[ { "50.00 percentile latency (ns)": 1345396, "90.00 percentile latency (ns)": 1374171, "90th percentile latency (ns)": 1374171, "95.00 percentile latency (ns)": 1381678, "97.00 percentile latency (ns)": 1387733, "99.00 percentile latency (ns)": 1400925, "99.90 percentile latency (ns)": 1531196, "Max latency (ns)": 6577354, "Mean latency (ns)": 1343589, "Min duration satisfied": "Yes", "Min latency (ns)": 1260162, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 742.03, "QPS w/o loadgen overhead": 744.28, "Result is": "VALID", "SUT name": "CSUT", "Scenario": "singlestream", "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.90th_percentile_latency_ms": 1.374171, "characteristics.90th_percentile_latency_ns": 1374171.0, "characteristics.90th_percentile_latency_s": 0.001374171, "characteristics.90th_percentile_latency_us": 1374.171, "characteristics.accuracy": 76.124, "characteristics.good": 38062, "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.90th_percentile_latency_ms", "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": 1024, "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)": 0, "target_qps": 1000, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "3b1499c87a802727", "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)": 477106, "90.00 percentile latency (ns)": 493552, "90th percentile latency (ns)": 493552, "95.00 percentile latency (ns)": 499135, "97.00 percentile latency (ns)": 507329, "99.00 percentile latency (ns)": 521966, "99.90 percentile latency (ns)": 539748, "Max latency (ns)": 2163028, "Mean latency (ns)": 474612, "Min duration satisfied": "Yes", "Min latency (ns)": 421161, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 2091.78, "QPS w/o loadgen overhead": 2106.98, "Result is": "VALID", "SUT name": "CSUT", "Scenario": "singlestream", "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.90th_percentile_latency_ms": 0.493552, "characteristics.90th_percentile_latency_ns": 493552.0, "characteristics.90th_percentile_latency_s": 0.000493552, "characteristics.90th_percentile_latency_us": 493.552, "characteristics.accuracy": 70.55, "characteristics.good": 35275, "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.90th_percentile_latency_ms", "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": 1024, "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)": 0, "target_qps": 2000, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "3e4fea9b629f0853", "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." } ]