentile latency (ns)": 47396863, "99.90 percentile": 1, "99.90 percentile latency (ns)": 47664086, "Max latency (ns)": 62902892, "Mean latency (ns)": 47010165, "Min duration satisfied": "Yes", "Min latency (ns)": 45366175, "Min queries satisfied": "Yes", "Mode": "Performance", "Per-sample latency": "", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per query": 100, "Scenario": "multistream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "Shared with host", "accelerator_memory_configuration": "SRAM", "accelerator_model_name": "NVIDIA Xavier", "accelerator_on-chip_memories": "1MB (128KB/SM) L1 + 512KB L2 + 4MB (DLA)", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "characteristics.accuracy": 76.042, "characteristics.good": 38021, "characteristics.samples_per_query": 100, "characteristics.samples_per_query.normalized_per_core": 100.0, "characteristics.samples_per_query.normalized_per_processor": 100.0, "characteristics.total": 50000, "ck_system": "Xavier", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_query", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "JetPack 4.3 DP, TensorRT 6.0, cuDNN 7.6.3, CUDA 10.0, cub 1.8.0", "host_memory_capacity": "16 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "8MB L2 (2MB per dual cluster)/4 MB L3 (shared)", "host_processor_core_count": 8, "host_processor_frequency": "2265.5 MHz", "host_processor_interconnect": "", "host_processor_model_name": "NVIDIA Carmel (ARMv8.2)", "host_processors_per_node": 1, "host_storage_capacity": "32 GB", "host_storage_type": "eMMC 5.1", "hw_notes": "GPU and both DLAs are used in Offline and MultiStream scenarios", "informal_model": "resnet", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "mlperf_version": 0.5, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/code", "note_details": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/results/Xavier", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.3", "other_software_stack": "pycuda 2019.1, pytorch 1.1, torchvision 0.2.2.post3", "performance_issue_same": true, "performance_issue_same_index": 0, "performance_issue_unique": true, "performance_sample_count": 1024, "print_timestamps": true, "problem": false, "qsl_rng_seed": 3133965575612453542, "retraining": "N", "sample_index_rng_seed": 665484352860916858, "samples_per_query": 100, "schedule_rng_seed": 3622009729038561421, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "NVIDIA", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/NVIDIA", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/Xavier", "system_name": "NVIDIA Jetson AGX Xavier", "target_latency (ns)": 50000000, "target_qps": 20, "task": "image classification", "task2": "image classification", "total_cores": 8, "uid": "ad4422a74ac495ea", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile": 1, "50.00 percentile latency (ns)": 47415281, "90.00 percentile": 1, "90.00 percentile latency (ns)": 47535154, "95.00 percentile": 1, "95.00 percentile latency (ns)": 47565714, "97.00 percentile": 1, "97.00 percentile latency (ns)": 47587821, "99.00 percentile": 1, "99.00 percentile latency (ns)": 47654172, "99.90 percentile": 1, "99.90 percentile latency (ns)": 47863202, "Max latency (ns)": 59914897, "Mean latency (ns)": 47389026, "Min duration satisfied": "Yes", "Min latency (ns)": 45817068, "Min queries satisfied": "Yes", "Mode": "Performance", "Per-sample latency": "", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per query": 302, "Scenario": "multistream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "Shared with host", "accelerator_memory_configuration": "SRAM", "accelerator_model_name": "NVIDIA Xavier", "accelerator_on-chip_memories": "1MB (128KB/SM) L1 + 512KB L2 + 4MB (DLA)", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "characteristics.accuracy": 70.774, "characteristics.good": 35387, "characteristics.samples_per_query": 302, "characteristics.samples_per_query.normalized_per_core": 302.0, "characteristics.samples_per_query.normalized_per_processor": 302.0, "characteristics.total": 50000, "ck_system": "Xavier", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_query", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "JetPack 4.3 DP, TensorRT 6.0, cuDNN 7.6.3, CUDA 10.0, cub 1.8.0", "host_memory_capacity": "16 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "8MB L2 (2MB per dual cluster)/4 MB L3 (shared)", "host_processor_core_count": 8, "host_processor_frequency": "2265.5 MHz", "host_processor_interconnect": "", "host_processor_model_name": "NVIDIA Carmel (ARMv8.2)", "host_processors_per_node": 1, "host_storage_capacity": "32 GB", "host_storage_type": "eMMC 5.1", "hw_notes": "GPU and both DLAs are used in Offline and MultiStream scenarios", "informal_model": "mobilenet", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "mlperf_version": 0.5, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/code", "note_details": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/results/Xavier", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.3", "other_software_stack": "pycuda 2019.1, pytorch 1.1, torchvision 0.2.2.post3", "performance_issue_same": true, "performance_issue_same_index": 0, "performance_issue_unique": true, "performance_sample_count": 1024, "print_timestamps": true, "problem": false, "qsl_rng_seed": 3133965575612453542, "retraining": "N", "sample_index_rng_seed": 665484352860916858, "samples_per_query": 302, "schedule_rng_seed": 3622009729038561421, "starting_weights_filename": "mobilenet_sym_no_bn.onnx", "status": "available", "submitter": "NVIDIA", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/NVIDIA", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/Xavier", "system_name": "NVIDIA Jetson AGX Xavier", "target_latency (ns)": 50000000, "target_qps": 20, "task": "image classification", "task2": "image classification", "total_cores": 8, "uid": "7eb04141caa4d385", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile": 1, "50.00 percentile latency (ns)": 27728718, "90.00 percentile": 1, "90.00 percentile latency (ns)": 44246053, "95.00 percentile": 1, "95.00 percentile latency (ns)": 44915552, "97.00 percentile": 1, "97.00 percentile latency (ns)": 45422225, "99.00 percentile": 1, "99.00 percentile latency (ns)": 46010909, "99.90 percentile": 1, "99.90 percentile latency (ns)": 46370739, "Max latency (ns)": 55394038, "Mean latency (ns)": 28260878, "Min duration satisfied": "Yes", "Min latency (ns)": 10246329, "Min queries satisfied": "Yes", "Mode": "Performance", "Per-sample latency": "", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per query": 1920, "Scenario": "multistream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "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, "characteristics.accuracy": 76.034, "characteristics.good": 38017, "characteristics.samples_per_query": 1920, "characteristics.samples_per_query.normalized_per_core": 240.0, "characteristics.samples_per_query.normalized_per_processor": 240.0, "characteristics.total": 50000, "ck_system": "T4x8", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_query", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 6.0, CUDA 10.1, cuDNN 7.6.3, 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": 28, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8280 CPU @ 2.70GHz", "host_processors_per_node": 2, "host_storage_capacity": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC off", "informal_model": "resnet", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "mlperf_version": 0.5, "normalize_cores": 8, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/code", "note_details": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/results/T4x8", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.3", "other_software_stack": "docker 18.09.2, 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": 1024, "print_timestamps": true, "problem": false, "qsl_rng_seed": 3133965575612453542, "retraining": "N", "sample_index_rng_seed": 665484352860916858, "samples_per_query": 1920, "schedule_rng_seed": 3622009729038561421, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "NVIDIA", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/NVIDIA", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/T4x8", "system_name": "Supermicro 4029GP-TRT-OTO-28 8xT4", "target_latency (ns)": 50000000, "target_qps": 20, "task": "image classification", "task2": "image classification", "total_cores": 56, "uid": "c8b93316a24bf848", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile": 1, "50.00 percentile latency (ns)": 24848143, "90.00 percentile": 1, "90.00 percentile latency (ns)": 40472075, "95.00 percentile": 1, "95.00 percentile latency (ns)": 42286277, "97.00 percentile": 1, "97.00 percentile latency (ns)": 42913989, "99.00 percentile": 1, "99.00 percentile latency (ns)": 44104209, "99.90 percentile": 1, "99.90 percentile latency (ns)": 45206244, "Max latency (ns)": 50980603, "Mean latency (ns)": 24729519, "Min duration satisfied": "Yes", "Min latency (ns)": 5639314, "Min queries satisfied": "Yes", "Mode": "Performance", "Per-sample latency": "", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per query": 6320, "Scenario": "multistream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "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, "characteristics.accuracy": 70.814, "characteristics.good": 35407, "characteristics.samples_per_query": 6320, "characteristics.samples_per_query.normalized_per_core": 790.0, "characteristics.samples_per_query.normalized_per_processor": 790.0, "characteristics.total": 50000, "ck_system": "T4x8", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_query", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 6.0, CUDA 10.1, cuDNN 7.6.3, 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": 28, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8280 CPU @ 2.70GHz", "host_processors_per_node": 2, "host_storage_capacity": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC off", "informal_model": "mobilenet", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "mlperf_version": 0.5, "normalize_cores": 8, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/code", "note_details": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/NVIDIA/results/T4x8", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.3", "other_software_stack": "docker 18.09.2, 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": 1024, "print_timestamps": true, "problem": false, "qsl_rng_seed": 3133965575612453542, "retraining": "N", "sample_index_rng_seed": 665484352860916858, "samples_per_query": 6320, "schedule_rng_seed": 3622009729038561421, "starting_weights_filename": "mobilenet_sym_no_bn.onnx", "status": "available", "submitter": "NVIDIA", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/NVIDIA", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/T4x8", "system_name": "Supermicro 4029GP-TRT-OTO-28 8xT4", "target_latency (ns)": 50000000, "target_qps": 20, "task": "image classification", "task2": "image classification", "total_cores": 56, "uid": "8e4bb9b6b4369d36", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" } ]