[ { "50.00 percentile latency (ns)": 304456498707, "90.00 percentile latency (ns)": 546926104653, "95.00 percentile latency (ns)": 577073734815, "97.00 percentile latency (ns)": 589227824139, "99.00 percentile latency (ns)": 601388883183, "99.90 percentile latency (ns)": 606959998207, "Max latency (ns)": 607342434045, "Mean latency (ns)": 81305396619, "Min duration satisfied": "Yes", "Min latency (ns)": 1499279417, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 136142, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A100-SXM4-40GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 75.738, "characteristics.good": 37869, "characteristics.samples_per_second": 136142, "characteristics.samples_per_second.normalized_per_core": 136142.0, "characteristics.samples_per_second.normalized_per_processor": 136142.0, "characteristics.total": 50000, "ck_system": "DGX-A100_A100-SXM4-40GBx1_TRT", "ck_used": true, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "resnet50", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_memory_capacity": "768 GB", "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": "3.6 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/open/Alibaba/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/open/Alibaba/results/DGX-A100_A100-SXM4-40GBx1_TRT", "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": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 460.32.03, DALI 0.30.0; GCC 7.5.0; Python 3.7.10", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 8640, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 82684800, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "code/resnet50/tensorrt/ofa_autosinian_is176.onnx", "status": "available", "submitter": "Alibaba", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Alibaba", "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/DGX-A100_A100-SXM4-40GBx1_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 125280, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "021fae24740a53cf", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 303516623760, "90.00 percentile latency (ns)": 545805844705, "95.00 percentile latency (ns)": 576114012085, "97.00 percentile latency (ns)": 588159866406, "99.00 percentile latency (ns)": 600331679517, "99.90 percentile latency (ns)": 605796310624, "Max latency (ns)": 606317810223, "Mean latency (ns)": 303490439322, "Min duration satisfied": "Yes", "Min latency (ns)": 580935626, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 69514, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "22GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A10", "accelerator_on-chip_memories": "", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 75.738, "characteristics.good": 37869, "characteristics.samples_per_second": 69514, "characteristics.samples_per_second.normalized_per_core": 69514.0, "characteristics.samples_per_second.normalized_per_processor": 69514.0, "characteristics.total": 50000, "ck_system": "A10x1_TRT", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "resnet50", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_memory_capacity": "768 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 32, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8357C CPU @ 2.70GHz", "host_processors_per_node": 2, "host_storage_capacity": "3.5 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/open/Alibaba/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/open/Alibaba/results/A10x1_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.4", "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 460.32.03, DALI 0.30.0", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 8640, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 42147600, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "code/resnet50/tensorrt/ofa_autosinian_is176.onnx", "status": "Preview", "submitter": "Alibaba", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Alibaba", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A10x1_TRT", "system_name": "Alibaba Cloud Sinian Platform A10 (1x A10)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 63860, "task": "image classification", "task2": "image classification", "total_cores": 64, "uid": "1e4b2091d2399212", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 301463802450, "90.00 percentile latency (ns)": 543611892506, "95.00 percentile latency (ns)": 573892760023, "97.00 percentile latency (ns)": 586002603264, "99.00 percentile latency (ns)": 598100464243, "99.90 percentile latency (ns)": 603539842265, "Max latency (ns)": 604130298373, "Mean latency (ns)": 301587853172, "Min duration satisfied": "Yes", "Min latency (ns)": 303252373, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 30414.6, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "Nvidia Tesla T4", "accelerator_on-chip_memories": "", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 75.742, "characteristics.good": 37871, "characteristics.samples_per_second": 30414.6, "characteristics.samples_per_second.normalized_per_core": 30414.6, "characteristics.samples_per_second.normalized_per_processor": 30414.6, "characteristics.total": 50000, "ck_system": "T4x1_TRT", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "resnet50", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2.2.3, CUDA 11.2", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 24, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8163 CPU @ 2.50GHz", "host_processors_per_node": 2, "host_storage_capacity": "3.5 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/open/Alibaba/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/open/Alibaba/results/T4x1_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.4", "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 460.32.03, DALI 0.30.0", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 2048, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 18374400, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "code/resnet50/tensorrt/ofa_autosinian_is176.onnx", "status": "available", "submitter": "Alibaba", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Alibaba", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/T4x1_TRT", "system_name": "Alibaba Cloud Sinian Platform T4 (1x T4)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 27840, "task": "image classification", "task2": "image classification", "total_cores": 48, "uid": "22479bb38fd81f90", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 318959001476, "90.00 percentile latency (ns)": 574058127273, "95.00 percentile latency (ns)": 605888405338, "97.00 percentile latency (ns)": 618583123641, "99.00 percentile latency (ns)": 631484968952, "99.90 percentile latency (ns)": 637078579092, "Max latency (ns)": 637690515500, "Mean latency (ns)": 319017299257, "Min duration satisfied": "Yes", "Min latency (ns)": 574989878, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 38087.4, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A100-SXM4-40GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.028, "characteristics.good": 38014, "characteristics.samples_per_second": 38087.4, "characteristics.samples_per_second.normalized_per_core": 38087.4, "characteristics.samples_per_second.normalized_per_processor": 38087.4, "characteristics.total": 50000, "ck_system": "DGX-A100_A100-SXM4-40GBx1_TRT", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_memory_capacity": "768 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 26, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8269CY CPU @ 2.50GHz", "host_processors_per_node": 2, "host_storage_capacity": "3.6 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Alibaba/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Alibaba/results/DGX-A100_A100-SXM4-40GBx1_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.4", "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 460.32.03, DALI 0.30.0", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 2048, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 24288000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "Alibaba", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Alibaba", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/DGX-A100_A100-SXM4-40GBx1_TRT", "system_name": "Alibaba Cloud Sinian Platform A100 (1x A100-SXM4-40GB)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 36800, "task": "image classification", "task2": "image classification", "total_cores": 52, "uid": "4224bc6e69809122", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 317813626365, "90.00 percentile latency (ns)": 572889625936, "95.00 percentile latency (ns)": 604703243901, "97.00 percentile latency (ns)": 617486965658, "99.00 percentile latency (ns)": 630271358095, "99.90 percentile latency (ns)": 635977298980, "Max latency (ns)": 636579123787, "Mean latency (ns)": 317913504645, "Min duration satisfied": "Yes", "Min latency (ns)": 156959933, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 13478.3, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "22GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A10", "accelerator_on-chip_memories": "", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.028, "characteristics.good": 38014, "characteristics.samples_per_second": 13478.3, "characteristics.samples_per_second.normalized_per_core": 13478.3, "characteristics.samples_per_second.normalized_per_processor": 13478.3, "characteristics.total": 50000, "ck_system": "A10x1_TRT", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_memory_capacity": "768 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 32, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8357C CPU @ 2.70GHz", "host_processors_per_node": 2, "host_storage_capacity": "3.5 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Alibaba/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Alibaba/results/A10x1_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.4", "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 460.32.03, DALI 0.30.0", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 2048, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 8580000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "resnet50_v1.onnx", "status": "Preview", "submitter": "Alibaba", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Alibaba", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/A10x1_TRT", "system_name": "Alibaba Cloud Sinian Platform A10 (1x A10)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 13000, "task": "image classification", "task2": "image classification", "total_cores": 64, "uid": "4a214353d4038a97", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" } ]