Created
September 9, 2019 17:59
-
-
Save guschmue/13a7dcfea4038acf0d1c41845401eadd to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
{ | |
"cells": [ | |
{ | |
"cell_type": "markdown", | |
"metadata": {}, | |
"source": [ | |
"## End to End Mlperf Submission example\n", | |
"\n", | |
"This is following the [General MLPerf Submission Rules](https://docs.google.com/document/d/1Kt79dyaM9w1ckHJoLS6dBxfCqf-LABMrH2JRZm1p_bk/edit#heading=h.k6uj9xqxzwoe).\n" | |
] | |
}, | |
{ | |
"cell_type": "markdown", | |
"metadata": {}, | |
"source": [ | |
"### Get the mlperf source code" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": null, | |
"metadata": {}, | |
"outputs": [], | |
"source": [ | |
"!git clone https://github.com/mlperf/inference.git --depth 1" | |
] | |
}, | |
{ | |
"cell_type": "markdown", | |
"metadata": {}, | |
"source": [ | |
"### Build loadgen" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 1, | |
"metadata": {}, | |
"outputs": [ | |
{ | |
"name": "stdout", | |
"output_type": "stream", | |
"text": [ | |
"/home/gs/submission/inference\n" | |
] | |
} | |
], | |
"source": [ | |
"%cd inference" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": null, | |
"metadata": {}, | |
"outputs": [], | |
"source": [ | |
"# build loadgen\n", | |
"!pip install pybind11\n", | |
"!cd loadgen; CFLAGS=\"-std=c++14 -O3\" python setup.py develop" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": null, | |
"metadata": { | |
"scrolled": true | |
}, | |
"outputs": [], | |
"source": [ | |
"%cd v0.5/classification_and_detection\n", | |
"!python setup.py develop" | |
] | |
}, | |
{ | |
"cell_type": "markdown", | |
"metadata": {}, | |
"source": [ | |
"### Download data" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": null, | |
"metadata": {}, | |
"outputs": [], | |
"source": [ | |
"%%bash\n", | |
"\n", | |
"## download imagenet and coco.\n", | |
"## we keep a copy under /data so just symlink it.\n", | |
"mkdir data\n", | |
"ln -s /data/imagenet2012 data/\n", | |
"ln -s /data/coco data/" | |
] | |
}, | |
{ | |
"cell_type": "markdown", | |
"metadata": {}, | |
"source": [ | |
"### Download models" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": null, | |
"metadata": {}, | |
"outputs": [], | |
"source": [ | |
"%%bash\n", | |
"\n", | |
"mkdir models\n", | |
"\n", | |
"# resnet50\n", | |
"wget -q https://zenodo.org/record/2535873/files/resnet50_v1.pb -O models/resnet50_v1.pb \n", | |
"wget -q https://zenodo.org/record/2592612/files/resnet50_v1.onnx -O models/resnet50_v1.onnx\n", | |
"\n", | |
"# mobilenet\n", | |
"wget -q https://zenodo.org/record/2269307/files/mobilenet_v1_1.0_224.tgz -O models/mobilenet_v1_1.0_224.tgz\n", | |
"cd models; tar zxvf mobilenet_v1_1.0_224.tgz\n", | |
"wget -q https://zenodo.org/record/3157894/files/mobilenet_v1_1.0_224.onnx -O models/mobilenet_v1_1.0_224.onnx\n", | |
"\n", | |
"# ssd-mobilenet\n", | |
"wget -q http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_coco_2018_01_28.tar.gz -O models/ssd_mobilenet_v1_coco_2018_01_28.tar.gz\n", | |
"cd models; tar zxvf ssd_mobilenet_v1_coco_2018_01_28.tar.gz; mv ssd_mobilenet_v1_coco_2018_01_28/frozen_inference_graph.pb ssd_mobilenet_v1_coco_2018_01_28.pb\n", | |
"wget -q https://zenodo.org/record/3163026/files/ssd_mobilenet_v1_coco_2018_01_28.onnx -O models/ssd_mobilenet_v1_coco_2018_01_28.onnx \n", | |
"\n", | |
"# ssd-resnet34\n", | |
"wget -q https://zenodo.org/record/3345892/files/tf_ssd_resnet34_22.1.zip -O models/tf_ssd_resnet34_22.1.zip\n", | |
"cd models; unzip tf_ssd_resnet34_22.1.zip; mv tf_ssd_resnet34_22.1/resnet34_tf.22.1.pb .\n", | |
"wget -q https://zenodo.org/record/3228411/files/resnet34-ssd1200.onnx -O models/resnet34-ssd1200.onnx" | |
] | |
}, | |
{ | |
"cell_type": "markdown", | |
"metadata": {}, | |
"source": [ | |
"### Run benchmarks using the reference implementation\n", | |
"\n", | |
"Lets pretent we have 2 platforms (tf-gpu and onnxruntime-cpu) and want to submit the following:\n", | |
"```\n", | |
"onnxruntime-cpu: mobilenet\n", | |
"tf-gpu: resnet50\n", | |
"tf gpu: ssd-mobilenet\n", | |
"tf gpu: ssd-resnet34\n", | |
"```\n", | |
"\n", | |
"The following script will run those combinations.\n", | |
"To make this a little faster we do accuracy runs only with 500 images (which would not be a valid submission) and since I run this on a small server some of the latencies don't meet the mlperf rules (which would not be a valid submission)." | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 45, | |
"metadata": {}, | |
"outputs": [], | |
"source": [ | |
"import logging\n", | |
"import os\n", | |
"logger = logging.getLogger()\n", | |
"logger.setLevel(logging.CRITICAL)\n", | |
"\n", | |
"os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' \n", | |
"os.environ['CUDA_VISIBLE_DEVICES'] = \"0\"\n", | |
"\n", | |
"# final results go here\n", | |
"SUBMISSION_DIR = \"/tmp/Mlperf\"\n", | |
"os.environ['SUBMISSION_DIR'] = SUBMISSION_DIR" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 31, | |
"metadata": { | |
"scrolled": false | |
}, | |
"outputs": [ | |
{ | |
"name": "stdout", | |
"output_type": "stream", | |
"text": [ | |
"====== mobilenet/SingleStream =====\n", | |
"TestScenario.SingleStream qps=21.98, mean=0.0452, time=22.744, acc=72.600%, queries=500, tiles=50.0:0.0463,80.0:0.0465,90.0:0.0466,95.0:0.0468,99.0:0.0578,99.9:0.1085\n", | |
"accuracy=72.600%, good=363, total=500\n", | |
"TestScenario.SingleStream qps=22.83, mean=0.0437, time=21.905, queries=500, tiles=50.0:0.0462,80.0:0.0464,90.0:0.0465,95.0:0.0468,99.0:0.0497,99.9:0.0805\n", | |
"====== mobilenet/MultiStream =====\n", | |
"TestScenario.MultiStream qps=1.00, mean=0.0920, time=250.125, acc=72.600%, queries=250, tiles=50.0:0.0924,80.0:0.0985,90.0:0.1052,95.0:0.1103,99.0:0.1146,99.9:0.1260\n", | |
"accuracy=72.600%, good=363, total=500\n", | |
"TestScenario.MultiStream qps=2.00, mean=0.0848, time=500.131, queries=1000, tiles=50.0:0.0858,80.0:0.0955,90.0:0.0996,95.0:0.1049,99.0:0.1140,99.9:0.1512\n", | |
"====== mobilenet/Server =====\n", | |
"TestScenario.Server qps=0.99, mean=0.0417, time=504.482, acc=72.600%, queries=500, tiles=50.0:0.0417,80.0:0.0499,90.0:0.0564,95.0:0.0667,99.0:0.0847,99.9:0.1275\n", | |
"accuracy=72.600%, good=363, total=500\n", | |
"TestScenario.Server qps=21.01, mean=0.0237, time=0.048, queries=1, tiles=50.0:0.0237,80.0:0.0237,90.0:0.0237,95.0:0.0237,99.0:0.0237,99.9:0.0237\n", | |
"====== mobilenet/Offline =====\n", | |
"TestScenario.Offline qps=2.28, mean=3.3837, time=7.024, acc=72.600%, queries=16, tiles=50.0:3.7338,80.0:4.5426,90.0:5.1020,95.0:5.1272,99.0:5.1410,99.9:5.1441\n", | |
"accuracy=72.600%, good=363, total=500\n", | |
"TestScenario.Offline qps=72.16, mean=3.8105, time=6.930, queries=500, tiles=50.0:3.6139,80.0:5.9200,90.0:6.8850,95.0:6.9063,99.0:6.9063,99.9:6.9063\n", | |
"====== resnet50/SingleStream =====\n", | |
"TestScenario.SingleStream qps=101.70, mean=0.0095, time=4.916, acc=76.000%, queries=500, tiles=50.0:0.0096,80.0:0.0103,90.0:0.0104,95.0:0.0105,99.0:0.0106,99.9:0.0115\n", | |
"accuracy=76.000%, good=380, total=500\n", | |
"TestScenario.SingleStream qps=105.73, mean=0.0093, time=4.729, queries=500, tiles=50.0:0.0089,80.0:0.0103,90.0:0.0107,95.0:0.0109,99.0:0.0110,99.9:0.0114\n", | |
"====== resnet50/MultiStream =====\n", | |
"TestScenario.MultiStream qps=1.00, mean=0.0135, time=250.043, acc=76.000%, queries=250, tiles=50.0:0.0128,80.0:0.0129,90.0:0.0130,95.0:0.0130,99.0:0.0132,99.9:0.1519\n", | |
"accuracy=76.000%, good=380, total=500\n", | |
"TestScenario.MultiStream qps=2.00, mean=0.0138, time=500.037, queries=1000, tiles=50.0:0.0134,80.0:0.0135,90.0:0.0135,95.0:0.0136,99.0:0.0142,99.9:0.2063\n", | |
"====== resnet50/Server =====\n", | |
"TestScenario.Server qps=0.99, mean=0.0108, time=504.436, acc=76.000%, queries=500, tiles=50.0:0.0109,80.0:0.0110,90.0:0.0111,95.0:0.0114,99.0:0.0140,99.9:0.0193\n", | |
"accuracy=76.000%, good=380, total=500\n", | |
"TestScenario.Server qps=24.50, mean=0.0095, time=0.041, queries=1, tiles=50.0:0.0095,80.0:0.0095,90.0:0.0095,95.0:0.0095,99.0:0.0095,99.9:0.0095\n", | |
"====== resnet50/Offline =====\n", | |
"TestScenario.Offline qps=7.66, mean=0.9524, time=2.088, acc=76.000%, queries=16, tiles=50.0:0.9529,80.0:1.1776,90.0:1.2437,95.0:1.3075,99.0:1.3305,99.9:1.3357\n", | |
"accuracy=76.000%, good=380, total=500\n", | |
"TestScenario.Offline qps=240.24, mean=1.1732, time=2.081, queries=500, tiles=50.0:1.0933,80.0:1.6170,90.0:1.8518,95.0:1.8518,99.0:2.0509,99.9:2.0509\n", | |
"====== ssd-mobilenet/SingleStream =====\n", | |
"loading annotations into memory...\n", | |
"Done (t=0.46s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Loading and preparing results...\n", | |
"Converting ndarray to lists...\n", | |
"(2106, 7)\n", | |
"0/2106\n", | |
"DONE (t=0.01s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Running per image evaluation...\n", | |
"Evaluate annotation type *bbox*\n", | |
"DONE (t=1.02s).\n", | |
"Accumulating evaluation results...\n", | |
"DONE (t=0.61s).\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.256\n", | |
" Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.380\n", | |
" Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.280\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.019\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.180\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.557\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.231\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.283\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.284\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.022\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.189\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.597\n", | |
"TestScenario.SingleStream qps=216.16, mean=0.0276, time=2.318, acc=92.165%, mAP=25.649%, queries=501, tiles=50.0:0.0273,80.0:0.0283,90.0:0.0289,95.0:0.0301,99.0:0.0334,99.9:0.0373\n", | |
"mAP=25.649%\n", | |
"TestScenario.SingleStream qps=35.88, mean=0.0278, time=13.965, queries=501, tiles=50.0:0.0275,80.0:0.0284,90.0:0.0292,95.0:0.0302,99.0:0.0341,99.9:0.0403\n", | |
"====== ssd-mobilenet/MultiStream =====\n", | |
"loading annotations into memory...\n", | |
"Done (t=0.49s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Loading and preparing results...\n", | |
"Converting ndarray to lists...\n", | |
"(2107, 7)\n", | |
"0/2107\n", | |
"DONE (t=0.01s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Running per image evaluation...\n", | |
"Evaluate annotation type *bbox*\n", | |
"DONE (t=1.04s).\n", | |
"Accumulating evaluation results...\n", | |
"DONE (t=0.48s).\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.256\n", | |
" Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.380\n", | |
" Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.280\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.019\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.180\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.557\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.231\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.283\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.284\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.022\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.189\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.597\n", | |
"TestScenario.MultiStream qps=75.48, mean=0.0892, time=3.325, acc=92.169%, mAP=25.648%, queries=251, tiles=50.0:0.0896,80.0:0.0968,90.0:0.0992,95.0:0.1015,99.0:0.1788,99.9:0.2639\n", | |
"mAP=25.649%\n", | |
"TestScenario.MultiStream qps=2.00, mean=0.0878, time=501.111, queries=1002, tiles=50.0:0.0885,80.0:0.0957,90.0:0.0980,95.0:0.0994,99.0:0.1075,99.9:0.3103\n", | |
"====== ssd-mobilenet/Server =====\n", | |
"loading annotations into memory...\n", | |
"Done (t=0.50s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Loading and preparing results...\n", | |
"Converting ndarray to lists...\n", | |
"(2106, 7)\n", | |
"0/2106\n", | |
"DONE (t=0.01s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Running per image evaluation...\n", | |
"Evaluate annotation type *bbox*\n", | |
"DONE (t=1.05s).\n", | |
"Accumulating evaluation results...\n", | |
"DONE (t=0.48s).\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.256\n", | |
" Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.380\n", | |
" Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.280\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.019\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.180\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.557\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.231\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.283\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.284\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.022\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.189\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.597\n", | |
"TestScenario.Server qps=217.79, mean=0.0618, time=2.300, acc=92.165%, mAP=25.649%, queries=501, tiles=50.0:0.0611,80.0:0.0659,90.0:0.0690,95.0:0.0845,99.0:0.1041,99.9:0.1311\n", | |
"mAP=25.649%\n", | |
"TestScenario.Server qps=16.50, mean=0.0317, time=0.061, queries=1, tiles=50.0:0.0317,80.0:0.0317,90.0:0.0317,95.0:0.0317,99.0:0.0317,99.9:0.0317\n", | |
"====== ssd-mobilenet/Offline =====\n", | |
"loading annotations into memory...\n", | |
"Done (t=0.46s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Loading and preparing results...\n", | |
"Converting ndarray to lists...\n", | |
"(2106, 7)\n", | |
"0/2106\n", | |
"DONE (t=0.01s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Running per image evaluation...\n", | |
"Evaluate annotation type *bbox*\n", | |
"DONE (t=1.00s).\n", | |
"Accumulating evaluation results...\n", | |
"DONE (t=0.57s).\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.256\n", | |
" Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.380\n", | |
" Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.280\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.019\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.180\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.557\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.231\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.283\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.284\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.022\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.189\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.597\n", | |
"TestScenario.Offline qps=7.57, mean=4.6831, time=2.245, acc=92.165%, mAP=25.649%, queries=17, tiles=50.0:5.5412,80.0:6.3013,90.0:6.6706,95.0:6.8977,99.0:7.3473,99.9:7.4485\n", | |
"mAP=25.649%\n", | |
"TestScenario.Offline qps=51.87, mean=5.5460, time=9.658, queries=501, tiles=50.0:5.3666,80.0:8.3505,90.0:9.2078,95.0:9.2078,99.0:9.6345,99.9:9.6345\n", | |
"====== ssd-resnet34/SingleStream =====\n", | |
"loading annotations into memory...\n", | |
"Done (t=0.44s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Loading and preparing results...\n", | |
"Converting ndarray to lists...\n", | |
"(89779, 7)\n", | |
"0/89779\n", | |
"DONE (t=0.40s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Running per image evaluation...\n", | |
"Evaluate annotation type *bbox*\n", | |
"DONE (t=5.32s).\n", | |
"Accumulating evaluation results...\n", | |
"DONE (t=1.51s).\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.253\n", | |
" Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.449\n", | |
" Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.258\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.178\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.349\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.298\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.236\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.373\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.391\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.221\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.486\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.436\n", | |
"TestScenario.SingleStream qps=43.03, mean=0.1077, time=11.643, acc=33.184%, mAP=25.346%, queries=501, tiles=50.0:0.1061,80.0:0.1118,90.0:0.1184,95.0:0.1229,99.0:0.1302,99.9:0.1489\n", | |
"mAP=0.430%\n", | |
"TestScenario.SingleStream qps=9.26, mean=0.1079, time=54.120, queries=501, tiles=50.0:0.1059,80.0:0.1141,90.0:0.1198,95.0:0.1231,99.0:0.1319,99.9:0.1508\n", | |
"====== ssd-resnet34/MultiStream =====\n", | |
"loading annotations into memory...\n", | |
"Done (t=0.44s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Loading and preparing results...\n", | |
"Converting ndarray to lists...\n", | |
"(89936, 7)\n", | |
"0/89936\n", | |
"DONE (t=0.40s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Running per image evaluation...\n", | |
"Evaluate annotation type *bbox*\n", | |
"DONE (t=6.00s).\n", | |
"Accumulating evaluation results...\n", | |
"DONE (t=1.52s).\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.253\n", | |
" Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.449\n", | |
" Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.258\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.178\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.349\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.298\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.236\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.373\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.391\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.221\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.486\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.436\n", | |
"TestScenario.MultiStream qps=18.66, mean=0.2203, time=13.449, acc=33.271%, mAP=25.340%, queries=251, tiles=50.0:0.2154,80.0:0.2271,90.0:0.2323,95.0:0.2387,99.0:0.2789,99.9:0.6477\n", | |
"mAP=0.430%\n", | |
"TestScenario.MultiStream qps=2.00, mean=0.2281, time=501.643, queries=1002, tiles=50.0:0.2252,80.0:0.2334,90.0:0.2426,95.0:0.2490,99.0:0.2974,99.9:0.7694\n", | |
"====== ssd-resnet34/Server =====\n", | |
"loading annotations into memory...\n", | |
"Done (t=0.45s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Loading and preparing results...\n", | |
"Converting ndarray to lists...\n", | |
"(89779, 7)\n", | |
"0/89779\n", | |
"DONE (t=0.40s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Running per image evaluation...\n", | |
"Evaluate annotation type *bbox*\n", | |
"DONE (t=5.28s).\n", | |
"Accumulating evaluation results...\n", | |
"DONE (t=1.44s).\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.253\n", | |
" Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.449\n", | |
" Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.258\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.178\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.349\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.298\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.236\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.373\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.391\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.221\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.486\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.436\n", | |
"TestScenario.Server qps=43.52, mean=0.1348, time=11.511, acc=33.184%, mAP=25.346%, queries=501, tiles=50.0:0.1264,80.0:0.1401,90.0:0.1869,95.0:0.2038,99.0:0.2186,99.9:0.2462\n", | |
"mAP=0.430%\n", | |
"TestScenario.Server qps=7.08, mean=0.1139, time=0.141, queries=1, tiles=50.0:0.1139,80.0:0.1139,90.0:0.1139,95.0:0.1139,99.0:0.1139,99.9:0.1139\n", | |
"====== ssd-resnet34/Offline =====\n", | |
"loading annotations into memory...\n", | |
"Done (t=0.44s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Loading and preparing results...\n", | |
"Converting ndarray to lists...\n", | |
"(89778, 7)\n", | |
"0/89778\n", | |
"DONE (t=0.40s)\n", | |
"creating index...\n", | |
"index created!\n", | |
"Running per image evaluation...\n", | |
"Evaluate annotation type *bbox*\n", | |
"DONE (t=5.44s).\n", | |
"Accumulating evaluation results...\n", | |
"DONE (t=1.46s).\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.253\n", | |
" Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.449\n", | |
" Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.258\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.178\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.349\n", | |
" Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.298\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.236\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.373\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.391\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.221\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.486\n", | |
" Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.436\n", | |
"TestScenario.Offline qps=1.42, mean=24.5135, time=11.936, acc=33.184%, mAP=25.346%, queries=17, tiles=50.0:27.1318,80.0:32.9352,90.0:36.7293,95.0:38.2327,99.0:38.6533,99.9:38.7480\n", | |
"mAP=0.430%\n", | |
"TestScenario.Offline qps=9.29, mean=31.5222, time=53.929, queries=501, tiles=50.0:28.9873,80.0:46.5904,90.0:53.2797,95.0:53.5744,99.0:53.5744,99.9:53.5744\n" | |
] | |
}, | |
{ | |
"name": "stderr", | |
"output_type": "stream", | |
"text": [ | |
] | |
} | |
], | |
"source": [ | |
"%%bash\n", | |
"\n", | |
"# where to find stuff\n", | |
"export DATA_ROOT=`pwd`/data\n", | |
"export MODEL_DIR=`pwd`/models\n", | |
"\n", | |
"# options for official runs\n", | |
"gopt=\"--max-batchsize 32 --samples-per-query 2 --threads 2\"\n", | |
"gopt=\"$gopt --count 500\"\n", | |
"\n", | |
"rm -rf output\n", | |
"result=output/results.csv\n", | |
"\n", | |
"\n", | |
"function one_run {\n", | |
" # args: mode framework device model ...\n", | |
" scenario=$1; shift\n", | |
" model=$3\n", | |
" output_dir=output/$1-$2/$model\n", | |
" echo \"====== $model/$scenario =====\"\n", | |
"\n", | |
" # accuracy run\n", | |
" ./run_local.sh $@ --scenario $scenario --accuracy --output $output_dir/$scenario.Accuracy\n", | |
" if [ $model == \"resnet50\" -o $model == \"mobilenet\" ] ; then\n", | |
" python tools/accuracy-imagenet.py --mlperf-accuracy-file $output_dir/$scenario.Accuracy/mlperf_log_accuracy.json \\\n", | |
" --imagenet-val-file $DATA_ROOT/imagenet2012/val_map.txt > $output_dir/$scenario.Accuracy/Accuracy.txt\n", | |
" else\n", | |
" python tools/accuracy-coco.py --mlperf-accuracy-file $output_dir/$scenario.Accuracy/mlperf_log_accuracy.json \\\n", | |
" --coco-dir $DATA_ROOT/coco | grep mAP > $output_dir/$scenario.Accuracy/Accuracy.txt\n", | |
" fi\n", | |
" cat $output_dir/$scenario.Accuracy/Accuracy.txt\n", | |
"\n", | |
" # performance run\n", | |
" ./run_local.sh $@ --scenario $scenario --output $output_dir/$scenario\n", | |
"\n", | |
" # summary to csv\n", | |
" python tools/lglog2csv.py --input $output_dir/$scenario/mlperf_log_summary.txt --runtime \"$1-$2\" --machine $HOSTNAME --model $3 --name $1-$2-py >> $result\n", | |
"}\n", | |
"\n", | |
"function one_model {\n", | |
" # args: framework device model ...\n", | |
" one_run SingleStream $@\n", | |
" one_run MultiStream $@\n", | |
" one_run Server $@\n", | |
" one_run Offline $@\n", | |
"}\n", | |
"\n", | |
"\n", | |
"mkdir output\n", | |
"echo \"build,date,machine,runtime,model,mode,qps,mean,latency_90,latency_99\" > $result\n", | |
"\n", | |
"# run image classifier benchmarks \n", | |
"export DATA_DIR=$DATA_ROOT/imagenet2012\n", | |
"one_model onnxruntime cpu mobilenet $gopt\n", | |
"one_model tf gpu resnet50 $gopt\n", | |
"\n", | |
"# run ssd benchmarks\n", | |
"export DATA_DIR=$DATA_ROOT/coco\n", | |
"one_model tf gpu ssd-mobilenet $gopt\n", | |
"one_model tf gpu ssd-resnet34 $gopt\n" | |
] | |
}, | |
{ | |
"cell_type": "markdown", | |
"metadata": {}, | |
"source": [ | |
"### Create submission directory\n", | |
"\n", | |
"The mlperf submission needs to contain a json file that describes the systems used.\n" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 36, | |
"metadata": {}, | |
"outputs": [], | |
"source": [ | |
"%%bash\n", | |
"\n", | |
"mkdir -p ${SUBMISSION_DIR}/results ${SUBMISSION_DIR}/benchmarks ${SUBMISSION_DIR}/systems\n", | |
"\n", | |
"# copy results\n", | |
"cp -r output/tf-gpu ${SUBMISSION_DIR}/results\n", | |
"cp -r output/onnxruntime-cpu ${SUBMISSION_DIR}/results\n", | |
"\n", | |
"# copy/idenify benchmark source code\n", | |
"echo \"git clone https://github.com/mlperf/inference.git\" > ${SUBMISSION_DIR}/benchmarks/VERSION.txt\n", | |
"git rev-parse HEAD >> ${SUBMISSION_DIR}/benchmarks/VERSION.txt\n", | |
"\n", | |
"cat > ${SUBMISSION_DIR}/systems/tf-gpu.json <<EOF\n", | |
"{\n", | |
" \"system_name\": \"tf-gpu\",\n", | |
" \"submission\": {\n", | |
" \"submitter\": \"Mlperf\",\n", | |
" \"division\": \"closed\",\n", | |
" \"status\": \"cloud\"\n", | |
" },\n", | |
" \"sw\": {\n", | |
" \"framework\": \"tf-14.1\",\n", | |
" \"cublas_version\": \"10.2.0.163\",\n", | |
" \"cuda_version\": \"10.1.163\",\n", | |
" \"cudnn_version\": \"7.6.0.64\",\n", | |
" \"container_base\": \"Ubuntu-16.04\",\n", | |
" \"cuda_driver_version\": \"418.67\",\n", | |
" \"os\": \"Ubuntu 16.04.2 LTS\" \n", | |
" },\n", | |
" \"notes\": \"\",\n", | |
" \"hw\": {\n", | |
" \"sys_storage_type\": \"SATA SSD\",\n", | |
" \"cpu\": \"2x Intel(R) Xeon(R) CPU E5-2698 v4 @2.20GHz\",\n", | |
" \"host_processors_per_node\": \"1\",\n", | |
" \"host_processor_model_name\": \"Intel(R) Xeon(R) CPU E5-1620 v3 @ 3.50GH\",\n", | |
" \"host_processor_core_count\": \"4\",\n", | |
" \"host_processor_vcpu_count\": \"8\",\n", | |
" \"host_processor_frequency\": \"3.50GH\",\n", | |
" \"host_processor_caches\": \"10240KB\",\n", | |
" \"host_processor_interconnect\": \"-\",\n", | |
" \"host_memory_capacity\": \"32GB\",\n", | |
" \"host_storage_type\": \"SSD\",\n", | |
" \"host_storage_capacity\": \"200GB\",\n", | |
" \"host_networking\": \"-\",\n", | |
" \"host_networking_topology\": \"-\",\n", | |
" \"host_memory_configuration\": \"-\",\n", | |
" \"accelerator\": \"Tesla V100-SXM2-16GB\",\n", | |
" \"accelerator_model_name\": \"GTX-1080\",\n", | |
" \"accelerator_host_interconnect\": \"-\",\n", | |
" \"accelerator_core_count\": \"2560\",\n", | |
" \"accelerator_frequency\": \"1733MHz\",\n", | |
" \"accelerator_on-chip_memories\": \"1\",\n", | |
" \"accelerator_memory_configuration\": \"HBM\",\n", | |
" \"accelerator_memory_capacity\": \"8 GB\",\n", | |
" \"num_accelerators\": \"1\",\n", | |
" \"num_nodes\": \"1\",\n", | |
" \"accelerators_per_node\": \"1\",\n", | |
" \"hw_notes\": \"\"\n", | |
" }\n", | |
"}\n", | |
"EOF\n", | |
"\n", | |
"cat > ${SUBMISSION_DIR}/systems/onnxruntime-cpu.json <<EOF\n", | |
"{\n", | |
" \"system_name\": \"onnxruntime-cpu\",\n", | |
" \"submission\": {\n", | |
" \"submitter\": \"Mlperf\",\n", | |
" \"division\": \"closed\",\n", | |
" \"status\": \"cloud\"\n", | |
" },\n", | |
" \"sw\": {\n", | |
" \"framework\": \"tf-14.1\",\n", | |
" \"os\": \"Ubuntu 16.04.2 LTS\" \n", | |
" },\n", | |
" \"notes\": \"\",\n", | |
" \"hw\": {\n", | |
" \"sys_storage_type\": \"SATA SSD\",\n", | |
" \"cpu\": \"2x Intel(R) Xeon(R) CPU E5-2698 v4 @2.20GHz\",\n", | |
" \"host_processors_per_node\": \"1\",\n", | |
" \"host_processor_model_name\": \"Intel(R) Xeon(R) CPU E5-1620 v3 @ 3.50GH\",\n", | |
" \"host_processor_core_count\": \"4\",\n", | |
" \"host_processor_vcpu_count\": \"8\",\n", | |
" \"host_processor_frequency\": \"3.50GH\",\n", | |
" \"host_processor_caches\": \"10240KB\",\n", | |
" \"host_processor_interconnect\": \"-\",\n", | |
" \"host_memory_capacity\": \"32GB\",\n", | |
" \"host_storage_type\": \"SSD\",\n", | |
" \"host_storage_capacity\": \"200GB\",\n", | |
" \"host_networking\": \"-\",\n", | |
" \"host_networking_topology\": \"-\",\n", | |
" \"host_memory_configuration\": \"-\",\n", | |
" \"accelerator\": \"-\",\n", | |
" \"num_accelerators\": \"0\",\n", | |
" \"num_nodes\": \"1\",\n", | |
" \"accelerators_per_node\": \"1\",\n", | |
" \"hw_notes\": \"\"\n", | |
" }\n", | |
"}\n", | |
"EOF" | |
] | |
}, | |
{ | |
"cell_type": "markdown", | |
"metadata": {}, | |
"source": [ | |
"### What's in the submission directory now ?\n" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 41, | |
"metadata": {}, | |
"outputs": [ | |
{ | |
"name": "stdout", | |
"output_type": "stream", | |
"text": [ | |
"/tmp/Mlperf\r\n", | |
"/tmp/Mlperf/systems\r\n", | |
"/tmp/Mlperf/benchmarks\r\n", | |
"/tmp/Mlperf/results\r\n", | |
"/tmp/Mlperf/results/tf-gpu\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-mobilenet\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-mobilenet/Server.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-mobilenet/Offline.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-mobilenet/SingleStream\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-mobilenet/SingleStream.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-mobilenet/MultiStream\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-mobilenet/Offline\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-mobilenet/MultiStream.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-mobilenet/Server\r\n", | |
"/tmp/Mlperf/results/tf-gpu/resnet50\r\n", | |
"/tmp/Mlperf/results/tf-gpu/resnet50/Server.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/resnet50/Offline.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/resnet50/SingleStream\r\n", | |
"/tmp/Mlperf/results/tf-gpu/resnet50/SingleStream.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/resnet50/MultiStream\r\n", | |
"/tmp/Mlperf/results/tf-gpu/resnet50/Offline\r\n", | |
"/tmp/Mlperf/results/tf-gpu/resnet50/MultiStream.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/resnet50/Server\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-resnet34\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-resnet34/Server.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-resnet34/Offline.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-resnet34/SingleStream\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-resnet34/SingleStream.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-resnet34/MultiStream\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-resnet34/Offline\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-resnet34/MultiStream.Accuracy\r\n", | |
"/tmp/Mlperf/results/tf-gpu/ssd-resnet34/Server\r\n", | |
"/tmp/Mlperf/results/onnxruntime-cpu\r\n", | |
"/tmp/Mlperf/results/onnxruntime-cpu/mobilenet\r\n", | |
"/tmp/Mlperf/results/onnxruntime-cpu/mobilenet/Server.Accuracy\r\n", | |
"/tmp/Mlperf/results/onnxruntime-cpu/mobilenet/Offline.Accuracy\r\n", | |
"/tmp/Mlperf/results/onnxruntime-cpu/mobilenet/SingleStream\r\n", | |
"/tmp/Mlperf/results/onnxruntime-cpu/mobilenet/SingleStream.Accuracy\r\n", | |
"/tmp/Mlperf/results/onnxruntime-cpu/mobilenet/MultiStream\r\n", | |
"/tmp/Mlperf/results/onnxruntime-cpu/mobilenet/Offline\r\n", | |
"/tmp/Mlperf/results/onnxruntime-cpu/mobilenet/MultiStream.Accuracy\r\n", | |
"/tmp/Mlperf/results/onnxruntime-cpu/mobilenet/Server\r\n" | |
] | |
} | |
], | |
"source": [ | |
"!find {SUBMISSION_DIR} -type d" | |
] | |
}, | |
{ | |
"cell_type": "markdown", | |
"metadata": {}, | |
"source": [ | |
"And if we look at the files for a specific model/platform/scenario" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 63, | |
"metadata": {}, | |
"outputs": [ | |
{ | |
"name": "stdout", | |
"output_type": "stream", | |
"text": [ | |
"\n", | |
"\u001b[31mFor example: tf-gpu resnet50 SingleStream\u001b[0m\n", | |
"\n", | |
"mlperf_log_accuracy.json mlperf_log_summary.txt results.json\n", | |
"mlperf_log_detail.txt\t mlperf_log_trace.json\n", | |
"\n", | |
"Accuracy in results/tf-gpu/resnet50/SingleStream.Accuracy/Accuracy.txt\n", | |
"accuracy=76.000%, good=380, total=500\n", | |
"\n", | |
"Summary in results/tf-gpu/resnet50/SingleStream/mlperf_log_summary.txt\n", | |
"================================================\n", | |
"MLPerf Results Summary\n", | |
"================================================\n", | |
"SUT name : PySUT\n", | |
"Scenario : Single Stream\n", | |
"Mode : Performance\n", | |
"90th percentile latency (ns) : 10690255\n", | |
"Result is : INVALID\n", | |
" Min duration satisfied : NO\n", | |
" Min queries satisfied : Yes\n", | |
"Recommendations:\n", | |
" * Decrease the expected latency so the loadgen pre-generates more queries.\n", | |
"\n", | |
"================================================\n", | |
"Additional Stats\n", | |
"================================================\n", | |
"QPS w/ loadgen overhead : 106.82\n", | |
"QPS w/o loadgen overhead : 107.18\n", | |
"\n", | |
"Min latency (ns) : 8248658\n", | |
"Max latency (ns) : 11679843\n", | |
"Mean latency (ns) : 9330370\n", | |
"50.00 percentile latency (ns) : 8944492\n", | |
"90.00 percentile latency (ns) : 10690255\n", | |
"95.00 percentile latency (ns) : 10913400\n", | |
"97.00 percentile latency (ns) : 10957330\n", | |
"99.00 percentile latency (ns) : 11057972\n", | |
"99.90 percentile latency (ns) : 11679843\n", | |
"\n", | |
"================================================\n", | |
"Test Parameters Used\n", | |
"================================================\n", | |
"samples_per_query : 1\n", | |
"target_qps : 100\n", | |
"target_latency (ns): 0\n", | |
"max_async_queries : 1\n", | |
"min_duration (ms): 60000\n", | |
"max_duration (ms): 0\n", | |
"min_query_count : 500\n", | |
"max_query_count : 500\n", | |
"qsl_rng_seed : 0\n", | |
"sample_index_rng_seed : 0\n", | |
"schedule_rng_seed : 0\n", | |
"accuracy_log_rng_seed : 0\n", | |
"accuracy_log_probability : 0\n", | |
"\n", | |
"1 ERROR encountered. See detailed log.\n", | |
"\n", | |
"The system info that goes with tf-gpu\n", | |
"{\n", | |
" \"system_name\": \"tf-gpu\",\n", | |
" \"submission\": {\n", | |
" \"submitter\": \"Mlperf\",\n", | |
" \"division\": \"closed\",\n", | |
" \"status\": \"cloud\"\n", | |
" },\n", | |
" \"sw\": {\n", | |
" \"framework\": \"tf-14.1\",\n", | |
" \"cublas_version\": \"10.2.0.163\",\n", | |
" \"cuda_version\": \"10.1.163\",\n", | |
" \"cudnn_version\": \"7.6.0.64\",\n", | |
" \"container_base\": \"Ubuntu-16.04\",\n", | |
" \"cuda_driver_version\": \"418.67\",\n", | |
" \"os\": \"Ubuntu 16.04.2 LTS\" \n", | |
" },\n", | |
" \"notes\": \"\",\n", | |
" \"hw\": {\n", | |
" \"sys_storage_type\": \"SATA SSD\",\n", | |
" \"cpu\": \"2x Intel(R) Xeon(R) CPU E5-2698 v4 @2.20GHz\",\n", | |
" \"host_processors_per_node\": \"1\",\n", | |
" \"host_processor_model_name\": \"Intel(R) Xeon(R) CPU E5-1620 v3 @ 3.50GH\",\n", | |
" \"host_processor_core_count\": \"4\",\n", | |
" \"host_processor_vcpu_count\": \"8\",\n", | |
" \"host_processor_frequency\": \"3.50GH\",\n", | |
" \"host_processor_caches\": \"10240KB\",\n", | |
" \"host_processor_interconnect\": \"-\",\n", | |
" \"host_memory_capacity\": \"32GB\",\n", | |
" \"host_storage_type\": \"SSD\",\n", | |
" \"host_storage_capacity\": \"200GB\",\n", | |
" \"host_networking\": \"-\",\n", | |
" \"host_networking_topology\": \"-\",\n", | |
" \"host_memory_configuration\": \"-\",\n", | |
" \"accelerator\": \"Tesla V100-SXM2-16GB\",\n", | |
" \"accelerator_model_name\": \"GTX-1080\",\n", | |
" \"accelerator_host_interconnect\": \"-\",\n", | |
" \"accelerator_core_count\": \"2560\",\n", | |
" \"accelerator_frequency\": \"1733MHz\",\n", | |
" \"accelerator_on-chip_memories\": \"1\",\n", | |
" \"accelerator_memory_configuration\": \"HBM\",\n", | |
" \"accelerator_memory_capacity\": \"8 GB\",\n", | |
" \"num_accelerators\": \"1\",\n", | |
" \"num_nodes\": \"1\",\n", | |
" \"accelerators_per_node\": \"1\",\n", | |
" \"hw_notes\": \"\"\n", | |
" }\n", | |
"}\n" | |
] | |
} | |
], | |
"source": [ | |
"print(\"\\n\\x1b[31mFor example: tf-gpu resnet50 SingleStream\\x1b[0m\\n\")\n", | |
"\n", | |
"!ls {SUBMISSION_DIR}/results/tf-gpu/resnet50/SingleStream\n", | |
"\n", | |
"print(\"\\nAccuracy in results/tf-gpu/resnet50/SingleStream.Accuracy/Accuracy.txt\")\n", | |
"!cat {SUBMISSION_DIR}/results/tf-gpu/resnet50/SingleStream.Accuracy/Accuracy.txt\n", | |
"\n", | |
"print(\"\\nSummary in results/tf-gpu/resnet50/SingleStream/mlperf_log_summary.txt\")\n", | |
"!cat {SUBMISSION_DIR}/results/tf-gpu/resnet50/SingleStream/mlperf_log_summary.txt\n", | |
"\n", | |
"print(\"\\nThe system info that goes with tf-gpu\")\n", | |
"!cat {SUBMISSION_DIR}/systems/tf-gpu.json\n" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": null, | |
"metadata": {}, | |
"outputs": [], | |
"source": [] | |
} | |
], | |
"metadata": { | |
"kernelspec": { | |
"display_name": "Python 3", | |
"language": "python", | |
"name": "python3" | |
}, | |
"language_info": { | |
"codemirror_mode": { | |
"name": "ipython", | |
"version": 3 | |
}, | |
"file_extension": ".py", | |
"mimetype": "text/x-python", | |
"name": "python", | |
"nbconvert_exporter": "python", | |
"pygments_lexer": "ipython3", | |
"version": "3.7.3" | |
} | |
}, | |
"nbformat": 4, | |
"nbformat_minor": 2 | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment