[ { "50.00 percentile latency (ns)": 671981687521, "90.00 percentile latency (ns)": 671981687521, "95.00 percentile latency (ns)": 671981687521, "97.00 percentile latency (ns)": 671981687521, "99.00 percentile latency (ns)": 671981687521, "99.90 percentile latency (ns)": 671981687521, "Max latency (ns)": 671981687521, "Mean latency (ns)": 671981687521, "Min duration satisfied": "Yes", "Min latency (ns)": 671981687521, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "SUT", "Samples per second": 4223.33, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.292, "characteristics.good": 38146, "characteristics.samples_per_second": 4223.33, "characteristics.samples_per_second.normalized_per_core": 52.791624999999996, "characteristics.samples_per_second.normalized_per_processor": 2111.665, "characteristics.total": 50000, "ck_system": "1-node-2S-ICX-OpenVINO-INT8", "ck_used": false, "cooling": "Air", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "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": "OpenVINO", "host_memory_capacity": "1024GB", "host_memory_configuration": "8 slots / 64GB each / 3200 MT/s per socket", "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": "", "host_storage_type": "", "hw_notes": "Intel Devlopment Platform (2U Server)", "informal_model": "resnet50", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.1, "normalize_cores": 80, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Intel/results/1-node-2S-ICX-OpenVINO-INT8", "number_of_nodes": 1, "operating_system": "Ubuntu 20.04.1 LTS", "other_software_stack": "5.4.0-45-generic", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 2838000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "available", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/1-node-2S-ICX-OpenVINO-INT8", "system_name": "1-node-2S-ICX-OpenVINO-INT8", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 4300, "task": "image classification", "task2": "image classification", "total_cores": 80, "uid": "5982c371f0581985", "use_accelerator": false, "weight_data_types": "No change.", "weight_transformations": "https://docs.openvinotoolkit.org/latest/openvino_docs_IE_DG_Bfloat16Inference.html" }, { "50.00 percentile latency (ns)": 443733772806, "90.00 percentile latency (ns)": 796340395007, "95.00 percentile latency (ns)": 840492434452, "97.00 percentile latency (ns)": 858160038218, "99.00 percentile latency (ns)": 875902595769, "99.90 percentile latency (ns)": 883758430110, "Max latency (ns)": 884621807448, "Mean latency (ns)": 443699175029, "Min duration satisfied": "Yes", "Min latency (ns)": 2671143931, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "PySUT", "Samples per second": 4103.45, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 75.788, "characteristics.good": 37894, "characteristics.samples_per_second": 4103.45, "characteristics.samples_per_second.normalized_per_core": 51.293124999999996, "characteristics.samples_per_second.normalized_per_processor": 2051.725, "characteristics.total": 50000, "ck_system": "1-node-2S-ICX-PyTorch-INT8", "ck_used": false, "cooling": "Air", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "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": "PyTorch", "host_memory_capacity": "1024GB", "host_memory_configuration": "8 slots / 64GB each / 3200 MT/s per socket", "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": "", "host_storage_type": "", "hw_notes": "Intel Devlopment Platform (2U Server)", "informal_model": "resnet50", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.1, "normalize_cores": 80, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Intel/results/1-node-2S-ICX-PyTorch-INT8", "number_of_nodes": 1, "operating_system": "Ubuntu 20.04.1 LTS", "other_software_stack": "5.4.0-45-generic", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 3630000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/4588417/files/resnet50-19c8e357.pth", "status": "available", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/1-node-2S-ICX-PyTorch-INT8", "system_name": "1-node-2S-ICX-PyTorch-INT8", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 5500, "task": "image classification", "task2": "image classification", "total_cores": 80, "uid": "d580ff48c411ca47", "use_accelerator": false, "weight_data_types": "int8", "weight_transformations": "We transform float32 weight into int8 with IPEX auto-mix-precision." }, { "50.00 percentile latency (ns)": 686393608067, "90.00 percentile latency (ns)": 686393608067, "95.00 percentile latency (ns)": 686393608067, "97.00 percentile latency (ns)": 686393608067, "99.00 percentile latency (ns)": 686393608067, "99.90 percentile latency (ns)": 686393608067, "Max latency (ns)": 686393608067, "Mean latency (ns)": 686393608067, "Min duration satisfied": "Yes", "Min latency (ns)": 686393608067, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "SUT", "Samples per second": 4038.5, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.292, "characteristics.good": 38146, "characteristics.samples_per_second": 4038.5, "characteristics.samples_per_second.normalized_per_core": 50.48125, "characteristics.samples_per_second.normalized_per_processor": 2019.25, "characteristics.total": 50000, "ck_system": "1-node-2S-ICX-OpenVINO-INT8", "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_second", "dim_x_maximize": true, "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": "OpenVINO", "host_memory_capacity": "1024GB", "host_memory_configuration": "8 slots / 64GB each / 3200 MT/s per socket", "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": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 80, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Intel/results/1-node-2S-ICX-OpenVINO-INT8", "number_of_nodes": 1, "operating_system": "Ubuntu 20.04.1 LTS", "other_software_stack": "5.4.0-45-generic", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 2772000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "preview", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/1-node-2S-ICX-OpenVINO-INT8", "system_name": "1-node-2S-ICX-OpenVINO-INT8", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 4200, "task": "image classification", "task2": "image classification", "total_cores": 80, "uid": "f4c88f50624b54b4", "use_accelerator": false, "weight_data_types": "No change.", "weight_transformations": "https://docs.openvinotoolkit.org/latest/openvino_docs_IE_DG_Bfloat16Inference.html" }, { "50.00 percentile latency (ns)": 1380363740542, "90.00 percentile latency (ns)": 1380363740542, "95.00 percentile latency (ns)": 1380363740542, "97.00 percentile latency (ns)": 1380363740542, "99.00 percentile latency (ns)": 1380363740542, "99.90 percentile latency (ns)": 1380363740542, "Max latency (ns)": 1380363740542, "Mean latency (ns)": 1380363740542, "Min duration satisfied": "Yes", "Min latency (ns)": 1380363740542, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "CSUT", "Samples per second": 5259.48, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.438, "characteristics.good": 38219, "characteristics.samples_per_second": 5259.48, "characteristics.samples_per_second.normalized_per_core": 5.478624999999999, "characteristics.samples_per_second.normalized_per_processor": 657.435, "characteristics.total": 50000, "ck_system": "1-node-8S-CPX-TensorFlow-BF16", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "TensorFlow source from (https://github.com/Intel-tensorflow/tensorflow/tree/tf2_lars)", "host_memory_capacity": "3072GB", "host_memory_configuration": "6 slots / 64GB each / 3200 MT/s per socket", "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": 8, "host_storage_capacity": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 960, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Intel/results/1-node-8S-CPX-TensorFlow-BF16", "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": "5.4.0-66-generic; GCC 7.5.0; Python 3.7.10", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 7260000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "available", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-8S-CPX-TensorFlow-BF16", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 11000, "task": "image classification", "task2": "image classification", "total_cores": 960, "uid": "4ff9a18c1d4f7da6", "use_accelerator": false, "weight_data_types": "Google bfloat16.", "weight_transformations": "We transform float32 weight into bfloat16 using TensorFlow's Cast operation." }, { "50.00 percentile latency (ns)": 13540734003402, "90.00 percentile latency (ns)": 13540734003402, "95.00 percentile latency (ns)": 13540734003402, "97.00 percentile latency (ns)": 13540734003402, "99.00 percentile latency (ns)": 13540734003402, "99.90 percentile latency (ns)": 13540734003402, "Max latency (ns)": 13540734003402, "Mean latency (ns)": 124920131613, "Min duration satisfied": "Yes", "Min latency (ns)": 13540734003402, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "SUT", "Samples per second": 2437.09, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.474, "characteristics.good": 38237, "characteristics.samples_per_second": 2437.09, "characteristics.samples_per_second.normalized_per_core": 5.077270833333333, "characteristics.samples_per_second.normalized_per_processor": 609.2725, "characteristics.total": 50000, "ck_system": "1-node-4S-CPX-OpenVINO-BF16", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "OpenVINO", "host_memory_capacity": "1536GB", "host_memory_configuration": "6 slots / 64GB each / 3200 MT/s per socket", "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": 4, "host_storage_capacity": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 480, "normalize_processors": 4, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Intel/results/1-node-4S-CPX-OpenVINO-BF16", "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": "5.4.0-65-generic; GCC 7.5.0; Python 3.7.10", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 33000000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "available", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-4S-CPX-OpenVINO-BF16", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 50000, "task": "image classification", "task2": "image classification", "total_cores": 480, "uid": "74dc1d440f78c01e", "use_accelerator": false, "weight_data_types": "No change.", "weight_transformations": "https://docs.openvinotoolkit.org/latest/openvino_docs_IE_DG_Bfloat16Inference.html" }, { "50.00 percentile latency (ns)": 5957770333724, "90.00 percentile latency (ns)": 5957770333724, "95.00 percentile latency (ns)": 5957770333724, "97.00 percentile latency (ns)": 5957770333724, "99.00 percentile latency (ns)": 5957770333724, "99.90 percentile latency (ns)": 5957770333724, "Max latency (ns)": 5957770333724, "Mean latency (ns)": 367847887145, "Min duration satisfied": "Yes", "Min latency (ns)": 5957770333724, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "SUT", "Samples per second": 5538.98, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.292, "characteristics.good": 38146, "characteristics.samples_per_second": 5538.98, "characteristics.samples_per_second.normalized_per_core": 11.539541666666667, "characteristics.samples_per_second.normalized_per_processor": 1384.745, "characteristics.total": 50000, "ck_system": "1-node-4S-CPX-OpenVINO-INT8", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "OpenVINO", "host_memory_capacity": "1536GB", "host_memory_configuration": "6 slots / 64GB each / 3200 MT/s per socket", "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": 4, "host_storage_capacity": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 480, "normalize_processors": 4, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Intel/results/1-node-4S-CPX-OpenVINO-INT8", "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": "5.4.0-65-generic; GCC 7.5.0; Python 3.7.10", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 33000000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "available", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-4S-CPX-OpenVINO-INT8", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 50000, "task": "image classification", "task2": "image classification", "total_cores": 480, "uid": "37f5fc6645ea4c16", "use_accelerator": false, "weight_data_types": "No change.", "weight_transformations": "https://docs.openvinotoolkit.org/latest/openvino_docs_IE_DG_Bfloat16Inference.html" }, { "50.00 percentile latency (ns)": 685409467155, "90.00 percentile latency (ns)": 685409467155, "95.00 percentile latency (ns)": 685409467155, "97.00 percentile latency (ns)": 685409467155, "99.00 percentile latency (ns)": 685409467155, "99.90 percentile latency (ns)": 685409467155, "Max latency (ns)": 685409467155, "Mean latency (ns)": 685409467155, "Min duration satisfied": "Yes", "Min latency (ns)": 685409467155, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "CSUT", "Samples per second": 10592.2, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.046, "characteristics.good": 38023, "characteristics.samples_per_second": 10592.2, "characteristics.samples_per_second.normalized_per_core": 11.033541666666668, "characteristics.samples_per_second.normalized_per_processor": 1324.025, "characteristics.total": 50000, "ck_system": "1-node-8S-CPX-TensorFlow-INT8", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "TensorFlow v2.3.0", "host_memory_capacity": "3072GB", "host_memory_configuration": "6 slots / 64GB each / 3200 MT/s per socket", "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": 8, "host_storage_capacity": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 960, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Intel/results/1-node-8S-CPX-TensorFlow-INT8", "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": "5.4.0-66-generic; GCC 7.5.0; Python 3.7.10", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 7260000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "available", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-8S-CPX-TensorFlow-INT8", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 11000, "task": "image classification", "task2": "image classification", "total_cores": 960, "uid": "bff3e708ff8541af", "use_accelerator": false, "weight_data_types": "Signed 8-bit integer.", "weight_transformations": "We transform the original fp32 weight to int8 weight using symmetric quantization." }, { "50.00 percentile latency (ns)": 33191316691, "90.00 percentile latency (ns)": 59738525490, "95.00 percentile latency (ns)": 63061796986, "97.00 percentile latency (ns)": 64654289964, "99.00 percentile latency (ns)": 66686226836, "99.90 percentile latency (ns)": 68897803458, "Max latency (ns)": 69699384232, "Mean latency (ns)": 33222029280, "Min duration satisfied": "Yes", "Min latency (ns)": 133667321, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "CSUT", "Samples per second": 5965.62, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "characteristics.accuracy": 75.766, "characteristics.good": 37883, "characteristics.samples_per_second": 5965.62, "characteristics.samples_per_second.normalized_per_core": 24.856749999999998, "characteristics.samples_per_second.normalized_per_processor": 2982.81, "characteristics.total": 50000, "ck_system": "clx-9282-2s_pytorch-caffe2", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "PyTorch", "host_memory_capacity": "768GB", "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": "422G", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "resnet", "input_data_types": "We transform the fp32 input data to int8 data.", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1, "mlperf_version": 0.5, "normalize_cores": 240, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/Intel/results/clx-9282-2s_pytorch-caffe2", "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": "Deep Learning Math Kernel Research(Euler) v0.0.3, MKL (2019.4.243); GCC 7.5.0; Python 3.7.10", "performance_issue_same": true, "performance_issue_same_index": 0, "performance_issue_unique": true, "performance_sample_count": 50000, "print_timestamps": true, "problem": false, "qsl_rng_seed": 3133965575612453542, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 665484352860916858, "samples_per_query": 415800, "schedule_rng_seed": 3622009729038561421, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/2541184/files/resnet50_v1.onnx and the transformed int8 weight named init_net_int8_euler.pb", "status": "available", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/clx-9282-2s_pytorch-caffe2", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "target_latency (ns)": 0, "target_qps": 6300, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "8fa4515b0e458e98", "use_accelerator": false, "weight_data_types": "We transform the original fp32 weight to int8 weight.", "weight_transformations": "We transform the weight from fp32 datatype in onnx file to int8 datatype in pb file." }, { "50.00 percentile latency (ns)": 33999928499, "90.00 percentile latency (ns)": 60818617570, "95.00 percentile latency (ns)": 64169460494, "97.00 percentile latency (ns)": 65630653880, "99.00 percentile latency (ns)": 67624975688, "99.90 percentile latency (ns)": 69309504315, "Max latency (ns)": 70060493552, "Mean latency (ns)": 34010139387, "Min duration satisfied": "Yes", "Min latency (ns)": 390561960, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "CSUT", "Samples per second": 29203.3, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "characteristics.accuracy": 70.484, "characteristics.good": 35242, "characteristics.samples_per_second": 29203.3, "characteristics.samples_per_second.normalized_per_core": 121.68041666666666, "characteristics.samples_per_second.normalized_per_processor": 14601.65, "characteristics.total": 50000, "ck_system": "clx-9282-2s_pytorch-caffe2", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "PyTorch", "host_memory_capacity": "768GB", "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": "422G", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "mobilenet", "input_data_types": "We transform the fp32 input data to int8 data.", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1, "mlperf_version": 0.5, "normalize_cores": 240, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/Intel/results/clx-9282-2s_pytorch-caffe2", "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": "Deep Learning Math Kernel Research(Euler) v0.0.3, MKL (2019.4.243); GCC 7.5.0; Python 3.7.10", "performance_issue_same": true, "performance_issue_same_index": 0, "performance_issue_unique": true, "performance_sample_count": 50000, "print_timestamps": true, "problem": false, "qsl_rng_seed": 3133965575612453542, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 665484352860916858, "samples_per_query": 2046000, "schedule_rng_seed": 3622009729038561421, "starting_weights_filename": "For mobilenet model, the original weight filename: https://zenodo.org/record/3157894/files/mobilenet_v1_1.0_224.onnx. The transformed int8 weight named init_net_int8_euler.pb", "status": "available", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/clx-9282-2s_pytorch-caffe2", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "target_latency (ns)": 0, "target_qps": 31000, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "c7440e7ecc4bd583", "use_accelerator": false, "weight_data_types": "We transform the original fp32 weight to int8 weight.", "weight_transformations": "We transform the weight from fp32 datatype in onnx file to int8 datatype in pb file." }, { "50.00 percentile latency (ns)": 89328006043, "90.00 percentile latency (ns)": 159612962582, "95.00 percentile latency (ns)": 168483560472, "97.00 percentile latency (ns)": 172043493862, "99.00 percentile latency (ns)": 175522652201, "99.90 percentile latency (ns)": 176919059754, "Max latency (ns)": 177081859423, "Mean latency (ns)": 89211866418, "Min duration satisfied": "Yes", "Min latency (ns)": 1146471724, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "PyFastSUT", "Samples per second": 11181.3, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.samples_per_second": 11181.3, "characteristics.samples_per_second.normalized_per_core": 11.6471875, "characteristics.samples_per_second.normalized_per_processor": 1397.6625, "ck_system": "1-node-8s-cpx-mx-int8", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "MXNet", "host_memory_capacity": "1536GB", "host_memory_configuration": "6 slots / 32GB each / 3200 MT/s per socket", "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": 8, "host_storage_capacity": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet", "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": 960, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/results/1-node-8s-cpx-mx-int8", "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": "5.4.0-45-generic; GCC 7.5.0; Python 3.7.10", "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": 12786827339337101903, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 1980000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/2592612/files/resnet50_v1.onnx", "status": "preview", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-8s-cpx-mx-int8", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 30000, "task": "image classification", "task2": "image classification", "total_cores": 960, "uid": "a7624e594004645b", "use_accelerator": false, "weight_data_types": "FP32 + Int8", "weight_transformations": "We transfer the weight from fp32 datatype in onnx file to mxnet files then quatize to fp32+int8 datatype file by Intel® Low Precision Optimization Tool." }, { "50.00 percentile latency (ns)": 74242821972, "90.00 percentile latency (ns)": 74242821972, "95.00 percentile latency (ns)": 74242821972, "97.00 percentile latency (ns)": 74242821972, "99.00 percentile latency (ns)": 74242821972, "99.90 percentile latency (ns)": 74242821972, "Max latency (ns)": 74242821972, "Mean latency (ns)": 74242821972, "Min duration satisfied": "Yes", "Min latency (ns)": 74242821972, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "CSUT", "Samples per second": 10667.7, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.046, "characteristics.good": 38023, "characteristics.samples_per_second": 10667.7, "characteristics.samples_per_second.normalized_per_core": 11.112187500000001, "characteristics.samples_per_second.normalized_per_processor": 1333.4625, "characteristics.total": 50000, "ck_system": "1-node-8s-cpx-tf-int8", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "TensorFlow v2.3.0", "host_memory_capacity": "1536GB", "host_memory_configuration": "6 slots / 32GB each / 3200 MT/s per socket", "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": 8, "host_storage_capacity": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "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": 960, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/results/1-node-8s-cpx-tf-int8", "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": "5.4.0-45-generic; GCC 7.5.0; Python 3.7.10", "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": 12786827339337101903, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 792000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "preview", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-8s-cpx-tf-int8", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 12000, "task": "image classification", "task2": "image classification", "total_cores": 960, "uid": "486cbb24478f9b9e", "use_accelerator": false, "weight_data_types": "Signed 8-bit integer.", "weight_transformations": "We transform the original fp32 weight to int8 weight using symmetric quantization." }, { "50.00 percentile latency (ns)": 69452255639, "90.00 percentile latency (ns)": 69452255639, "95.00 percentile latency (ns)": 69452255639, "97.00 percentile latency (ns)": 69452255639, "99.00 percentile latency (ns)": 69452255639, "99.90 percentile latency (ns)": 69452255639, "Max latency (ns)": 69452255639, "Mean latency (ns)": 69452255639, "Min duration satisfied": "Yes", "Min latency (ns)": 69452255639, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "SUT\u0000", "Samples per second": 2375.73, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.474, "characteristics.good": 38237, "characteristics.samples_per_second": 2375.73, "characteristics.samples_per_second.normalized_per_core": 4.9494375, "characteristics.samples_per_second.normalized_per_processor": 593.9325, "characteristics.total": 50000, "ck_system": "1-node-4s-cpx-ov-bf16", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "OpenVINO", "host_memory_capacity": "768GB", "host_memory_configuration": "6 slots / 32GB each / 3200 MT/s per socket", "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": 4, "host_storage_capacity": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "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": 480, "normalize_processors": 4, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/results/1-node-4s-cpx-ov-bf16", "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": "5.4.0-45-generic; GCC 7.5.0; Python 3.7.10", "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": 12786827339337101903, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 165000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "preview", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-4s-cpx-ov-bf16", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 2500, "task": "image classification", "task2": "image classification", "total_cores": 480, "uid": "b34607bc1064e099", "use_accelerator": false, "weight_data_types": "No change.", "weight_transformations": "https://docs.openvinotoolkit.org/latest/openvino_docs_IE_DG_Bfloat16Inference.html" }, { "50.00 percentile latency (ns)": 72950348834, "90.00 percentile latency (ns)": 72950348834, "95.00 percentile latency (ns)": 72950348834, "97.00 percentile latency (ns)": 72950348834, "99.00 percentile latency (ns)": 72950348834, "99.90 percentile latency (ns)": 72950348834, "Max latency (ns)": 72950348834, "Mean latency (ns)": 72950348834, "Min duration satisfied": "Yes", "Min latency (ns)": 72950348834, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "CSUT", "Samples per second": 2714.17, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.046, "characteristics.good": 38023, "characteristics.samples_per_second": 2714.17, "characteristics.samples_per_second.normalized_per_core": 11.309041666666667, "characteristics.samples_per_second.normalized_per_processor": 1357.085, "characteristics.total": 50000, "ck_system": "1-node-2s-clx-tf-int8", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "TensorFlow v2.3.0", "host_memory_capacity": "384GB", "host_memory_configuration": "6 slots / 32GB each / 2934 MT/s per socket", "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": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "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": 240, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/results/1-node-2s-clx-tf-int8", "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": "5.4.0-45-generic; GCC 7.5.0; Python 3.7.10", "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": 12786827339337101903, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 198000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "available", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-2s-clx-tf-int8", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 3000, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "3424ab688ef21e9b", "use_accelerator": false, "weight_data_types": "Signed 8-bit integer.", "weight_transformations": "We transform the original fp32 weight to int8 weight using symmetric quantization." }, { "50.00 percentile latency (ns)": 116430668426, "90.00 percentile latency (ns)": 116430668426, "95.00 percentile latency (ns)": 116430668426, "97.00 percentile latency (ns)": 116430668426, "99.00 percentile latency (ns)": 116430668426, "99.90 percentile latency (ns)": 116430668426, "Max latency (ns)": 116430668426, "Mean latency (ns)": 116430668426, "Min duration satisfied": "Yes", "Min latency (ns)": 116430668426, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "SUT\u0000", "Samples per second": 5668.61, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.292, "characteristics.good": 38146, "characteristics.samples_per_second": 5668.61, "characteristics.samples_per_second.normalized_per_core": 11.809604166666666, "characteristics.samples_per_second.normalized_per_processor": 1417.1525, "characteristics.total": 50000, "ck_system": "1-node-4s-cpx-ov-int8", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "OpenVINO", "host_memory_capacity": "768GB", "host_memory_configuration": "6 slots / 32GB each / 3200 MT/s per socket", "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": 4, "host_storage_capacity": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "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": 480, "normalize_processors": 4, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/results/1-node-4s-cpx-ov-int8", "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": "5.4.0-45-generic; GCC 7.5.0; Python 3.7.10", "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": 12786827339337101903, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 660000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "preview", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-4s-cpx-ov-int8", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 10000, "task": "image classification", "task2": "image classification", "total_cores": 480, "uid": "17ad673a7f574947", "use_accelerator": false, "weight_data_types": "Signed 8-bit integer.", "weight_transformations": "We transform the original fp32 weight to int8 weight using symmetric quantization." }, { "50.00 percentile latency (ns)": 47395212782, "90.00 percentile latency (ns)": 84667262402, "95.00 percentile latency (ns)": 89418903219, "97.00 percentile latency (ns)": 90962357187, "99.00 percentile latency (ns)": 93206238391, "99.90 percentile latency (ns)": 93772855859, "Max latency (ns)": 93869686488, "Mean latency (ns)": 47428077549, "Min duration satisfied": "Yes", "Min latency (ns)": 1268091406, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "PyFastSUT", "Samples per second": 2812.41, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.086, "characteristics.good": 38043, "characteristics.samples_per_second": 2812.41, "characteristics.samples_per_second.normalized_per_core": 11.718375, "characteristics.samples_per_second.normalized_per_processor": 1406.205, "characteristics.total": 50000, "ck_system": "1-node-2s-clx-mx-int8", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "MXNet", "host_memory_capacity": "384GB", "host_memory_configuration": "6 slots / 32GB each / 2934 MT/s per socket", "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": "", "host_storage_type": "", "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": 1, "mlperf_version": 0.7, "normalize_cores": 240, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/results/1-node-2s-clx-mx-int8", "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": "5.4.0-45-generic; GCC 7.5.0; Python 3.7.10", "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": 12786827339337101903, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 264000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/2592612/files/resnet50_v1.onnx", "status": "available", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-2s-clx-mx-int8", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 4000, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "4dc0950b3dbd4ca0", "use_accelerator": false, "weight_data_types": "FP32 + Int8", "weight_transformations": "We transfer the weight from fp32 datatype in onnx file to mxnet files then quatize to fp32+int8 datatype file by Intel® Low Precision Optimization Tool." }, { "50.00 percentile latency (ns)": 74443945711, "90.00 percentile latency (ns)": 74443945711, "95.00 percentile latency (ns)": 74443945711, "97.00 percentile latency (ns)": 74443945711, "99.00 percentile latency (ns)": 74443945711, "99.90 percentile latency (ns)": 74443945711, "Max latency (ns)": 74443945711, "Mean latency (ns)": 74443945711, "Min duration satisfied": "Yes", "Min latency (ns)": 74443945711, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "CSUT", "Samples per second": 5319.44, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.438, "characteristics.good": 38219, "characteristics.samples_per_second": 5319.44, "characteristics.samples_per_second.normalized_per_core": 5.541083333333333, "characteristics.samples_per_second.normalized_per_processor": 664.93, "characteristics.total": 50000, "ck_system": "1-node-8s-cpx-tf-bf16", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "TensorFlow source from (https://github.com/Intel-tensorflow/tensorflow/tree/tf2_lars)", "host_memory_capacity": "1536GB", "host_memory_configuration": "6 slots / 32GB each / 3200 MT/s per socket", "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": 8, "host_storage_capacity": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "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": 960, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/results/1-node-8s-cpx-tf-bf16", "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": "5.4.0-45-generic; GCC 7.5.0; Python 3.7.10", "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": 12786827339337101903, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 396000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "preview", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-8s-cpx-tf-bf16", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 6000, "task": "image classification", "task2": "image classification", "total_cores": 960, "uid": "db96172fd132651e", "use_accelerator": false, "weight_data_types": "Google bfloat16.", "weight_transformations": "We transform float32 weight into bfloat16 using TensorFlow's Cast operation." }, { "50.00 percentile latency (ns)": 234798174887, "90.00 percentile latency (ns)": 234798174887, "95.00 percentile latency (ns)": 234798174887, "97.00 percentile latency (ns)": 234798174887, "99.00 percentile latency (ns)": 234798174887, "99.90 percentile latency (ns)": 234798174887, "Max latency (ns)": 234798174887, "Mean latency (ns)": 234798174887, "Min duration satisfied": "Yes", "Min latency (ns)": 234798174887, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "SUT\u0000", "Samples per second": 2810.92, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.292, "characteristics.good": 38146, "characteristics.samples_per_second": 2810.92, "characteristics.samples_per_second.normalized_per_core": 11.712166666666667, "characteristics.samples_per_second.normalized_per_processor": 1405.46, "characteristics.total": 50000, "ck_system": "1-node-2s-clx-ov-int8", "ck_used": true, "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_second", "dim_x_maximize": true, "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": "OpenVINO", "host_memory_capacity": "384GB", "host_memory_configuration": "6 slots / 32GB each / 2934 MT/s per socket", "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": "", "host_storage_type": "", "hw_notes": "", "informal_model": "resnet", "input_data_types": "No change.", "key.accuracy": "characteristics.accuracy", "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": 240, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Intel/results/1-node-2s-clx-ov-int8", "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": "5.4.0-45-generic; GCC 7.5.0; Python 3.7.10", "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": 12786827339337101903, "retraining": "We don’t retrain the model weight.", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 660000, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "available", "submitter": "Intel", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Intel", "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/1-node-2s-clx-ov-int8", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 10000, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "47907c438db7af58", "use_accelerator": false, "weight_data_types": "No change.", "weight_transformations": "Please refer to close/Intel/calibration/OpenVINO" } ]