[ { "path": "LICENSE", "content": "The MIT License (MIT)\n\nCopyright (c) 2016 Piotr\n\nPermission is hereby granted, free of charge, to any person obtaining a copy\nof this software and associated documentation files (the \"Software\"), to deal\nin the Software without restriction, including without limitation the rights\nto use, copy, modify, merge, publish, distribute, sublicense, and/or sell\ncopies of the Software, and to permit persons to whom the Software is\nfurnished to do so, subject to the following conditions:\n\nThe above copyright notice and this permission notice shall be included in all\ncopies or substantial portions of the Software.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\nIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\nAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\nLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\nOUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\nSOFTWARE.\n" }, { "path": "README.md", "content": "# keras2cpp\n\nThis is a bunch of code to port Keras neural network model into pure C++. Neural network weights and architecture are stored in plain text file and input is presented as `vector > >` in case of image. The code is prepared to support simple Convolutional network (from MNIST example) but can be easily extended. There are implemented only ReLU and Softmax activations.\n\nIt is working with the Theano backend.\n\n## Usage\n\n 1. Save your network weights and architecture.\n 2. Dump network structure to plain text file with `dump_to_simple_cpp.py` script.\n 3. Use network with code from `keras_model.h` and `keras_model.cc` files - see example below.\n\n## Example\n\n 1. Run one iteration of simple CNN on MNIST data with `example/mnist_cnn_one_iteration.py` script. It will produce files with architecture `example/my_nn_arch.json` and weights in HDF5 format `example/my_nn_weights.h5`.\n 2. Dump network to plain text file `python dump_to_simple_cpp.py -a example/my_nn_arch.json -w example/my_nn_weights.h5 -o example/dumped.nnet`.\n 3. Compile example `g++ -std=c++11 keras_model.cc example_main.cc` - see code in `example_main.cc`.\n 4. Run binary `./a.out` - you shoul get the same output as in step one from Keras.\n\n## Testing\n\nIf you want to test dumping for your network, please use `test_run.sh` script. Please provide there your network architecture and weights. The script do following job:\n\n 1. Dump network into text file.\n 2. Generate random sample.\n 3. Compute predictions from keras and keras2cpp on generated sample.\n 4. Compare predictions.\n\n## Similar repositories\n\n- Keras to C++ with usage of Tensorflow C API https://github.com/aljabr0/from-keras-to-c\n" }, { "path": "dump_to_simple_cpp.py", "content": "import numpy as np\nnp.random.seed(1337)\nfrom keras.models import Sequential, model_from_json\nimport json\nimport argparse\n\nnp.set_printoptions(threshold=np.inf)\nparser = argparse.ArgumentParser(description='This is a simple script to dump Keras model into simple format suitable for porting into pure C++ model')\n\nparser.add_argument('-a', '--architecture', help=\"JSON with model architecture\", required=True)\nparser.add_argument('-w', '--weights', help=\"Model weights in HDF5 format\", required=True)\nparser.add_argument('-o', '--output', help=\"Ouput file name\", required=True)\nparser.add_argument('-v', '--verbose', help=\"Verbose\", required=False)\nargs = parser.parse_args()\n\nprint 'Read architecture from', args.architecture\nprint 'Read weights from', args.weights\nprint 'Writing to', args.output\n\narch = open(args.architecture).read()\nmodel = model_from_json(arch)\nmodel.load_weights(args.weights)\nmodel.compile(loss='categorical_crossentropy', optimizer='adadelta')\narch = json.loads(arch)\n\nwith open(args.output, 'w') as fout:\n fout.write('layers ' + str(len(model.layers)) + '\\n')\n\n layers = []\n for ind, l in enumerate(arch[\"config\"]):\n if args.verbose:\n print ind, l\n fout.write('layer ' + str(ind) + ' ' + l['class_name'] + '\\n')\n\n if args.verbose:\n print str(ind), l['class_name']\n layers += [l['class_name']]\n if l['class_name'] == 'Convolution2D':\n #fout.write(str(l['config']['nb_filter']) + ' ' + str(l['config']['nb_col']) + ' ' + str(l['config']['nb_row']) + ' ')\n\n #if 'batch_input_shape' in l['config']:\n # fout.write(str(l['config']['batch_input_shape'][1]) + ' ' + str(l['config']['batch_input_shape'][2]) + ' ' + str(l['config']['batch_input_shape'][3]))\n #fout.write('\\n')\n\n W = model.layers[ind].get_weights()[0]\n if args.verbose:\n print W.shape\n fout.write(str(W.shape[0]) + ' ' + str(W.shape[1]) + ' ' + str(W.shape[2]) + ' ' + str(W.shape[3]) + ' ' + l['config']['border_mode'] + '\\n')\n\n for i in range(W.shape[0]):\n for j in range(W.shape[1]):\n for k in range(W.shape[2]):\n fout.write(str(W[i,j,k]) + '\\n')\n fout.write(str(model.layers[ind].get_weights()[1]) + '\\n')\n\n if l['class_name'] == 'Activation':\n fout.write(l['config']['activation'] + '\\n')\n if l['class_name'] == 'MaxPooling2D':\n fout.write(str(l['config']['pool_size'][0]) + ' ' + str(l['config']['pool_size'][1]) + '\\n')\n #if l['class_name'] == 'Flatten':\n # print l['config']['name']\n if l['class_name'] == 'Dense':\n #fout.write(str(l['config']['output_dim']) + '\\n')\n W = model.layers[ind].get_weights()[0]\n if args.verbose:\n print W.shape\n fout.write(str(W.shape[0]) + ' ' + str(W.shape[1]) + '\\n')\n\n\n for w in W:\n fout.write(str(w) + '\\n')\n fout.write(str(model.layers[ind].get_weights()[1]) + '\\n')\n" }, { "path": "example/dumped.nnet", "content": "layers 12\nlayer 0 Convolution2D\n4 1 3 3 same\n[ 0.13240439 -0.39469701 -0.05886053]\n[-0.16273086 0.27228925 0.70811206]\n[-0.0382412 0.91429299 0.72160912]\n[ 0.41376248 -0.13718128 -0.55548537]\n[ 0.06595665 -0.43903521 0.04035483]\n[ 0.34481722 0.63220054 0.07781094]\n[ 0.41419399 -0.13306008 0.72000468]\n[ 0.28247169 0.10319189 0.50113165]\n[-0.14225948 0.33037826 -0.4037863 ]\n[ 0.21228614 -0.69273335 0.45505175]\n[ 0.63414061 -0.75804955 -0.62078679]\n[-0.5811435 -0.20503305 0.65040439]\n[ 0.04864343 -0.02648391 -0.00558628 0.06567492]\nlayer 1 Activation\nrelu\nlayer 2 Convolution2D\n4 4 3 3 same\n[-0.10421877 -0.02785102 0.38203281]\n[ 0.2905634 0.15197834 0.04869499]\n[-0.09692651 -0.30311567 -0.26422366]\n[-0.00276754 -0.19394568 -0.39375487]\n[ 0.06125383 0.13863547 0.20455685]\n[ 0.01164767 -0.23414099 -0.00473231]\n[-0.06518493 -0.19672309 0.14905775]\n[ 0.15651399 0.2891407 0.38651308]\n[ 0.1900083 -0.37353948 -0.21590027]\n[-0.03375062 -0.04178833 -0.2734594 ]\n[-0.45986223 -0.19825426 -0.0696836 ]\n[ 0.05288092 0.21991098 -0.36576423]\n[-0.01610908 0.11497457 -0.61316365]\n[-0.01072869 0.14255261 -0.4019627 ]\n[ 0.38004246 0.19172479 0.06467063]\n[ 0.59239095 0.29410571 0.20647256]\n[ 0.09890065 0.51240301 0.27240333]\n[-0.102652 -0.17938389 0.4384366 ]\n[ 0.08311749 -0.21888258 -0.61088938]\n[ 0.26572421 0.08327031 0.01123066]\n[-0.13556297 -0.17235672 -0.00744808]\n[-0.58282799 -0.2371998 -0.02106583]\n[-0.374268 -0.43444765 -0.41671872]\n[-0.48330817 -0.07184441 -0.24444637]\n[-0.60360354 -0.47681275 -0.34594122]\n[ 0.20064028 0.17831512 -0.02612569]\n[ 0.45589581 0.4922587 -0.06326807]\n[-0.29128221 0.09931409 0.24801998]\n[-0.06581956 0.19527058 0.2431576 ]\n[ 0.45251682 -0.0852524 0.35469437]\n[ 0.11536153 -0.55935317 0.05506861]\n[-0.17236997 -0.33490059 -0.68386561]\n[-0.19986324 0.06262327 -0.37508488]\n[ 0.13902199 -0.30721149 0.18993607]\n[-0.13360961 -0.13865279 -0.38592908]\n[-0.03786371 0.2630626 0.14453529]\n[-0.06631006 -0.4063293 -0.2363375 ]\n[ 0.19549578 -0.00670204 -0.32390189]\n[-0.09911008 -0.36811453 -0.29338896]\n[-0.27110478 0.43789902 0.40986046]\n[-0.2690371 0.42322639 0.17070536]\n[ 0.21714966 0.11392772 -0.21328911]\n[ 0.3063466 0.04640941 -0.02356246]\n[ 0.11048113 0.24086125 -0.04183687]\n[ 0.14844583 0.25013149 -0.21424711]\n[ 0.37809837 0.09550973 0.29315791]\n[-0.05270186 0.12204736 0.32245034]\n[ 0.35050899 0.36823237 -0.04089836]\n[-0.12815939 -0.17673898 0.0116122 -0.07211183]\nlayer 3 Activation\nrelu\nlayer 4 MaxPooling2D\n2 2\nlayer 5 Dropout\nlayer 6 Flatten\nlayer 7 Dense\n784 6\n[-0.03906742 0.06698083 -0.08050424 -0.00319093 0.01833775 0.00991163]\n[-0.01495006 0.03293357 0.0197208 -0.04033855 -0.03371154 -0.04916968]\n[-0.0560285 -0.06582659 0.00927044 -0.02810042 -0.04538313 -0.01811428]\n[-0.04213877 -0.06204742 0.00786292 -0.08450737 0.06187303 0.02497161]\n[ 0.04896151 0.02935885 0.09386393 -0.04570646 0.05239167 -0.08469187]\n[ 0.01835042 0.04916618 0.12470287 0.00891534 -0.04932225 0.04197544]\n[ 0.06273581 0.02127488 0.0401251 -0.06060422 -0.13253379 0.05847419]\n[ 0.12982321 0.1010593 0.2134286 -0.08732843 -0.12440202 -0.0320878 ]\n[ 0.07623585 0.0563375 0.07737941 -0.0296917 -0.14714883 -0.02166067]\n[ 0.16959247 0.11041992 0.19505848 0.00311863 -0.04162547 0.04950376]\n[ 0.0729512 0.07153659 0.07287914 -0.09814236 -0.11967845 -0.01919617]\n[ 0.03707012 0.09931304 -0.03107041 0.02655093 -0.01263645 -0.08081397]\n[ 0.04423667 -0.00867222 0.01667016 -0.04046617 -0.05045989 -0.07810739]\n[ 0.00049676 0.03291544 -0.05188211 -0.03523895 -0.0487449 0.00134809]\n[ 0.01925284 0.04327489 0.07676191 -0.06858168 0.05246273 0.08637736]\n[-0.03913829 0.04089814 -0.04338754 -0.06232202 0.03502903 -0.00492453]\n[-0.01337897 0.08926103 -0.04653127 0.00476701 0.08402187 -0.0500262 ]\n[-0.0044559 0.00838482 -0.04894469 0.06226534 0.03247495 0.05359224]\n[ 0.11980098 -0.08101542 0.07371085 0.10186744 0.0188344 0.02669876]\n[ 0.19923729 0.04178629 0.10129342 0.00315007 -0.02827051 -0.03766487]\n[ 0.07194332 -0.06360228 -0.02254308 0.08250529 -0.15655711 0.02502301]\n[ 0.09679929 0.05437199 0.04281991 -0.00609906 -0.16727532 0.03721406]\n[ 0.00803989 0.08278463 0.06190167 -0.00692337 -0.08304318 0.02458697]\n[ 0.08096614 0.16214547 0.11075521 -0.05861256 -0.0690367 -0.03747399]\n[ 0.21584366 0.20894378 0.20348011 -0.03105162 -0.08168395 -0.10425199]\n[ 0.10633728 0.10531231 0.05532461 0.00492275 -0.01648618 -0.02858951]\n[ 0.07343981 0.0795544 0.00406283 -0.1350463 0.09071051 -0.06666093]\n[-0.04112834 0.0801154 -0.02094875 0.04578911 -0.08394425 0.04739556]\n[ 0.04340751 -0.04323349 -0.03977324 0.029741 -0.03849278 0.09510033]\n[-0.07096316 0.07336214 0.01099949 0.04040894 0.07377584 -0.00709548]\n[ 0.07882775 -0.07016723 -0.07656394 -0.06225074 0.03239924 0.0848458 ]\n[-0.03906078 -0.07604289 -0.07036381 0.05022649 0.06826855 0.16075383]\n[ 0.05868277 -0.0029423 -0.09052019 0.08513133 0.06857975 0.0548847 ]\n[ 0.07210821 0.028421 0.03118186 0.02777781 -0.06508817 0.03011957]\n[ 0.02577195 -0.0021653 -0.04538134 0.03983229 -0.04444971 0.02814031]\n[ 0.00989027 -0.01719115 -0.05992389 0.04526545 -0.06981328 0.05890024]\n[ 0.01345126 -0.03591003 -0.01485783 0.00824048 -0.06265052 0.08473093]\n[ 0.02871292 0.02361192 0.04635924 -0.00073332 -0.11096975 0.03686479]\n[-0.05461308 0.13399795 -0.04151184 -0.01471005 -0.03385759 -0.02356914]\n[-0.07658672 0.08487934 0.06285738 -0.11803006 0.02261236 0.00323268]\n[-0.05935308 0.1054464 0.07836062 -0.06944308 0.02371275 -0.17641465]\n[-0.06491436 0.06459507 0.05346636 -0.02125865 -0.00230505 -0.03430885]\n[-0.02145559 0.04702919 0.04046459 -0.03417374 0.00791908 -0.05758655]\n[-0.12696458 -0.08001456 -0.08423816 -0.03338545 0.03814339 0.04256224]\n[-0.06056114 -0.00051539 -0.04495636 -0.00070647 0.0030581 0.02429797]\n[-0.07350745 -0.02889902 -0.04967271 -0.07581174 -0.04283836 0.03076509]\n[-0.02569569 -0.09465789 -0.00548012 0.04420721 -0.02736876 -0.00089581]\n[-0.05284071 -0.07590015 0.02237807 -0.02403216 0.00089716 0.00054248]\n[-0.014353 -0.11837757 -0.01349876 0.00988213 -0.01841527 0.00264359]\n[-0.05818343 -0.09152924 -0.03935111 0.0623118 -0.00294162 -0.01540728]\n[-0.03307559 -0.10525655 -0.02241764 0.02028148 0.02561427 0.00384658]\n[-0.02743094 -0.05859468 -0.01221527 -0.00461497 0.027101 0.02186194]\n[-0.10393144 0.00701244 -0.02056715 -0.04508474 -0.06323299 0.04851193]\n[-0.12013954 -0.03573343 0.04121689 -0.06994867 -0.01653158 -0.02954953]\n[-0.11202008 0.00544483 0.04861621 -0.11717936 0.00877294 0.05966207]\n[-0.10761394 -0.12025759 0.09506065 -0.09846672 -0.06385981 -0.00706128]\n[-0.04523962 -0.05461986 0.00158466 0.05548372 -0.00193037 -0.05045158]\n[-0.06085176 -0.05391665 -0.01233329 -0.01210866 -0.08770373 0.03171697]\n[-0.13669041 -0.10556038 -0.0083451 -0.03938511 -0.14809524 -0.0092752 ]\n[-0.03547235 0.00632177 -0.10543172 0.01324754 -0.05205651 -0.05102921]\n[-0.044781 -0.06328677 0.03889268 -0.01418422 -0.01351375 -0.02115388]\n[-0.02551365 -0.05330036 -0.06344371 -0.07193805 0.02519399 0.01447381]\n[-0.0156537 -0.06582764 -0.03264342 -0.00252054 0.03280666 -0.03166911]\n[-0.04884146 -0.09733901 -0.07943926 -0.00302826 -0.04236387 -0.06903206]\n[-0.09653606 -0.0540468 -0.05871857 0.01528029 -0.0458906 -0.00639062]\n[-0.12743062 -0.06528249 -0.15508506 -0.10875963 -0.02613571 0.03022596]\n[-0.10855629 -0.07260169 -0.07472604 0.00578795 0.02758324 0.06854044]\n[-0.02120412 -0.05679055 0.09603125 0.0175001 -0.09669294 -0.07805043]\n[-0.09110473 -0.06463159 0.06585391 0.00330143 -0.08988428 -0.01761026]\n[ 0.02224932 -0.0066196 0.13158537 -0.09908723 -0.10756366 -0.02889675]\n[-0.0445697 -0.0108994 -0.08682603 0.05705853 -0.00622619 0.00903386]\n[-0.12899421 -0.01250589 -0.11728453 -0.10747638 -0.0724315 0.06194278]\n[-0.08139972 0.02385198 -0.01709059 -0.01669967 -0.1100228 -0.00961974]\n[-0.0432987 -0.11630281 0.11211184 -0.07596958 -0.00239594 -0.06798765]\n[-0.07364786 -0.04755174 -0.02912027 -0.00833957 0.04815742 -0.18895435]\n[-0.0347453 0.01868596 -0.0147765 -0.0446967 0.0478966 -0.12075885]\n[-0.06177768 -0.04607397 -0.01522131 -0.02732979 0.04139465 -0.02353299]\n[-0.133301 -0.00655711 -0.03392529 -0.11689362 -0.0883434 0.07187817]\n[-0.06189934 0.0524984 -0.11142992 -0.1036801 0.02576936 0.07690011]\n[ 0.02531234 -0.01139269 -0.12611471 -0.09724379 0.04059971 0.04154509]\n[-0.01147614 -0.04213808 -0.11520851 -0.01658878 0.08237385 0.04917121]\n[-0.0494155 -0.05302509 0.06465708 0.05598482 0.00447672 -0.09627396]\n[-0.04844014 -0.00902561 0.18258154 0.11114652 -0.15210725 -0.18368633]\n[ 0.04977537 -0.04202075 0.04880532 0.09028839 -0.07794982 -0.0135695 ]\n[ 0.0737197 0.02134143 0.03950929 -0.0462447 -0.06542954 -0.04404024]\n[-0.11784814 0.05312875 -0.12023604 0.03744375 0.01453782 0.04277182]\n[-0.0362859 0.04788436 -0.03567404 -0.05353031 0.10789248 -0.11958148]\n[-0.03160278 0.0486227 0.07829643 -0.02335911 -0.00032572 -0.19590537]\n[-0.09877094 0.05633143 0.03377883 0.03512298 0.05049034 -0.13365418]\n[ 0.00762662 0.01745776 0.09184489 -0.02832299 0.03992055 0.0698674 ]\n[-0.00207301 0.07375712 0.02974467 -0.07909034 0.04247067 0.10213128]\n[ 0.00723242 0.01831096 -0.08343405 -0.15346451 0.04382225 0.01230231]\n[ 0.10217623 0.01167687 0.00540794 -0.02553524 0.05552289 -0.01947977]\n[ 0.10517532 -0.00784067 -0.08542134 -0.11327109 0.09972257 -0.03549052]\n[ 0.05870317 0.11366854 -0.07332972 -0.04271904 0.11874869 -0.13107273]\n[ 0.06873229 0.09667565 0.03740055 -0.01186924 0.06541399 -0.15281917]\n[-0.00027914 -0.02707825 0.12996851 0.02023274 -0.10725679 -0.12667494]\n[ 0.05440604 0.03363384 0.12773462 -0.0225022 -0.11709606 -0.08246228]\n[-0.07478495 -0.01820215 -0.06289016 0.00661487 -0.02890409 0.07456721]\n[ 0.05315993 -0.00774269 -0.01374839 0.0367431 -0.00135519 -0.06750096]\n[-0.03885182 0.07873096 0.02371075 0.02000866 0.16509888 -0.07409362]\n[ 0.02051857 0.0620492 -0.00213286 0.10097467 0.06414621 -0.13120499]\n[ 0.06285419 0.05325737 0.04649446 0.14846779 0.04864058 -0.0202658 ]\n[ 0.13457911 -0.05248253 -0.05584732 0.0658607 -0.00522313 -0.05087093]\n[ 0.12267354 0.05394118 -0.06588896 -0.01165381 0.04335695 -0.0288139 ]\n[ 0.14390378 0.10954358 -0.04838676 -0.02016603 0.03786809 -0.05666351]\n[ 0.14462852 0.06843998 0.02745682 -0.05879001 0.04786909 -0.07078739]\n[ 0.12860374 0.0628712 -0.04283768 -0.02921913 0.08406699 0.00698227]\n[ 0.00307443 0.07492833 0.00517122 0.02832262 0.0014128 -0.03315274]\n[-0.0366415 0.18652259 0.09298028 0.0114279 -0.05174063 -0.01810233]\n[ 0.06353159 0.02525968 0.10372189 0.02731322 -0.09390891 0.06205925]\n[ 0.07241279 0.01047202 0.08987788 0.02021185 0.02901615 -0.00858951]\n[-0.06564526 0.02128737 -0.02443759 -0.07290538 -0.01963917 -0.05623203]\n[-0.10410903 0.04836389 -0.07412664 -0.01642562 0.01443276 0.02051037]\n[ 0.00953599 -0.03415989 0.00981615 0.04472892 0.06925561 0.10029602]\n[ 0.01554269 -0.03229945 -0.01203572 0.12803426 0.10587206 0.03710327]\n[ 0.11929016 -0.05155375 -0.00730276 0.05756969 0.05656901 -0.10001393]\n[ 0.13459331 -0.028423 -0.16193444 0.02671673 0.04575602 -0.06020094]\n[ 0.03446808 0.06116208 -0.10255814 -0.0269738 0.06089609 -0.05435224]\n[ 0.15360135 0.07809708 -0.07819074 -0.0441137 0.10515328 -0.04624659]\n[ 0.12455522 -0.00274724 -0.03359824 -0.06094872 0.06885184 0.01613074]\n[ 0.05241013 0.03465469 0.0544696 0.02889933 -0.07364102 0.01123893]\n[-0.01323363 0.05176437 0.02000826 0.11144416 -0.13222131 0.04050557]\n[ 0.01396541 0.13095745 0.02711691 -0.03186902 -0.10700434 0.05891183]\n[ 0.03830038 -0.10976595 0.05091108 0.07822291 0.00051992 0.07743204]\n[ 0.15484604 -0.01745577 -0.12746601 0.04840966 -0.00309741 0.03579752]\n[ 0.02922957 0.05066233 0.04701658 -0.05638792 0.05541473 0.03763079]\n[-0.04514857 0.02194637 -0.02062281 0.03265412 -0.07245879 0.08555832]\n[ 0.06741564 -0.08316622 -0.03128064 0.04865873 0.01566836 0.12975709]\n[-0.02963226 -0.10759581 -0.01098184 -0.02771188 -0.02080985 0.13046084]\n[ 0.03023411 -0.07968377 -0.01689306 -0.01264298 -0.10188069 -0.02617635]\n[-0.0503275 0.06735608 -0.08736092 0.05937959 -0.00524964 -0.0202806 ]\n[-0.0342137 0.10130821 -0.06394593 0.0444742 -0.00839805 -0.03509626]\n[ 0.03010297 -0.05151312 -0.04433104 0.03548456 0.03975218 0.007585 ]\n[ 0.00854354 -0.02776004 0.05805048 0.06378321 -0.08829919 0.03157675]\n[ 0.04181591 0.00249214 0.08059046 0.09673768 -0.19465317 0.04820249]\n[ 0.0700449 0.03535937 0.05061055 0.09118696 -0.21229593 -0.02301093]\n[ 0.03088937 0.01046988 0.03040583 0.04241165 -0.17937416 0.03388511]\n[ 0.11100613 -0.0860538 -0.04198904 0.1901321 0.00919448 0.04254381]\n[ 0.13026188 -0.11809216 0.03116922 0.14583004 0.05218115 0.03234338]\n[ 0.05639847 -0.0844645 -0.08791742 0.06192544 -0.03236663 0.07176448]\n[-0.03110873 0.03326483 -0.09342306 0.00719841 -0.02181715 0.13209145]\n[ 0.02926968 -0.06007582 0.01810193 -0.021138 -0.10274869 0.14875817]\n[-0.01407827 -0.03772659 0.00997541 0.00261242 -0.04543411 0.14778784]\n[-0.00864227 -0.01674397 0.01345735 0.03611076 -0.13224488 0.03311513]\n[-0.06951351 -0.0015919 0.02223685 0.01239971 -0.04158155 0.00971505]\n[-0.09682007 0.01379257 -0.00054124 0.00312394 -0.01066519 -0.03672417]\n[ 0.05549898 -0.08749791 0.04858857 0.06707751 0.01995481 -0.0036293 ]\n[ 0.09252531 -0.01970577 0.08262435 0.02110152 -0.10228407 0.05909617]\n[-0.05498303 -0.02594341 -0.0035772 0.05643748 -0.17565481 -0.02449322]\n[ 0.04412676 -0.06540321 0.10419527 0.11233973 -0.19555743 0.10897983]\n[ 0.15166388 0.01486791 0.01769827 0.12870763 0.00991944 0.00663956]\n[ 0.10962525 -0.00414004 -0.09437564 0.15096906 0.08112747 -0.01150652]\n[ 0.02475246 -0.03095607 -0.00947252 0.1118011 -0.02173202 0.07033352]\n[ 0.04599787 0.06665192 -0.06099332 -0.06268867 0.05021445 0.01119815]\n[-0.02999268 0.04205724 -0.11774008 -0.03122759 -0.04538905 0.16326749]\n[ 0.04934228 -0.0953588 0.01141531 0.06725168 -0.11418029 0.04892549]\n[-0.06355075 -0.1086745 -0.00195693 0.14663446 -0.07739544 0.10336433]\n[-0.02612478 -0.01453772 -0.03547798 0.044515 -0.02110003 0.10322534]\n[-0.05150209 -0.0180339 0.00939813 -0.01876376 -0.05246141 0.04234632]\n[-0.08762245 -0.11673234 0.05118603 0.01297363 -0.01001009 0.10524361]\n[-0.08567178 -0.14249459 0.13776262 0.01439468 -0.14940049 0.07299089]\n[-0.06279238 -0.10569572 0.09934897 0.04906239 -0.16361082 0.06625141]\n[ 0.05229718 -0.09052116 0.13404104 0.04930389 -0.17941342 0.04353089]\n[-0.02686829 -0.12129239 0.04753835 0.15620217 -0.10866011 -0.02461487]\n[ 0.03397669 0.06308611 0.01684876 0.11957835 0.01646716 -0.12103692]\n[ 0.13224643 -0.04496242 0.05914445 0.12491267 0.07710338 0.05791619]\n[ 0.0412988 -0.0768477 -0.06245616 0.04562742 0.06254521 -0.06498001]\n[ 0.03063324 0.03106824 -0.02559988 0.06763268 -0.0651234 0.05155628]\n[ 0.0418902 -0.07840691 0.02341092 0.035845 0.03472321 0.05307869]\n[ 0.01396064 -0.00728493 0.01153168 -0.07698955 0.05941084 0.13373546]\n[-0.06957011 0.07046435 -0.11192854 -0.06017029 -0.05490563 0.03300064]\n[-0.14343464 0.08020713 0.01546191 -0.04416481 -0.00722431 -0.01022952]\n[-0.09709038 0.01148857 -0.09649117 -0.05650258 -0.0684473 0.04289802]\n[-0.09163192 -0.00832356 -0.00961536 -0.04997683 -0.08259019 0.01100062]\n[-0.17413211 -0.03966004 0.00783409 -0.00765945 -0.00658657 -0.04461117]\n[-0.07055373 0.024131 0.03841666 -0.11819246 -0.166894 0.05523549]\n[-0.07620355 -0.07380182 0.01279752 -0.12449677 -0.01981781 -0.0294143 ]\n[ 0.04515965 -0.12187946 -0.09689888 -0.13681799 0.06321786 -0.12640236]\n[-0.04023914 0.00545566 -0.07380436 -0.03177956 0.04133968 -0.09878676]\n[ 0.03876712 0.04780795 0.06613705 -0.0798581 -0.07041597 0.02009777]\n[-0.01466717 -0.04095791 -0.06117681 0.017744 -0.02853148 0.02935289]\n[-0.06306019 -0.0588049 -0.07670173 -0.02984815 -0.04393519 -0.01071882]\n[-0.01691479 -0.03707276 -0.04221364 -0.03778722 -0.00076601 0.06874673]\n[-0.07801819 -0.06891399 0.06429972 -0.00189043 0.03712595 0.0220933 ]\n[ 0.02979842 -0.11590263 -0.14524511 -0.03976246 -0.0851257 0.03161072]\n[-0.00075146 -0.05422279 -0.16947828 -0.09824593 0.07455433 -0.09075713]\n[-0.03539257 -0.0077349 -0.07038279 -0.02245235 0.05691826 0.03691422]\n[-0.10431595 -0.06712238 -0.18116839 0.01729527 0.05096266 -0.01168822]\n[-0.00102106 -0.12141351 -0.15861224 -0.10181027 0.07429059 -0.10902476]\n[ 0.00775752 -0.04790719 -0.07350548 -0.06977136 0.08977472 -0.10640195]\n[-0.01784684 0.01781068 -0.10203142 0.04107929 -0.02521095 -0.05587351]\n[-0.05395092 0.01020161 0.06169144 -0.0250671 -0.01564677 -0.03906799]\n[ 0.04229081 0.01736699 -0.04419591 0.02952868 -0.06702921 0.0220938 ]\n[ 0.06377815 -0.01866848 -0.00872547 0.05797683 0.05467021 0.0118074 ]\n[-0.02050604 -0.0856597 -0.07597944 -0.00963104 -0.04709589 -0.00769084]\n[-0.00340328 0.07873099 -0.0219969 -0.05687838 0.06580792 0.00297902]\n[ 0.04022289 0.00317212 0.04879058 0.03770782 -0.03577699 -0.03475757]\n[-0.00706839 0.00817533 -0.03345953 0.08339515 0.06095733 -0.08682188]\n[ 0.02722257 0.02425286 -0.0345313 0.08355992 0.0522751 -0.0449243 ]\n[ 0.0099801 -0.01105665 -0.0232964 -0.02217771 0.03242119 0.06061608]\n[ 0.08399611 -0.02406195 0.02454117 0.07061335 -0.03615051 -0.02929047]\n[-0.03101282 0.02415061 -0.03012299 0.04266731 0.02002877 0.00579005]\n[ 0.08879535 0.02304467 -0.02114121 0.04518966 -0.06163095 -0.0645394 ]\n[ 0.0874828 -0.01495793 0.02253911 0.07239142 -0.00243468 -0.05371052]\n[-0.02884641 -0.04605388 0.08200453 0.04786541 0.01582255 -0.01059281]\n[-0.06646664 0.02130977 -0.02211558 -0.06502445 0.02839461 -0.00296204]\n[-0.0063191 0.02135059 -0.03310223 -0.0723725 -0.01088556 0.02342578]\n[ 0.03307476 -0.00205816 -0.06744242 -0.00648706 0.0194298 0.0109417 ]\n[ 0.07308613 -0.04531072 -0.06914005 -0.02438866 0.03880366 0.0083587 ]\n[ 0.02373182 0.04986393 0.0046019 -0.05207612 -0.00573088 0.07560703]\n[-0.00279333 -0.00783953 0.08252762 0.07336356 0.00140709 -0.0343411 ]\n[-0.04199142 -0.00874342 -0.03191872 0.05077265 -0.02291996 0.06468597]\n[ 0.03074639 0.00323415 0.07363073 0.01344377 -0.0344721 -0.07490882]\n[ 0.05864273 0.06454518 -0.06300256 0.03239371 -0.00249341 0.02659346]\n[-0.0344055 0.06376082 -0.02887102 0.08223609 -0.03262004 0.02870097]\n[ 0.01010473 -0.0090417 -0.01612562 0.08783603 -0.08671906 -0.02188782]\n[ 0.07311525 -0.02435566 0.06883435 0.03148881 -0.11076438 -0.04429966]\n[ 0.06303325 0.01264509 0.14135678 0.01972907 0.00872923 -0.07572689]\n[ 0.04516434 0.01409933 0.12629771 -0.07746307 -0.11591296 0.01541979]\n[ 0.11733777 0.05183874 0.02602149 -0.09596024 0.00783244 -0.01830325]\n[-0.05513093 -0.0005425 -0.05468076 -0.05329437 -0.02899629 -0.03885645]\n[-0.02990598 0.05032407 -0.02790986 0.01764168 0.07079449 0.06969906]\n[ 0.07217807 -0.0403676 0.04199512 -0.04355459 -0.00424795 0.03450686]\n[ 0.01203644 -0.07774626 0.05507602 -0.04055535 -0.07002122 0.01328743]\n[-0.04818821 -0.01758004 0.05088845 -0.00749813 0.04058031 0.10315325]\n[ 0.06443591 -0.05923147 0.02365677 0.00316905 0.03135337 0.13334037]\n[ 0.06809917 0.01315044 -0.01585105 0.04918423 0.00850209 0.00988356]\n[ 0.0736881 0.01316321 -0.09731589 0.01891903 0.00801433 0.12068653]\n[ 0.09107485 -0.05991253 -0.04163438 0.01160709 -0.10218006 0.13151526]\n[ 0.00759288 -0.08449205 -0.16844754 0.01386801 -0.09983791 0.08254603]\n[ 0.00546961 -0.12577212 -0.2370593 0.00632897 -0.05404766 0.21107449]\n[ 0.04833199 -0.15973663 -0.12746678 0.01278205 -0.00261998 0.09800395]\n[ 0.08713182 -0.03916506 -0.06439253 -0.0138543 0.05141735 -0.08583855]\n[ 0.05299447 0.14401291 0.01017158 -0.04149881 0.0276635 -0.0918515 ]\n[ 0.00038663 0.02805049 0.06326583 -0.07504413 -0.04173167 -0.10145886]\n[ 0.05976153 -0.01742465 0.06341676 0.05602439 -0.07059325 -0.09018681]\n[ 0.01108022 -0.03864975 0.02851808 -0.06720294 -0.0641618 0.0656224 ]\n[ 0.02856216 0.07055982 -0.0516327 0.06126515 -0.06335178 0.06992488]\n[-0.02264741 0.01859816 -0.08388671 0.01682108 -0.0259085 0.08928236]\n[-0.1091122 0.01057275 0.0226247 -0.03718273 0.00135818 -0.03461534]\n[-0.04477059 -0.07221311 -0.03108195 -0.00183363 0.01866931 0.03458804]\n[ 0.00115266 -0.06592103 -0.13075645 -0.07985996 -0.10431947 0.06145226]\n[-0.01576591 0.00800673 -0.09192831 -0.01888154 0.02111752 0.06716861]\n[-0.03456827 0.06472316 -0.21096472 -0.03178708 -0.04704238 0.12340169]\n[-0.00635043 -0.05227181 -0.09891172 0.05748896 0.02813034 0.09038568]\n[ 0.01646154 -0.15380681 -0.06830674 0.00648034 -0.06387944 0.01933319]\n[-0.06033361 -0.09666492 -0.0347028 0.06034534 -0.0975316 0.0525537 ]\n[-0.03687568 -0.09934153 -0.08847664 0.08158989 0.00141661 -0.01972735]\n[-0.00287533 0.04690229 -0.02634276 0.00471038 -0.01305853 -0.07336445]\n[-0.08552708 -0.08550128 0.14672536 -0.14150299 -0.09082367 -0.01985498]\n[ 0.02511594 0.03920835 -0.08183341 -0.00175885 -0.02925517 -0.06423401]\n[ 0.04209926 -0.08868414 0.04160867 -0.02163641 0.02538342 -0.07373472]\n[-0.1010308 -0.09864283 0.00304896 0.04347901 -0.06171658 0.0357643 ]\n[ 0.00675949 0.02017584 -0.11227395 -0.07224586 -0.06970908 0.02627237]\n[-0.01575015 -0.0068797 -0.02497365 -0.05909395 -0.11599399 0.00620197]\n[-0.00582996 -0.10990653 0.00341506 -0.07777496 -0.12476213 -0.09949174]\n[-0.1146011 -0.07505476 -0.09607791 -0.03088069 -0.16680256 -0.03121797]\n[-0.0869001 -0.0560614 -0.13494053 -0.02908197 -0.10780253 0.03288222]\n[-0.04528867 -0.10762697 -0.08480581 0.07087144 -0.02079575 -0.04462719]\n[-0.15111348 -0.16765293 0.02092774 0.0520954 0.08606736 0.00502715]\n[-0.07411221 -0.10411596 0.03319677 0.08239169 0.10579952 0.02374489]\n[-0.00763384 -0.12334859 0.08693864 0.10098316 -0.03707935 -0.00459614]\n[-0.07790284 -0.07505847 0.11518796 0.02719008 -0.0498294 0.02567977]\n[-0.01085752 -0.02108232 0.1166356 -0.07947682 -0.03691655 -0.08533392]\n[ 0.01284698 0.04319083 0.09651497 -0.11431589 0.01230511 -0.00311542]\n[ 0.01906602 -0.08105581 -0.05718938 0.03989894 -0.00492739 -0.07123473]\n[-0.01209347 0.05918923 -0.00647449 -0.02324463 -0.04660026 -0.0172042 ]\n[-0.03998321 -0.05774198 0.00606982 -0.01544447 -0.01927958 0.0078071 ]\n[-0.16435644 -0.04937918 0.0327552 -0.08888155 -0.04324656 -0.10150621]\n[-0.1195085 -0.09279716 0.05081519 -0.07332635 -0.01160537 -0.11926854]\n[-0.1242918 -0.0516049 0.09920215 -0.15690064 -0.01110534 -0.16299208]\n[-0.11686368 -0.03242994 -0.01932491 -0.10553797 -0.06686535 -0.1331961 ]\n[-0.14782013 -0.08558808 -0.0254719 0.05517821 -0.10985656 -0.06093979]\n[-0.12791874 -0.18109022 -0.01371315 0.1069378 0.01094283 -0.05869334]\n[-0.07214297 -0.1171359 0.04347658 0.11057278 -0.00463003 -0.08695465]\n[-0.12860602 -0.10304763 0.12716153 0.18252239 0.04018192 -0.07817193]\n[-0.01159176 -0.02519655 0.16372627 0.01329133 -0.02916689 -0.09670461]\n[-0.12614983 -0.07885019 0.18875229 -0.17262574 -0.08638599 0.03756535]\n[ 0.00875268 -0.10847837 0.06044701 -0.09786192 0.00160857 -0.02058518]\n[ 0.06909539 -0.04950211 -0.00288423 0.02066973 -0.05376922 -0.07762389]\n[-0.0743156 -0.0328237 -0.049174 -0.05797269 -0.07078674 -0.00918986]\n[-0.08920635 -0.0545126 0.00686068 -0.05586634 0.04931222 -0.03772788]\n[-0.10032107 0.05828014 0.06278351 -0.07885226 0.09543183 -0.19087875]\n[-0.11047766 0.00621466 0.07614939 -0.1074705 0.09285117 -0.13230141]\n[-0.10786755 -0.10159543 0.07893702 -0.16805226 0.11080868 -0.1021629 ]\n[-0.20107403 0.01537381 0.05700956 -0.21065055 -0.1344308 -0.0115444 ]\n[-0.18496245 0.08148756 -0.15201768 -0.13335752 -0.02321573 -0.011833 ]\n[-0.11402065 -0.06084957 -0.03581726 -0.03327167 0.03821984 -0.04533262]\n[-0.01624171 -0.0779098 0.04551651 0.01745802 -0.02982147 -0.12480359]\n[ 0.01769885 -0.04275391 0.15404399 0.16352718 -0.00707891 -0.2204005 ]\n[ 0.01117152 0.05379866 0.24829371 0.17850298 -0.02290346 -0.08285065]\n[-0.041535 0.02230919 0.11049987 -0.09049127 -0.13301167 -0.0365572 ]\n[-0.03630528 0.01687351 0.00366041 -0.0079483 0.01041406 0.03271633]\n[-0.04517176 0.06207936 0.03069755 0.06896499 0.01395736 -0.04435897]\n[-0.09688669 -0.08676144 -0.05385501 -0.12382041 0.08324644 -0.03318273]\n[-0.13742276 0.02406721 0.04640939 -0.01621931 0.06602018 -0.05023581]\n[-0.07473897 0.04026653 0.06028711 -0.10495933 0.1385542 -0.1276679 ]\n[-0.10651979 0.03429695 0.04538415 -0.10229702 0.08979802 -0.04196034]\n[-0.13032119 -0.04043769 0.11250041 -0.1974812 0.14628579 0.02734789]\n[-0.22818984 0.11965807 -0.03218153 -0.2079038 0.16446579 0.06393486]\n[-0.07006577 0.10467544 0.02071472 -0.18452962 0.06919638 0.03106326]\n[-0.03559869 0.02725672 0.03675464 -0.20351875 0.0232321 -0.04322459]\n[-0.05047734 0.0274868 -0.00218126 -0.08532499 -0.10196719 -0.06309893]\n[-0.04747172 0.14822915 0.23301753 0.04975881 -0.0221993 -0.13814203]\n[ 0.11581982 0.1191036 0.3581551 0.00564652 0.05509095 -0.15107015]\n[ 0.10870871 0.06199756 -0.00569341 -0.00700572 0.01480637 -0.01425133]\n[ 0.07747076 -0.09602888 0.04439135 -0.01127195 0.03467308 -0.07501526]\n[ 0.02426844 -0.05566151 0.00642693 -0.05767361 -0.08362609 0.02029405]\n[-0.09816962 -0.02658927 -0.01496939 -0.10540568 0.0186364 0.08163155]\n[-0.12812944 0.06017309 0.05948135 0.0422884 0.02789039 -0.04322787]\n[ 0.00902574 0.07005555 0.04662625 -0.00793589 0.11771545 -0.09565743]\n[ 0.00804521 0.05894405 0.09210335 -0.01404389 0.10033886 -0.00780966]\n[ 0.05670874 -0.05379384 0.08547706 -0.01229658 0.12814467 0.03154401]\n[ 0.07123813 0.15409248 -0.08224371 -0.16866289 0.06741117 0.09393467]\n[ 0.0585053 0.13926491 0.00874373 -0.17216343 0.01771168 0.05755534]\n[-0.08185123 -0.05465635 0.02915151 -0.10413721 0.05234181 0.07561123]\n[-0.04534185 0.03318586 0.09185749 -0.09925884 -0.00177414 0.09427547]\n[-0.00602874 0.0983023 0.16971703 -0.05565074 -0.06988478 -0.0038655 ]\n[-0.06408705 0.06344847 0.18636818 -0.01032426 -0.0274127 0.03033029]\n[ 0.05786566 -0.03625295 -0.00456099 0.07288183 0.06470373 -0.05072778]\n[ 8.11063126e-02 -5.44092171e-02 5.30217998e-02 -5.36613427e-02\n -9.26311463e-02 -5.50363911e-05]\n[ -3.53034362e-02 7.34884739e-02 -2.10967101e-02 8.20038840e-05\n -5.33712730e-02 9.64356810e-02]\n[ 0.05584727 -0.07564959 -0.15007757 0.05104481 0.02999229 0.12481848]\n[-0.0232985 -0.06109238 0.03707458 0.08997541 -0.02453282 -0.06030597]\n[-0.00600583 0.03801842 0.00207624 -0.00619094 0.08539897 -0.02032562]\n[ 0.05329441 0.01990469 0.07880558 0.03468892 0.08287051 -0.12279335]\n[ 0.17008583 -0.02420184 0.10956061 0.03707284 0.08192559 -0.13542266]\n[ 0.00689564 0.16682151 -0.0718221 -0.01286904 -0.01786523 -0.05141586]\n[ 0.01732595 0.02000139 -0.09730851 -0.13509241 0.01723163 -0.04194546]\n[-0.00428402 -0.08683665 -0.01248466 -0.1095397 -0.03168366 0.00840248]\n[ 0.03923729 -0.03976212 0.06191822 0.02809906 -0.00890674 0.06491708]\n[-0.00244019 -0.02708972 0.04302517 0.02165691 -0.11158521 0.06028541]\n[ 0.06191679 0.0346404 -0.01934946 0.05341758 -0.10207887 -0.02903006]\n[ 0.18077523 0.04433471 0.00292609 0.08369821 -0.06648031 0.00327894]\n[ 0.06779093 -0.02453737 0.03696441 0.10139662 0.0057073 0.06533521]\n[-0.08305582 0.00392416 0.03467768 0.04304846 -0.02512644 0.00148449]\n[ 0.0237844 -0.044475 -0.13727956 0.03278318 -0.02362916 0.23397957]\n[-0.09663746 -0.11980815 0.00401888 0.04943704 -0.07798276 0.11937261]\n[ 0.01367956 -0.04242942 0.05436371 0.06366228 -0.10846025 0.01615124]\n[ 0.04725935 -0.03958574 0.09791607 0.0699836 -0.05288411 0.01660228]\n[ 0.06723899 0.05828278 0.13796987 0.12947816 -0.0848249 -0.07867571]\n[ 0.09004601 0.11790036 0.07024731 0.03698653 -0.08884452 -0.05824902]\n[ 0.0608949 0.12994705 -0.03104216 0.04728473 -0.00590138 -0.06956135]\n[ 0.10000875 0.0184104 -0.05483288 0.05167157 -0.01115394 0.08897478]\n[ 0.09150535 0.02289857 -0.0336893 0.03180741 -0.0349074 0.06696738]\n[ 0.11248624 -0.12722918 -0.04288004 0.0402756 -0.02649304 0.07953005]\n[ 0.12766029 -0.04065711 -0.04956887 0.10743555 -0.02775034 -0.04729911]\n[ 0.15002055 -0.06368937 -0.10485335 0.0755353 -0.06773707 0.02015029]\n[-0.02768589 -0.03655818 0.02317311 0.00530616 0.08075526 0.02742437]\n[-0.05100085 -0.01623904 -0.02910977 0.02831373 -0.00286221 -0.06059257]\n[-0.01781596 -0.07093476 -0.05223223 0.06764248 0.05424672 0.09459551]\n[-0.0613246 -0.06580082 -0.01659304 0.02639491 0.01538513 0.02701483]\n[-0.00202238 -0.11767528 0.02013254 0.00271924 -0.05775877 0.08950236]\n[-0.02002476 -0.02891426 0.05198368 0.04681492 -0.141717 0.04950737]\n[-0.00093731 -0.00549905 0.07312111 0.04672351 -0.0818921 0.02488073]\n[ 0.04781969 0.06862619 -0.00383235 0.00413803 -0.07412501 -0.01267266]\n[ 0.05001335 0.13050383 -0.05133763 0.08162124 -0.04806522 -0.08597045]\n[ 0.04842478 0.0747736 -0.0505995 0.03143326 0.00115336 -0.07796593]\n[ 0.08966991 0.00203328 -0.15589206 0.03598305 0.08351294 0.04372056]\n[ 0.06886587 -0.06364698 -0.1483838 0.10187615 0.07967211 0.00440404]\n[ 0.1710425 -0.09286346 -0.08643477 0.07932941 0.00098841 -0.02698534]\n[-0.02433348 0.04172625 0.05226521 0.00533783 0.05660186 -0.03382249]\n[ 0.02411989 0.06922956 0.00014566 -0.03333139 -0.01545451 -0.03078027]\n[-0.07563546 -0.0329041 -0.0369438 0.01354204 -0.04418798 0.07539021]\n[ 0.05991993 -0.01107481 -0.07549372 -0.05445573 -0.03596645 0.16255426]\n[-0.01958104 -0.09585484 0.01032085 -0.00696564 0.00264835 0.00019101]\n[-0.05710143 -0.07729269 0.01816921 0.12392949 -0.05271572 0.03375493]\n[-0.06439681 -0.04730009 0.01827502 0.01467785 -0.06931528 0.06886269]\n[-0.12057208 -0.04666622 0.05454067 -0.00199253 -0.02283627 -0.01716528]\n[-0.08040415 -0.02855227 0.01631452 -0.01662366 -0.02278931 -0.01928622]\n[-0.12489829 -0.00914164 -0.03296525 -0.08766359 -0.06126074 -0.08506618]\n[-0.08119691 0.07733595 -0.01320531 -0.05800965 0.06421506 -0.12369088]\n[ 0.02596982 0.07830025 -0.08262688 -0.04046429 0.10851005 -0.03363471]\n[-0.00607743 0.03464349 -0.00704505 0.06975123 0.17377134 -0.04914957]\n[ 0.04593702 -0.08941098 -0.08495176 -0.02143899 0.03919432 -0.02518678]\n[-0.03638298 -0.0792419 -0.08857468 0.05829025 0.02684897 -0.04104811]\n[-0.00682425 -0.08390704 -0.0186637 -0.06851941 0.00654578 -0.0746432 ]\n[-0.01915401 -0.0474208 0.05034585 -0.03819183 -0.01545872 0.0060174 ]\n[-0.00467668 -0.07371803 0.04443135 0.02383116 -0.01439601 -0.05091343]\n[-0.0306581 0.01656961 0.04239885 -0.09936392 0.03832189 0.10983934]\n[-0.10524937 -0.09753571 -0.01143074 -0.10657636 -0.03071666 0.01464091]\n[-0.05763919 -0.04856155 -0.02525212 -0.17501558 -0.02490427 0.01284005]\n[-0.13940194 -0.11334521 -0.0966932 -0.03793513 -0.05040701 0.01652947]\n[-0.15320279 -0.06436833 -0.0271381 -0.02419725 -0.03192935 0.00675967]\n[-0.128921 -0.246473 -0.00221251 0.01958969 -0.04113596 -0.06354882]\n[-0.00943925 -0.13827267 -0.10736376 -0.11224121 -0.03527088 0.01452933]\n[-0.05145054 -0.07337625 -0.09671575 -0.12290615 0.08592106 -0.05064091]\n[-0.12557201 -0.06504656 -0.11401266 0.01201283 0.10882793 -0.045607 ]\n[ 0.00047309 0.05889156 -0.07642809 -0.05545363 0.04504377 -0.06428661]\n[ 3.79813090e-02 7.34318001e-03 -3.93701484e-05 4.99136522e-02\n -4.93356064e-02 2.73512211e-02]\n[-0.03118901 -0.01200502 -0.01040418 0.03281386 0.07756227 0.0802449 ]\n[ 0.01769503 0.05020934 -0.04753393 0.05950855 -0.07089911 -0.04519154]\n[ 0.0443761 0.00151219 -0.07124611 -0.00134795 -0.02688657 0.06165159]\n[ 0.04120163 -0.02048206 -0.00353376 0.02275546 0.04434854 -0.01490622]\n[ 0.08151054 -0.0254981 -0.04692166 -0.07468412 0.06345793 0.0210215 ]\n[-0.03123961 -0.08669706 0.0343455 0.07274178 0.01763093 0.0685271 ]\n[-0.06190425 0.01233341 -0.05290114 -0.03021464 -0.07366998 -0.02637992]\n[ 0.10249367 -0.02711898 0.04624803 0.06023959 0.00280379 0.05387839]\n[-0.00551062 -0.01144353 0.00691785 0.07808731 0.00190173 0.03206439]\n[ 0.08520673 0.04504014 0.00060269 0.01337086 -0.02961163 -0.00405518]\n[ 0.08741169 -0.00704759 0.10708498 -0.05164173 -0.01963125 -0.08817019]\n[ 0.05587859 0.04446787 0.04108514 0.0769354 0.04998524 -0.06743234]\n[-0.02597478 0.04292867 0.04225266 -0.01232554 0.00249567 -0.04766686]\n[-0.04557613 -0.06089809 0.00499027 0.01751164 -0.03621696 -0.03983154]\n[ 0.01469613 -0.04527127 -0.06079877 -0.00719674 -0.07625341 -0.01494268]\n[-0.08249182 0.04329181 0.07580782 -0.07192215 -0.05699712 -0.08432283]\n[ 0.07269385 0.03542061 -0.00866005 0.07044493 0.00493744 0.03871492]\n[-0.00480785 0.04063975 0.04584704 0.03682789 -0.07058839 -0.06206352]\n[-0.0248164 0.07458694 -0.07981014 -0.00428526 0.05047368 -0.03993575]\n[-0.05692981 0.01625342 -0.12072356 -0.03726922 -0.03386045 0.00283743]\n[-0.02498297 0.07712871 0.01916334 0.03441772 0.06204141 -0.08501631]\n[-0.1055309 0.04614376 0.0120049 0.12080041 -0.04429146 -0.03026236]\n[ 0.04406772 0.02364833 0.08039239 0.03211306 -0.06560377 -0.0514333 ]\n[-0.06159844 0.17557676 0.03713775 0.05793579 -0.05753848 -0.04570324]\n[-0.06796383 0.2176924 0.02069011 0.00361046 -0.08295631 -0.05174478]\n[ 0.02155103 0.1026372 -0.00995247 0.03972985 -0.08807784 -0.06252069]\n[-0.09874402 0.04289055 0.05443001 -0.05626391 -0.03924071 -0.00148924]\n[ 0.05142415 0.10364208 -0.06108316 0.04274806 -0.04291839 -0.02966257]\n[ 0.07196663 0.00560528 -0.04168062 0.00132644 0.07660192 0.00333562]\n[ 0.00013542 0.03840497 -0.05121634 -0.03069201 -0.06030351 -0.03031237]\n[-0.04191826 -0.05238188 0.03718313 -0.08851282 0.05312361 0.09282121]\n[ 0.02798723 -0.0438569 -0.02080902 -0.01092144 -0.05213049 0.01647401]\n[-0.0165485 -0.0365198 -0.09573365 0.04300163 -0.09133686 0.04295415]\n[-0.00216795 -0.09183475 0.04201121 0.07137991 -0.02369085 0.06095942]\n[-0.01534411 0.01977289 -0.00932657 0.07173731 0.04840126 0.04417225]\n[-0.074004 -0.16776744 -0.07289654 0.06248488 -0.04314394 0.04261868]\n[-0.03569035 -0.08790234 -0.00229335 0.01375446 0.02520715 0.07956246]\n[-0.04862782 0.06932572 -0.01669175 -0.01396045 -0.04241837 0.03965073]\n[ 0.05289885 0.117036 0.03062664 -0.03531889 0.04497956 0.01854477]\n[-0.00769493 0.04494339 -0.03419326 -0.04736501 -0.04951311 -0.03950291]\n[ 0.02906445 -0.04711517 -0.0454364 -0.07976387 0.04484801 0.03170849]\n[-0.04130537 -0.03226221 -0.03359691 -0.08511429 -0.00897318 -0.0983621 ]\n[ 0.03924594 -0.02916579 -0.06255648 -0.02740895 -0.0377712 0.05868147]\n[-0.0329107 -0.01577795 -0.00565499 0.01373254 -0.00711384 -0.08724231]\n[-0.00495188 -0.00154927 -0.03164452 0.05839494 -0.01135823 0.07375232]\n[-0.04731026 0.01787225 -0.12093417 0.03104019 0.0728953 0.09238551]\n[ 0.0130134 -0.06200435 -0.10512212 0.05057555 -0.06693162 0.10173878]\n[ 0.05626862 -0.18613601 -0.23153642 -0.03318107 0.04771827 0.1001261 ]\n[ 0.06938198 -0.18837951 -0.189422 0.05777368 0.00080361 0.16981642]\n[-0.00789124 -0.20842065 -0.17656729 0.01014619 -0.1255115 0.1304343 ]\n[-0.04647372 -0.26197448 -0.08452164 0.02480382 -0.16322449 0.02428417]\n[ 0.02942501 -0.23362935 0.06261629 0.0492407 -0.06489585 -0.02855127]\n[ 0.10123306 -0.0937614 0.11829423 -0.05942305 -0.0638321 -0.00767653]\n[ 0.03515032 -0.06457788 0.27184325 -0.08996352 -0.16159408 -0.01089757]\n[-0.07372177 -0.0090925 0.15810782 -0.1326541 -0.03745582 -0.00502236]\n[-0.02260247 -0.06947938 0.15360887 -0.00746727 -0.05963999 -0.11189066]\n[-0.09204135 0.02806653 -0.01532012 0.03676529 -0.00136964 0.05402471]\n[-0.04058576 0.06207782 0.00441554 -0.06069969 0.06625783 -0.08017676]\n[-0.03886466 0.06770531 -0.10361184 -0.07814255 0.06383438 -0.06547377]\n[-0.07178563 -0.06216462 -0.0826185 -0.01841482 0.00875618 0.08441661]\n[ 0.00514941 -0.13517484 -0.14302835 -0.05045532 -0.16351539 0.03881734]\n[ 0.06038361 -0.10357808 -0.26816589 0.02287425 -0.10810837 0.10897718]\n[-0.00469722 -0.16116744 -0.23600103 -0.01966639 -0.06966556 0.09714681]\n[ 0.00323771 -0.26162848 -0.12383876 -0.01363802 -0.01340331 0.07708447]\n[ 0.03174062 -0.31414208 0.01348246 -0.01330318 -0.04444208 0.08569098]\n[-0.0372158 -0.29225239 0.08628132 -0.01368504 -0.0604109 -0.00511595]\n[-0.03769875 -0.13700223 0.16922884 0.02004614 -0.1250495 -0.10593883]\n[ 0.00898364 -0.05084655 0.23351337 -0.06909657 -0.18425065 -0.07234947]\n[ 0.01060263 -0.07058497 0.27787721 -0.23051545 -0.13470556 -0.03363731]\n[ 0.05730317 -0.09995969 0.20089324 -0.17654829 -0.08293732 0.04954606]\n[ 4.92709056e-02 3.30025293e-02 -4.64077712e-05 -9.06658247e-02\n -1.36768771e-02 -6.39049038e-02]\n[ 0.01839902 -0.06888157 -0.03553672 0.0634606 0.01123822 0.00511846]\n[ 0.04469841 0.02552879 -0.03906132 -0.07057896 -0.00561385 -0.05002847]\n[-0.0266485 -0.04922157 -0.12058857 -0.03268656 0.00513991 0.06110254]\n[ 0.01578491 -0.09232703 -0.12197913 0.03432452 -0.07400176 0.13910373]\n[ 0.0259084 -0.13964759 -0.27677381 0.05767421 -0.15414582 0.19262943]\n[ 0.02410723 -0.20292684 -0.20245947 -0.02236412 -0.10315355 0.16249098]\n[ 0.0020604 -0.1993482 -0.19578572 0.02237512 -0.02299276 0.11855898]\n[-0.01057898 -0.33508417 -0.06773041 -0.00834153 -0.08541429 -0.0064417 ]\n[-0.04980646 -0.15825674 0.09098524 0.01356584 -0.11530922 -0.05147254]\n[-0.0138434 -0.14391924 0.18955825 -0.00921918 -0.1295817 0.0038868 ]\n[-0.11109211 -0.05249695 0.19168107 -0.05689886 -0.09197777 0.07958837]\n[-0.11979652 -0.05910861 0.26285806 -0.1363706 -0.13857168 0.03443174]\n[-0.07265623 -0.08622662 0.28751302 -0.08210479 -0.12759681 0.02263789]\n[-0.08330109 -0.09738417 -0.01186525 0.01900351 -0.03624653 -0.0469721 ]\n[-0.06011249 0.00646945 -0.02051205 -0.08167917 -0.00371891 -0.02599418]\n[ 0.00410909 -0.08946645 -0.05230439 -0.07754093 -0.03394083 0.04376241]\n[-0.08882749 -0.05604849 -0.06795415 -0.03447219 -0.06124071 0.03394322]\n[-0.01954887 -0.15327506 -0.11060452 0.07712906 -0.06319876 0.00760951]\n[ 0.00530279 -0.192081 -0.15943152 -0.00636008 -0.06855785 0.09275523]\n[-0.00188405 -0.17166963 -0.20277663 -0.05625352 -0.12082655 0.01291196]\n[-0.07836711 -0.20520724 -0.11228264 -0.01602561 -0.17870305 0.1060649 ]\n[-0.10098472 -0.2474573 0.08438078 -0.02799721 -0.20860061 0.06785383]\n[-0.05343942 -0.09845677 0.15171385 -0.01145344 -0.17055483 0.04457689]\n[-0.03875558 -0.12813036 0.15463637 -0.03472103 -0.14800641 -0.0263137 ]\n[-0.0291168 -0.03924006 0.16405782 -0.09091195 -0.05825635 0.02404475]\n[-0.12416057 -0.06044127 -0.01125654 -0.04537313 -0.06262229 0.13447532]\n[-0.09889081 -0.1556879 0.02548652 -0.02218553 -0.05300447 0.01801485]\n[ 0.06525732 0.03582808 0.10920662 0.0644175 -0.0022815 -0.07423581]\n[ 0.07823211 -0.02774064 -0.03544968 0.06075738 0.0553913 -0.05242233]\n[ 0.00123041 -0.08780491 -0.03489149 -0.08350885 -0.04109978 -0.0534561 ]\n[-0.1149976 0.0177169 -0.10534903 -0.04952307 0.02255956 -0.05585439]\n[-0.0871532 -0.12651753 -0.02503573 -0.04426679 -0.10098818 -0.12002897]\n[-0.11770146 -0.12337702 -0.13315602 -0.13782211 -0.03842288 0.03779097]\n[-0.15434819 -0.19969378 -0.0533765 -0.09697416 0.03262857 0.13533725]\n[-0.15810129 -0.03373437 -0.01062032 -0.11541918 0.09179009 0.17744531]\n[-0.02556911 -0.02208714 -0.02074615 -0.17632903 0.13939235 0.05378716]\n[-0.00859233 0.0665143 0.08448694 -0.15354057 -0.11360283 0.07342695]\n[ 0.03170392 0.01226973 0.06161458 -0.10677943 -0.09538125 0.02955556]\n[ 0.01637571 0.02291869 -0.00486508 -0.10978864 0.03196978 -0.00119864]\n[-0.01123583 -0.10243765 -0.10375879 0.03156911 -0.00312853 -0.10900303]\n[ 0.01390369 -0.04178841 -0.08232404 -0.01377539 -0.04055586 0.06383023]\n[ 0.08507586 -0.06774803 -0.04459132 0.03536529 -0.06980737 -0.08377846]\n[ 1.83119476e-02 -2.20378788e-05 -4.70558703e-02 -7.15567619e-02\n -5.27981557e-02 -9.16424915e-02]\n[-0.03484254 0.04082989 0.00369357 -0.10849848 0.03957698 -0.05245356]\n[-0.09881279 0.01745884 -0.11592711 -0.09311091 0.10624602 -0.08507068]\n[-0.11038984 -0.06384718 -0.16289976 -0.1089168 -0.03206481 0.03839753]\n[ 0.02752457 -0.00907848 -0.12478393 -0.10536714 0.07663055 0.11762825]\n[-0.08858506 -0.06922511 -0.05703901 -0.04144585 0.10630862 0.13778536]\n[-0.02427102 -0.0106544 -0.08189183 -0.11311279 0.13016868 0.1357471 ]\n[-0.02505998 0.08577369 -0.00567453 -0.14956892 0.02537528 0.12645851]\n[ 0.06954842 0.10040838 0.00845767 -0.18121123 -0.11243245 0.05786859]\n[-0.04285837 -0.02493123 -0.0229837 -0.16576533 -0.03796538 -0.05438682]\n[ 0.05705964 -0.04449693 -0.11074064 -0.04516108 -0.01780244 -0.12226619]\n[ 0.03289556 0.04605675 -0.04799105 0.04821578 0.0250967 -0.08030977]\n[ 0.06883781 -0.07500488 -0.11142156 0.17076261 0.00628945 -0.01696148]\n[ 0.06842478 -0.11456108 0.0262215 -0.01819566 0.06242902 -0.03736234]\n[-0.08205154 -0.07980271 -0.02614263 0.00192974 -0.05448024 -0.05101029]\n[-0.07054337 0.02778933 -0.05840603 -0.02397775 0.06783634 -0.02285083]\n[ 0.03394033 0.02750584 -0.06408812 -0.06291281 0.03478998 -0.02012466]\n[ 0.00486957 0.0586816 -0.11000957 -0.11570497 0.14010166 0.01395124]\n[-0.01512818 0.03020184 -0.08333265 -0.12020911 0.09289031 0.11576611]\n[-0.04718693 -0.10320611 -0.12724544 -0.13121018 0.16500162 0.16109243]\n[ 0.0115124 -0.02343783 -0.11267095 -0.15439481 0.16024344 0.12133294]\n[-0.00924116 -0.03999001 -0.02101704 -0.05275147 0.05602563 0.15410069]\n[ 0.049056 0.01651078 0.00811297 0.04126582 0.02335404 0.05333965]\n[ 5.76205105e-02 9.63858794e-03 8.02360773e-02 5.70908412e-02\n 1.22992657e-02 -6.92026006e-05]\n[ 0.14396952 0.00972819 -0.04323345 0.06040709 0.04499682 0.01710355]\n[ 0.08502851 0.08263845 -0.08156404 0.06395511 0.00099856 -0.10946572]\n[ 0.10113055 -0.10002855 0.01485239 0.09571408 -0.01877613 0.03666028]\n[ 0.01146812 -0.06143152 -0.09492616 0.12954885 -0.00065532 -0.04009742]\n[ 0.01266196 0.0819984 0.047096 -0.02935063 0.02346306 -0.0196944 ]\n[-0.05512476 -0.01332361 -0.06817065 0.02236891 -0.10267596 0.09189562]\n[-0.083919 -0.03605526 -0.02932199 -0.02719379 0.03852296 0.07188715]\n[ 0.00600483 -0.02655954 0.06322473 -0.15873766 0.04515571 0.09676984]\n[-0.01862936 0.03461575 0.03958502 -0.02610168 0.01909079 0.11907221]\n[ 0.05270803 -0.01999879 0.08816964 -0.03710591 0.03879414 0.11751094]\n[ 0.05399841 -0.02176004 0.02660166 0.07033206 -0.00021014 0.06643411]\n[ 0.08539724 -0.04394952 -0.01331997 0.04805927 -0.00060164 0.12453128]\n[ 0.15583102 -0.01959829 0.11361287 0.13277112 -0.05170732 0.08145317]\n[ 0.17910284 0.05089704 0.00528917 0.11413533 -0.05781568 0.10377657]\n[ 0.08245552 0.10090733 -0.03479505 0.07159761 -0.05795054 0.01136379]\n[ 0.0389607 0.05964078 -0.07295477 0.1281978 -0.02180543 0.04948157]\n[ 0.1229486 -0.07319279 -0.05997281 0.0844834 -0.10655858 0.0894669 ]\n[ 0.10541876 -0.05354195 0.00028371 0.08102857 0.04574807 0.05318305]\n[ -6.65398408e-03 7.25593865e-02 -2.95798054e-05 -2.36942172e-02\n -1.78858750e-02 -5.77889122e-02]\n[-0.05108573 -0.12077842 -0.00262608 0.04244742 -0.05529458 0.04374898]\n[ 0.01275766 -0.08263015 -0.00399919 0.06124596 0.02406947 0.07931511]\n[-0.02696937 -0.12176222 0.03671923 0.00481847 -0.1173773 0.06500722]\n[ 0.00104482 -0.02869925 0.02321416 -0.00014381 -0.08755617 0.08489241]\n[ 0.03704666 -0.00397471 0.06798993 0.05955146 -0.12600242 -0.00396218]\n[ 0.0788386 0.01444275 0.10939837 0.01515214 -0.15604103 0.00290358]\n[ 0.09905066 0.01313802 0.08741435 0.0919837 -0.13316198 0.04869229]\n[ 0.13331567 0.03671496 0.07464004 0.01444172 -0.09916789 0.03276154]\n[ 0.08488057 0.00595172 0.00338763 0.12884003 -0.15842536 0.03127812]\n[ 0.07457466 -0.08717185 -0.04588981 0.08657996 -0.05788819 0.08176958]\n[ 0.07824977 -0.06394845 -0.04273864 0.10819834 -0.00835586 0.01183544]\n[ 0.02201765 0.02059264 0.00864112 0.06972165 0.05695596 0.08902124]\n[-0.01544811 -0.04681647 -0.04810363 0.10080346 -0.02184247 0.02373272]\n[ 0.01104253 0.02739184 0.08262788 0.02268107 0.04819253 -0.00672995]\n[-0.07060029 -0.0334276 0.02591504 -0.00728519 0.04244732 0.02189313]\n[ 0.01874634 0.03752931 -0.12131198 0.05733044 -0.1290354 0.10302898]\n[-0.0681495 0.04893824 0.01150581 -0.02956781 -0.12430092 0.04321539]\n[ 0.01895567 -0.03944237 0.07842116 0.05792803 -0.0446355 0.06940101]\n[-0.02470793 0.02999764 0.0454248 0.02464112 -0.03739325 0.08014808]\n[-0.09868927 -0.01630697 0.08584633 0.05076698 -0.0462268 0.09092548]\n[-0.00414749 0.07093184 0.02632718 0.07258747 -0.07827809 0.04215058]\n[-0.01547547 -0.00078162 0.04392213 0.0094031 -0.0615246 0.02153521]\n[ 0.01396228 0.08888188 -0.01357533 0.02293648 0.00327463 -0.02963009]\n[ 0.08573545 0.00517351 0.00172319 0.15778035 0.05273307 0.04189184]\n[ 0.02804283 0.02621911 0.00711008 0.03531394 0.10086582 0.05177085]\n[-0.03802873 -0.04024641 0.02080545 0.10669859 -0.02659208 0.08730263]\n[-0.00162489 -0.01129435 -0.02461695 0.00053571 0.003333 0.02515261]\n[-0.05390078 0.13838756 -0.03254241 -0.07774678 -0.01354587 -0.06902446]\n[ 0.00385146 0.00226919 -0.03537548 -0.05522361 0.02386935 0.01565781]\n[-0.01194499 0.06804617 -0.08183085 0.01053591 -0.087484 -0.04261578]\n[-0.05986556 -0.01158776 0.00818894 -0.02896479 0.04920187 0.02398697]\n[-0.16522004 0.01502649 -0.03463824 -0.14270571 0.03288426 0.03527961]\n[-0.0461283 -0.10288361 -0.00793323 0.00420822 -0.02979923 0.03530048]\n[-0.15853932 -0.06390239 0.02377473 -0.04934523 -0.00271464 0.01473798]\n[-0.11196984 -0.01549585 0.00038989 -0.05385035 -0.06825149 -0.0008967 ]\n[-0.13769379 -0.02528614 -0.04569443 -0.02795361 0.00405588 -0.08145855]\n[-0.05022296 0.01584918 -0.01828052 -0.0534338 0.02049299 -0.04450252]\n[-0.0434696 0.04685648 0.0577606 -0.09859888 0.0667012 -0.12948473]\n[ 0.02107473 0.10285812 -0.07627611 -0.03532748 0.11291386 -0.11796299]\n[ 0.01511768 0.01312367 -0.03995832 -0.0690352 0.06163998 -0.08577722]\n[-0.04931452 0.0947738 0.02728596 -0.0716171 -0.08045644 -0.0952751 ]\n[ 0.01380203 -0.06715139 -0.00573948 0.02941873 -0.02511296 -0.06128195]\n[ 0.02050374 -0.05374135 -0.00234578 0.0395844 -0.00243795 0.08625361]\n[ 0.04770144 0.08085485 -0.07950226 0.01149448 0.05483034 0.08702312]\n[ 0.01026897 0.05670335 0.087634 0.10011385 0.04838173 -0.04394186]\n[ 0.14458971 -0.07542491 0.02961972 0.03574933 -0.1111038 0.03269218]\n[ 0.11596359 -0.03247185 0.05235395 -0.0600836 -0.06662521 0.0349807 ]\n[ 0.10323525 0.01158974 0.1745612 0.00085632 -0.13926363 0.02639571]\n[ 0.00666897 0.07704429 0.11052977 0.04496747 -0.12271487 -0.01487018]\n[ 0.08590381 0.06928587 0.13328457 -0.03483072 -0.16092479 -0.08771002]\n[ 0.15286379 0.06087631 0.14771245 -0.09948061 -0.07283711 -0.13333012]\n[ 0.20099333 0.09366754 0.08945104 -0.05917395 -0.08368565 -0.00934734]\n[ 0.04065054 0.00358804 -0.01808859 -0.02077919 0.0212387 -0.08033948]\n[ 0.08313444 0.08640381 -0.00160073 -0.02923647 -0.03433564 0.06350448]\n[-0.02020158 -0.05308037 0.02985067 -0.03921992 -0.03118743 0.03228926]\n[-0.06683315 0.03690299 -0.00696427 0.03112697 -0.02011655 0.01792667]\n[ 0.05010377 -0.07447167 -0.00803838 -0.01391037 0.05990709 -0.04307606]\n[ 0.00203963 -0.00067048 0.05554594 0.07511309 0.06928112 -0.04575794]\n[ 0.03747524 -0.06519578 -0.04098183 -0.040111 0.00267218 0.0914906 ]\n[ 0.17611992 0.0696538 0.07026698 0.08044767 -0.04419657 0.10615183]\n[ 0.06416123 0.0410319 -0.05785089 0.09637652 -0.0672601 0.07314874]\n[ 0.05260746 0.04596101 -0.00825373 0.00708481 -0.1175796 0.06192071]\n[ 0.07304839 0.14654826 0.0156169 0.05453049 -0.1017371 0.02541858]\n[ 0.02187769 0.07340283 0.08459067 0.025456 -0.16767789 0.00702915]\n[ 0.09360152 0.17671131 0.08358616 -0.07923687 0.02500546 -0.04286424]\n[-0.0473164 0.1387646 -0.00739351 0.00696312 -0.06374584 -0.07322901]\n[ 0.00139427 0.1119215 0.06723893 0.03898719 0.07849869 -0.04315067]\n[-0.06237566 0.1041767 0.075775 -0.08432952 0.00538813 -0.08120499]\n[-0.05326359 0.04191865 -0.05832578 -0.0892785 -0.01206145 -0.02470583]\n[ 0.02981657 0.0460008 0.04505556 -0.06547883 0.06372488 0.03433738]\n[ 0.02344203 0.07300235 0.03939307 -0.03525153 -0.00249784 0.06868308]\n[-0.02707121 -0.01241245 0.03510102 -0.10612556 -0.05960274 -0.05593782]\n[ 0.03699509 0.06634322 -0.0383116 -0.0911908 -0.02236472 0.08784393]\n[-0.03526923 -0.00932716 0.03537239 -0.02251676 0.08861982 0.03043855]\n[-0.02106779 0.10961135 0.07732417 -0.02145649 0.0029918 -0.02276084]\n[-0.01878962 0.07995333 -0.02555212 0.05929609 -0.03645477 -0.03330115]\n[-0.00434183 0.05053777 0.01905963 0.03820831 0.00653686 -0.00958953]\n[-0.05529717 0.02882817 0.04341602 -0.02230118 0.05950479 0.02378485]\n[-0.04353496 0.04051781 0.00497058 0.01333212 0.12444106 -0.06053591]\n[-0.03716992 0.05778565 -0.03715798 -0.13301019 0.10687227 -0.08828854]\n[-0.04016401 -0.00630084 0.00936579 0.01140711 -0.03411772 0.00126085]\n[-0.06845575 0.0690188 0.09891833 -0.13741623 0.00302645 -0.05983498]\n[-0.10868749 -0.04376632 -0.02390319 0.01988708 0.03176391 0.00045184]\n[-0.04874357 -0.00131884 0.06689507 -0.00192092 -0.0231897 -0.04948901]\n[-0.04148264 0.03610456 -0.07548529 0.02080825 -0.10622376 -0.00255498]\n[-0.13710162 -0.04210557 -0.02226113 -0.07067752 0.01055378 0.01697848]\n[ 0.02569921 0.06085956 0.09240148 -0.02902308 -0.00673824 -0.05969966]\n[-0.00299338 0.04747175 0.11120199 -0.0211431 -0.04272969 -0.00016657]\n[-0.02298886 0.00853862 0.06887483 -0.04512478 0.0101483 -0.08736951]\n[-0.0625679 0.06749941 -0.01505921 0.07652339 -0.02619012 -0.06052216]\n[ -2.54116729e-02 8.55673701e-02 -8.18568049e-04 -1.87312253e-05\n 6.77334964e-02 -1.17370918e-01]\n[-0.12737264 0.17444804 0.00369026 -0.09669803 0.10768119 -0.10697158]\n[-0.14050037 0.18770652 -0.09177223 -0.02691669 0.01427296 -0.09950489]\n[-0.10784652 0.11378074 -0.03330221 -0.13316873 0.08811253 0.00634981]\n[-0.03255222 0.07455693 0.08990514 -0.02002538 -0.08740292 -0.01427763]\n[-0.13758343 0.00388414 0.05839653 -0.06891268 -0.09607254 0.00150393]\n[-0.00805338 -0.00641551 0.10732695 0.00825777 -0.09013713 -0.09018692]\n[ 0.00634707 0.03748396 -0.04069734 -0.01871664 0.02729673 -0.07527903]\n[-0.1054848 -0.05112318 0.00019624 0.002072 -0.06174811 -0.06164182]\n[-0.13917063 -0.06504253 0.03217362 0.08001704 0.01362474 -0.00809617]\n[-0.1513183 0.03943324 0.09324314 0.02170712 0.0524634 -0.12414398]\n[ 0.01536033 -0.02915555 0.06137927 -0.04326545 0.05121922 -0.05635652]\n[-0.01775774 0.00028998 0.15172511 -0.04325743 0.08811694 -0.10613346]\n[ 0.02413365 0.10830846 -0.0262336 -0.07824454 -0.05339748 -0.05170796]\n[-0.00627984 0.19934681 -0.13096158 -0.14640474 -0.05769848 -0.03953461]\n[ 0.01103548 0.16765593 -0.09106661 -0.14263006 0.11891822 -0.07948778]\n[-0.05621858 0.13540868 -0.10386599 0.02738478 0.04738396 0.00300363]\n[ 0.03961071 0.05910172 -0.01274582 -0.05281831 0.0127644 -0.08616414]\n[-0.0197908 0.04225454 0.02377734 -0.03850439 -0.17292088 -0.13920163]\n[-0.07787354 -0.11750272 0.10362615 -0.05911998 -0.1646657 -0.1171345 ]\n[-0.05528364 -0.06021412 0.05757664 0.03915655 -0.02556406 0.00693995]\n[ 0.03379917 -0.00753838 -0.00580442 0.05051072 0.06037804 -0.05273503]\n[-0.01339275 -0.10471368 0.00201666 -0.04782136 0.0287921 -0.07339031]\n[-0.06092492 0.00549107 0.05160923 -0.02539773 0.01820634 -0.12731265]\n[-0.07634158 -0.00296087 0.07508548 -0.03017025 0.04381514 -0.20406139]\n[-0.03333542 0.11771797 0.10484347 -0.03779281 0.1043502 -0.15738863]\n[ 0.00725201 0.00533028 0.12898615 0.01564245 0.100384 0.0501099 ]\n[-0.0838731 0.2192701 0.04951849 -0.16150431 -0.08024314 -0.00244305]\n[-0.05749775 0.13449143 -0.17180091 0.01945794 0.06970881 -0.01357811]\n[ 0.09988805 0.12277343 -0.1578104 -0.02371587 0.15475827 0.05720674]\n[ 0.0939876 0.09390745 -0.08190562 -0.01558519 0.16216686 0.0577699 ]\n[ 0.02930515 0.14202896 -0.0498645 0.09410011 -0.0905753 -0.09945953]\n[-0.02268049 0.02339502 0.03950382 -0.0517937 -0.11349598 -0.08833327]\n[ 0.08516674 -0.01251649 0.06132861 -0.03194287 -0.08439503 0.04220618]\n[-0.01899105 -0.1045885 0.077949 -0.01958727 0.02808487 0.05675743]\n[-0.02527532 -0.05968197 -0.00986177 -0.09452567 0.04833149 -0.03405481]\n[-0.00804637 0.04327626 -0.00891493 0.0404686 0.02445388 0.02132383]\n[-0.10686646 0.16725922 0.05255186 0.07999355 0.08804691 -0.23960486]\n[-0.06965735 0.09843512 0.07647634 0.03474835 0.07674371 -0.15451641]\n[ 0.03860416 0.07051504 0.10006531 0.13294117 0.05803839 -0.05538081]\n[ 0.06737103 0.03752292 0.04055005 0.06125275 -0.01747492 0.11705406]\n[-0.08042438 0.22490498 -0.07153002 -0.00587896 -0.03975414 0.0298453 ]\n[ 0.11747472 0.07065479 -0.18580684 0.01055077 -0.00294106 -0.09013598]\n[ 0.16344641 0.09852085 -0.02280455 0.06350011 0.10644891 -0.07841627]\n[ 0.04090171 0.0539231 0.03191345 0.08975121 0.07354202 -0.05085464]\n[ 0.06949573 0.02994953 0.11174608 0.14790806 0.01851488 -0.03612482]\n[ 0.00102438 -0.03996995 0.04026007 0.00085368 -0.09988602 0.07002623]\n[ 0.09183715 0.06067352 -0.05980634 -0.01438377 -0.09251653 0.11715078]\n[-0.00386922 -0.10703073 0.07787523 0.02702085 0.08162405 -0.0717581 ]\n[-0.07778512 -0.06356189 0.05408532 -0.07408235 -0.09237519 0.0507946 ]\n[-0.07788907 0.04865259 -0.1389758 0.0692102 0.05811313 0.07969563]\n[ 0.07354986 -0.03002382 0.0805547 0.2104647 0.02959449 -0.17809092]\n[ 0.08451089 0.06469505 0.09785924 0.11381891 -0.0139894 -0.03917938]\n[ 0.13918306 -0.00193001 0.09200746 0.16640072 -0.06110357 -0.10224576]\n[ 0.16877943 -0.02561516 0.00987721 0.08890694 -0.10235725 -0.02215788]\n[ 0.02761451 0.1507888 -0.18001552 0.00864334 -0.06495474 -0.03670621]\n[ 0.11245865 0.04967188 -0.07628254 0.03296975 0.04564534 -0.17311384]\n[ 0.07160819 0.09903321 -0.04564312 0.02617879 0.00071424 -0.09162144]\n[ 0.03731268 0.07952689 0.08011412 0.1758664 0.07917697 -0.14210425]\n[ 0.00843566 -0.01549498 0.10385882 0.12704365 -0.00305744 -0.13962168]\n[ 0.03056654 0.01458561 0.05228808 0.03149624 -0.14009424 0.07575957]\n[ 0.03985079 -0.02655031 -0.12152576 0.07518446 0.03604449 0.06937316]\n[-0.04599643 -0.05598318 -0.07937098 -0.00095775 0.03727443 0.01474119]\n[ 0.04567473 0.01753204 -0.06741867 -0.06794508 -0.10253347 0.04230488]\n[ 0.04462462 0.01107366 -0.08735126 0.06885814 0.03471483 0.01292751]\n[ 0.00354577 -0.10288887 0.04419841 0.070103 0.09000523 -0.00721082]\n[ 0.03796019 0.00535762 -0.0348375 0.12293369 -0.00022215 -0.04892039]\n[ 0.1114773 -0.02669583 0.02275812 0.13235387 -0.06369523 -0.14881867]\n[ 0.06753314 0.09824273 -0.04225203 -0.00379097 -0.0951763 -0.12207072]\n[-0.03115888 0.20993769 0.03000314 -0.09485421 -0.0499488 -0.14387868]\n[-0.04423478 -0.00696093 0.03348728 0.03983349 0.04265827 -0.11249048]\n[-0.04539219 0.01402776 0.01623834 0.12802741 0.04629254 -0.07377387]\n[-0.06900846 -0.00894165 0.1098303 0.11562915 -0.05746885 -0.01682089]\n[ 0.01336491 -0.06197079 0.0806931 0.17345402 0.00053334 0.02890994]\n[ 0.09189578 0.014551 -0.01489017 0.10595137 -0.00745042 -0.03500335]\n[ 0.12131631 -0.04129434 -0.02708133 0.20030521 -0.01266863 -0.00490865]\n[ 0.06412533 -0.07589801 -0.07554442 0.09276785 0.0441382 0.0288954 ]\n[ 0.07487261 -0.05734253 -0.00203311 -0.0025931 -0.01919963 0.11948007]\n[ 0.06480353 -0.07415973 -0.1177642 0.01564089 0.06338034 0.09492156]\n[ 0.00965498 -0.02282054 0.06747879 0.05969336 0.01784871 0.02390275]\n[-0.02003992 0.00335185 -0.04478868 0.03818144 -0.0128112 0.01287017]\n[-0.04065459 -0.02603371 -0.02571965 0.03965176 -0.09204208 -0.03935078]\n[-0.04766786 0.1082798 -0.01631091 0.00945273 -0.05310407 -0.02537967]\n[-0.12904282 0.07982964 -0.02384206 -0.03279427 0.04572337 -0.15440276]\n[-0.09009448 0.01916551 0.05756131 0.02075357 0.03775262 -0.06985762]\n[-0.00639182 0.01354402 0.11129715 0.1652976 -0.0169815 0.00177091]\n[ 0.01051939 0.01672464 0.01024784 0.10512043 -0.04685514 -0.01185065]\n[ 0.12266921 -0.05120954 0.09076186 0.11903457 -0.02679647 0.06785123]\n[ 0.19854587 -0.117159 -0.06767677 0.107568 -0.08477832 0.02565571]\n[ 0.14433281 -0.03654353 -0.03057992 0.13306198 0.02715246 0.01270683]\n[-0.04808082 0.03179479 0.03737011 0.04716397 0.00207957 -0.00929967]\n[-0.02339549 -0.06695335 -0.0845779 -0.05967411 -0.00260638 0.13282208]\n[ 0.06550924 0.00990529 -0.06323579 0.06805041 -0.01212648 0.18428057]\n[-0.01365907 -0.04004657 -0.05474987 0.00434948 -0.12250189 0.1412147 ]\n[-0.07735828 0.00387685 -0.02766323 0.026003 -0.02776339 0.04378674]\n[-0.03873186 -0.02842149 -0.00020299 -0.05387518 -0.02083273 0.06436604]\n[-0.04522833 0.0598625 -0.04362006 -0.019587 0.02535262 -0.00996901]\n[-0.21301249 0.00997172 -0.02179454 -0.06571198 0.01145528 -0.10527243]\n[-0.10750125 -0.08637699 0.02337158 0.04361615 0.00198748 0.0242748 ]\n[-0.0601953 0.02206491 0.14650838 0.02445182 -0.06929463 0.04244595]\n[-0.07250794 0.00332874 0.1676188 -0.04659299 -0.01623251 -0.00857764]\n[ 0.12495214 -0.06444436 0.15579903 0.12007456 0.02670288 -0.0063992 ]\n[ 0.08592647 -0.08190694 -0.00897303 0.0336335 0.04025805 -0.02552569]\n[ 0.10188218 -0.00358185 -0.05461309 0.01297983 0.04758385 -0.06159959]\n[ 0.01163076 -0.04872379 -0.06203174 0.02760573 -0.05980317 -0.04863126]\n[ 0.03646083 -0.07539698 0.01410214 -0.0709289 -0.08435534 0.04741009]\n[ 0.00470874 -0.06506602 -0.08467245 -0.02716547 -0.01306493 0.01949632]\n[-0.00986365 -0.05028993 0.01529244 -0.07219611 -0.0805352 0.00282447]\n[ 0.00424298 0.07829684 0.05605292 -0.01308339 0.0289993 0.02352933]\n[-0.01316212 -0.00870331 -0.06443617 -0.05398602 0.03620742 0.03515423]\n[-0.10057551 0.1030671 -0.13465314 -0.0894946 0.00834181 -0.08463646]\n[-0.14764331 0.12393782 -0.07032378 -0.22490115 0.00489908 -0.04571831]\n[-0.07405813 0.03586433 0.01903501 -0.14590773 0.05224146 0.02445566]\n[-0.12631416 0.06733298 0.08068271 -0.05959218 0.06698295 0.08723918]\n[-0.00939874 0.11931802 -0.01462062 -0.04392191 0.02908198 -0.00908836]\n[ 0.07750757 0.06759952 -0.06454051 0.07551301 0.0246576 -0.05272722]\n[-0.03650379 0.00216671 0.05520288 0.05594345 0.10305379 -0.01561263]\n[ 0.06591907 -0.00408788 -0.00857377 0.08742893 -0.06154362 -0.00725429]\n[ 0.0087805 -0.04396345 0.03727139 -0.06167185 0.06584529 0.05273139]\n[ 0.02052966 -0.05413838 -0.07269059 -0.07978906 0.00905581 0.08197328]\n[-0.013017 -0.08677133 -0.03951392 -0.05108723 0.03192541 0.00130963]\n[ 0.02195974 0.01471327 -0.041569 -0.12173618 0.05705425 0.04977322]\n[ 0.03240337 0.02462504 -0.04132796 -0.1322763 -0.02947292 0.06270567]\n[-0.13428003 -0.05802382 -0.05476358 -0.08751541 -0.08120316 -0.06807917]\n[-0.11075317 0.03772194 -0.14089853 -0.15388365 0.0321135 -0.04277958]\n[-0.00949441 -0.06845162 -0.2019366 -0.07418422 0.01424402 -0.12506512]\n[-0.0238974 -0.02651614 -0.18607937 -0.04616588 0.05341812 -0.04659864]\n[-0.05222721 -0.03627942 -0.03076171 -0.07781958 -0.03211349 -0.07249965]\n[-0.03002287 0.0497722 0.03813708 -0.03756813 0.09162919 0.0187449 ]\n[-0.08644168 -0.07621741 -0.01266451 -0.10259596 -0.05855577 -0.07815471]\n[-0.08951465 0.0759464 -0.02579235 -0.0925295 -0.04764029 0.07599392]\n[ 0.01192029 0.07872926 -0.00607419 -0.03133556 -0.04657785 -0.02395956]\n[-0.01545587 0.05089727 -0.05282832 0.07024089 0.02201224 0.00413834]\n[-0.02432071 -0.08790541 0.04755025 -0.08306738 -0.001583 0.0578929 ]\n[ 0.00583684 0.02882309 0.0197444 0.06982818 -0.03355337 -0.00238172]\n[ 0.01387064 0.06662367 0.0507287 0.00070776 0.06749933 -0.02065942]\n[ 0.02084344 0.01149368 -0.07983389 -0.05000533 -0.00404061 -0.06852283]\n[-0.0201003 -0.03622016 -0.03217192 -0.04475833 -0.08523814 0.00879569]\n[-0.0176454 -0.07809188 -0.06454822 -0.02917211 0.03038744 -0.09238425]\n[-0.00707183 -0.06940052 -0.14832528 -0.01780364 -0.00063082 -0.07328011]\n[-0.0002371 0.05508889 -0.12647893 -0.03493948 0.10342947 -0.07538689]\n[ 0.04803584 0.02707655 -0.00957456 0.05970713 0.0101542 0.00793176]\n[-0.09571744 -0.07614208 -0.05025434 -0.02070732 0.00741233 -0.07638752]\n[-0.05636562 -0.01113752 -0.05606962 -0.01220792 0.09568168 0.09122528]\n[ 0.06123243 0.03073761 0.05206808 -0.02998531 0.05541601 0.0519431 ]\n[ 0.01458779 -0.05481991 -0.03217587 0.04852049 -0.02968468 0.01751289]\n[-0.08481023 -0.03858817 -0.08736905 -0.07551552 -0.0264373 -0.00050544]\n[-0.0509212 0.21145831 0.02917097 -0.01802795 -0.04289549 -0.11567215]\nlayer 8 Activation\nrelu\nlayer 9 Dropout\nlayer 10 Dense\n6 10\n[-0.23694457 -0.26670206 0.5543111 -0.21170883 0.14601716 -0.41281283\n 0.5825159 -0.61968309 -0.19427252 -0.33199763]\n[-0.52713776 0.65657002 -0.45421809 -0.71776617 0.57379711 -0.51777285\n 0.51330757 -0.64001513 -0.2265407 -0.15891935]\n[ 0.31625733 -0.79113078 -0.53154278 -0.37023902 -0.448488 0.24989481\n 0.26083246 -0.65155941 0.05267058 -0.56062621]\n[ 0.49898064 -0.5559637 0.3755132 -0.38629225 -0.48913252 -0.5279811\n 0.27443361 -0.63590944 -0.41930044 -0.63219696]\n[-0.16015369 -0.84459484 -0.03653246 -0.20003419 0.74036539 -0.1439731\n -0.30219018 -0.09053656 -0.10922217 0.69687885]\n[-0.44375637 -0.60822946 0.24794576 0.37305298 -0.44981813 0.12429843\n -0.7043792 -0.54847002 -0.0192339 -0.58583635]\n[-0.64684528 0.31195766 -0.46403262 0.07613805 -0.43853578 0.09954185\n -0.82225633 0.90443391 0.60570228 0.24354704]\nlayer 11 Activation\nsoftmax\n" }, { "path": "example/mnist_cnn_one_iteration.py", "content": "'''\nSave CNN network and one sample of train data.\n\nRun one iteration of training of convnet on the MNIST dataset.\n'''\n\nfrom __future__ import print_function\nimport numpy as np\nnp.random.seed(1337) # for reproducibility\n\nfrom keras.datasets import mnist\nfrom keras.models import Sequential\nfrom keras.layers.core import Dense, Dropout, Activation, Flatten\nfrom keras.layers.convolutional import Convolution2D, MaxPooling2D\nfrom keras.utils import np_utils\n\nbatch_size = 128\nnb_classes = 10\nnb_epoch = 1\n\n# input image dimensions\nimg_rows, img_cols = 28, 28\n# number of convolutional filters to use\nnb_filters = 4\n# size of pooling area for max pooling\nnb_pool = 2\n# convolution kernel size\nnb_conv = 3\n\n# the data, shuffled and split between train and test sets\n(X_train, y_train), (X_test, y_test) = mnist.load_data()\n\nX_train = X_train.reshape(X_train.shape[0], 1, img_rows, img_cols)\nX_test = X_test.reshape(X_test.shape[0], 1, img_rows, img_cols)\nX_train = X_train.astype('float32')\nX_test = X_test.astype('float32')\nX_train /= 255\nX_test /= 255\nprint('X_train shape:', X_train.shape)\nprint(X_train.shape[0], 'train samples')\nprint(X_test.shape[0], 'test samples')\n\n# convert class vectors to binary class matrices\nY_train = np_utils.to_categorical(y_train, nb_classes)\nY_test = np_utils.to_categorical(y_test, nb_classes)\n\nmodel = Sequential()\n\nmodel.add(Convolution2D(nb_filters, nb_conv, nb_conv, border_mode='same',\n input_shape=(1, img_rows, img_cols)))\nmodel.add(Activation('relu'))\nmodel.add(Convolution2D(nb_filters, nb_conv, nb_conv, border_mode='same'))\nmodel.add(Activation('relu'))\nmodel.add(MaxPooling2D(pool_size=(nb_pool, nb_pool)))\nmodel.add(Dropout(0.25))\n\nmodel.add(Flatten())\nmodel.add(Dense(6))\nmodel.add(Activation('relu'))\nmodel.add(Dropout(0.5))\nmodel.add(Dense(nb_classes))\nmodel.add(Activation('softmax'))\n\nmodel.compile(loss='categorical_crossentropy',\n optimizer='adadelta')\n\nmodel.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch,\n verbose=1, validation_data=(X_test, Y_test))\n\n# store model\nwith open('./my_nn_arch.json', 'w') as fout:\n fout.write(model.to_json())\nmodel.save_weights('./my_nn_weights.h5', overwrite=True)\n\n# store one sample in text file\nwith open(\"./sample_mnist.dat\", \"w\") as fin:\n fin.write(\"1 28 28\\n\")\n a = X_train[0,0]\n for b in a:\n fin.write(str(b)+'\\n')\n\n# get prediction on saved sample\n# c++ output should be the same ;)\nprint('Prediction on saved sample:')\nprint(str(model.predict(X_train[:1])))\n# on my pc I got:\n#[[ 0.03729606 0.00783805 0.06588034 0.21728528 0.01093729 0.34730983\n# 0.01350389 0.02174525 0.26624694 0.01195715]]\n\n" }, { "path": "example/my_nn_arch.json", "content": "{\"class_name\": \"Sequential\", \"config\": [{\"class_name\": \"Convolution2D\", \"config\": {\"b_regularizer\": null, \"W_constraint\": null, \"b_constraint\": null, \"name\": \"convolution2d_1\", \"activity_regularizer\": null, \"trainable\": true, \"dim_ordering\": \"th\", \"nb_col\": 3, \"subsample\": [1, 1], \"init\": \"glorot_uniform\", \"nb_filter\": 4, \"input_dtype\": \"float32\", \"border_mode\": \"same\", \"batch_input_shape\": [null, 1, 28, 28], \"W_regularizer\": null, \"activation\": \"linear\", \"nb_row\": 3}}, {\"class_name\": \"Activation\", \"config\": {\"activation\": \"relu\", \"trainable\": true, \"name\": \"activation_1\"}}, {\"class_name\": \"Convolution2D\", \"config\": {\"W_constraint\": null, \"b_constraint\": null, \"name\": \"convolution2d_2\", \"activity_regularizer\": null, \"trainable\": true, \"dim_ordering\": \"th\", \"nb_col\": 3, \"subsample\": [1, 1], \"init\": \"glorot_uniform\", \"nb_filter\": 4, \"border_mode\": \"same\", \"b_regularizer\": null, \"W_regularizer\": null, \"activation\": \"linear\", \"nb_row\": 3}}, {\"class_name\": \"Activation\", \"config\": {\"activation\": \"relu\", \"trainable\": true, \"name\": \"activation_2\"}}, {\"class_name\": \"MaxPooling2D\", \"config\": {\"name\": \"maxpooling2d_1\", \"trainable\": true, \"dim_ordering\": \"th\", \"pool_size\": [2, 2], \"strides\": [2, 2], \"border_mode\": \"valid\"}}, {\"class_name\": \"Dropout\", \"config\": {\"p\": 0.25, \"trainable\": true, \"name\": \"dropout_1\"}}, {\"class_name\": \"Flatten\", \"config\": {\"trainable\": true, \"name\": \"flatten_1\"}}, {\"class_name\": \"Dense\", \"config\": {\"W_constraint\": null, \"b_constraint\": null, \"name\": \"dense_1\", \"activity_regularizer\": null, \"trainable\": true, \"init\": \"glorot_uniform\", \"bias\": true, \"input_dim\": null, \"b_regularizer\": null, \"W_regularizer\": null, \"activation\": \"linear\", \"output_dim\": 6}}, {\"class_name\": \"Activation\", \"config\": {\"activation\": \"relu\", \"trainable\": true, \"name\": \"activation_3\"}}, {\"class_name\": \"Dropout\", \"config\": {\"p\": 0.5, \"trainable\": true, \"name\": \"dropout_2\"}}, {\"class_name\": \"Dense\", \"config\": {\"W_constraint\": null, \"b_constraint\": null, \"name\": \"dense_2\", \"activity_regularizer\": null, \"trainable\": true, \"init\": \"glorot_uniform\", \"bias\": true, \"input_dim\": null, \"b_regularizer\": null, \"W_regularizer\": null, \"activation\": \"linear\", \"output_dim\": 10}}, {\"class_name\": \"Activation\", \"config\": {\"activation\": \"softmax\", \"trainable\": true, \"name\": \"activation_4\"}}]}" }, { "path": "example_main.cc", "content": "#include \"keras_model.h\"\n\n#include \n\nusing namespace std;\nusing namespace keras;\n\n// Step 1\n// Dump keras model and input sample into text files\n// python dump_to_simple_cpp.py -a example/my_nn_arch.json -w example/my_nn_weights.h5 -o example/dumped.nnet\n// Step 2\n// Use text files in c++ example. To compile:\n// g++ keras_model.cc example_main.cc\n// To execute:\n// a.out\n\nint main() {\n cout << \"This is simple example with Keras neural network model loading into C++.\\n\"\n << \"Keras model will be used in C++ for prediction only.\" << endl;\n\n DataChunk *sample = new DataChunk2D();\n sample->read_from_file(\"./example/sample_mnist.dat\");\n std::cout << sample->get_3d().size() << std::endl;\n KerasModel m(\"./example/dumped.nnet\", true);\n m.compute_output(sample);\n delete sample;\n\n return 0;\n}\n" }, { "path": "keras_model.cc", "content": "#include \"keras_model.h\"\n\n#include \n#include \n#include \n#include \nusing namespace std;\n\n\nstd::vector keras::read_1d_array(std::ifstream &fin, int cols) {\n vector arr;\n float tmp_float;\n char tmp_char;\n fin >> tmp_char; // for '['\n for(int n = 0; n < cols; ++n) {\n fin >> tmp_float;\n arr.push_back(tmp_float);\n }\n fin >> tmp_char; // for ']'\n return arr;\n}\n\nvoid keras::DataChunk2D::read_from_file(const std::string &fname) {\n ifstream fin(fname.c_str());\n fin >> m_depth >> m_rows >> m_cols;\n\n for(int d = 0; d < m_depth; ++d) {\n vector > tmp_single_depth;\n for(int r = 0; r < m_rows; ++r) {\n vector tmp_row = keras::read_1d_array(fin, m_cols);\n tmp_single_depth.push_back(tmp_row);\n }\n data.push_back(tmp_single_depth);\n }\n fin.close();\n}\n\n\nvoid keras::LayerConv2D::load_weights(std::ifstream &fin) {\n char tmp_char = ' ';\n string tmp_str = \"\";\n float tmp_float;\n bool skip = false;\n fin >> m_kernels_cnt >> m_depth >> m_rows >> m_cols >> m_border_mode;\n if (m_border_mode == \"[\") { m_border_mode = \"valid\"; skip = true; }\n\n //cout << \"LayerConv2D \" << m_kernels_cnt << \"x\" << m_depth << \"x\" << m_rows <<\n // \"x\" << m_cols << \" border_mode \" << m_border_mode << endl;\n // reading kernel weights\n for(int k = 0; k < m_kernels_cnt; ++k) {\n vector > > tmp_depths;\n for(int d = 0; d < m_depth; ++d) {\n vector > tmp_single_depth;\n for(int r = 0; r < m_rows; ++r) {\n if (!skip) { fin >> tmp_char; } // for '['\n else { skip = false; }\n vector tmp_row;\n for(int c = 0; c < m_cols; ++c) {\n fin >> tmp_float;\n tmp_row.push_back(tmp_float);\n }\n fin >> tmp_char; // for ']'\n tmp_single_depth.push_back(tmp_row);\n }\n tmp_depths.push_back(tmp_single_depth);\n }\n m_kernels.push_back(tmp_depths);\n }\n // reading kernel biases\n fin >> tmp_char; // for '['\n for(int k = 0; k < m_kernels_cnt; ++k) {\n fin >> tmp_float;\n m_bias.push_back(tmp_float);\n }\n fin >> tmp_char; // for ']'\n\n}\n\nvoid keras::LayerActivation::load_weights(std::ifstream &fin) {\n fin >> m_activation_type;\n //cout << \"Activation type \" << m_activation_type << endl;\n}\n\nvoid keras::LayerMaxPooling::load_weights(std::ifstream &fin) {\n fin >> m_pool_x >> m_pool_y;\n //cout << \"MaxPooling \" << m_pool_x << \"x\" << m_pool_y << endl;\n}\n\nvoid keras::LayerDense::load_weights(std::ifstream &fin) {\n fin >> m_input_cnt >> m_neurons;\n float tmp_float;\n char tmp_char = ' ';\n for(int i = 0; i < m_input_cnt; ++i) {\n vector tmp_n;\n fin >> tmp_char; // for '['\n for(int n = 0; n < m_neurons; ++n) {\n fin >> tmp_float;\n tmp_n.push_back(tmp_float);\n }\n fin >> tmp_char; // for ']'\n m_weights.push_back(tmp_n);\n }\n //cout << \"weights \" << m_weights.size() << endl;\n fin >> tmp_char; // for '['\n for(int n = 0; n < m_neurons; ++n) {\n fin >> tmp_float;\n m_bias.push_back(tmp_float);\n }\n fin >> tmp_char; // for ']'\n //cout << \"bias \" << m_bias.size() << endl;\n\n}\n\nkeras::KerasModel::KerasModel(const string &input_fname, bool verbose)\n : m_verbose(verbose) {\n load_weights(input_fname);\n}\n\n\nkeras::DataChunk* keras::LayerFlatten::compute_output(keras::DataChunk* dc) {\n vector > > im = dc->get_3d();\n\n size_t csize = im[0].size();\n size_t rsize = im[0][0].size();\n size_t size = im.size() * csize * rsize;\n keras::DataChunkFlat *out = new DataChunkFlat(size);\n float * y_ret = out->get_1d_rw().data();\n for(size_t i = 0, dst = 0; i < im.size(); ++i) {\n for(size_t j = 0; j < csize; ++j) {\n float * row = im[i][j].data();\n for(size_t k = 0; k < rsize; ++k) {\n y_ret[dst++] = row[k];\n }\n }\n }\n\n return out;\n}\n\n\nkeras::DataChunk* keras::LayerMaxPooling::compute_output(keras::DataChunk* dc) {\n vector > > im = dc->get_3d();\n vector > > y_ret;\n for(unsigned int i = 0; i < im.size(); ++i) {\n vector > tmp_y;\n for(unsigned int j = 0; j < (unsigned int)(im[0].size()/m_pool_x); ++j) {\n tmp_y.push_back(vector((int)(im[0][0].size()/m_pool_y), 0.0));\n }\n y_ret.push_back(tmp_y);\n }\n for(unsigned int d = 0; d < y_ret.size(); ++d) {\n for(unsigned int x = 0; x < y_ret[0].size(); ++x) {\n unsigned int start_x = x*m_pool_x;\n unsigned int end_x = start_x + m_pool_x;\n for(unsigned int y = 0; y < y_ret[0][0].size(); ++y) {\n unsigned int start_y = y*m_pool_y;\n unsigned int end_y = start_y + m_pool_y;\n\n vector values;\n for(unsigned int i = start_x; i < end_x; ++i) {\n for(unsigned int j = start_y; j < end_y; ++j) {\n values.push_back(im[d][i][j]);\n }\n }\n y_ret[d][x][y] = *max_element(values.begin(), values.end());\n }\n }\n }\n keras::DataChunk *out = new keras::DataChunk2D();\n out->set_data(y_ret);\n return out;\n}\n\nvoid keras::missing_activation_impl(const string &act) {\n cout << \"Activation \" << act << \" not defined!\" << endl;\n cout << \"Please add its implementation before use.\" << endl;\n exit(1);\n}\n\nkeras::DataChunk* keras::LayerActivation::compute_output(keras::DataChunk* dc) {\n\n if (dc->get_data_dim() == 3) {\n vector > > y = dc->get_3d();\n if(m_activation_type == \"relu\") {\n for(unsigned int i = 0; i < y.size(); ++i) {\n for(unsigned int j = 0; j < y[0].size(); ++j) {\n for(unsigned int k = 0; k < y[0][0].size(); ++k) {\n if(y[i][j][k] < 0) y[i][j][k] = 0;\n }\n }\n }\n keras::DataChunk *out = new keras::DataChunk2D();\n out->set_data(y);\n return out;\n } else {\n keras::missing_activation_impl(m_activation_type);\n }\n } else if (dc->get_data_dim() == 1) { // flat data, use 1D\n vector y = dc->get_1d();\n if(m_activation_type == \"relu\") {\n for(unsigned int k = 0; k < y.size(); ++k) {\n if(y[k] < 0) y[k] = 0;\n }\n } else if(m_activation_type == \"softmax\") {\n float sum = 0.0;\n for(unsigned int k = 0; k < y.size(); ++k) {\n y[k] = exp(y[k]);\n sum += y[k];\n }\n for(unsigned int k = 0; k < y.size(); ++k) {\n y[k] /= sum;\n }\n } else if(m_activation_type == \"sigmoid\") {\n for(unsigned int k = 0; k < y.size(); ++k) {\n y[k] = 1/(1+exp(-y[k]));\n }\n } else if(m_activation_type == \"tanh\") {\n for(unsigned int k = 0; k < y.size(); ++k) {\n y[k] = tanh(y[k]);\n }\n } else {\n keras::missing_activation_impl(m_activation_type);\n }\n\n keras::DataChunk *out = new DataChunkFlat();\n out->set_data(y);\n return out;\n } else { throw \"data dim not supported\"; }\n\n return dc;\n}\n\n\n// with border mode = valid\nstd::vector< std::vector > keras::conv_single_depth_valid(\n\tstd::vector< std::vector > const & im,\n\tstd::vector< std::vector > const & k)\n{\n size_t k1_size = k.size(), k2_size = k[0].size();\n unsigned int st_x = (k1_size - 1) >> 1;\n unsigned int st_y = (k2_size - 1) >> 1;\n\n std::vector< std::vector > y(im.size() - 2*st_x, vector(im[0].size() - 2*st_y, 0));\n\n for(unsigned int i = st_x; i < im.size()-st_x; ++i) {\n for(unsigned int j = st_y; j < im[0].size()-st_y; ++j) {\n\n float sum = 0;\n for(unsigned int k1 = 0; k1 < k.size(); ++k1) {\n //const float * k_data = k[k1_size-k1-1].data();\n //const float * im_data = im[i-st_x+k1].data();\n for(unsigned int k2 = 0; k2 < k[0].size(); ++k2) {\n sum += k[k1_size-k1-1][k2_size-k2-1] * im[i-st_x+k1][j-st_y+k2];\n }\n }\n y[i-st_x][j-st_y] = sum;\n }\n }\n return y;\n}\n\n\n// with border mode = same\nstd::vector< std::vector > keras::conv_single_depth_same(\n\tstd::vector< std::vector > const & im,\n\tstd::vector< std::vector > const & k)\n{\n size_t k1_size = k.size(), k2_size = k[0].size();\n unsigned int st_x = (k1_size - 1) >> 1;\n unsigned int st_y = (k2_size - 1) >> 1;\n\n size_t max_imc = im.size() - 1;\n size_t max_imr = im[0].size() - 1;\n std::vector< std::vector > y(im.size(), vector(im[0].size(), 0));\n\n for(unsigned int i = 0; i < im.size(); ++i) {\n for(unsigned int j = 0; j < im[0].size(); ++j) {\n float sum = 0;\n for(unsigned int k1 = 0; k1 < k.size(); ++k1) {\n //const float * k_data = k[k1_size-k1-1].data(); // it is not working ...\n //const float * im_data = im[i-st_x+k1].data();\n for(unsigned int k2 = 0; k2 < k[0].size(); ++k2) {\n if(i-st_x+k1 < 0) continue;\n if(i-st_x+k1 > max_imc) continue;\n if(j-st_y+k2 < 0) continue;\n if(j-st_y+k2 > max_imr) continue;\n\n sum += k[k1_size-k1-1][k2_size-k2-1] * im[i-st_x+k1][j-st_y+k2];\n }\n }\n y[i][j] = sum;\n }\n }\n return y;\n}\n\n\nkeras::DataChunk* keras::LayerConv2D::compute_output(keras::DataChunk* dc) {\n\n unsigned int st_x = (m_kernels[0][0].size()-1) >> 1;\n unsigned int st_y = (m_kernels[0][0][0].size()-1) >> 1;\n vector< vector< vector > > y_ret;\n auto const & im = dc->get_3d();\n\n size_t size_x = (m_border_mode == \"valid\")? im[0].size() - 2 * st_x : im[0].size();\n size_t size_y = (m_border_mode == \"valid\")? im[0][0].size() - 2 * st_y: im[0][0].size();\n for(unsigned int i = 0; i < m_kernels.size(); ++i) { // depth\n vector > tmp;\n tmp.reserve(size_x);\n for(unsigned int j = 0; j < size_x; ++j) { // rows\n tmp.emplace_back(vector(size_y, 0.0));\n }\n y_ret.push_back(tmp);\n }\n\n for(unsigned int j = 0; j < m_kernels.size(); ++j) { // loop over kernels\n for(unsigned int m = 0; m < im.size(); ++m) { // loope over image depth\n\n vector > tmp_w = (m_border_mode == \"valid\")?\n keras::conv_single_depth_valid(im[m], m_kernels[j][m]) :\n keras::conv_single_depth_same(im[m], m_kernels[j][m]);\n\n for(unsigned int x = 0; x < tmp_w.size(); ++x) {\n for(unsigned int y = 0; y < tmp_w[0].size(); ++y) {\n y_ret[j][x][y] += tmp_w[x][y];\n }\n }\n }\n\n for(unsigned int x = 0; x < y_ret[0].size(); ++x) {\n for(unsigned int y = 0; y < y_ret[0][0].size(); ++y) {\n y_ret[j][x][y] += m_bias[j];\n }\n }\n }\n\n keras::DataChunk *out = new keras::DataChunk2D();\n out->set_data(y_ret);\n return out;\n}\n\nkeras::DataChunk* keras::LayerDense::compute_output(keras::DataChunk* dc) {\n //cout << \"weights: input size \" << m_weights.size() << endl;\n //cout << \"weights: neurons size \" << m_weights[0].size() << endl;\n //cout << \"bias \" << m_bias.size() << endl;\n size_t size = m_weights[0].size();\n size_t size8 = size >> 3;\n keras::DataChunkFlat *out = new DataChunkFlat(size, 0);\n float * y_ret = out->get_1d_rw().data();\n\n auto const & im = dc->get_1d();\n\n for (size_t j = 0; j < m_weights.size(); ++j) { // iter over input\n const float * w = m_weights[j].data();\n float p = im[j];\n size_t k = 0;\n for (size_t i = 0; i < size8; ++i) { // iter over neurons\n y_ret[k] += w[k] * p; // vectorize if you can\n y_ret[k+1] += w[k+1] * p;\n y_ret[k+2] += w[k+2] * p;\n y_ret[k+3] += w[k+3] * p;\n y_ret[k+4] += w[k+4] * p;\n y_ret[k+5] += w[k+5] * p;\n y_ret[k+6] += w[k+6] * p;\n y_ret[k+7] += w[k+7] * p;\n k += 8;\n }\n while (k < size) { y_ret[k] += w[k] * p; ++k; }\n }\n for (size_t i = 0; i < size; ++i) { // add biases\n y_ret[i] += m_bias[i];\n }\n\n return out;\n}\n\n\nstd::vector keras::KerasModel::compute_output(keras::DataChunk *dc) {\n //cout << endl << \"KerasModel compute output\" << endl;\n //cout << \"Input data size:\" << endl;\n dc->show_name();\n\n keras::DataChunk *inp = dc;\n keras::DataChunk *out = 0;\n for(int l = 0; l < (int)m_layers.size(); ++l) {\n //cout << \"Processing layer \" << m_layers[l]->get_name() << endl;\n out = m_layers[l]->compute_output(inp);\n\n //cout << \"Input\" << endl;\n //inp->show_name();\n //cout << \"Output\" << endl;\n //out->show_name();\n if(inp != dc) delete inp;\n //delete inp;\n inp = 0L;\n inp = out;\n }\n\n std::vector flat_out = out->get_1d();\n out->show_values();\n delete out;\n\n return flat_out;\n}\n\nvoid keras::KerasModel::load_weights(const string &input_fname) {\n if(m_verbose) cout << \"Reading model from \" << input_fname << endl;\n ifstream fin(input_fname.c_str());\n string layer_type = \"\";\n string tmp_str = \"\";\n int tmp_int = 0;\n\n fin >> tmp_str >> m_layers_cnt;\n if(m_verbose) cout << \"Layers \" << m_layers_cnt << endl;\n\n for(int layer = 0; layer < m_layers_cnt; ++layer) { // iterate over layers\n fin >> tmp_str >> tmp_int >> layer_type;\n if(m_verbose) cout << \"Layer \" << tmp_int << \" \" << layer_type << endl;\n\n Layer *l = 0L;\n if(layer_type == \"Convolution2D\") {\n l = new LayerConv2D();\n } else if(layer_type == \"Activation\") {\n l = new LayerActivation();\n } else if(layer_type == \"MaxPooling2D\") {\n l = new LayerMaxPooling();\n } else if(layer_type == \"Flatten\") {\n l = new LayerFlatten();\n } else if(layer_type == \"Dense\") {\n l = new LayerDense();\n } else if(layer_type == \"Dropout\") {\n continue; // we dont need dropout layer in prediciton mode\n }\n if(l == 0L) {\n cout << \"Layer is empty, maybe it is not defined? Cannot define network.\" << endl;\n return;\n }\n l->load_weights(fin);\n m_layers.push_back(l);\n }\n\n fin.close();\n}\n\nkeras::KerasModel::~KerasModel() {\n for(int i = 0; i < (int)m_layers.size(); ++i) {\n delete m_layers[i];\n }\n}\n\nint keras::KerasModel::get_output_length() const\n{\n int i = m_layers.size() - 1;\n while ((i > 0) && (m_layers[i]->get_output_units() == 0)) --i;\n return m_layers[i]->get_output_units();\n}\n" }, { "path": "keras_model.h", "content": "#ifndef KERAS_MODEL__H\n#define KERAS_MODEL__H\n\n#include \n#include \n#include \n#include \n\nnamespace keras\n{\n\tstd::vector read_1d_array(std::ifstream &fin, int cols);\n\tvoid missing_activation_impl(const std::string &act);\n\tstd::vector< std::vector > conv_single_depth_valid(std::vector< std::vector > const & im, std::vector< std::vector > const & k);\n\tstd::vector< std::vector > conv_single_depth_same(std::vector< std::vector > const & im, std::vector< std::vector > const & k);\n\n\tclass DataChunk;\n\tclass DataChunk2D;\n\tclass DataChunkFlat;\n\n\tclass Layer;\n\tclass LayerFlatten;\n\tclass LayerMaxPooling;\n\tclass LayerActivation;\n\tclass LayerConv2D;\n\tclass LayerDense;\n\n\tclass KerasModel;\n}\n\nclass keras::DataChunk {\npublic:\n virtual ~DataChunk() {}\n virtual size_t get_data_dim(void) const { return 0; }\n virtual std::vector const & get_1d() const { throw \"not implemented\"; };\n virtual std::vector > > const & get_3d() const { throw \"not implemented\"; };\n virtual void set_data(std::vector > > const &) {};\n virtual void set_data(std::vector const &) {};\n //virtual unsigned int get_count();\n virtual void read_from_file(const std::string &fname) {};\n virtual void show_name() = 0;\n virtual void show_values() = 0;\n};\n\nclass keras::DataChunk2D : public keras::DataChunk {\npublic:\n std::vector< std::vector< std::vector > > const & get_3d() const { return data; };\n virtual void set_data(std::vector > > const & d) { data = d; };\n size_t get_data_dim(void) const { return 3; }\n\n void show_name() {\n std::cout << \"DataChunk2D \" << data.size() << \"x\" << data[0].size() << \"x\" << data[0][0].size() << std::endl;\n }\n\n void show_values() {\n std::cout << \"DataChunk2D values:\" << std::endl;\n for(size_t i = 0; i < data.size(); ++i) {\n std::cout << \"Kernel \" << i << std::endl;\n for(size_t j = 0; j < data[0].size(); ++j) {\n for(size_t k = 0; k < data[0][0].size(); ++k) {\n std::cout << data[i][j][k] << \" \";\n }\n std::cout << std::endl;\n }\n }\n }\n //unsigned int get_count() {\n // return data.size()*data[0].size()*data[0][0].size();\n //}\n\n void read_from_file(const std::string &fname);\n std::vector > > data; // depth, rows, cols\n\n int m_depth;\n int m_rows;\n int m_cols;\n};\n\nclass keras::DataChunkFlat : public keras::DataChunk {\npublic:\n DataChunkFlat(size_t size) : f(size) { }\n DataChunkFlat(size_t size, float init) : f(size, init) { }\n DataChunkFlat(void) { }\n\n std::vector f;\n std::vector & get_1d_rw() { return f; }\n std::vector const & get_1d() const { return f; }\n void set_data(std::vector const & d) { f = d; };\n size_t get_data_dim(void) const { return 1; }\n\n void show_name() {\n std::cout << \"DataChunkFlat \" << f.size() << std::endl;\n }\n void show_values() {\n std::cout << \"DataChunkFlat values:\" << std::endl;\n for(size_t i = 0; i < f.size(); ++i) std::cout << f[i] << \" \";\n std::cout << std::endl;\n }\n void read_from_file(const std::string &fname) {};\n //unsigned int get_count() { return f.size(); }\n};\n\nclass keras::Layer {\npublic:\n virtual void load_weights(std::ifstream &fin) = 0;\n virtual keras::DataChunk* compute_output(keras::DataChunk*) = 0;\n\n Layer(std::string name) : m_name(name) {}\n virtual ~Layer() {}\n\n virtual unsigned int get_input_rows() const = 0;\n virtual unsigned int get_input_cols() const = 0;\n virtual unsigned int get_output_units() const = 0;\n\n std::string get_name() { return m_name; }\n std::string m_name;\n};\n\n\nclass keras::LayerFlatten : public Layer {\npublic:\n LayerFlatten() : Layer(\"Flatten\") {}\n void load_weights(std::ifstream &fin) {};\n keras::DataChunk* compute_output(keras::DataChunk*);\n\n virtual unsigned int get_input_rows() const { return 0; } // look for the value in the preceding layer\n virtual unsigned int get_input_cols() const { return 0; } // same as for rows\n virtual unsigned int get_output_units() const { return 0; }\n};\n\n\nclass keras::LayerMaxPooling : public Layer {\npublic:\n LayerMaxPooling() : Layer(\"MaxPooling2D\") {};\n\n void load_weights(std::ifstream &fin);\n keras::DataChunk* compute_output(keras::DataChunk*);\n\n virtual unsigned int get_input_rows() const { return 0; } // look for the value in the preceding layer\n virtual unsigned int get_input_cols() const { return 0; } // same as for rows\n virtual unsigned int get_output_units() const { return 0; }\n\n int m_pool_x;\n int m_pool_y;\n\n};\n\nclass keras::LayerActivation : public Layer {\npublic:\n LayerActivation() : Layer(\"Activation\") {}\n void load_weights(std::ifstream &fin);\n keras::DataChunk* compute_output(keras::DataChunk*);\n\n virtual unsigned int get_input_rows() const { return 0; } // look for the value in the preceding layer\n virtual unsigned int get_input_cols() const { return 0; } // same as for rows\n virtual unsigned int get_output_units() const { return 0; }\n\n std::string m_activation_type;\n};\n\nclass keras::LayerConv2D : public Layer {\npublic:\n LayerConv2D() : Layer(\"Conv2D\") {}\n\n void load_weights(std::ifstream &fin);\n keras::DataChunk* compute_output(keras::DataChunk*);\n std::vector > > > m_kernels; // kernel, depth, rows, cols\n std::vector m_bias; // kernel\n\n virtual unsigned int get_input_rows() const { return m_rows; }\n virtual unsigned int get_input_cols() const { return m_cols; }\n virtual unsigned int get_output_units() const { return m_kernels_cnt; }\n\n std::string m_border_mode;\n int m_kernels_cnt;\n int m_depth;\n int m_rows;\n int m_cols;\n};\n\nclass keras::LayerDense : public Layer {\npublic:\n LayerDense() : Layer(\"Dense\") {}\n\n void load_weights(std::ifstream &fin);\n keras::DataChunk* compute_output(keras::DataChunk*);\n std::vector > m_weights; //input, neuron\n std::vector m_bias; // neuron\n\n virtual unsigned int get_input_rows() const { return 1; } // flat, just one row\n virtual unsigned int get_input_cols() const { return m_input_cnt; }\n virtual unsigned int get_output_units() const { return m_neurons; }\n\n int m_input_cnt;\n int m_neurons;\n};\n\nclass keras::KerasModel {\npublic:\n KerasModel(const std::string &input_fname, bool verbose);\n ~KerasModel();\n std::vector compute_output(keras::DataChunk *dc);\n\n unsigned int get_input_rows() const { return m_layers.front()->get_input_rows(); }\n unsigned int get_input_cols() const { return m_layers.front()->get_input_cols(); }\n int get_output_length() const;\n\nprivate:\n\n void load_weights(const std::string &input_fname);\n int m_layers_cnt; // number of layers\n std::vector m_layers; // container with layers\n bool m_verbose;\n\n};\n\n#endif\n" }, { "path": "test_compare.py", "content": "import numpy as np\nnp.random.seed(1336)\nimport json\nimport argparse\nimport sys\n\ndef get_numbers_from_file(fname):\n r = []\n with open(fname) as fin:\n for line in fin:\n tmp = line.split()\n r = [float(t) for t in tmp]\n return np.array(r)\n\nnp.set_printoptions(threshold=np.inf)\nparser = argparse.ArgumentParser(description='This is a simple script compare predictions from keras and keras2cpp.')\n\nparser.add_argument('-k', '--keras_response', help=\"Response from Keras (test_run_cnn.py)\", required=True)\nparser.add_argument('-c', '--keras2cpp_response', help=\"Response from Keras2cpp (test_run_cnn.cc)\", required=True)\nargs = parser.parse_args()\n\n\nkeras_output = get_numbers_from_file(args.keras_response)\nkeras2cpp_output = get_numbers_from_file(args.keras2cpp_response)\n\nif len(keras_output) != len(keras2cpp_output):\n print \"Different output dimensions\"\n sys.exit(1)\n\nsub = np.sum(np.abs(keras_output - keras2cpp_output))\n\nif sub < 1e-6:\n print 'Test: [DONE]'\n print 'Dump is working correctly.'\n sys.exit(0)\nelse:\n print 'Test: [ERROR]'\n print 'The output from Keras and Keras2cpp are different.'\n print 'Difference value:', sub\n sys.exit(1)\n" }, { "path": "test_run.sh", "content": "#!/bin/bash\n\necho 'Test for CNN dumping'\n\n# Parameters\nINPUT_ARCH=\"example/my_nn_arch.json\"\nINPUT_WEIGHTS=\"example/my_nn_weights.h5\"\n\nDUMPED_CNN=\"test_cnn.dumped\"\nDATA_SAMPLE=\"test_random_input.dat\"\nKERAS_OUTPUT=\"test_keras_output.dat\"\nKERAS2CPP_OUTPUT=\"test_keras2cpp_output.dat\"\nTEST_BIN=\"test_bin\"\n\necho 'Test, step 1'\necho 'Dump network into plain text file' $DUMPED_CNN\npython dump_to_simple_cpp.py -a $INPUT_ARCH -w $INPUT_WEIGHTS -o $DUMPED_CNN\n\necho 'Test, step 2'\necho 'Generate random input sample and save in' $DATA_SAMPLE\necho 'Compute ouput on generated sample with Keras and store predictions for comparison'\npython test_run_cnn.py -a $INPUT_ARCH -w $INPUT_WEIGHTS -d $DATA_SAMPLE -o $KERAS_OUTPUT\n\necho 'Test, step 3'\necho 'Compile keras2cpp code'\ng++ -std=c++11 test_run_cnn.cc keras_model.cc -o $TEST_BIN\necho 'Run predictions with dumped network and random data sample from step 2'\n./$TEST_BIN $DUMPED_CNN $DATA_SAMPLE $KERAS2CPP_OUTPUT\n\necho 'Test, step 4'\necho 'Compare Keras and Keras2cpp outputs'\npython test_compare.py --keras_response $KERAS_OUTPUT --keras2cpp_response $KERAS2CPP_OUTPUT\n\n# Clean\necho 'Cleaning after test'\nrm $DUMPED_CNN\nrm $DATA_SAMPLE\nrm $KERAS_OUTPUT\nrm $KERAS2CPP_OUTPUT\nrm $TEST_BIN\n# used only if you log hidden layers output in test_run_cnn.py file\n#rm test_layer_*.output\n" }, { "path": "test_run_cnn.cc", "content": "#include \"keras_model.h\"\n\n#include \n\nusing namespace std;\nusing namespace keras;\n\n\n\nint main(int argc, char *argv[]) {\n\n if(argc != 4) {\n cout << \"Wrong input, going to exit.\" << endl;\n cout << \"There should be arguments: dumped_cnn_file input_sample output_file.\" << endl;\n return -1;\n }\n string dumped_cnn = argv[1];\n string input_data = argv[2];\n string response_file = argv[3];\n\n cout << \"Testing network from \" << dumped_cnn << \" on data from \" << input_data << endl;\n\n // Input data sample\n DataChunk *sample = new DataChunk2D();\n sample->read_from_file(input_data);\n\n // Construct network\n KerasModel m(dumped_cnn, false);\n std::vector response = m.compute_output(sample);\n\n // clean sample\n delete sample;\n\n // save response into file\n ofstream fout(response_file);\n for(unsigned int i = 0; i < response.size(); i++) {\n fout << response[i] << \" \";\n }\n fout.close();\n return 0;\n}\n" }, { "path": "test_run_cnn.py", "content": "import numpy as np\nnp.random.seed(1336)\nfrom keras.models import Sequential, model_from_json\nimport json\nimport argparse\nfrom keras import backend as K\n\nnp.set_printoptions(threshold=np.inf)\nparser = argparse.ArgumentParser(description='This is a simple script to run Keras model from saved architecture and weights.\\\n This script also creates a input data sample for c++ run_net.')\n\nparser.add_argument('-a', '--architecture', help=\"JSON with model architecture\", required=True)\nparser.add_argument('-w', '--weights', help=\"Model weights in HDF5 format\", required=True)\nparser.add_argument('-d', '--data_sample', help=\"File where to write random data sample\", required=True)\nparser.add_argument('-o', '--output', help=\"File where to write network outpu\", required=True)\nparser.add_argument('-v', '--verbose', help=\"Verbose\", required=False)\nargs = parser.parse_args()\n\nprint 'Verbose', args.verbose\nprint 'Read architecture from', args.architecture\nprint 'Read weights from', args.weights\n\narch = open(args.architecture).read()\nmodel = model_from_json(arch)\nmodel.load_weights(args.weights)\nmodel.compile(loss='categorical_crossentropy', optimizer='adadelta')\narch = json.loads(arch)\nprint 'There are', str(len(model.layers)), 'layers in your network, is it good?'\nprint 'I think yes :)'\n\n\nfirst_layer = arch[\"config\"][0][\"config\"]\ninput_shape = first_layer[\"batch_input_shape\"]\nprint \"Input shape of your network\", input_shape\n\nprint \"Generate random input for testing purposes\"\nrandom_input = np.random.rand(1, input_shape[1], input_shape[2], input_shape[3])\nprint \"Random input shape\", random_input.shape\nresponse = model.predict(random_input)[0]\nif args.verbose:\n print '-'*50\n print 'Prediction from Keras'\n print response\n print '-'*50\n# save response to the file\nwith open(args.output, \"w\") as fin:\n fin.write(' '.join([str(r) for r in response]))\n# store one sample in text file\n# this code is working for input_shape[1] == 1\nif input_shape[1] != 1:\n print '-'*50\n print 'Sorry but below code can be a buggy for image depth > 1 !!!'\n print '-'*50\n\n# save random data sample into file\nwith open(args.data_sample, \"w\") as fin:\n print \"Save to\", args.data_sample, \"sample shape\", str(input_shape[1]) + \" \" + str(input_shape[2]) + \" \" + str(input_shape[3])\n fin.write(str(input_shape[1]) + \" \" + str(input_shape[2]) + \" \" + str(input_shape[3]) + \"\\n\")\n a = random_input[0,0]\n for b in a:\n fin.write(str(b)+'\\n')\n\n# Get layers output (for debuging)\n'''\nfor l in xrange(len(model.layers)):\n with open('test_layer_' + str(l) + '.output', 'w') as fout:\n\n get_layer_output = K.function([model.layers[0].input, K.learning_phase()],\n [model.layers[l].output])\n layer_output = get_layer_output([random_input, 0])\n\n print 'Layer', l, layer_output[0].shape\n if l > 10:\n print layer_output\n fout.write(str(layer_output[0].shape) + '\\n')\n fout.write(str(layer_output) + '\\n')\n'''\n#print 'input?'\n#get_layer_output = K.function([model.layers[0].input, K.learning_phase()],\n# [model.layers[0].input])\n#layer_output = get_layer_output([random_input, 0])\n#print layer_output\n" } ]