[ { "50.00 percentile latency (ns)": 307533442056, "90.00 percentile latency (ns)": 546156973385, "95.00 percentile latency (ns)": 576006506356, "97.00 percentile latency (ns)": 587926144714, "99.00 percentile latency (ns)": 599890078371, "99.90 percentile latency (ns)": 605226776150, "Max latency (ns)": 605795433820, "Mean latency (ns)": 25013621272, "Min duration satisfied": "Yes", "Min latency (ns)": 9296251450, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 1077800.0, "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": 8, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 75.736, "characteristics.good": 37868, "characteristics.samples_per_second": 1077800.0, "characteristics.samples_per_second.normalized_per_core": 134725.0, "characteristics.samples_per_second.normalized_per_processor": 134725.0, "characteristics.total": 50000, "ck_system": "DGX-A100_A100-SXM4-40GBx8_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": 8, "normalize_processors": 8, "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-SXM4-40GBx8_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": 652924800, "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-SXM4-40GBx8_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 989280, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "45f78c53fc85f54f", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 309804130936, "90.00 percentile latency (ns)": 555481989327, "95.00 percentile latency (ns)": 586151209553, "97.00 percentile latency (ns)": 598425719102, "99.00 percentile latency (ns)": 610639364162, "99.90 percentile latency (ns)": 616196130529, "Max latency (ns)": 616745968983, "Mean latency (ns)": 90475794657, "Min duration satisfied": "Yes", "Min latency (ns)": 1012276826, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 136677, "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": 2, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 75.74, "characteristics.good": 37870, "characteristics.samples_per_second": 136677, "characteristics.samples_per_second.normalized_per_core": 68338.5, "characteristics.samples_per_second.normalized_per_processor": 68338.5, "characteristics.total": 50000, "ck_system": "A10x2_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": "3.5 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": 2, "normalize_processors": 2, "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/A10x2_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": 84295200, "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": "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/A10x2_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 127720, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "ee3d09a4714a2914", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 300964325729, "90.00 percentile latency (ns)": 540297338367, "95.00 percentile latency (ns)": 570228289686, "97.00 percentile latency (ns)": 582195095621, "99.00 percentile latency (ns)": 594159615784, "99.90 percentile latency (ns)": 599525067287, "Max latency (ns)": 600123779090, "Mean latency (ns)": 144998157855, "Min duration satisfied": "Yes", "Min latency (ns)": 1761829924, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 197079, "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": 8, "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": 197079, "characteristics.samples_per_second.normalized_per_core": 24634.875, "characteristics.samples_per_second.normalized_per_processor": 24634.875, "characteristics.total": 50000, "ck_system": "T4x8_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 7.2.2.3, CUDA 11.2", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 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": "7 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": 8, "normalize_processors": 8, "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/T4x8_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": 118272000, "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/T4x8_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 179200, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "770ba4fef8bf8cb6", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 323416258112, "90.00 percentile latency (ns)": 580872573534, "95.00 percentile latency (ns)": 613106496820, "97.00 percentile latency (ns)": 625982246537, "99.00 percentile latency (ns)": 638850614323, "99.90 percentile latency (ns)": 644672640734, "Max latency (ns)": 645297916740, "Mean latency (ns)": 38823638099, "Min duration satisfied": "Yes", "Min latency (ns)": 2896442303, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 301107, "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": 8, "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": 301107, "characteristics.samples_per_second.normalized_per_core": 37638.375, "characteristics.samples_per_second.normalized_per_processor": 37638.375, "characteristics.total": 50000, "ck_system": "DGX-A100_A100-SXM4-40GBx8_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": 8, "normalize_processors": 8, "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-SXM4-40GBx8_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": 194304000, "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-SXM4-40GBx8_TRT", "system_name": "Alibaba Cloud Sinian Platform A100 (8x A100-SXM4-40GB)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 294400, "task": "image classification", "task2": "image classification", "total_cores": 52, "uid": "48c9c3af13b4b91a", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 312989108725, "90.00 percentile latency (ns)": 562032502361, "95.00 percentile latency (ns)": 593126212439, "97.00 percentile latency (ns)": 605613740550, "99.00 percentile latency (ns)": 618038428505, "99.90 percentile latency (ns)": 623660508105, "Max latency (ns)": 624234175178, "Mean latency (ns)": 312590926995, "Min duration satisfied": "Yes", "Min latency (ns)": 246953908, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 27489.7, "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": 2, "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": 27489.7, "characteristics.samples_per_second.normalized_per_core": 13744.85, "characteristics.samples_per_second.normalized_per_processor": 13744.85, "characteristics.total": 50000, "ck_system": "A10x2_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": 2, "normalize_processors": 2, "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/A10x2_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": 17160000, "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/A10x2_TRT", "system_name": "Alibaba Cloud Sinian Platform A10 (2x A10)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 26000, "task": "image classification", "task2": "image classification", "total_cores": 64, "uid": "c5ce96a04ddbab3b", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 32516648295, "90.00 percentile latency (ns)": 59083964256, "95.00 percentile latency (ns)": 62456887113, "97.00 percentile latency (ns)": 63838958486, "99.00 percentile latency (ns)": 65176213076, "99.90 percentile latency (ns)": 65798536314, "Max latency (ns)": 65855871251, "Mean latency (ns)": 32642289328, "Min duration satisfied": "Yes", "Min latency (ns)": 42277021, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 5540.1, "Scenario": "offline", "accelerator_memory_capacity": "16GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "Nvidia Tesla T4", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "characteristics.accuracy": 76.034, "characteristics.good": 38017, "characteristics.samples_per_second": 5540.1, "characteristics.samples_per_second.normalized_per_core": 5540.1, "characteristics.samples_per_second.normalized_per_processor": 5540.1, "characteristics.total": 50000, "ck_system": "alibaba_cloud_t4", "ck_used": true, "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 6.0, CUDA 10.1, cuDNN 7.6.3, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "93GB", "host_processor_core_count": 120, "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processor_vcpu_count": 24, "host_processors_per_node": 1, "host_storage_capacity": "500GB", "host_storage_type": "SSD", "informal_model": "resnet", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1, "mlperf_version": 0.5, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/Alibaba/code", "note_details": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/Alibaba/results/alibaba_cloud_t4", "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": "docker 18.09.2, python 3.6.8,gcc 5.5.0,onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; 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": 3133965575612453542, "retraining": "N", "sample_index_rng_seed": 665484352860916858, "samples_per_query": 364848, "schedule_rng_seed": 3622009729038561421, "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": "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/alibaba_cloud_t4", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "target_latency (ns)": 0, "target_qps": 5528, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "0428c91f791c8269", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 33566677939, "90.00 percentile latency (ns)": 61040149602, "95.00 percentile latency (ns)": 64556032567, "97.00 percentile latency (ns)": 65967608302, "99.00 percentile latency (ns)": 67381367035, "99.90 percentile latency (ns)": 68012692641, "Max latency (ns)": 68083718492, "Mean latency (ns)": 33692696182, "Min duration satisfied": "Yes", "Min latency (ns)": 23593699, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 17473.6, "Scenario": "offline", "accelerator_memory_capacity": "16GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "Nvidia Tesla T4", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "characteristics.accuracy": 70.814, "characteristics.good": 35407, "characteristics.samples_per_second": 17473.6, "characteristics.samples_per_second.normalized_per_core": 17473.6, "characteristics.samples_per_second.normalized_per_processor": 17473.6, "characteristics.total": 50000, "ck_system": "alibaba_cloud_t4", "ck_used": true, "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 6.0, CUDA 10.1, cuDNN 7.6.3, libjemalloc2, cub 1.8.0, tensorrt-laboratory mlperf branch", "host_memory_capacity": "93GB", "host_processor_core_count": 120, "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processor_vcpu_count": 24, "host_processors_per_node": 1, "host_storage_capacity": "500GB", "host_storage_type": "SSD", "informal_model": "mobilenet", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1, "mlperf_version": 0.5, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/Alibaba/code", "note_details": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/Alibaba/results/alibaba_cloud_t4", "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": "docker 18.09.2, python 3.6.8,gcc 5.5.0,onnx 1.3.0, tensorflow 1.13.1, pytorch 1.1.0, torchvision 0.3.0, pycuda 2019.1, sacrebleu 1.3.3, SimpleJSON, OpenCV 4.1.1; 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": 3133965575612453542, "retraining": "N", "sample_index_rng_seed": 665484352860916858, "samples_per_query": 1189669, "schedule_rng_seed": 3622009729038561421, "starting_weights_filename": "mobilenet_sym_no_bn.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/alibaba_cloud_t4", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "target_latency (ns)": 0, "target_qps": 18025.3, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "e76c4947a3e55636", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 32879933123, "90.00 percentile latency (ns)": 59147406022, "95.00 percentile latency (ns)": 62423840838, "97.00 percentile latency (ns)": 63740886389, "99.00 percentile latency (ns)": 65058643407, "99.90 percentile latency (ns)": 65660673606, "Max latency (ns)": 65708059266, "Mean latency (ns)": 32872736822, "Min duration satisfied": "Yes", "Min latency (ns)": 36138239, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "", "Samples per second": 69306.6, "Scenario": "offline", "accelerator_memory_capacity": 0, "accelerator_memory_configuration": "n/a", "accelerator_model_name": "HanGuang 800", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "characteristics.accuracy": 76.038, "characteristics.good": 38019, "characteristics.samples_per_second": 69306.6, "characteristics.samples_per_second.normalized_per_core": 69306.6, "characteristics.samples_per_second.normalized_per_processor": 69306.6, "characteristics.total": 50000, "ck_system": "alibaba_HanGuang", "ck_used": true, "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": "TensorFlow 1.13.1 6612da89516247503f03ef76e974b51a434fb52e", "host_memory_capacity": "512GB", "host_processor_core_count": 120, "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "12TB", "host_storage_type": "SSD", "informal_model": "resnet", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1, "mlperf_version": 0.5, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/Alibaba/code", "note_details": "https://github.com/mlcommons/inference_results_v0.5/tree/master/closed/Alibaba/results/alibaba_HanGuang", "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": "HanGuangAI 1.0.3, python 3.5.2; GCC 7.5.0; Python 3.7.10", "performance_issue_same": true, "performance_issue_same_index": 0, "performance_issue_unique": true, "performance_sample_count": 1600, "print_timestamps": true, "problem": false, "qsl_rng_seed": 3133965575612453542, "retraining": "none", "sample_index_rng_seed": 665484352860916858, "samples_per_query": 4554000, "schedule_rng_seed": 3622009729038561421, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "preview", "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/alibaba_HanGuang", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "target_latency (ns)": 0, "target_qps": 69000, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "4345169c8258ef9e", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization" } ]