[ { "50.00 percentile latency (ns)": 898891901428, "90.00 percentile latency (ns)": 1264578500197, "95.00 percentile latency (ns)": 1294575522888, "97.00 percentile latency (ns)": 1305218319957, "99.00 percentile latency (ns)": 1314896299221, "99.90 percentile latency (ns)": 1318643584935, "Max latency (ns)": 1318935818370, "Mean latency (ns)": 826249673628, "Min duration satisfied": "Yes", "Min latency (ns)": 1976747908, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "RNNT SERVER", "Samples per second": 5704.6, "Scenario": "offline", "accelerator_cooling_type": "passive", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA T4", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "1.1.3", "characteristics.accuracy": 92.5545108119687, "characteristics.power": 856.7994692949202, "characteristics.power.normalized_per_core": 214.19986732373005, "characteristics.power.normalized_per_processor": 214.19986732373005, "characteristics.samples_per_second": 5704.6, "characteristics.samples_per_second.normalized_per_core": 1426.15, "characteristics.samples_per_second.normalized_per_processor": 1426.15, "characteristics.word error rate": 7.4454891880312974, "ck_system": "XE2420_edge_T4x4_TRT", "ck_used": true, "cooling": "", "dataset": "LibriSpeech", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "filesystem": "ext3/ext4", "formal_model": "rnn-t", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_cooling_type": "passive", "host_memory_capacity": "384 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": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC on", "informal_model": "rnnt", "input_data_types": "fp16", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "4.40.10.00", "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": 4, "normalize_processors": 4, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/DellEMC/results/XE2420_edge_T4x4_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": 2513, "power_settings": "closed/DellEMC/measurements/XE2420_edge_T4x4_TRT/XE2420_T4x4_power_settings.adoc", "print_timestamps": 0, "problem": false, "psu_details": "2x2000W", "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 7524000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "DistributedDataParallel_1576581068.9962234-epoch-100.pt", "status": "available", "submitter": "DellEMC", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/DellEMC", "sw_notes": "Powered by CK v2.5.8 (https://github.com/ctuning/ck)", "system_cooling_type": "air", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/XE2420_edge_T4x4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 11400, "task": "speech recognition", "task2": "speech recognition", "total_cores": 240, "uid": "7482da3b84d4b8bb", "use_accelerator": true, "weight_data_types": "fp16", "weight_transformations": "quantization, affine fusion" } ]