[ { "50.00 percentile latency (ns)": 318005, "90.00 percentile latency (ns)": 324052, "90th percentile latency (ns)": 324052, "95.00 percentile latency (ns)": 328379, "97.00 percentile latency (ns)": 330524, "99.00 percentile latency (ns)": 338085, "99.90 percentile latency (ns)": 785193, "Max latency (ns)": 100234513, "Mean latency (ns)": 322108, "Min duration satisfied": "Yes", "Min latency (ns)": 281312, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 3045.91, "QPS w/o loadgen overhead": 3104.55, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "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.90th_percentile_latency_ms": 0.324052, "characteristics.90th_percentile_latency_ns": 324052.0, "characteristics.90th_percentile_latency_s": 0.000324052, "characteristics.90th_percentile_latency_us": 324.052, "characteristics.accuracy": 75.776, "characteristics.good": 37888, "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.90th_percentile_latency_ms", "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": 1024, "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": 2048, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 1, "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": 2857.14, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "46f5affe8b63caea", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 280373, "90.00 percentile latency (ns)": 282475, "90th percentile latency (ns)": 282475, "95.00 percentile latency (ns)": 285207, "97.00 percentile latency (ns)": 287732, "99.00 percentile latency (ns)": 289431, "99.90 percentile latency (ns)": 292053, "Max latency (ns)": 3154185, "Mean latency (ns)": 280911, "Min duration satisfied": "Yes", "Min latency (ns)": 258323, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 3501.47, "QPS w/o loadgen overhead": 3559.85, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "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.90th_percentile_latency_ms": 0.282475, "characteristics.90th_percentile_latency_ns": 282475.0, "characteristics.90th_percentile_latency_s": 0.000282475, "characteristics.90th_percentile_latency_us": 282.475, "characteristics.accuracy": 75.776, "characteristics.good": 37888, "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.90th_percentile_latency_ms", "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": 1024, "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": 2048, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 1, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "build/models/ResNet50/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": 3333.33, "task": "image classification", "task2": "image classification", "total_cores": 64, "uid": "380b9bbfd52b7c38", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 464831, "90.00 percentile latency (ns)": 479590, "90th percentile latency (ns)": 479590, "95.00 percentile latency (ns)": 481697, "97.00 percentile latency (ns)": 483122, "99.00 percentile latency (ns)": 487419, "99.90 percentile latency (ns)": 497479, "Max latency (ns)": 4592143, "Mean latency (ns)": 467578, "Min duration satisfied": "Yes", "Min latency (ns)": 438992, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 2085.95, "QPS w/o loadgen overhead": 2138.68, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "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.90th_percentile_latency_ms": 0.47959, "characteristics.90th_percentile_latency_ns": 479590.0, "characteristics.90th_percentile_latency_s": 0.00047959, "characteristics.90th_percentile_latency_us": 479.59, "characteristics.accuracy": 75.776, "characteristics.good": 37888, "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.90th_percentile_latency_ms", "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": 1024, "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": 1, "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": 2040.82, "task": "image classification", "task2": "image classification", "total_cores": 48, "uid": "50435501874e42f9", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 478617, "90.00 percentile latency (ns)": 486693, "90th percentile latency (ns)": 486693, "95.00 percentile latency (ns)": 489944, "97.00 percentile latency (ns)": 492372, "99.00 percentile latency (ns)": 500047, "99.90 percentile latency (ns)": 1136387, "Max latency (ns)": 100189964, "Mean latency (ns)": 484108, "Min duration satisfied": "Yes", "Min latency (ns)": 440286, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 2039.24, "QPS w/o loadgen overhead": 2065.65, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "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.90th_percentile_latency_ms": 0.486693, "characteristics.90th_percentile_latency_ns": 486693.0, "characteristics.90th_percentile_latency_s": 0.000486693, "characteristics.90th_percentile_latency_us": 486.693, "characteristics.accuracy": 76.162, "characteristics.good": 38081, "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.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": "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": 1024, "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": 1, "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": 1515.15, "task": "image classification", "task2": "image classification", "total_cores": 52, "uid": "e4f1683445dda25f", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 425801, "90.00 percentile latency (ns)": 427415, "90th percentile latency (ns)": 427415, "95.00 percentile latency (ns)": 429796, "97.00 percentile latency (ns)": 433078, "99.00 percentile latency (ns)": 434855, "99.90 percentile latency (ns)": 436965, "Max latency (ns)": 4124456, "Mean latency (ns)": 426283, "Min duration satisfied": "Yes", "Min latency (ns)": 402730, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 2320.72, "QPS w/o loadgen overhead": 2345.86, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "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.90th_percentile_latency_ms": 0.427415, "characteristics.90th_percentile_latency_ns": 427415.0, "characteristics.90th_percentile_latency_s": 0.000427415, "characteristics.90th_percentile_latency_us": 427.415, "characteristics.accuracy": 76.032, "characteristics.good": 38016, "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.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": "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": 1024, "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": 1, "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": 2222.22, "task": "image classification", "task2": "image classification", "total_cores": 64, "uid": "bdbbac04f38c4a6a", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" } ]