[ { "50.00 percentile latency (ns)": 18330795, "90.00 percentile latency (ns)": 20072468, "95.00 percentile latency (ns)": 20776045, "97.00 percentile latency (ns)": 21234820, "99.00 percentile latency (ns)": 22101459, "99.90 percentile latency (ns)": 23760356, "Completed samples per second": 20245.06, "Max latency (ns)": 46287298, "Mean latency (ns)": 18497785, "Min duration satisfied": "Yes", "Min latency (ns)": 14502638, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PyFastSUT", "Scenario": "server", "Scheduled samples per second": 20245.82, "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.AUC": 80.21599672162321, "characteristics.scheduled_queries_per_second": 20245.82, "characteristics.scheduled_queries_per_second.normalized_per_core": 253.07274999999998, "characteristics.scheduled_queries_per_second.normalized_per_processor": 10122.91, "ck_system": "1-node-2S-ICX-PyTorch-INT8", "ck_used": false, "cooling": "Air", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.scheduled_queries_per_second", "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "division": "closed", "formal_model": "dlrm", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "dlrm-99.9", "input_data_types": "No change.", "key.accuracy": "characteristics.AUC", "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": 204800, "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://dlrm.s3-us-west-1.amazonaws.com/models/tb00_40M.pt", "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)": 30000000, "target_qps": 20250, "task": "recommendation", "task2": "recommendation", "total_cores": 80, "uid": "1beeb1378c8d3b06", "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)": 18330795, "90.00 percentile latency (ns)": 20072468, "95.00 percentile latency (ns)": 20776045, "97.00 percentile latency (ns)": 21234820, "99.00 percentile latency (ns)": 22101459, "99.90 percentile latency (ns)": 23760356, "Completed samples per second": 20245.06, "Max latency (ns)": 46287298, "Mean latency (ns)": 18497785, "Min duration satisfied": "Yes", "Min latency (ns)": 14502638, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PyFastSUT", "Scenario": "server", "Scheduled samples per second": 20245.82, "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.AUC": 80.21599672162321, "characteristics.scheduled_queries_per_second": 20245.82, "characteristics.scheduled_queries_per_second.normalized_per_core": 253.07274999999998, "characteristics.scheduled_queries_per_second.normalized_per_processor": 10122.91, "ck_system": "1-node-2S-ICX-PyTorch-INT8", "ck_used": false, "cooling": "Air", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.scheduled_queries_per_second", "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "division": "closed", "formal_model": "dlrm", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "dlrm-99", "input_data_types": "No change.", "key.accuracy": "characteristics.AUC", "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": 204800, "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://dlrm.s3-us-west-1.amazonaws.com/models/tb00_40M.pt", "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)": 30000000, "target_qps": 20250, "task": "recommendation", "task2": "recommendation", "total_cores": 80, "uid": "d54f40598153bf14", "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)": 15116446, "90.00 percentile latency (ns)": 19210168, "95.00 percentile latency (ns)": 20732956, "97.00 percentile latency (ns)": 21817188, "99.00 percentile latency (ns)": 24285577, "99.90 percentile latency (ns)": 33519939, "Completed samples per second": 5798.47, "Max latency (ns)": 551365337, "Mean latency (ns)": 15614043, "Min duration satisfied": "Yes", "Min latency (ns)": 9431679, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PyFastSUT", "Scenario": "server", "Scheduled samples per second": 5801.47, "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.AUC": 79.3158953866414, "characteristics.scheduled_queries_per_second": 5801.47, "characteristics.scheduled_queries_per_second.normalized_per_core": 24.17279166666667, "characteristics.scheduled_queries_per_second.normalized_per_processor": 2900.735, "ck_system": "1-node-2S-ICX-PyTorch-FP32", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.scheduled_queries_per_second", "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "division": "open", "formal_model": "dlrm-99", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": 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": "dlrm-99", "input_data_types": "No change.", "key.accuracy": "characteristics.AUC", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 6144000, "mlperf_version": 1.0, "normalize_cores": 240, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/open/Intel/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/open/Intel/results/1-node-2S-ICX-PyTorch-FP32", "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": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 204800, "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": "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-2S-ICX-PyTorch-FP32", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 5800, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "0d4847842368d382", "use_accelerator": false, "weight_data_types": "fp32", "weight_transformations": "no need to transform" }, { "50.00 percentile latency (ns)": 19495310, "90.00 percentile latency (ns)": 21825766, "95.00 percentile latency (ns)": 22816550, "97.00 percentile latency (ns)": 23608891, "99.00 percentile latency (ns)": 26194390, "99.90 percentile latency (ns)": 149409922, "Completed samples per second": 23970.86, "Max latency (ns)": 703762678, "Mean latency (ns)": 20292756, "Min duration satisfied": "Yes", "Min latency (ns)": 15055442, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PyFastSUT", "Scenario": "server", "Scheduled samples per second": 23995.5, "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.AUC": 80.26857042258752, "characteristics.scheduled_queries_per_second": 23995.5, "characteristics.scheduled_queries_per_second.normalized_per_core": 24.9953125, "characteristics.scheduled_queries_per_second.normalized_per_processor": 2999.4375, "ck_system": "1-node-8S-CPX-PyTorch-BF16", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.scheduled_queries_per_second", "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "division": "closed", "formal_model": "dlrm", "formal_model_accuracy": 99.9, "formal_model_link": "", "framework": "PyTorch", "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": "dlrm-99.9", "input_data_types": "No change.", "key.accuracy": "characteristics.AUC", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 16384000, "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-PyTorch-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": 204800, "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-PyTorch-BF16", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 24000, "task": "recommendation", "task2": "recommendation", "total_cores": 960, "uid": "bb3bb3e10c924b2e", "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)": 19495310, "90.00 percentile latency (ns)": 21825766, "95.00 percentile latency (ns)": 22816550, "97.00 percentile latency (ns)": 23608891, "99.00 percentile latency (ns)": 26194390, "99.90 percentile latency (ns)": 149409922, "Completed samples per second": 23970.86, "Max latency (ns)": 703762678, "Mean latency (ns)": 20292756, "Min duration satisfied": "Yes", "Min latency (ns)": 15055442, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PyFastSUT", "Scenario": "server", "Scheduled samples per second": 23995.5, "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.AUC": 80.26857042258752, "characteristics.scheduled_queries_per_second": 23995.5, "characteristics.scheduled_queries_per_second.normalized_per_core": 24.9953125, "characteristics.scheduled_queries_per_second.normalized_per_processor": 2999.4375, "ck_system": "1-node-8S-CPX-PyTorch-BF16", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.scheduled_queries_per_second", "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "division": "closed", "formal_model": "dlrm", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "PyTorch", "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": "dlrm-99", "input_data_types": "No change.", "key.accuracy": "characteristics.AUC", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 16384000, "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-PyTorch-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": 204800, "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-PyTorch-BF16", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 24000, "task": "recommendation", "task2": "recommendation", "total_cores": 960, "uid": "4557bd92d79f1386", "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)": 22438507, "90.00 percentile latency (ns)": 24828576, "95.00 percentile latency (ns)": 25743345, "97.00 percentile latency (ns)": 26417310, "99.00 percentile latency (ns)": 27889722, "99.90 percentile latency (ns)": 38230809, "Completed samples per second": 12185.58, "Max latency (ns)": 773036573, "Mean latency (ns)": 22667878, "Min duration satisfied": "Yes", "Min latency (ns)": 17726803, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PyFastSUT", "Scenario": "server", "Scheduled samples per second": 12194.87, "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.AUC": 80.26857042258752, "characteristics.scheduled_queries_per_second": 12194.87, "characteristics.scheduled_queries_per_second.normalized_per_core": 25.40597916666667, "characteristics.scheduled_queries_per_second.normalized_per_processor": 3048.7175, "ck_system": "1-node-4S-CPX-PyTorch-BF16", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.scheduled_queries_per_second", "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "division": "closed", "formal_model": "dlrm", "formal_model_accuracy": 99.9, "formal_model_link": "", "framework": "PyTorch", "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": "dlrm-99.9", "input_data_types": "No change.", "key.accuracy": "characteristics.AUC", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 12288000, "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-PyTorch-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": 204800, "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-4S-CPX-PyTorch-BF16", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 12200, "task": "recommendation", "task2": "recommendation", "total_cores": 480, "uid": "95d82634b77b447a", "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)": 22438507, "90.00 percentile latency (ns)": 24828576, "95.00 percentile latency (ns)": 25743345, "97.00 percentile latency (ns)": 26417310, "99.00 percentile latency (ns)": 27889722, "99.90 percentile latency (ns)": 38230809, "Completed samples per second": 12185.58, "Max latency (ns)": 773036573, "Mean latency (ns)": 22667878, "Min duration satisfied": "Yes", "Min latency (ns)": 17726803, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PyFastSUT", "Scenario": "server", "Scheduled samples per second": 12194.87, "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.AUC": 80.26857042258752, "characteristics.scheduled_queries_per_second": 12194.87, "characteristics.scheduled_queries_per_second.normalized_per_core": 25.40597916666667, "characteristics.scheduled_queries_per_second.normalized_per_processor": 3048.7175, "ck_system": "1-node-4S-CPX-PyTorch-BF16", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.scheduled_queries_per_second", "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "division": "closed", "formal_model": "dlrm", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "PyTorch", "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": "dlrm-99", "input_data_types": "No change.", "key.accuracy": "characteristics.AUC", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 12288000, "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-PyTorch-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": 204800, "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-4S-CPX-PyTorch-BF16", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 12200, "task": "recommendation", "task2": "recommendation", "total_cores": 480, "uid": "c2f5db4476e4e4c9", "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)": 11767617, "90.00 percentile latency (ns)": 16776433, "95.00 percentile latency (ns)": 19003059, "97.00 percentile latency (ns)": 21054441, "99.00 percentile latency (ns)": 28127247, "99.90 percentile latency (ns)": 48170583, "Completed samples per second": 8853.55, "Max latency (ns)": 58670563, "Mean latency (ns)": 12300247, "Min duration satisfied": "Yes", "Min latency (ns)": 4974994, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PyFastSUT", "Scenario": "server", "Scheduled samples per second": 8853.93, "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.AUC": 80.26857129194714, "characteristics.scheduled_queries_per_second": 8853.93, "characteristics.scheduled_queries_per_second.normalized_per_core": 9.222843750000001, "characteristics.scheduled_queries_per_second.normalized_per_processor": 1106.74125, "ck_system": "1-node-8s-cpx-pt-bf16", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.scheduled_queries_per_second", "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "division": "closed", "formal_model": "dlrm", "formal_model_accuracy": 99.9, "formal_model_link": "", "framework": "PyTorch", "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": "dlrm-99.9", "input_data_types": "No change.", "key.accuracy": "characteristics.AUC", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1622080, "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-pt-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": 204800, "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-pt-bf16", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 8850, "task": "recommendation", "task2": "recommendation", "total_cores": 960, "uid": "03ebfc3474090c42", "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)": 11767617, "90.00 percentile latency (ns)": 16776433, "95.00 percentile latency (ns)": 19003059, "97.00 percentile latency (ns)": 21054441, "99.00 percentile latency (ns)": 28127247, "99.90 percentile latency (ns)": 48170583, "Completed samples per second": 8853.55, "Max latency (ns)": 58670563, "Mean latency (ns)": 12300247, "Min duration satisfied": "Yes", "Min latency (ns)": 4974994, "Min queries satisfied": "Yes", "Mode": "Performance", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PyFastSUT", "Scenario": "server", "Scheduled samples per second": 8853.93, "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.AUC": 80.26857129194714, "characteristics.scheduled_queries_per_second": 8853.93, "characteristics.scheduled_queries_per_second.normalized_per_core": 9.222843750000001, "characteristics.scheduled_queries_per_second.normalized_per_processor": 1106.74125, "ck_system": "1-node-8s-cpx-pt-bf16", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.scheduled_queries_per_second", "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "division": "closed", "formal_model": "dlrm", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "PyTorch", "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": "dlrm-99", "input_data_types": "No change.", "key.accuracy": "characteristics.AUC", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1622080, "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-pt-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": 204800, "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-pt-bf16", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 30000000, "target_qps": 8850, "task": "recommendation", "task2": "recommendation", "total_cores": 960, "uid": "734d1274218a9612", "use_accelerator": false, "weight_data_types": "Google bfloat16.", "weight_transformations": "We transform float32 weight into bfloat16 using TensorFlow's Cast operation." } ]