[ { "50.00 percentile latency (ns)": 476754982268, "90.00 percentile latency (ns)": 659693325134, "95.00 percentile latency (ns)": 675329146577, "97.00 percentile latency (ns)": 680832173750, "99.00 percentile latency (ns)": 685724984333, "99.90 percentile latency (ns)": 687549640480, "Max latency (ns)": 687670310964, "Mean latency (ns)": 434644748827, "Min duration satisfied": "Yes", "Min latency (ns)": 614306310, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "RNNT SERVER", "Samples per second": 1343.67, "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 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": 92.5680398656114, "characteristics.samples_per_second": 1343.67, "characteristics.samples_per_second.normalized_per_core": 1343.67, "characteristics.samples_per_second.normalized_per_processor": 1343.67, "characteristics.word error rate": 7.4319601343886, "ck_system": "T4x1_TRT", "ck_used": false, "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", "formal_model": "rnn-t", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "TensorRT 8.0.2, CUDA 11.3", "host_memory_capacity": "128 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 4, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) D-2123IT CPU @ 2.20GHz", "host_processors_per_node": 1, "host_storage_capacity": "2 TB", "host_storage_type": "M.2", "hw_notes": "ECC on", "informal_model": "rnnt", "input_data_types": "fp16", "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": 1.1, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Lenovo/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Lenovo/results/T4x1_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 20.04.2", "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.57.02", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 2513, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 924000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "DistributedDataParallel_1576581068.9962234-epoch-100.pt", "status": "available", "submitter": "Lenovo", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Lenovo", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/T4x1_TRT", "system_name": "Lenovo ThinkSystem SE350 server", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1400, "task": "speech recognition", "task2": "speech recognition", "total_cores": 4, "uid": "d8196d4040800c40", "use_accelerator": true, "weight_data_types": "fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 488951752833, "90.00 percentile latency (ns)": 686419787193, "95.00 percentile latency (ns)": 703028897461, "97.00 percentile latency (ns)": 708969056370, "99.00 percentile latency (ns)": 714355986866, "99.90 percentile latency (ns)": 716427186163, "Max latency (ns)": 716558317684, "Mean latency (ns)": 446337899947, "Min duration satisfied": "Yes", "Min latency (ns)": 1051068319, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "RNNT SERVER", "Samples per second": 1289.5, "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 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": 92.5545108119687, "characteristics.samples_per_second": 1289.5, "characteristics.samples_per_second.normalized_per_core": 1289.5, "characteristics.samples_per_second.normalized_per_processor": 1289.5, "characteristics.word error rate": 7.4454891880312974, "ck_system": "T4x1_TRT", "ck_used": false, "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", "formal_model": "rnn-t", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_memory_capacity": "128 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 4, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) D-2123IT CPU @ 2.20GHz", "host_processors_per_node": 1, "host_storage_capacity": "2 TB", "host_storage_type": "M.2", "hw_notes": "ECC on", "informal_model": "rnnt", "input_data_types": "fp16", "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/Lenovo/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Lenovo/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": 2513, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 924000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "DistributedDataParallel_1576581068.9962234-epoch-100.pt", "status": "available", "submitter": "Lenovo", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Lenovo", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/T4x1_TRT", "system_name": "Lenovo ThinkSystem SE350 server", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1400, "task": "speech recognition", "task2": "speech recognition", "total_cores": 4, "uid": "e977844b25827114", "use_accelerator": true, "weight_data_types": "fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 43371022863, "90.00 percentile latency (ns)": 60702197798, "95.00 percentile latency (ns)": 62027051952, "97.00 percentile latency (ns)": 62544523850, "99.00 percentile latency (ns)": 62968782735, "99.90 percentile latency (ns)": 62986108210, "Max latency (ns)": 62986140498, "Mean latency (ns)": 40411586825, "Min duration satisfied": "Yes", "Min latency (ns)": 1690762425, "Min queries satisfied": "Yes", "Mode": "Performance", "Result is": "VALID", "SUT name": "RNNT SERVER", "Samples per second": 1466.99, "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 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": 92.5567656542425, "characteristics.samples_per_second": 1466.99, "characteristics.samples_per_second.normalized_per_core": 1466.99, "characteristics.samples_per_second.normalized_per_processor": 1466.99, "characteristics.word error rate": 7.4432343457575145, "ck_system": "T4x1_TRT", "ck_used": false, "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", "formal_model": "rnn-t", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "TensorRT 7.2, CUDA 11.0 Update 1", "host_memory_capacity": "128 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 4, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) D-2123IT CPU @ 2.20GHz", "host_processors_per_node": 1, "host_storage_capacity": "2 TB", "host_storage_type": "M.2", "hw_notes": "ECC off", "informal_model": "rnnt", "input_data_types": "fp16", "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": 0.7, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Lenovo/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Lenovo/results/T4x1_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.4", "other_software_stack": "TensorRT 7.2, CUDA 11.0 Update 1, cuDNN 8.0.2, DALI 0.25.0", "performance_issue_same": true, "performance_issue_same_index": 0, "performance_issue_unique": true, "performance_sample_count": 2513, "print_timestamps": true, "problem": false, "qsl_rng_seed": 12786827339337101903, "retraining": "N", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 92400, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "DistributedDataParallel_1576581068.9962234-epoch-100.pt", "status": "available", "submitter": "Lenovo", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Lenovo", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/T4x1_TRT", "system_name": "LenovoSE350 (1x T4)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1400, "task": "speech recognition", "task2": "speech recognition", "total_cores": 4, "uid": "ea6d96db7fc7f271", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" } ]