TensorFlow-20160622

C-LIS CO., LTD.

5FOTPS'MPXͰ  झຯͷը૾ऩूαʔόʔΛ࡞Δ ݄̒߸ #dltfb

C-LIS CO., LTD. ༗ࢁܓೋʢ,FJKJ"3*:"."ʣ C-LIS CO., LTD. AndroidΞϓϦ։ൃऀ ػցֶशॳ৺ऀ
΍ͬͯ·ͤΜ Photo : Koji MORIGUCHI (AUN CREATIVE FIRM) ͱɺݴ͍ͳ͕Βૣ൒೥……

޷Έͷ؟ڸ່ͬը૾ΛࣗಈͰऩू͍ͨ͠

4PGUXBSF%FTJHO೥݄߸ʰ"OESPJE4UVEJPͷελΠϧͰޮ཰ΞοϓʱΑΓ © ࠜઇΕ͍ ؟ ڸ ͬ ່

C-LIS CO., LTD. ̎Ϋϥε෼ྨ 1 0

C-LIS CO., LTD. ࡞੒σʔλ಺༁ ؟ڸ່ͬը૾ຕ  ඇ؟ڸ່ͬը૾ຕ ؟ڸ່ͬ ඇ؟ڸ່ͬ

5FOTPS'MPXͰ  झຯͷը૾ऩूαʔόʔΛ࡞Δ ݄̐߸ #dltfb https://goo.gl/MX72Bc

C-LIS CO., LTD. ໰୊ൃੜ ΦϑΟεͷαʔόʔϚγϯͷ(16͕࢖͑ͳ͘ͳΔ $ python3 megane_co/cifar10_train.py I
tensorflow/stream_executor/dso_loader.cc:105] successfully opened CUDA library libcublas.so locally I tensorflow/stream_executor/dso_loader.cc:105] successfully opened CUDA library libcudnn.so locally I tensorflow/stream_executor/dso_loader.cc:105] successfully opened CUDA library libcufft.so locally I tensorflow/stream_executor/dso_loader.cc:105] successfully opened CUDA library libcuda.so.1 locally I tensorflow/stream_executor/dso_loader.cc:105] successfully opened CUDA library libcurand.so locally Filling queue with 400 CIFAR images before starting to train. This will take a few minutes. E tensorflow/stream_executor/cuda/cuda_driver.cc:481] failed call to cuInit: CUDA_ERROR_NO_DEVICE

C-LIS CO., LTD. 5FOTPS'MPXͰ(16͕࢖͑ͳ͍ https://blog.keiji.io/2016/05/cuda_error_no_device.html ֓ཁ ɹGPUϚγϯͰTensorFlowΛ࣮ߦ͢ΔࡍɺCUDA΋cuDNN΋ਖ਼ৗʹηοτΞοϓͰ͖͍ͯ ΔͷʹɺTensorFlow͕GPUΛೝࣝ͠ͳ͍ʢCUDA_ERROR_NO_DEVICE΍No GPU
devices available on machine.͕දࣔ͞ΕΔʣ৔߹ɺਖ਼͍͠σΟεϓϨΠΧʔυɾυϥΠό͕Πϯε τʔϧ͞Ε͍ͯͳ͍Մೳੑ͕͋Γ·͢ɻ ɹචऀͷ؀ڥͰ͸ɺطଘͷυϥΠόʔΛΞϯΠϯετʔϧʢsudo apt-get remove --purge nvidia*ʣͯ͠ɺNVIDIAͷαΠτ͔ΒυϥΠόʔΛμ΢ϯϩʔυɾΠϯετʔϧ͢Δ͜ͱͰ GPUΛೝࣝ͢ΔΑ͏ʹͳΓ·ͨ͠ɻ

C-LIS CO., LTD. ซͤͯಡΈ͍ͨ 5FOTPS'MPXPO%PDLFSͰ(16Λ࢖͑ΔΑ͏ʹ͢Δํ๏]גࣜձࣾΧϒΫ  IUUQXXXLBCVLVDPKQEFWFMPQFSTFSSPSTXJUIUFOTPSqPXPOHQV /7*%*"%PDLFSͰ؆୯ʹ(16ରԠͷ5FOTPS'MPXೖΓίϯςφΛ࡞Δํ๏ ]NVPOPUFT  IUUQXXXNVPKQOWJEJBEPDLFSUFOTPSqPXIUNM

ಉਓࢽΛॻͩ͘

ݸਓαʔΫϧʰΊ͕ͶΛ͔͚ΔΜͩʱ

https://techbookfest.org/#A-28

ࣥච্ͷ՝୊

C-LIS CO., LTD. ը૾ͷݖརؔ܎ ͦΕͧΕͷΠϥετʹஶ࡞ݖ͕ଘࡏ ࢿྉͰ࢖͏ʹ͸ڐ୚ΛऔΔඞཁ͕͋Δ D ࠜઇΕ͍ ʢ˞ݸਓͷݟղͰ͢ʣ

4PGUXBSF%FTJHO೥݄߸ʰ"OESPJE4UVEJPͷελΠϧͰޮ཰ΞοϓʱΑΓ © ࠜઇΕ͍ ؟ ڸ ͬ ່

$*'"3ʹؐΔ

C-LIS CO., LTD. $*'"3 Ϋϥε෼ྨ ֶशσʔλ   ςετσʔλ
IUUQTXXXDTUPSPOUPFEV_LSJ[DJGBSIUNM

5FOTPS'MPX5VUPSJBM$*'"3

C-LIS CO., LTD. νϡʔτϦΞϧʁ Ϟσϧ͸DVEBDPOWOFUϕʔε  ৞ΈࠐΈɾϓʔϦϯά̎૚  ہॴԠ౴ਖ਼نԽ૚  શ݁߹૚̎૚ %BUB"VHNFOUBUJPOʢΫϦοϓɾࠨӈ൓సɾً౓ɾίϯτϥετʣ  ന৭Խʢ8IJUFOJOHʣ
ֶश཰ͱڞ༗ॏΈͷݮਰ ϛχόονֶशʢόοναΠζ γϟοϑϧόονʣ

͔̍Β࡞ͬͯΈΑ͏ʂ

C-LIS CO., LTD. ͞·͟·ͳࢪࡦͰ্͕Δೝࣝਫ਼౓ ہॴԠ౴ਖ਼نԽʢ-3/ʣ૚ͷ௥Ճ %BUBBVHNFOUBUJPOɾന৭Խʢ8IJUFOJOHʣ υϩοϓΞ΢τΛઃఆ͢Δ ϞσϧΛଟ૚ʹ͢Δ

C-LIS CO., LTD. ࣮ࡍʹ͸ʜʜ 0.10 0.10 0.10 0.10 0.10
0.10 0.10 0.10 0.10 0.74 0.41

C-LIS CO., LTD.

C-LIS CO., LTD. def _export_graph(sess, epoch): constant_graph_def = graph_util.convert_variables_to_constants( sess,
sess.graph_def, ["output/logits"]) file_path = os.path.join(FLAGS.graph_dir, 'graph_%02d_epoch.pb' % epoch) with gfile.FastGFile(file_path, 'wb') as f: f.write(constant_graph_def.SerializeToString()) άϥϑʴύϥϝʔλʔ with tf.gfile.FastGFile(graph_file, 'rb') as f: graph_def = tf.GraphDef() graph_def.ParseFromString(f.read()) _ = tf.import_graph_def(graph_def, name='')

ॳߘͷ׬੒

C-LIS CO., LTD. ϨϏϡΞʔͷօ͞ΜͷࢦఠΛ΋ͱʹ௚͢ ʮ5FOTPS'MPXͷಛ௃ʹฒྻ෼ࢄॲཧΛؚΊΔͷ͕ྑ͍ͱࢥ͍·͢ʯ ʮ͜ͷຊΛಡΉʹ౰ͨͬͯɺಡऀ͕஌͓͔ͬͯͳ͚Ε͹ͳΒͳ͍͜ͱ͸ͳΜͰ͠ΐ͏ ͔ʯ ʮ1ZUIPOܥͱޓ׵ੑΛอͭͨΊʹɺ͜ͷهड़Λ଍ͨ͠ํ͕ʜʜʯ ʮষͷऴ൫ͷҾ͖ͱɺ࣍ͷষͰ࿩୊͕ܨ͕ͬͯͳ͍Ͱ͢ʯ ʮ͕Μ͹Εʂ͕Μ͹Εʂͱ੠͕͚͢ΔͱɺԹ͔Έͷ͋ΔωοτϫʔΫʹͳΔ͸ͣ
Ͱ͢ʯ FUDʜ

C-LIS CO., LTD.

C-LIS CO., LTD. 3$͕Ͱͨ https://github.com/tensorﬂow/tensorﬂow/releases/tag/v0.9.0rc0

C-LIS CO., LTD. ೝࣝਫ਼౓͸޲্͠ͳ͔͚ͬͨͲࢼͨ͜͠ͱ ࠷దԽΞϧΰϦζϜΛมߋ  ʢ(SBEJFOU%FDFOU͔Β"EBN΁ʣ όοναΠζΛมߋ

C-LIS CO., LTD. ਂࠁͳܭࢉ࣌ؒͷ૿Ճ

ϛχόονֶश

C-LIS CO., LTD. ܭࢉ͕࣌ؒେ෯ʹ୹ॖ

C-LIS CO., LTD. όοναΠζ Every 2.0s: nvidia-smi -l Thu
Jun 16 17:23:25 2016 Thu Jun 16 17:23:25 2016 +------------------------------------------------------+ | NVIDIA-SMI 361.42 Driver Version: 361.42 | |-------------------------------+----------------------+----------------------+ | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | |===============================+======================+======================| | 0 GeForce GTX 960 Off | 0000:01:00.0 On | N/A | | 38% 57C P2 34W / 120W | 1916MiB / 2044MiB | 21% Default | +-------------------------------+----------------------+----------------------+ +-----------------------------------------------------------------------------+ | Processes: GPU Memory | | GPU PID Type Process name Usage | |=============================================================================| | 0 832 G /usr/bin/X 34MiB | | 0 18183 C python3 1867MiB | +-----------------------------------------------------------------------------+ Thu Jun 16 17:23:30 2016 +------------------------------------------------------+ | NVIDIA-SMI 361.42 Driver Version: 361.42 | |-------------------------------+----------------------+----------------------+

C-LIS CO., LTD. όοναΠζ Every 2.0s: nvidia-smi -l Fri
Jun 17 20:40:45 2016 Fri Jun 17 20:40:45 2016 +------------------------------------------------------+ | NVIDIA-SMI 361.42 Driver Version: 361.42 | |-------------------------------+----------------------+----------------------+ | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | |===============================+======================+======================| | 0 GeForce GTX 960 Off | 0000:01:00.0 On | N/A | | 45% 73C P2 103W / 120W | 1917MiB / 2044MiB | 93% Default | +-------------------------------+----------------------+----------------------+ +-----------------------------------------------------------------------------+ | Processes: GPU Memory | | GPU PID Type Process name Usage | |=============================================================================| | 0 832 G /usr/bin/X 34MiB | | 0 24204 C python3 1867MiB | +-----------------------------------------------------------------------------+ Fri Jun 17 20:40:50 2016 +------------------------------------------------------+ | NVIDIA-SMI 361.42 Driver Version: 361.42 | |-------------------------------+----------------------+----------------------+

(16ɺ࢓ࣄͯ͠ͳ͔ͬͨ

݄̒25೔ʢ౔ʣ ळ༿ݪ ௨ӡձؗ ʰٕज़ॻయʱʹͯ൦෍༧ఆʂ ຊจ64p / ൦෍Ձ֨1,000ԁ https://techbookfest.org/#A-28 TensorFlowͷجૅ
TensorFlowͱ͸ σʔλϑϩʔάϥϑ ม਺ͱϓϨʔεϗϧμʔ Tensorʢςϯιϧʣ TensorBoardʹΑΔάϥϑͷՄࢹԽ CIFAR-10ͷֶशͱධՁ σʔλͷಡΈࠐΈ ਪ࿦ʢinferenceʣ ֶशʢlearnʣ ධՁʢevaluateʣ σʔλอଘͱϑΟϧλʔͷՄࢹԽ ֶशσʔλͷอଘ ϑΟϧλʔͷՄࢹԽ CIFAR-10ฃಆه ਖ਼نԽ૚Λ௥Ճ͢Δ ܇࿅σʔλΛՃ޻͢Δ υϩοϓΞ΢τΛઃఆ͢Δ VGGNetͷϞσϧΛ࢖͏ ࠷దԽΞϧΰϦζϜΛมߋ͢Δ όοναΠζΛ૿΍͢ TensorFlow͸͡Ί·ͨ͠  CIFAR-10 ฃಆه

C-LIS CO., LTD. 5FOTPS'MPXΛ࢖ͬͨػցֶश͜ͱ͸͡ΊCZ5PSV6&/0:"."  IUUQXXXTMJEFTIBSFOFU5PSV6FOPZBNBUFOTPSqPXHEH σΟʔϓϥʔχϯάͰ͓ͦদ͞Μͷ࿡ͭࢠ͸ݟ෼͚ΒΕΔͷ͔ʁɹʙ४උฤʙ  IUUQCPIFNJBIBUFOBCMPHDPNFOUSZ ͗͢ΌʔΜϝϞ  IUUQNFNPTVHZBODPN  5FOTPS'MPX 
IUUQTHJUIVCDPNUFOTPSqPXUFOTPSqPX ࢀߟ

C-LIS CO., LTD. C-LIS CO., LTD.

ຊࢿྉ͸ɺ༗ݶձࣾγʔϦεͷஶ࡞෺Ͱ͢ɻܝࡌ͞Ε͍ͯΔΠϥετ͸ɺಛʹهࡌ͕ͳ͍৔߹͸ࠜઇΕ͍ͷஶ࡞෺Ͱ͢ɻ ຊࢿྉͷશ෦ɺ·ͨ͸Ұ෦ʹ͍ͭͯɺஶ࡞ऀ͔ΒจॻʹΑΔڐ୚Λಘͣʹෳ੡͢Δ͜ͱ͸ې͡ΒΕ͍ͯ·͢ɻ 5IF"OESPJE4UVEJPJDPOJTSFQSPEVDFEPSNPEJpFEGSPNXPSLDSFBUFEBOETIBSFECZ(PPHMFBOEVTFE BDDPSEJOHUPUFSNTEFTDSJCFEJOUIF$SFBUJWF$PNNPOT"UUSJCVUJPO-JDFOTF ֤੡඼໊ɾϒϥϯυ໊ɺձ໊ࣾͳͲ͸ɺҰൠʹ֤ࣾͷ঎ඪ·ͨ͸ొ࿥঎ඪͰ͢ɻຊࢿྉதͰ͸ɺɺɺäΛׂѪͯ͠ ͍·͢ɻ 5IF"OESPJESPCPUJTSFQSPEVDFEPSNPEJpFEGSPNXPSLDSFBUFEBOETIBSFECZ(PPHMFBOEVTFEBDDPSEJOH UPUFSNTEFTDSJCFEJOUIF$SFBUJWF$PNNPOT"UUSJCVUJPO-JDFOTF

TensorFlow-20160622

TensorFlow-20160622

More Decks by ARIYAMA Keiji

Other Decks in Technology

Featured

Transcript