[ { "50.00 percentile latency (ns)": 1476220, "90.00 percentile latency (ns)": 1996424, "90th percentile latency (ns)": 1996424, "95.00 percentile latency (ns)": 2660948, "97.00 percentile latency (ns)": 3279232, "99.00 percentile latency (ns)": 4735099, "99.90 percentile latency (ns)": 20120640, "Max latency (ns)": 58347527, "Mean latency (ns)": 1683990, "Min duration satisfied": "Yes", "Min latency (ns)": 1338328, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 591.95, "QPS w/o loadgen overhead": 593.83, "Result is": "VALID", "SUT name": "TFLite_SUT", "Scenario": "singlestream", "accelerator_frequency": "-", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "-", "accelerator_memory_configuration": "-", "accelerator_model_name": "-", "accelerator_on-chip_memories": "-", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 1.996424, "characteristics.90th_percentile_latency_ns": 1996424.0, "characteristics.90th_percentile_latency_s": 0.001996424, "characteristics.90th_percentile_latency_us": 1996.424, "characteristics.accuracy": 44.116, "characteristics.good": 22058, "characteristics.total": 50000, "ck_system": "xavier-tflite-v2.2.0-ruy", "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.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "mobilenet-v1-0.25-128-non-quantized", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TFLite v2.2.0 (ruy)", "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": "", "informal_model": "mobilenet-v1-0.25-128-non-quantized", "input_data_types": "fp32", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 120, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/results/xavier-tflite-v2.2.0-ruy", "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://zenodo.org/record/2269307/files/mobilenet_v1_0.25_128.tgz", "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-tflite-v2.2.0-ruy", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 333.333, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "b58c0a0a971457eb", "use_accelerator": false, "weight_data_types": "fp32", "weight_transformations": "TFLite" }, { "50.00 percentile latency (ns)": 2251810, "90.00 percentile latency (ns)": 2290468, "90th percentile latency (ns)": 2290468, "95.00 percentile latency (ns)": 2306533, "97.00 percentile latency (ns)": 2320229, "99.00 percentile latency (ns)": 2369575, "99.90 percentile latency (ns)": 2512845, "Max latency (ns)": 3917323, "Mean latency (ns)": 2259561, "Min duration satisfied": "Yes", "Min latency (ns)": 2198848, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 439.93, "QPS w/o loadgen overhead": 442.56, "Result is": "VALID", "SUT name": "PySUT", "Scenario": "singlestream", "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.90th_percentile_latency_ms": 2.290468, "characteristics.90th_percentile_latency_ns": 2290468.0, "characteristics.90th_percentile_latency_s": 0.002290468, "characteristics.90th_percentile_latency_us": 2290.468, "characteristics.accuracy": 76.258, "characteristics.good": 38129, "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.90th_percentile_latency_ms", "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": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://zenodo.org/record/3716059/files/resnet-SingleStream-gpu-b1-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)": 0, "target_qps": 500, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "c5e9f9b6d4697855", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "ONNX -> TensorRT" }, { "50.00 percentile latency (ns)": 15880113, "90.00 percentile latency (ns)": 23006950, "90th percentile latency (ns)": 23006950, "95.00 percentile latency (ns)": 25921285, "97.00 percentile latency (ns)": 27670929, "99.00 percentile latency (ns)": 34621696, "99.90 percentile latency (ns)": 44322372, "Max latency (ns)": 79610432, "Mean latency (ns)": 17879075, "Min duration satisfied": "Yes", "Min latency (ns)": 14882406, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 55.86, "QPS w/o loadgen overhead": 55.93, "Result is": "VALID", "SUT name": "TFLite_SUT", "Scenario": "singlestream", "accelerator_frequency": "-", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "-", "accelerator_memory_configuration": "-", "accelerator_model_name": "-", "accelerator_on-chip_memories": "-", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 23.00695, "characteristics.90th_percentile_latency_ns": 23006950.0, "characteristics.90th_percentile_latency_s": 0.02300695, "characteristics.90th_percentile_latency_us": 23006.95, "characteristics.accuracy": 76.642, "characteristics.good": 38321, "characteristics.total": 50000, "ck_system": "xavier-tflite-v2.2.0-ruy", "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.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "efficientnet-lite1-non-quantized", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TFLite v2.2.0 (ruy)", "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": "", "informal_model": "efficientnet-lite1-non-quantized", "input_data_types": "fp32", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 120, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/results/xavier-tflite-v2.2.0-ruy", "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://storage.googleapis.com/cloud-tpu-checkpoints/efficientnet/lite/efficientnet-lite1.tar.gz", "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-tflite-v2.2.0-ruy", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 50, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "a59fd5724f489d36", "use_accelerator": false, "weight_data_types": "fp32", "weight_transformations": "TFLite" }, { "50.00 percentile latency (ns)": 21205849, "90.00 percentile latency (ns)": 28491733, "90th percentile latency (ns)": 28491733, "95.00 percentile latency (ns)": 31140808, "97.00 percentile latency (ns)": 33566705, "99.00 percentile latency (ns)": 40379512, "99.90 percentile latency (ns)": 55318657, "Max latency (ns)": 60898867, "Mean latency (ns)": 23053920, "Min duration satisfied": "Yes", "Min latency (ns)": 19848990, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 43.33, "QPS w/o loadgen overhead": 43.38, "Result is": "VALID", "SUT name": "TFLite_SUT", "Scenario": "singlestream", "accelerator_frequency": "-", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "-", "accelerator_memory_configuration": "-", "accelerator_model_name": "-", "accelerator_on-chip_memories": "-", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 28.491733, "characteristics.90th_percentile_latency_ns": 28491733.0, "characteristics.90th_percentile_latency_s": 0.028491733, "characteristics.90th_percentile_latency_us": 28491.733, "characteristics.accuracy": 77.542, "characteristics.good": 38771, "characteristics.total": 50000, "ck_system": "xavier-tflite-v2.2.0-ruy", "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.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "efficientnet-lite2-non-quantized", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TFLite v2.2.0 (ruy)", "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": "", "informal_model": "efficientnet-lite2-non-quantized", "input_data_types": "fp32", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 120, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/results/xavier-tflite-v2.2.0-ruy", "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://storage.googleapis.com/cloud-tpu-checkpoints/efficientnet/lite/efficientnet-lite2.tar.gz", "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-tflite-v2.2.0-ruy", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 50, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "5a91a1041528b36b", "use_accelerator": false, "weight_data_types": "fp32", "weight_transformations": "TFLite" }, { "50.00 percentile latency (ns)": 31983467, "90.00 percentile latency (ns)": 41380131, "90th percentile latency (ns)": 41380131, "95.00 percentile latency (ns)": 45465492, "97.00 percentile latency (ns)": 48418452, "99.00 percentile latency (ns)": 55658382, "99.90 percentile latency (ns)": 79705121, "Max latency (ns)": 119179089, "Mean latency (ns)": 33971103, "Min duration satisfied": "Yes", "Min latency (ns)": 29749674, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 29.42, "QPS w/o loadgen overhead": 29.44, "Result is": "VALID", "SUT name": "TFLite_SUT", "Scenario": "singlestream", "accelerator_frequency": "-", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "-", "accelerator_memory_configuration": "-", "accelerator_model_name": "-", "accelerator_on-chip_memories": "-", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 41.380131, "characteristics.90th_percentile_latency_ns": 41380131.0, "characteristics.90th_percentile_latency_s": 0.041380131, "characteristics.90th_percentile_latency_us": 41380.131, "characteristics.accuracy": 79.268, "characteristics.good": 39634, "characteristics.total": 50000, "ck_system": "xavier-tflite-v2.2.0-ruy", "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.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "efficientnet-lite3-non-quantized", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TFLite v2.2.0 (ruy)", "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": "", "informal_model": "efficientnet-lite3-non-quantized", "input_data_types": "fp32", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 120, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/results/xavier-tflite-v2.2.0-ruy", "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://storage.googleapis.com/cloud-tpu-checkpoints/efficientnet/lite/efficientnet-lite3.tar.gz", "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-tflite-v2.2.0-ruy", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 50, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "f84186b64284a32e", "use_accelerator": false, "weight_data_types": "fp32", "weight_transformations": "TFLite" }, { "50.00 percentile latency (ns)": 51581469, "90.00 percentile latency (ns)": 59164897, "90th percentile latency (ns)": 59164897, "95.00 percentile latency (ns)": 61717850, "97.00 percentile latency (ns)": 63778906, "99.00 percentile latency (ns)": 71195800, "99.90 percentile latency (ns)": 79590307, "Max latency (ns)": 80749338, "Mean latency (ns)": 46844037, "Min duration satisfied": "Yes", "Min latency (ns)": 32054981, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 21.34, "QPS w/o loadgen overhead": 21.35, "Result is": "VALID", "SUT name": "ArmNN_TFLite_SUT", "Scenario": "singlestream", "accelerator_frequency": "-", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "-", "accelerator_memory_configuration": "-", "accelerator_model_name": "-", "accelerator_on-chip_memories": "-", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 59.164897, "characteristics.90th_percentile_latency_ns": 59164897.0, "characteristics.90th_percentile_latency_s": 0.059164897, "characteristics.90th_percentile_latency_us": 59164.897, "characteristics.accuracy": 79.268, "characteristics.good": 39634, "characteristics.total": 50000, "ck_system": "xavier-armnn-v20.08-neon", "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.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "efficientnet-lite3-non-quantized", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "ArmNN v20.08 (Neon)", "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": "", "informal_model": "efficientnet-lite3-non-quantized", "input_data_types": "fp32", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 120, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/results/xavier-armnn-v20.08-neon", "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://storage.googleapis.com/cloud-tpu-checkpoints/efficientnet/lite/efficientnet-lite3.tar.gz", "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-armnn-v20.08-neon", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 20, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "c2811a4c564c407d", "use_accelerator": false, "weight_data_types": "fp32", "weight_transformations": "TFLite -> ArmNN" }, { "50.00 percentile latency (ns)": 51659808, "90.00 percentile latency (ns)": 66791119, "90th percentile latency (ns)": 66791119, "95.00 percentile latency (ns)": 69789362, "97.00 percentile latency (ns)": 73545652, "99.00 percentile latency (ns)": 83659562, "99.90 percentile latency (ns)": 95937694, "Max latency (ns)": 96031939, "Mean latency (ns)": 55595999, "Min duration satisfied": "Yes", "Min latency (ns)": 49652713, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 17.98, "QPS w/o loadgen overhead": 17.99, "Result is": "VALID", "SUT name": "TFLite_SUT", "Scenario": "singlestream", "accelerator_frequency": "-", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "-", "accelerator_memory_configuration": "-", "accelerator_model_name": "-", "accelerator_on-chip_memories": "-", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 66.791119, "characteristics.90th_percentile_latency_ns": 66791119.0, "characteristics.90th_percentile_latency_s": 0.066791119, "characteristics.90th_percentile_latency_us": 66791.119, "characteristics.accuracy": 80.676, "characteristics.good": 40338, "characteristics.total": 50000, "ck_system": "xavier-tflite-v2.2.0-ruy", "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.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "efficientnet-lite4-non-quantized", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TFLite v2.2.0 (ruy)", "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": "", "informal_model": "efficientnet-lite4-non-quantized", "input_data_types": "fp32", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 120, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/results/xavier-tflite-v2.2.0-ruy", "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://storage.googleapis.com/cloud-tpu-checkpoints/efficientnet/lite/efficientnet-lite4.tar.gz", "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-tflite-v2.2.0-ruy", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 50, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "f4ce8842d54343e6", "use_accelerator": false, "weight_data_types": "fp32", "weight_transformations": "TFLite" }, { "50.00 percentile latency (ns)": 52284156, "90.00 percentile latency (ns)": 88623563, "90th percentile latency (ns)": 88623563, "95.00 percentile latency (ns)": 93061148, "97.00 percentile latency (ns)": 95222626, "99.00 percentile latency (ns)": 108417743, "99.90 percentile latency (ns)": 113785131, "Max latency (ns)": 119064036, "Mean latency (ns)": 61307335, "Min duration satisfied": "Yes", "Min latency (ns)": 50270589, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 16.3, "QPS w/o loadgen overhead": 16.31, "Result is": "VALID", "SUT name": "ArmNN_TFLite_SUT", "Scenario": "singlestream", "accelerator_frequency": "-", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "-", "accelerator_memory_configuration": "-", "accelerator_model_name": "-", "accelerator_on-chip_memories": "-", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 88.623563, "characteristics.90th_percentile_latency_ns": 88623563.0, "characteristics.90th_percentile_latency_s": 0.088623563, "characteristics.90th_percentile_latency_us": 88623.563, "characteristics.accuracy": 80.676, "characteristics.good": 40338, "characteristics.total": 50000, "ck_system": "xavier-armnn-v20.08-neon", "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.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "efficientnet-lite4-non-quantized", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "ArmNN v20.08 (Neon)", "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": "", "informal_model": "efficientnet-lite4-non-quantized", "input_data_types": "fp32", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 120, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/results/xavier-armnn-v20.08-neon", "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://storage.googleapis.com/cloud-tpu-checkpoints/efficientnet/lite/efficientnet-lite4.tar.gz", "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-armnn-v20.08-neon", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 20, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "0ef1e7d5646438f9", "use_accelerator": false, "weight_data_types": "fp32", "weight_transformations": "TFLite -> ArmNN" }, { "50.00 percentile latency (ns)": 97506718, "90.00 percentile latency (ns)": 113259554, "90th percentile latency (ns)": 113259554, "95.00 percentile latency (ns)": 122721896, "97.00 percentile latency (ns)": 127113205, "99.00 percentile latency (ns)": 141496133, "99.90 percentile latency (ns)": 180276888, "Max latency (ns)": 255672625, "Mean latency (ns)": 102362712, "Min duration satisfied": "Yes", "Min latency (ns)": 94462620, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 9.77, "QPS w/o loadgen overhead": 9.77, "Result is": "VALID", "SUT name": "TFLite_SUT", "Scenario": "singlestream", "accelerator_frequency": "-", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "-", "accelerator_memory_configuration": "-", "accelerator_model_name": "-", "accelerator_on-chip_memories": "-", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 113.259554, "characteristics.90th_percentile_latency_ns": 113259554.0, "characteristics.90th_percentile_latency_s": 0.113259554, "characteristics.90th_percentile_latency_us": 113259.554, "characteristics.accuracy": 80.676, "characteristics.good": 40338, "characteristics.total": 50000, "ck_system": "xavier-tflite-v2.3.0-ruy", "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.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "efficientnet-lite4-non-quantized", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TFLite v2.3.0 (ruy)", "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": "", "informal_model": "efficientnet-lite4-non-quantized", "input_data_types": "fp32", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 120, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/results/xavier-tflite-v2.3.0-ruy", "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://storage.googleapis.com/cloud-tpu-checkpoints/efficientnet/lite/efficientnet-lite4.tar.gz", "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-tflite-v2.3.0-ruy", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 50, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "5ad98d0b43a1aef9", "use_accelerator": false, "weight_data_types": "fp32", "weight_transformations": "TFLite" }, { "50.00 percentile latency (ns)": 367999010, "90.00 percentile latency (ns)": 369180242, "90th percentile latency (ns)": 369180242, "95.00 percentile latency (ns)": 369547457, "97.00 percentile latency (ns)": 369983217, "99.00 percentile latency (ns)": 372941026, "99.90 percentile latency (ns)": 390337106, "Max latency (ns)": 433434032, "Mean latency (ns)": 368240673, "Min duration satisfied": "Yes", "Min latency (ns)": 366244086, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 2.72, "QPS w/o loadgen overhead": 2.72, "Result is": "VALID", "SUT name": "ArmNN_TFLite_SUT", "Scenario": "singlestream", "accelerator_frequency": "-", "accelerator_host_interconnect": "-", "accelerator_interconnect": "-", "accelerator_interconnect_topology": "-", "accelerator_memory_capacity": "-", "accelerator_memory_configuration": "-", "accelerator_model_name": "-", "accelerator_on-chip_memories": "-", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 369.180242, "characteristics.90th_percentile_latency_ns": 369180242.0, "characteristics.90th_percentile_latency_s": 0.369180242, "characteristics.90th_percentile_latency_us": 369180.242, "characteristics.accuracy": 80.676, "characteristics.good": 40338, "characteristics.total": 50000, "ck_system": "firefly-armnn-v20.08-neon", "ck_used": true, "cooling": "on-board fan", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "efficientnet-lite4-non-quantized", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "ArmNN v20.08 (Neon)", "host_memory_capacity": "4 GiB", "host_memory_configuration": "-", "host_networking": "-", "host_networking_topology": "-", "host_processor_caches": "L1I$ 48 KiB, L1D$ 32 KiB, L2$ 1 MiB (big); L1I$ 32 KiB, L1D$ 32 KiB, L2$ 512 KiB (LITTLE)", "host_processor_core_count": 120, "host_processor_frequency": "1800 MHz (big), 1400 MHz (LITTLE)", "host_processor_interconnect": "-", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 1, "host_storage_capacity": "128 GiB", "host_storage_type": "SanDisk Extreme microSD", "hw_notes": "http://en.t-firefly.com/product/rk3399/; http://opensource.rock-chips.com/wiki_RK3399", "informal_model": "efficientnet-lite4-non-quantized", "input_data_types": "fp32", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 120, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/results/firefly-armnn-v20.08-neon", "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://storage.googleapis.com/cloud-tpu-checkpoints/efficientnet/lite/efficientnet-lite4.tar.gz", "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/firefly-armnn-v20.08-neon", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 2, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "ec83a9ee10024e2d", "use_accelerator": false, "weight_data_types": "fp32", "weight_transformations": "TFLite -> ArmNN" }, { "50.00 percentile latency (ns)": 431489077, "90.00 percentile latency (ns)": 433977908, "90th percentile latency (ns)": 433977908, "95.00 percentile latency (ns)": 435326403, "97.00 percentile latency (ns)": 438555092, "99.00 percentile latency (ns)": 441824494, "99.90 percentile latency (ns)": 453841955, "Max latency (ns)": 454127263, "Mean latency (ns)": 431215013, "Min duration satisfied": "Yes", "Min latency (ns)": 416268346, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 2.32, "QPS w/o loadgen overhead": 2.32, "Result is": "VALID", "SUT name": "ArmNN_TFLite_SUT", "Scenario": "singlestream", "accelerator_frequency": "800 MHz", "accelerator_host_interconnect": "-", "accelerator_interconnect": "-", "accelerator_interconnect_topology": "-", "accelerator_memory_capacity": "4 GiB (shared with host)", "accelerator_memory_configuration": "-", "accelerator_model_name": "Arm Mali-T860 MP4", "accelerator_on-chip_memories": "-", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 433.977908, "characteristics.90th_percentile_latency_ns": 433977908.0, "characteristics.90th_percentile_latency_s": 0.433977908, "characteristics.90th_percentile_latency_us": 433977.908, "characteristics.accuracy": 80.676, "characteristics.good": 40338, "characteristics.total": 50000, "ck_system": "firefly-armnn-v20.08-opencl", "ck_used": true, "cooling": "on-board fan", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "efficientnet-lite4-non-quantized", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "ArmNN v20.08 (OpenCL)", "host_memory_capacity": "4 GiB", "host_memory_configuration": "-", "host_networking": "-", "host_networking_topology": "-", "host_processor_caches": "L1I$ 48 KiB, L1D$ 32 KiB, L2$ 1 MiB (big); L1I$ 32 KiB, L1D$ 32 KiB, L2$ 512 KiB (LITTLE)", "host_processor_core_count": 120, "host_processor_frequency": "1800 MHz (big), 1400 MHz (LITTLE)", "host_processor_interconnect": "-", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 1, "host_storage_capacity": "128 GiB", "host_storage_type": "SanDisk Extreme microSD", "hw_notes": "http://en.t-firefly.com/product/rk3399/; http://opensource.rock-chips.com/wiki_RK3399", "informal_model": "efficientnet-lite4-non-quantized", "input_data_types": "fp32", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/results/firefly-armnn-v20.08-opencl", "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://storage.googleapis.com/cloud-tpu-checkpoints/efficientnet/lite/efficientnet-lite4.tar.gz", "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/firefly-armnn-v20.08-opencl", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 2, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "1c8a2b294d7e71ae", "use_accelerator": true, "weight_data_types": "fp32", "weight_transformations": "TFLite -> ArmNN" }, { "50.00 percentile latency (ns)": 493629230, "90.00 percentile latency (ns)": 542549738, "90th percentile latency (ns)": 542549738, "95.00 percentile latency (ns)": 554348211, "97.00 percentile latency (ns)": 562793187, "99.00 percentile latency (ns)": 574680043, "99.90 percentile latency (ns)": 582569642, "Max latency (ns)": 591846994, "Mean latency (ns)": 485360219, "Min duration satisfied": "Yes", "Min latency (ns)": 302836009, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 2.06, "QPS w/o loadgen overhead": 2.06, "Result is": "VALID", "SUT name": "TFLite_SUT", "Scenario": "singlestream", "accelerator_frequency": "-", "accelerator_host_interconnect": "-", "accelerator_interconnect": "-", "accelerator_interconnect_topology": "-", "accelerator_memory_capacity": "-", "accelerator_memory_configuration": "-", "accelerator_model_name": "-", "accelerator_on-chip_memories": "-", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 542.549738, "characteristics.90th_percentile_latency_ns": 542549738.0, "characteristics.90th_percentile_latency_s": 0.542549738, "characteristics.90th_percentile_latency_us": 542549.738, "characteristics.accuracy": 80.676, "characteristics.good": 40338, "characteristics.total": 50000, "ck_system": "firefly-tflite-v2.2.0-ruy", "ck_used": true, "cooling": "on-board fan", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "non-quantized-efficientnet-lite4", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TFLite v2.2.0 (ruy)", "host_memory_capacity": "4 GiB", "host_memory_configuration": "-", "host_networking": "-", "host_networking_topology": "-", "host_processor_caches": "L1I$ 48 KiB, L1D$ 32 KiB, L2$ 1 MiB (big); L1I$ 32 KiB, L1D$ 32 KiB, L2$ 512 KiB (LITTLE)", "host_processor_core_count": 120, "host_processor_frequency": "1800 MHz (big), 1400 MHz (LITTLE)", "host_processor_interconnect": "-", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 1, "host_storage_capacity": "128 GiB", "host_storage_type": "SanDisk Extreme microSD", "hw_notes": "http://en.t-firefly.com/product/rk3399/; http://opensource.rock-chips.com/wiki_RK3399", "informal_model": "non-quantized-efficientnet-lite4", "input_data_types": "fp32", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 120, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/dividiti/results/firefly-tflite-v2.2.0-ruy", "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": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "https://storage.googleapis.com/cloud-tpu-checkpoints/efficientnet/lite/efficientnet-lite4.tar.gz", "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/firefly-tflite-v2.2.0-ruy", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 25, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "09834c7dba40faee", "use_accelerator": false, "weight_data_types": "fp32", "weight_transformations": "TFLite" } ]