[ { "50.00 percentile latency (ns)": 939449, "90.00 percentile latency (ns)": 1026322, "90th percentile latency (ns)": 1026322, "95.00 percentile latency (ns)": 1087465, "97.00 percentile latency (ns)": 1201368, "99.00 percentile latency (ns)": 1662187, "99.90 percentile latency (ns)": 12860892, "Max latency (ns)": 39793212, "Mean latency (ns)": 998943, "Min duration satisfied": "Yes", "Min latency (ns)": 769298, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 932.63, "QPS w/o loadgen overhead": 1001.06, "Result is": "VALID", "SUT name": "QAIC_SUT", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "PCIe 4 lane Gen4", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "8 GB", "accelerator_memory_configuration": "LPDDR4x", "accelerator_model_name": "QUALCOMM Cloud AI 100 DM.2e", "accelerator_on-chip_memories": "72 MB", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 1.026322, "characteristics.90th_percentile_latency_ns": 1026322.0, "characteristics.90th_percentile_latency_s": 0.001026322, "characteristics.90th_percentile_latency_us": 1026.322, "characteristics.accuracy": 75.976, "characteristics.good": 37988, "characteristics.power": 0.019102949696573952, "characteristics.power.normalized_per_core": 0.019102949696573952, "characteristics.power.normalized_per_processor": 0.019102949696573952, "characteristics.total": 50000, "ck_system": "aedk_15w-qaic-v1.4.66-aic100", "ck_used": true, "cooling": "passive", "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": "Qualcomm Cloud AI SDK v1.4.66", "host_memory_capacity": "12 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "1x (L1d: 64K; L1i: 64K; L2: 512K); 3x (L1d: 64K; L1i: 64K; L2: 256K); 4x (L1d: 64K; L1i: 64K; L2: 128K); L3: 4M", "host_processor_core_count": 8, "host_processor_frequency": "1x (Cortex-A77 @ 2840 MHz); 3x (Cortex-A77 @ 2420 MHz); 4x (Cortex-A55 @ 1800 MHz)", "host_processor_interconnect": "", "host_processor_model_name": "Qualcomm Snapdragon 865", "host_processors_per_node": 1, "host_storage_capacity": "80 GB", "host_storage_type": "Internal flash", "hw_notes": "With 15W Accelerator TDP constraints", "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)": 600000, "min_query_count": 1024, "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/aedk_15w-qaic-v1.4.66-aic100", "number_of_nodes": 1, "operating_system": "CentOS Linux release 8.0.1905 (kernel: 4.19.81 #1 SMP PREEMPT Tue May 11 08:23:20 UTC 2021 aarch64 aarch64 aarch64 GNU/Linux)", "other_software_stack": "Collective Knowledge v1.55.5", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "No", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "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/aedk_15w-qaic-v1.4.66-aic100", "system_name": "AI Development Kit", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1000, "task": "image classification", "task2": "image classification", "total_cores": 8, "uid": "9e9c48429be0fd84", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "see calibration.md" }, { "50.00 percentile latency (ns)": 1567500, "90.00 percentile latency (ns)": 1618090, "90th percentile latency (ns)": 1618090, "95.00 percentile latency (ns)": 1632921, "97.00 percentile latency (ns)": 1642411, "99.00 percentile latency (ns)": 1660200, "99.90 percentile latency (ns)": 1689691, "Max latency (ns)": 10166187, "Mean latency (ns)": 1569070, "Min duration satisfied": "Yes", "Min latency (ns)": 1443169, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 632.17, "QPS w/o loadgen overhead": 637.32, "Result is": "VALID", "SUT name": "QAIC_SUT", "Scenario": "singlestream", "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.90th_percentile_latency_ms": 1.61809, "characteristics.90th_percentile_latency_ns": 1618090.0, "characteristics.90th_percentile_latency_s": 0.00161809, "characteristics.90th_percentile_latency_us": 1618.09, "characteristics.accuracy": 75.99, "characteristics.good": 37995, "characteristics.total": 50000, "ck_system": "r282_z93_q1-qaic-v1.5.6-aic100", "ck_used": true, "cooling": "air", "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": "Qualcomm Cloud AI SDK v1.5.6", "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": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "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-v1.5.6-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": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "No", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "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-v1.5.6-aic100", "system_name": "Gigabyte R282-Z93 1x QAIC100", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1000, "task": "image classification", "task2": "image classification", "total_cores": 32, "uid": "cf2ef0ca0744e3ae", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "see calibration.md" }, { "50.00 percentile latency (ns)": 1572017, "90.00 percentile latency (ns)": 1620436, "90th percentile latency (ns)": 1620436, "95.00 percentile latency (ns)": 1633396, "97.00 percentile latency (ns)": 1641557, "99.00 percentile latency (ns)": 1656583, "99.90 percentile latency (ns)": 1681376, "Max latency (ns)": 3545564, "Mean latency (ns)": 1572650, "Min duration satisfied": "Yes", "Min latency (ns)": 1450035, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 630.02, "QPS w/o loadgen overhead": 635.87, "Result is": "VALID", "SUT name": "QAIC_SUT", "Scenario": "singlestream", "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.90th_percentile_latency_ms": 1.620436, "characteristics.90th_percentile_latency_ns": 1620436.0, "characteristics.90th_percentile_latency_s": 0.001620436, "characteristics.90th_percentile_latency_us": 1620.436, "characteristics.accuracy": 75.99, "characteristics.good": 37995, "characteristics.power": 0.35295990603472915, "characteristics.power.normalized_per_core": 0.07059198120694583, "characteristics.power.normalized_per_processor": 0.07059198120694583, "characteristics.total": 50000, "ck_system": "r282_z93_q5-qaic-v1.5.6-aic100", "ck_used": true, "cooling": "air", "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": "Qualcomm Cloud AI SDK v1.5.6", "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": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "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-v1.5.6-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": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "No", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "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-v1.5.6-aic100", "system_name": "Gigabyte R282-Z93 5x QAIC100", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1000, "task": "image classification", "task2": "image classification", "total_cores": 32, "uid": "23a6389ef6e24018", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "see calibration.md" }, { "50.00 percentile latency (ns)": 940495, "90.00 percentile latency (ns)": 1024711, "90th percentile latency (ns)": 1024711, "95.00 percentile latency (ns)": 1081584, "97.00 percentile latency (ns)": 1165019, "99.00 percentile latency (ns)": 1386212, "99.90 percentile latency (ns)": 3246785, "Max latency (ns)": 30456970, "Mean latency (ns)": 961303, "Min duration satisfied": "Yes", "Min latency (ns)": 697168, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 971.74, "QPS w/o loadgen overhead": 1040.25, "Result is": "VALID", "SUT name": "QAIC_SUT", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "PCIe 4 lane Gen4", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "LPDDR4x", "accelerator_model_name": "QUALCOMM Cloud AI 100 DM.2", "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.90th_percentile_latency_ms": 1.024711, "characteristics.90th_percentile_latency_ns": 1024711.0, "characteristics.90th_percentile_latency_s": 0.001024711, "characteristics.90th_percentile_latency_us": 1024.711, "characteristics.accuracy": 75.976, "characteristics.good": 37988, "characteristics.power": 0.018047647371784506, "characteristics.power.normalized_per_core": 0.018047647371784506, "characteristics.power.normalized_per_processor": 0.018047647371784506, "characteristics.total": 50000, "ck_system": "aedk_20w-qaic-v1.4.66-aic100", "ck_used": true, "cooling": "passive", "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": "Qualcomm Cloud AI SDK v1.4.66", "host_memory_capacity": "12 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "1x (L1d: 64K; L1i: 64K; L2: 512K); 3x (L1d: 64K; L1i: 64K; L2: 256K); 4x (L1d: 64K; L1i: 64K; L2: 128K); L3: 4M", "host_processor_core_count": 8, "host_processor_frequency": "1x (Cortex-A77 @ 2840 MHz); 3x (Cortex-A77 @ 2420 MHz); 4x (Cortex-A55 @ 1800 MHz)", "host_processor_interconnect": "", "host_processor_model_name": "Qualcomm Snapdragon 865", "host_processors_per_node": 1, "host_storage_capacity": "80 GB", "host_storage_type": "Internal flash", "hw_notes": "With 20W Accelerator TDP constraints", "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)": 600000, "min_query_count": 1024, "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/aedk_20w-qaic-v1.4.66-aic100", "number_of_nodes": 1, "operating_system": "CentOS Linux release 8.0.1905 (kernel: 4.19.81 #1 SMP PREEMPT Tue May 11 08:23:20 UTC 2021 aarch64 aarch64 aarch64 GNU/Linux)", "other_software_stack": "Collective Knowledge v1.55.5", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "No", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "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/aedk_20w-qaic-v1.4.66-aic100", "system_name": "AI Development Kit", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1000, "task": "image classification", "task2": "image classification", "total_cores": 8, "uid": "4a5db002965430e1", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "see calibration.md" }, { "50.00 percentile latency (ns)": 835914, "90.00 percentile latency (ns)": 888152, "90th percentile latency (ns)": 888152, "95.00 percentile latency (ns)": 925755, "97.00 percentile latency (ns)": 975858, "99.00 percentile latency (ns)": 1169394, "99.90 percentile latency (ns)": 2765182, "Max latency (ns)": 24349517, "Mean latency (ns)": 847288, "Min duration satisfied": "Yes", "Min latency (ns)": 713417, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 1103.56, "QPS w/o loadgen overhead": 1180.24, "Result is": "VALID", "SUT name": "QAIC_SUT", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "PCIe 4 lane Gen4", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "LPDDR4x", "accelerator_model_name": "QUALCOMM Cloud AI 100 DM.2", "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.90th_percentile_latency_ms": 0.888152, "characteristics.90th_percentile_latency_ns": 888152.0, "characteristics.90th_percentile_latency_s": 0.000888152, "characteristics.90th_percentile_latency_us": 888.152, "characteristics.accuracy": 75.942, "characteristics.good": 37971, "characteristics.power": 0.02049657472180942, "characteristics.power.normalized_per_core": 0.02049657472180942, "characteristics.power.normalized_per_processor": 0.02049657472180942, "characteristics.total": 50000, "ck_system": "aedk_20w-qaic-v1.3.14-aic100", "ck_used": true, "cooling": "passive", "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": "Qualcomm Cloud AI SDK v1.3.14", "host_memory_capacity": "12 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "1x (L1d: 64K; L1i: 64K; L2: 512K); 3x (L1d: 64K; L1i: 64K; L2: 256K); 4x (L1d: 64K; L1i: 64K; L2: 128K); L3: 4M", "host_processor_core_count": 120, "host_processor_frequency": "1x (Cortex-A77 @ 2840 MHz); 3x (Cortex-A77 @ 2420 MHz); 4x (Cortex-A55 @ 1800 MHz)", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 1, "host_storage_capacity": "894 GB", "host_storage_type": "NVMe SSD", "hw_notes": "With 20W Accelerator TDP constraints", "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)": 600000, "min_query_count": 1024, "mlperf_version": 1.0, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Qualcomm/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Qualcomm/results/aedk_20w-qaic-v1.3.14-aic100", "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": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "No", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 1, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "available", "submitter": "Qualcomm", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Qualcomm", "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/aedk_20w-qaic-v1.3.14-aic100", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1000, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "b4121b15247612db", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "see calibration/" }, { "50.00 percentile latency (ns)": 769350, "90.00 percentile latency (ns)": 854660, "90th percentile latency (ns)": 854660, "95.00 percentile latency (ns)": 988874, "97.00 percentile latency (ns)": 1024134, "99.00 percentile latency (ns)": 1163869, "99.90 percentile latency (ns)": 4466263, "Max latency (ns)": 26444516, "Mean latency (ns)": 800449, "Min duration satisfied": "Yes", "Min latency (ns)": 596544, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 1161.28, "QPS w/o loadgen overhead": 1249.3, "Result is": "VALID", "SUT name": "QAIC_SUT", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "PCIe 4 lane Gen4", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "8 GB", "accelerator_memory_configuration": "LPDDR4x", "accelerator_model_name": "QUALCOMM Cloud AI 100 DM.2e", "accelerator_on-chip_memories": "72 MB", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 0.85466, "characteristics.90th_percentile_latency_ns": 854660.0, "characteristics.90th_percentile_latency_s": 0.00085466, "characteristics.90th_percentile_latency_us": 854.66, "characteristics.accuracy": 75.942, "characteristics.good": 37971, "characteristics.power": 0.018511757141544953, "characteristics.power.normalized_per_core": 0.018511757141544953, "characteristics.power.normalized_per_processor": 0.018511757141544953, "characteristics.total": 50000, "ck_system": "aedk_15w-qaic-v1.3.14-aic100", "ck_used": true, "cooling": "passive", "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": "Qualcomm Cloud AI SDK v1.3.14", "host_memory_capacity": "12 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "1x (L1d: 64K; L1i: 64K; L2: 512K); 3x (L1d: 64K; L1i: 64K; L2: 256K); 4x (L1d: 64K; L1i: 64K; L2: 128K); L3: 4M", "host_processor_core_count": 120, "host_processor_frequency": "1x (Cortex-A77 @ 2840 MHz); 3x (Cortex-A77 @ 2420 MHz); 4x (Cortex-A55 @ 1800 MHz)", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 1, "host_storage_capacity": "894 GB", "host_storage_type": "NVMe SSD", "hw_notes": "With 15W Accelerator TDP constraints", "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)": 600000, "min_query_count": 1024, "mlperf_version": 1.0, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Qualcomm/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Qualcomm/results/aedk_15w-qaic-v1.3.14-aic100", "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": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "No", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 1, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "https://zenodo.org/record/2535873/files/resnet50_v1.pb", "status": "available", "submitter": "Qualcomm", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Qualcomm", "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/aedk_15w-qaic-v1.3.14-aic100", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1000, "task": "image classification", "task2": "image classification", "total_cores": 120, "uid": "2890547ac8a6ab01", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "see calibration/" } ]