[ { "50.00 percentile": 1, "50.00 percentile latency (ns)": 48555858, "90.00 percentile": 1, "90.00 percentile latency (ns)": 48643782, "95.00 percentile": 1, "95.00 percentile latency (ns)": 48692395, "97.00 percentile": 1, "97.00 percentile latency (ns)": 48807736, "99.00 percentile": 1, "99.00 percentile latency (ns)": 49305533, "99.90 percentile": 1, "99.90 percentile latency (ns)": 49808527, "Max latency (ns)": 52252272, "Mean latency (ns)": 48575300, "Min duration satisfied": "Yes", "Min latency (ns)": 48311112, "Min queries satisfied": "Yes", "Mode": "Performance", "Per-sample latency": "", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "PySUT", "Samples per query": 70, "Scenario": "multistream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "Shared with host", "accelerator_memory_configuration": "SRAM", "accelerator_model_name": "NVIDIA Xavier", "accelerator_on-chip_memories": "1MB (128KB/SM) L1 + 512KB L2 + 4MB (DLA)", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.258, "characteristics.good": 38129, "characteristics.samples_per_query": 70, "characteristics.samples_per_query.normalized_per_core": 70.0, "characteristics.samples_per_query.normalized_per_processor": 70.0, "characteristics.total": 50000, "ck_system": "xavier-tensorrt-v6.0", "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_query", "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 v6.0", "host_memory_capacity": "32 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "8MB L2 (2MB per dual cluster)/4 MB L3 (shared)", "host_processor_core_count": 120, "host_processor_frequency": "2265.6 MHz", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 1, "host_storage_capacity": "400 GB", "host_storage_type": "SanDisk Extreme microSD", "hw_notes": "DLAs are not used in the Offline and MultiStream scenarios (GPU only)", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 272160, "min_duration (ms)": 60000, "min_query_count": 270336, "mlperf_version": 0.7, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/dividiti/results/xavier-tensorrt-v6.0", "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": "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": 12786827339337101903, "retraining": "no", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 70, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://zenodo.org/record/3716059/files/resnet-MultiStream-gpu-b70-int8-linear.plan", "status": "available", "submitter": "dividiti", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/dividiti", "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/xavier-tensorrt-v6.0", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 50000000, "target_qps": 20, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "3a1c39d6f9a8bb72", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "ONNX -> TensorRT" } ]