[ { "50.00 percentile latency (ns)": 5653802, "90.00 percentile latency (ns)": 8767717, "95.00 percentile latency (ns)": 10239784, "97.00 percentile latency (ns)": 11379808, "99.00 percentile latency (ns)": 14031886, "99.90 percentile latency (ns)": 18710822, "Completed samples per second": 3426.8, "Max latency (ns)": 23341945, "Mean latency (ns)": 6430095, "Min duration satisfied": "Yes", "Min latency (ns)": 4738524, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "SUT", "Scenario": "server", "Scheduled samples per second": 3426.84, "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.scheduled_queries_per_second": 3426.84, "characteristics.scheduled_queries_per_second.normalized_per_core": 42.8355, "characteristics.scheduled_queries_per_second.normalized_per_processor": 1713.42, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 3425, "task": "image classification", "task2": "image classification", "total_cores": 80, "uid": "22d90de1eee41ccf", "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)": 9268091, "90.00 percentile latency (ns)": 11524658, "95.00 percentile latency (ns)": 12459169, "97.00 percentile latency (ns)": 13166292, "99.00 percentile latency (ns)": 14696209, "99.90 percentile latency (ns)": 18279184, "Completed samples per second": 2601.05, "Max latency (ns)": 36008435, "Mean latency (ns)": 9790105, "Min duration satisfied": "Yes", "Min latency (ns)": 7980241, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PySUT", "Scenario": "server", "Scheduled samples per second": 2601.08, "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.scheduled_queries_per_second": 2601.08, "characteristics.scheduled_queries_per_second.normalized_per_core": 32.5135, "characteristics.scheduled_queries_per_second.normalized_per_processor": 1300.54, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 2600, "task": "image classification", "task2": "image classification", "total_cores": 80, "uid": "18306454b0aa92fb", "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)": 6016629, "90.00 percentile latency (ns)": 8715637, "95.00 percentile latency (ns)": 9938075, "97.00 percentile latency (ns)": 10867504, "99.00 percentile latency (ns)": 13019763, "99.90 percentile latency (ns)": 20069872, "Completed samples per second": 3051.08, "Max latency (ns)": 31508403, "Mean latency (ns)": 6704363, "Min duration satisfied": "Yes", "Min latency (ns)": 5109340, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "SUT", "Scenario": "server", "Scheduled samples per second": 3051.1, "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.scheduled_queries_per_second": 3051.1, "characteristics.scheduled_queries_per_second.normalized_per_core": 38.13875, "characteristics.scheduled_queries_per_second.normalized_per_processor": 1525.55, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 3050, "task": "image classification", "task2": "image classification", "total_cores": 80, "uid": "8f140a1407094e9a", "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)": 13423238, "90.00 percentile latency (ns)": 13790823, "95.00 percentile latency (ns)": 14042016, "97.00 percentile latency (ns)": 14328033, "99.00 percentile latency (ns)": 14931484, "99.90 percentile latency (ns)": 17907798, "Completed samples per second": 3001.44, "Max latency (ns)": 44307575, "Mean latency (ns)": 13484712, "Min duration satisfied": "Yes", "Min latency (ns)": 12560920, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "CSUT", "Scenario": "server", "Scheduled samples per second": 3001.51, "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.scheduled_queries_per_second": 3001.51, "characteristics.scheduled_queries_per_second.normalized_per_core": 3.126572916666667, "characteristics.scheduled_queries_per_second.normalized_per_processor": 375.18875, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 3000, "task": "image classification", "task2": "image classification", "total_cores": 960, "uid": "446d3880dc7266d3", "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)": 10245823, "90.00 percentile latency (ns)": 10788456, "95.00 percentile latency (ns)": 11029423, "97.00 percentile latency (ns)": 11241420, "99.00 percentile latency (ns)": 12056896, "99.90 percentile latency (ns)": 20117550, "Completed samples per second": 1701.37, "Max latency (ns)": 49299587, "Mean latency (ns)": 10326094, "Min duration satisfied": "Yes", "Min latency (ns)": 9036560, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "SUT", "Scenario": "server", "Scheduled samples per second": 1701.39, "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": 720381539243781796, "accuracy_log_sampling_target": 4096, "characteristics.accuracy": 76.474, "characteristics.good": 38237, "characteristics.scheduled_queries_per_second": 1701.39, "characteristics.scheduled_queries_per_second.normalized_per_core": 3.5445625, "characteristics.scheduled_queries_per_second.normalized_per_processor": 425.3475, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 1700, "task": "image classification", "task2": "image classification", "total_cores": 480, "uid": "0db05324f61c73b1", "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)": 5706323, "90.00 percentile latency (ns)": 8024894, "95.00 percentile latency (ns)": 9186113, "97.00 percentile latency (ns)": 10174290, "99.00 percentile latency (ns)": 14523229, "99.90 percentile latency (ns)": 133003039, "Completed samples per second": 4501.2, "Max latency (ns)": 161635114, "Mean latency (ns)": 6605919, "Min duration satisfied": "Yes", "Min latency (ns)": 4948955, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "SUT", "Scenario": "server", "Scheduled samples per second": 4501.24, "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.scheduled_queries_per_second": 4501.24, "characteristics.scheduled_queries_per_second.normalized_per_core": 9.377583333333332, "characteristics.scheduled_queries_per_second.normalized_per_processor": 1125.31, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 4500, "task": "image classification", "task2": "image classification", "total_cores": 480, "uid": "3981945ae52a83c0", "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)": 13254242, "90.00 percentile latency (ns)": 13632775, "95.00 percentile latency (ns)": 13853471, "97.00 percentile latency (ns)": 14106290, "99.00 percentile latency (ns)": 14876585, "99.90 percentile latency (ns)": 40483670, "Completed samples per second": 7303.45, "Max latency (ns)": 161033958, "Mean latency (ns)": 13400520, "Min duration satisfied": "Yes", "Min latency (ns)": 12446618, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "CSUT", "Scenario": "server", "Scheduled samples per second": 7303.61, "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.scheduled_queries_per_second": 7303.61, "characteristics.scheduled_queries_per_second.normalized_per_core": 7.607927083333333, "characteristics.scheduled_queries_per_second.normalized_per_processor": 912.95125, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 7300, "task": "image classification", "task2": "image classification", "total_cores": 960, "uid": "567d50f4454456eb", "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)": 10172879, "90.00 percentile latency (ns)": 11682102, "95.00 percentile latency (ns)": 12547859, "97.00 percentile latency (ns)": 13218888, "99.00 percentile latency (ns)": 14662988, "99.90 percentile latency (ns)": 20195091, "Completed samples per second": 4849.84, "Max latency (ns)": 26199661, "Mean latency (ns)": 10512080, "Min duration satisfied": "Yes", "Min latency (ns)": 9412009, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "CSUT", "Scenario": "server", "Scheduled samples per second": 4850.62, "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": 76.124, "characteristics.good": 38062, "characteristics.scheduled_queries_per_second": 4850.62, "characteristics.scheduled_queries_per_second.normalized_per_core": 20.210916666666666, "characteristics.scheduled_queries_per_second.normalized_per_processor": 2425.31, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 4850, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "737b18b6464ea399", "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)": 3840716, "90.00 percentile latency (ns)": 4223213, "95.00 percentile latency (ns)": 4563714, "97.00 percentile latency (ns)": 4913849, "99.00 percentile latency (ns)": 6060406, "99.90 percentile latency (ns)": 11977182, "Completed samples per second": 27243.17, "Max latency (ns)": 17283221, "Mean latency (ns)": 3951515, "Min duration satisfied": "Yes", "Min latency (ns)": 3394209, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "CSUT", "Scenario": "server", "Scheduled samples per second": 27244.81, "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": 70.494, "characteristics.good": 35247, "characteristics.scheduled_queries_per_second": 27244.81, "characteristics.scheduled_queries_per_second.normalized_per_core": 113.52004166666667, "characteristics.scheduled_queries_per_second.normalized_per_processor": 13622.405, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "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": 1, "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)": 10000000, "target_qps": 27200, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "e01f53c2414298f6", "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)": 11521674, "90.00 percentile latency (ns)": 12605303, "95.00 percentile latency (ns)": 13217656, "97.00 percentile latency (ns)": 13690492, "99.00 percentile latency (ns)": 14969117, "99.90 percentile latency (ns)": 25344053, "Completed samples per second": 7845.81, "Max latency (ns)": 55363087, "Mean latency (ns)": 11754402, "Min duration satisfied": "Yes", "Min latency (ns)": 10163202, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PyFastSUT", "Scenario": "server", "Scheduled samples per second": 7847.21, "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.scheduled_queries_per_second": 7847.21, "characteristics.scheduled_queries_per_second.normalized_per_core": 8.174177083333333, "characteristics.scheduled_queries_per_second.normalized_per_processor": 980.90125, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 7850, "task": "image classification", "task2": "image classification", "total_cores": 960, "uid": "c2a72a4cfd20c9b1", "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)": 13201295, "90.00 percentile latency (ns)": 13637076, "95.00 percentile latency (ns)": 13842332, "97.00 percentile latency (ns)": 14084012, "99.00 percentile latency (ns)": 14688585, "99.90 percentile latency (ns)": 16238165, "Completed samples per second": 7514.21, "Max latency (ns)": 29606563, "Mean latency (ns)": 13279231, "Min duration satisfied": "Yes", "Min latency (ns)": 12420796, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "CSUT", "Scenario": "server", "Scheduled samples per second": 7515.83, "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.scheduled_queries_per_second": 7515.83, "characteristics.scheduled_queries_per_second.normalized_per_core": 7.828989583333334, "characteristics.scheduled_queries_per_second.normalized_per_processor": 939.47875, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 7520, "task": "image classification", "task2": "image classification", "total_cores": 960, "uid": "654642cd8e02ca1b", "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)": 11454167, "90.00 percentile latency (ns)": 11854788, "95.00 percentile latency (ns)": 12121588, "97.00 percentile latency (ns)": 12600025, "99.00 percentile latency (ns)": 14170315, "99.90 percentile latency (ns)": 27552695, "Completed samples per second": 1748.31, "Max latency (ns)": 80416700, "Mean latency (ns)": 11572686, "Min duration satisfied": "Yes", "Min latency (ns)": 9792145, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "SUT\u0000", "Scenario": "server", "Scheduled samples per second": 1748.43, "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.scheduled_queries_per_second": 1748.43, "characteristics.scheduled_queries_per_second.normalized_per_core": 3.6425625, "characteristics.scheduled_queries_per_second.normalized_per_processor": 437.1075, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 1750, "task": "image classification", "task2": "image classification", "total_cores": 480, "uid": "7aba56403bba7462", "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)": 12162919, "90.00 percentile latency (ns)": 12451948, "95.00 percentile latency (ns)": 12931749, "97.00 percentile latency (ns)": 13530079, "99.00 percentile latency (ns)": 14651313, "99.90 percentile latency (ns)": 16715450, "Completed samples per second": 1728.31, "Max latency (ns)": 53559637, "Mean latency (ns)": 12277419, "Min duration satisfied": "Yes", "Min latency (ns)": 11514940, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "CSUT", "Scenario": "server", "Scheduled samples per second": 1728.45, "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.scheduled_queries_per_second": 1728.45, "characteristics.scheduled_queries_per_second.normalized_per_core": 7.201875, "characteristics.scheduled_queries_per_second.normalized_per_processor": 864.225, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 1730, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "a539604c11cc4404", "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)": 5668952, "90.00 percentile latency (ns)": 6731318, "95.00 percentile latency (ns)": 7395797, "97.00 percentile latency (ns)": 7995292, "99.00 percentile latency (ns)": 12693137, "99.90 percentile latency (ns)": 24312237, "Completed samples per second": 4295.73, "Max latency (ns)": 96458226, "Mean latency (ns)": 6022224, "Min duration satisfied": "Yes", "Min latency (ns)": 4934980, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "SUT\u0000", "Scenario": "server", "Scheduled samples per second": 4296.15, "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.scheduled_queries_per_second": 4296.15, "characteristics.scheduled_queries_per_second.normalized_per_core": 8.950312499999999, "characteristics.scheduled_queries_per_second.normalized_per_processor": 1074.0375, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 4300, "task": "image classification", "task2": "image classification", "total_cores": 480, "uid": "caba5e94ce382313", "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)": 11920747, "90.00 percentile latency (ns)": 12549075, "95.00 percentile latency (ns)": 13093170, "97.00 percentile latency (ns)": 13719998, "99.00 percentile latency (ns)": 14985745, "99.90 percentile latency (ns)": 17611970, "Completed samples per second": 1798.25, "Max latency (ns)": 34724108, "Mean latency (ns)": 12088624, "Min duration satisfied": "Yes", "Min latency (ns)": 10769269, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PyFastSUT", "Scenario": "server", "Scheduled samples per second": 1798.38, "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.scheduled_queries_per_second": 1798.38, "characteristics.scheduled_queries_per_second.normalized_per_core": 7.493250000000001, "characteristics.scheduled_queries_per_second.normalized_per_processor": 899.19, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 1800, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "3d87419c6f84822a", "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)": 13600390, "90.00 percentile latency (ns)": 14080700, "95.00 percentile latency (ns)": 14228413, "97.00 percentile latency (ns)": 14388022, "99.00 percentile latency (ns)": 14857748, "99.90 percentile latency (ns)": 16225421, "Completed samples per second": 3146.69, "Max latency (ns)": 18678644, "Mean latency (ns)": 13649972, "Min duration satisfied": "Yes", "Min latency (ns)": 12548432, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "CSUT", "Scenario": "server", "Scheduled samples per second": 3147.17, "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.scheduled_queries_per_second": 3147.17, "characteristics.scheduled_queries_per_second.normalized_per_core": 3.2783020833333336, "characteristics.scheduled_queries_per_second.normalized_per_processor": 393.39625, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 3150, "task": "image classification", "task2": "image classification", "total_cores": 960, "uid": "848f59c847c71952", "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)": 10790308, "90.00 percentile latency (ns)": 11882828, "95.00 percentile latency (ns)": 12834514, "97.00 percentile latency (ns)": 13514022, "99.00 percentile latency (ns)": 14830982, "99.90 percentile latency (ns)": 18020478, "Completed samples per second": 2077.98, "Max latency (ns)": 57344002, "Mean latency (ns)": 11015872, "Min duration satisfied": "Yes", "Min latency (ns)": 9213972, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "SUT\u0000", "Scenario": "server", "Scheduled samples per second": 2078.13, "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.scheduled_queries_per_second": 2078.13, "characteristics.scheduled_queries_per_second.normalized_per_core": 8.658875, "characteristics.scheduled_queries_per_second.normalized_per_processor": 1039.065, "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.scheduled_queries_per_second", "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": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 270336, "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": 1, "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)": 15000000, "target_qps": 2080, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "1fd491b9b17b0ba8", "use_accelerator": false, "weight_data_types": "No change.", "weight_transformations": "Please refer to close/Intel/calibration/OpenVINO" } ]