[ { "50.00 percentile latency (ns)": 319545718068, "90.00 percentile latency (ns)": 575526792717, "95.00 percentile latency (ns)": 607404820852, "97.00 percentile latency (ns)": 620221280900, "99.00 percentile latency (ns)": 633007959295, "99.90 percentile latency (ns)": 638666283210, "Max latency (ns)": 639431888892, "Mean latency (ns)": 319574872226, "Min duration satisfied": "Yes", "Min latency (ns)": 16613789, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "QAIC_SUT", "Samples per second": 670.908, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "1x PCIe 8 lane Gen4", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "LPDDR4x", "accelerator_model_name": "QUALCOMM Cloud AI 100 PCIe HHHL", "accelerator_on-chip_memories": "144 MB", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.samples_per_second": 670.908, "characteristics.samples_per_second.normalized_per_core": 670.908, "characteristics.samples_per_second.normalized_per_processor": 670.908, "ck_system": "r282_z93_q1-qaic-v1.5.9-aic100", "ck_used": true, "cooling": "air", "dataset": "SQuAD v1.1", "dataset_link": "", "dim_x_default": "seq_number", "dim_x_maximize": true, "dim_y_default": "characteristics.samples_per_second", "dim_y_maximize": false, "division": "closed", "formal_model": "bert", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "Qualcomm Cloud AI SDK v1.5.9", "host_memory_capacity": "8x 64 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "L1d: 32K; L1i: 32K; L2: 512K; L3: 16384K", "host_processor_core_count": 16, "host_processor_frequency": "2800 MHz", "host_processor_interconnect": "1x 16-20GT/s xGMI2", "host_processor_model_name": "AMD EPYC 7282 16-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1 TB", "host_storage_type": "1x SSD", "hw_notes": "With 75W Accelerator TDP constraints", "informal_model": "bert-99", "input_data_types": "int64", "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/Qualcomm/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Qualcomm/results/r282_z93_q1-qaic-v1.5.9-aic100", "number_of_nodes": 1, "operating_system": "CentOS 7 (Linux kernel: 5.4.1-1.el7.elrepo.x86_64 #1 SMP Fri Nov 29 10:21:13 EST 2019 x86_64 x86_64 x86_64 GNU/Linux)", "other_software_stack": "Docker 20.10.7; Collective Knowledge v1.55.5", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 10833, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "No", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 429000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "https://zenodo.org/record/3733868/files/model.ckpt-5474.data-00000-of-00001", "status": "available", "submitter": "Qualcomm", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Qualcomm", "sw_notes": "Powered by Collective Knowledge v1.55.5", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/r282_z93_q1-qaic-v1.5.9-aic100", "system_name": "Gigabyte R282-Z93 1x QAIC100", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 650, "task": "NLP", "task2": "nlp", "total_cores": 32, "uid": "2ce6fb85ece4631a", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "see calibration.md" }, { "50.00 percentile latency (ns)": 320980912354, "90.00 percentile latency (ns)": 579464681014, "95.00 percentile latency (ns)": 611775986452, "97.00 percentile latency (ns)": 624683960402, "99.00 percentile latency (ns)": 637598829685, "99.90 percentile latency (ns)": 643460346584, "Max latency (ns)": 644093581950, "Mean latency (ns)": 321370582108, "Min duration satisfied": "Yes", "Min latency (ns)": 14227931, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "QAIC_SUT", "Samples per second": 3279.03, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "5x PCIe 8 lane Gen4", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "LPDDR4x", "accelerator_model_name": "QUALCOMM Cloud AI 100 PCIe HHHL", "accelerator_on-chip_memories": "144 MB", "accelerators_per_node": 5, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.power": 552.4227950310554, "characteristics.power.normalized_per_core": 110.48455900621109, "characteristics.power.normalized_per_processor": 110.48455900621109, "characteristics.samples_per_second": 3279.03, "characteristics.samples_per_second.normalized_per_core": 655.806, "characteristics.samples_per_second.normalized_per_processor": 655.806, "ck_system": "r282_z93_q5-qaic-v1.5.9-aic100", "ck_used": true, "cooling": "air", "dataset": "SQuAD v1.1", "dataset_link": "", "dim_x_default": "seq_number", "dim_x_maximize": true, "dim_y_default": "characteristics.samples_per_second", "dim_y_maximize": false, "division": "closed", "formal_model": "bert", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "Qualcomm Cloud AI SDK v1.5.9", "host_memory_capacity": "8x 64 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "L1d: 32K; L1i: 32K; L2: 512K; L3: 16384K", "host_processor_core_count": 16, "host_processor_frequency": "2800 MHz", "host_processor_interconnect": "1x 16-20GT/s xGMI2", "host_processor_model_name": "AMD EPYC 7282 16-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1 TB", "host_storage_type": "1x SSD", "hw_notes": "With 75W Accelerator TDP constraints", "informal_model": "bert-99", "input_data_types": "int64", "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": 5, "normalize_processors": 5, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Qualcomm/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Qualcomm/results/r282_z93_q5-qaic-v1.5.9-aic100", "number_of_nodes": 1, "operating_system": "CentOS 7 (Linux kernel: 5.4.1-1.el7.elrepo.x86_64 #1 SMP Fri Nov 29 10:21:13 EST 2019 x86_64 x86_64 x86_64 GNU/Linux)", "other_software_stack": "Docker 20.10.7; Collective Knowledge v1.55.5", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 10833, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "No", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 2112000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "https://zenodo.org/record/3733868/files/model.ckpt-5474.data-00000-of-00001", "status": "available", "submitter": "Qualcomm", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Qualcomm", "sw_notes": "Powered by Collective Knowledge v1.55.5", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/r282_z93_q5-qaic-v1.5.9-aic100", "system_name": "Gigabyte R282-Z93 5x QAIC100", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 3200, "task": "NLP", "task2": "nlp", "total_cores": 32, "uid": "92324eb8714a2f9d", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "see calibration.md" } ]