【失敗版】Raspberry Piに TensorFlow Deep Learning Frameworkをインストールする方法

【ビルド版】Raspberry Piで DeepDreamを動かしてキモイ絵をモリモリ量産 Caffe Deep Learning Framework

【インストール版】Raspberry Piで DeepDreamを動かしてキモイ絵をモリモリ量産 Caffe Deep Learning

Orange Pi PC 2の 64bitのチカラで DeepDreamしてキモイ絵を高速でモリモリ量産してみるテスト

pi@raspberrypi:~/pytorch $ uname -a
Linux raspberrypi 4.14.50-v7+ #1122 SMP Tue Jun 19 12:26:26 BST 2018 armv7l GNU/Linux

【成功版】Raspberry Piに TensorFlow Deep Learning Frameworkをインストールする方法

# Installing TensorFlow on Raspbian

# お決まりの sudo apt-get updateで最新状態に更新する
sudo apt-get update

# Installing TensorFlow on Raspbian

# Python 3.4+（Python 3.4以上）
python3 -V
# Python 3.5.3

# for Python 3.n
pip3 -V
# pip 9.0.1 from /usr/lib/python3/dist-packages (python 3.5)

# for Python 3.n
sudo apt-get -y install python3-pip
sudo apt -y install libatlas-base-dev

# Python 3.n
pip3 install tensorflow
# Downloading https://www.piwheels.org/simple/tensorflow/tensorflow-1.9.0-cp35-none-linux_armv7l.whl (66.7MB)
# tensorflow-1.9.0

# Installing TensorFlow on Raspbian

# Python 2.7
python -V
# Python 2.7.13

# for Python 2.7
pip -V
# pip 9.0.1 from /usr/lib/python2.7/dist-packages (python 2.7)

# for Python 2.7
sudo apt-get -y install python-pip
sudo apt -y install libatlas-base-dev

# Python 2.7
pip install tensorflow
# Downloading https://www.piwheels.org/simple/tensorflow/tensorflow-1.9.0-cp27-none-linux_armv7l.whl (66.7MB)

pi@raspberrypi:~/tf/deepdream_highres $ pip install tensorflow
Collecting tensorflow
  Downloading https://www.piwheels.org/simple/tensorflow/tensorflow-1.9.0-cp27-none-linux_armv7l.whl (66.7MB)
    100% |████████████████████████████████| 66.7MB 1.1kB/s
...
Installing collected packages: termcolor, six, setuptools, protobuf, absl-py, enum34, futures, grpcio, gast, numpy, markdown, werkzeug, wheel, tensorboard, backports.weakref, astor, funcsigs, pbr, mock, tensorflow
Successfully installed absl-py-0.3.0 astor-0.7.1 backports.weakref-1.0.post1 enum34-1.1.6 funcsigs-1.0.2 futures-3.2.0 gast-0.2.0 grpcio-1.14.1 markdown-2.6.11 mock-2.0.0 numpy-1.15.0 pbr-4.2.0 protobuf-3.6.0 setuptools-39.1.0 six-1.11.0 tensorboard-1.9.0 tensorflow-1.9.0 termcolor-1.1.0 werkzeug-0.14.1 wheel-0.31.1

# Python
import tensorflow as tf
tf.enable_eager_execution()
hello = tf.constant('Hello, TensorFlow !')
print(hello)

pi@raspberrypi:~/bazel $ python3
Python 3.5.3 (default, Jan 19 2017, 14:11:04)
[GCC 6.3.0 20170124] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: compiletime version 3.4 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.5
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: builtins.type size changed, may indicate binary incompatibility. Expected 432, got 412
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 56, got 52
  return f(*args, **kwds)
>>> tf.enable_eager_execution()
>>> hello = tf.constant('Hello, TensorFlow !')
>>> print(hello)
tf.Tensor(b'Hello, TensorFlow !', shape=(), dtype=string)

>>> quit()

pi@raspberrypi:~/bazel $ pip3 install tensorflow==1.3.0
Collecting tensorflow==1.3.0
  Could not find a version that satisfies the requirement tensorflow==1.3.0 (from versions: 0.11.0, 1.8.0, 1.9.0)
No matching distribution found for tensorflow==1.3.0

pi@raspberrypi:~/bazel $ pip3 install tensorflow==1.8.0
Collecting tensorflow==1.8.0
  Downloading https://www.piwheels.org/simple/tensorflow/tensorflow-1.8.0-cp35-none-linux_armv7l.whl (67.0MB)

pi@raspberrypi:~/bazel $ python3
Python 3.5.3 (default, Jan 19 2017, 14:11:04)
[GCC 6.3.0 20170124] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: compiletime version 3.4 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.5
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: builtins.type size changed, may indicate binary incompatibility. Expected 432, got 412
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 56, got 52
  return f(*args, **kwds)
>>> tf.enable_eager_execution()
>>> hello = tf.constant('Hello, TensorFlow !')
>>> print(hello)
tf.Tensor(b'Hello, TensorFlow !', shape=(), dtype=string)

>>> quit()

pi@raspberrypi:~ $ pip install tensorflow # Python 2.7
Collecting tensorflow
  Downloading https://www.piwheels.org/simple/tensorflow/tensorflow-1.9.0-cp27-none-linux_armv7l.whl (66.7MB)
    99% |████████████████████████████████| 66.7MB 401kB/s eta 0:00:01Exception:
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/pip/basecommand.py", line 215, in main
    status = self.run(options, args)
  File "/usr/lib/python2.7/dist-packages/pip/commands/install.py", line 353, in run
    wb.build(autobuilding=True)
  File "/usr/lib/python2.7/dist-packages/pip/wheel.py", line 749, in build
    self.requirement_set.prepare_files(self.finder)
...
  File "/usr/share/python-wheels/CacheControl-0.11.7-py2.py3-none-any.whl/cachecontrol/controller.py", line 275, in cache_response
    self.serializer.dumps(request, response, body=body),
  File "/usr/share/python-wheels/CacheControl-0.11.7-py2.py3-none-any.whl/cachecontrol/serialize.py", line 87, in dumps
    ).encode("utf8"),
MemoryError

sudo sed -i -e "s/^CONF_SWAPSIZE=.*/CONF_SWAPSIZE=2048/g" /etc/dphys-swapfile
cat /etc/dphys-swapfile | grep CONF_SWAPSIZE
# CONF_SWAPSIZE=2048
sudo service dphys-swapfile restart

free
              total        used        free      shared  buff/cache   available
Mem:        1000184       30080      923984       12856       46120      915968
Swap:       2097148           0     2097148

pi@raspberrypi:~ $ python
Python 2.7.13 (default, Nov 24 2017, 17:33:09)
[GCC 6.3.0 20170516] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
/home/pi/.local/lib/python2.7/site-packages/tensorflow/python/framework/tensor_util.py:32: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 56, got 52
  from tensorflow.python.framework import fast_tensor_util
>>> tf.enable_eager_execution()
>>> hello = tf.constant('Hello, TensorFlow !')
>>> print(hello)
tf.Tensor(Hello, TensorFlow !, shape=(), dtype=string)
>>> quit()

# TensorFlowのサンプルをダウンロードするので Gitから Cloneでダウンロードします
cd
git clone https://github.com/tensorflow/tensorflow.git --depth 1

cd tensorflow
cd tensorflow/examples/tutorials/deepdream/

ls -l
# -rw-r--r-- 1 pi pi 993670 Aug 10 09:25 deepdream.ipynb
# -rw-r--r-- 1 pi pi 108340 Aug 10 09:25 pilatus800.jpg
# -rw-r--r-- 1 pi pi   1493 Aug 10 09:25 README.md

# deepdream.ipynbスクリプトを「信頼」する
ipython3 trust deepdream.ipynb
# Signing notebook: deepdream.ipynb

# IPythonを実行する
ipython3 notebook

# IPythonの HTTPサーバを外部からもアクセスできる様に IP制限をしない
# IPythonを起動時にブラウザを起動しない（そもそもコマンドラインなのでブラウザを起動できない）
ipython3 notebook --ip=* --no-browser

NameError: name 'graph' is not defined
pip3 install igraph

NameError: name 'np' is not defined
pip3 install numpy

ImportError: libf77blas.so.3: cannot open shared object file: No such file or directory
sudo apt-get -y install libatlas-base-dev

ImportError: No module named 'tensorflow'

Kernel Restarting
The kernel appears to have died. It will restart automatically.

cd
git clone https://github.com/tensorflow/tensorflow.git --depth 1

cd tensorflow
cd tensorflow/examples/tutorials/deepdream/

ls -l
# -rw-r--r-- 1 pi pi 993670 Aug 10 09:25 deepdream.ipynb
# -rw-r--r-- 1 pi pi 108340 Aug 10 09:25 pilatus800.jpg
# -rw-r--r-- 1 pi pi   1493 Aug 10 09:25 README.md

# deepdream.ipynbスクリプトを「信頼」する
ipython trust deepdream.ipynb

# IPythonを実行する
ipython notebook

# IPythonの HTTPサーバを外部からもアクセスできる様に IP制限をしない
# IPythonを起動時にブラウザを起動しない（そもそもコマンドラインなのでブラウザを起動できない）
ipython notebook --ip=* --no-browser

# boilerplate code
from __future__ import print_function
import os
from io import BytesIO
import numpy as np
from functools import partial
import PIL.Image
from IPython.display import clear_output, Image, display, HTML

import tensorflow as tf

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
<ipython-input-1-0b40bb9ab189> in <module>()
      8 from IPython.display import clear_output, Image, display, HTML
      9
---> 10 import tensorflow as tf

ImportError: No module named tensorflow

pi@raspberrypi:~ $ pip install tensorflow # Python 2.7
...
  Using cached https://files.pythonhosted.org/packages/8c/10/79282747f9169f21c053c562a0baa21815a8c7879be97abd930dbcf862e8/setuptools-39.1.0-py2.py3-none-any.whl
Collecting mock>=2.0.0 (from tensorflow)
Exception:
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/pip/basecommand.py", line 215, in main
    status = self.run(options, args)
  File "/usr/lib/python2.7/dist-packages/pip/commands/install.py", line 353, in run
    wb.build(autobuilding=True)
...
  File "/usr/share/python-wheels/urllib3-1.19.1-py2.py3-none-any.whl/urllib3/connectionpool.py", line 643, in urlopen
    _stacktrace=sys.exc_info()[2])
  File "/usr/share/python-wheels/urllib3-1.19.1-py2.py3-none-any.whl/urllib3/util/retry.py", line 315, in increment
    total -= 1
TypeError: unsupported operand type(s) for -=: 'Retry' and 'int'

# pip installで unsupported operandエラーが出る場合の対処方法 pipを upgrade
sudo python -m pip install --upgrade --force setuptools
sudo python -m pip install --upgrade --force pip

pi@raspberrypi:~/deepdream $ ipython trust dream.ipynb
[TerminalIPythonApp] WARNING | Subcommand `ipython trust` is deprecated and will be removed in future versions.
[TerminalIPythonApp] WARNING | You likely want to use `jupyter trust` in the future
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/usr/lib/python2.7/dist-packages/traitlets/config/application.py", line 657, in launch_instance
...
  File "/usr/lib/python2.7/dist-packages/traitlets/config/application.py", line 445, in initialize_subcommand
    subapp = import_item(subapp)
  File "/usr/lib/python2.7/dist-packages/ipython_genutils/importstring.py", line 31, in import_item
    module = __import__(package, fromlist=[obj])
ImportError: No module named nbformat.sign

sudo pip install jupyter_kernel_gateway

ImportError: No module named shutil_get_terminal_size

pip install backports.shutil_get_terminal_size

import backports
print(backports)

pi@raspberrypi:~/deepdream $ ipython3 trust dream.ipynb
[TerminalIPythonApp] WARNING | Subcommand `ipython trust` is deprecated and will be removed in future versions.
[TerminalIPythonApp] WARNING | You likely want to use `jupyter trust` in the future
[TrustNotebookApp] Writing notebook-signing key to /home/pi/.local/share/jupyter/notebook_secret
Signing notebook: dream.ipynb

cd
git clone https://github.com/hjptriplebee/deep_dream_tensorflow.git
cd deep_dream_tensorflow

python3 main.py --input nature_image.jpg --output nature_image_tf.jpg
python3 main.py --input paint.jpg --output paint_tf.jpg

pi@raspberrypi:~/deep_dream_tensorflow $ python3 main.py --input nature_image.jpg --output nature_image_tf.jpg
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: compiletime version 3.4 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.5
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: builtins.type size changed, may indicate binary incompatibility. Expected 432, got 412
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 56, got 52
  return f(*args, **kwds)
octave num 1/4...
Bus error

ImportError: No module named 'cv2'

sudo pip3 install opencv-python
# Successfully installed opencv-python-3.4.2.17

    from .cv2 import *
ImportError: libjasper.so.1: cannot open shared object file: No such file or directory

sudo apt-get -y install libjasper-dev
# Setting up libjasper1:armhf (1.900.1-debian1-2.4+deb8u1) ...
# Setting up libjasper-dev (1.900.1-debian1-2.4+deb8u1) ...

sudo apt-get -y install libatlas-base-dev
sudo apt-get -y install libqtgui4
sudo apt-get -y install python3-pyqt5
# Setting up python3-pyqt5 (5.7+dfsg-5) ...

Bus error

解決方法不明

cd
mkdir tf
cd tf
git clone https://github.com/EdCo95/deepdream.git
cd deepdream

pi@raspberrypi:~/tf/deepdream $ python simple_dreaming.py
  File "simple_dreaming.py", line 70
    print(score, end=" ")
                    ^
SyntaxError: invalid syntax

python3 simple_dreaming.py
Bus error

pi@raspberrypi:~/tf/deepdream $ python3 simple_dreaming.py
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: compiletime version 3.4 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.5
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: builtins.type size changed, may indicate binary incompatibility. Expected 432, got 412
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 56, got 52
  return f(*args, **kwds)
Bus error

Bus error

解決方法不明

cd
mkdir tf
cd tf
git clone https://github.com/mtyka/deepdream_highres.git
cd deepdream_highres

wget https://storage.googleapis.com/download.tensorflow.org/models/inception5h.zip
unzip inception5h.zip

wget https://github.com/google/deepdream/raw/master/flowers.jpg

# Python 2
python deepdream.py --input flowers.jpg --output output.jpg --layer import/mixed5a_5x5_pre_relu --frames 7 --octaves 10 --iterations 10
ImportError: No module named tensorflow

# Python 3
python3 deepdream.py --input flowers.jpg --output output.jpg --layer import/mixed5a_5x5_pre_relu --frames 7 --octaves 10 --iterations 10
SyntaxError: Missing parentheses in call to 'print'

SyntaxError: Missing parentheses in call to 'print'
print "hoge"
 to
print("hoge")

pi@raspberrypi:~/tf/deepdream_highres $ python deepdream.py --input flowers.jpg --output output.jpg --layer import/mixed5a_5x5_pre_relu --frames 7 --octaves 10 --iterations 10
/home/pi/.local/lib/python2.7/site-packages/tensorflow/python/framework/tensor_util.py:32: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 56, got 52
  from tensorflow.python.framework import fast_tensor_util
--- Available Layers: ---
import/conv2d0_w Features/Channels:  64
import/conv2d0_b Features/Channels:  64
import/conv2d1_w Features/Channels:  64
...
import/softmax1_b Features/Channels:  1008
import/softmax2_w Features/Channels:  1008
import/softmax2_b Features/Channels:  1008
...
import/mixed4d_3x3_bottleneck_pre_relu Features/Channels:  144
import/mixed4d_3x3_bottleneck Features/Channels:  144
import/mixed4d_3x3_pre_relu/conv Features/Channels:  288
...
import/mixed5b_3x3_bottleneck_pre_relu/conv Features/Channels:  192
import/mixed5b_3x3_bottleneck_pre_relu Features/Channels:  192
import/mixed5b_3x3_bottleneck Features/Channels:  192
...
import/output Features/Channels:  1008
import/output1 Features/Channels:  1008
import/output2 Features/Channels:  1008
Number of layers 360
Total number of feature channels: 87322
Chosen layer:
name: "import/mixed5a_5x5_pre_relu"
op: "BiasAdd"
input: "import/mixed5a_5x5_pre_relu/conv"
input: "import/mixed5a_5x5_b"
device: "/device:CPU:0"
attr {
  key: "T"
  value {
    type: DT_FLOAT
  }
}
attr {
  key: "data_format"
  value {
    s: "NHWC"
  }
}

Cycle 0  Res: (240, 320, 3)
 Octave:  0 Res:  (10, 14, 3)
Bus error

cd
cd tensorflow/
cd tensorflow/examples/tutorials/mnist/

pi@raspberrypi:~/tensorflow/tensorflow/examples/tutorials/mnist $ python fully_connected_feed.py
Traceback (most recent call last):
  File "fully_connected_feed.py", line 28, in <module>
    import tensorflow as tf
ImportError: No module named tensorflow

python3 fully_connected_feed.py

pi@raspberrypi:~/tensorflow/tensorflow/examples/tutorials/mnist $ python3 fully_connected_feed.py
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: compiletime version 3.4 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.5
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: builtins.type size changed, may indicate binary incompatibility. Expected 432, got 412
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 56, got 52
  return f(*args, **kwds)
WARNING:tensorflow:From fully_connected_feed.py:120: read_data_sets (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.
Instructions for updating:
Please use alternatives such as official/mnist/dataset.py from tensorflow/models.
WARNING:tensorflow:From /home/pi/.local/lib/python3.5/site-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:260: maybe_download (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.
Instructions for updating:
Please write your own downloading logic.
WARNING:tensorflow:From /home/pi/.local/lib/python3.5/site-packages/tensorflow/contrib/learn/python/learn/datasets/base.py:252: _internal_retry.<locals>.wrap.<locals>.wrapped_fn (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.
Instructions for updating:
Please use urllib or similar directly.
Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes.
WARNING:tensorflow:From /home/pi/.local/lib/python3.5/site-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:262: extract_images (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.
Instructions for updating:
Please use tf.data to implement this functionality.
Extracting /tmp/tensorflow/mnist/input_data/train-images-idx3-ubyte.gz
Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes.
WARNING:tensorflow:From /home/pi/.local/lib/python3.5/site-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:267: extract_labels (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.
Instructions for updating:
Please use tf.data to implement this functionality.
Extracting /tmp/tensorflow/mnist/input_data/train-labels-idx1-ubyte.gz
Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes.
Extracting /tmp/tensorflow/mnist/input_data/t10k-images-idx3-ubyte.gz
Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes.
Extracting /tmp/tensorflow/mnist/input_data/t10k-labels-idx1-ubyte.gz
WARNING:tensorflow:From /home/pi/.local/lib/python3.5/site-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:290: DataSet.__init__ (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.
Instructions for updating:
Please use alternatives such as official/mnist/dataset.py from tensorflow/models.
Step 0: loss = 2.30 (0.664 sec)
Step 100: loss = 2.21 (0.012 sec)
Step 200: loss = 1.96 (0.012 sec)
Step 300: loss = 1.66 (0.011 sec)
Step 400: loss = 1.48 (0.011 sec)
Step 500: loss = 1.09 (0.011 sec)
Step 600: loss = 0.97 (0.014 sec)
Step 700: loss = 0.73 (0.012 sec)
Step 800: loss = 0.67 (0.013 sec)
Step 900: loss = 0.81 (0.012 sec)
Training Data Eval:
Num examples: 55000  Num correct: 47249  Precision @ 1: 0.8591
Validation Data Eval:
Num examples: 5000  Num correct: 4333  Precision @ 1: 0.8666
Test Data Eval:
Num examples: 10000  Num correct: 8609  Precision @ 1: 0.8609
Step 1000: loss = 0.53 (0.054 sec)
Step 1100: loss = 0.54 (0.342 sec)
Step 1200: loss = 0.42 (0.012 sec)
Step 1300: loss = 0.53 (0.013 sec)
Step 1400: loss = 0.49 (0.012 sec)
Step 1500: loss = 0.42 (0.015 sec)
Step 1600: loss = 0.35 (0.015 sec)
Step 1700: loss = 0.34 (0.012 sec)
Step 1800: loss = 0.49 (0.013 sec)
Step 1900: loss = 0.25 (0.015 sec)
Training Data Eval:
Num examples: 55000  Num correct: 49152  Precision @ 1: 0.8937
Validation Data Eval:
Num examples: 5000  Num correct: 4500  Precision @ 1: 0.9000
Test Data Eval:
Num examples: 10000  Num correct: 8995  Precision @ 1: 0.8995

cd
cd tensorflow/
cd tensorflow/examples/tutorials/word2vec

python3 word2vec_basic.py

pi@raspberrypi:~/tensorflow/tensorflow/examples/tutorials/word2vec $ ls -l
-rw-r--r-- 1 pi pi   429 Aug 12 19:41 BUILD
-rw-r--r-- 1 pi pi     0 Aug 12 19:41 __init__.py
-rw-r--r-- 1 pi pi 12795 Aug 12 19:41 word2vec_basic.py

pi@raspberrypi:~/tensorflow/tensorflow/examples/tutorials/word2vec $ python3 word2vec_basic.py
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: compiletime version 3.4 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.5
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: builtins.type size changed, may indicate binary incompatibility. Expected 432, got 412
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 56, got 52
  return f(*args, **kwds)
Found and verified text8.zip
Data size 17005207
Most common words (+UNK) [['UNK', 418391], ('the', 1061396), ('of', 593677), ('and', 416629), ('one', 411764)]
Sample data [5242, 3082, 12, 6, 195, 2, 3137, 46, 59, 156] ['anarchism', 'originated', 'as', 'a', 'term', 'of', 'abuse', 'first', 'used', 'against']
3082 originated -> 5242 anarchism
3082 originated -> 12 as
12 as -> 6 a
12 as -> 3082 originated
6 a -> 195 term
6 a -> 12 as
195 term -> 2 of
195 term -> 6 a
Initialized
Average loss at step  0 :  317.18463134765625
Nearest to an: initiator, chinese, quadriceps, marwan, manchurian, frontline, federline, townspeople,
Nearest to more: thessalonica, abstractions, forensics, daimlerchrysler, aficionados, ideas, third, inaccuracies,
Nearest to be: fuerza, micrometers, ought, carpets, lao, conductivity, inference, gardeners,
Nearest to over: ramos, discussed, emigrate, terms, nchen, slough, tryptophan, battlefield,
Nearest to at: pls, manifests, intensive, basque, thoroughbreds, olav, mania, conveyor,
Nearest to first: thunderbirds, geo, supposes, disability, calendars, cindy, recipient, leeds,
Nearest to to: glycemic, impractical, prototype, ller, dramatically, cetacean, promo, flower,
Nearest to he: celebrities, generalisation, pines, bixby, emulators, import, floated, cmf,
Nearest to some: whitepapers, trowa, russa, antagonist, tunnelling, kitts, lozada, allan,
Nearest to been: icosidodecahedron, taiping, tranquility, contender, dioscorus, mawr, phoenicia, civilised,
Nearest to years: hashes, modernisation, soar, flutie, martens, chinon, said, upcoming,
Nearest to these: chandrasekhar, afar, corporation, woogie, pasolini, knicks, bicycling, handedness,
Nearest to states: pedals, intersecting, letterboxes, marcus, neopaganism, ravenna, drinks, hemorrhagic,
Nearest to united: ray, curran, didgeridoo, zealot, trusteeship, yarns, gur, anything,
Nearest to will: subversion, packaged, mondegreen, coro, onions, giambattista, motivational, moneo,
Nearest to from: nazarbayev, annapolis, sark, overcomes, fish, wadi, wrigley, evaded,
Average loss at step  2000 :  113.80575774669647
Average loss at step  4000 :  52.69416046977043
Average loss at step  6000 :  33.0515801718235
Average loss at step  8000 :  23.336391828536986
Average loss at step  10000 :  17.929039194107055
Nearest to an: the, sedatives, horror, gb, rights, zf, tricks, autism,
Nearest to more: third, deciduous, cc, daughters, deserted, bones, hydride, true,
Nearest to be: conventions, accessories, conductivity, little, revolution, meet, switch, boards,
Nearest to over: discussed, terms, invented, nine, levy, baldwin, hours, determination,
Nearest to at: wasn, pls, absence, in, odds, vocals, on, technische,
Nearest to first: mosque, dynasties, ruled, coke, closer, sender, and, molinari,
Nearest to to: and, in, coke, paradigms, grow, prizes, rogers, trial,
Nearest to he: it, been, starts, table, phi, celebrities, they, and,
Nearest to some: vaginal, incorporate, online, agave, indefinitely, hostility, agincourt, tonight,
Nearest to been: he, phi, grant, levitt, hurricane, coke, llewellyn, lymphoma,
Nearest to years: zero, nine, rotate, said, aberdeenshire, fao, film, boards,
Nearest to these: restricted, corporation, afar, cl, whale, teammate, aberdeenshire, performed,
Nearest to states: pedals, drinks, gb, two, lunar, badges, secretly, mathbf,
Nearest to united: ray, anything, climatic, ac, coke, assistive, omnibus, involving,
Nearest to will: macedonians, gland, amateurs, used, against, serious, archive, shield,
Nearest to from: in, of, and, with, child, der, omega, gland,
Average loss at step  12000 :  14.0876000957489
Average loss at step  14000 :  11.830730155587196
Average loss at step  16000 :  10.117526993513108
Average loss at step  18000 :  8.394923798322678
Average loss at step  20000 :  8.038336719870568
Nearest to an: the, dasyprocta, rights, gb, horror, initiator, sedatives, concern,
Nearest to more: third, deciduous, thessalonica, hydride, sufferers, daughters, bones, operatorname,
Nearest to be: have, conventions, is, conductivity, accessories, was, zealand, are,
Nearest to over: discussed, invented, machined, terms, agouti, levy, determination, five,
Nearest to at: in, and, wasn, on, from, filtration, for, peptide,
Nearest to first: agouti, dynasties, pelagius, equestrian, morphisms, mosque, coke, peptide,
Nearest to to: and, for, nine, with, in, not, of, by,
Nearest to he: it, they, been, she, starts, and, boulevard, also,
Nearest to some: the, many, this, vaginal, circ, peptide, tonight, yin,
Nearest to been: he, is, levitt, long, by, tal, providers, grant,
Nearest to years: operatorname, nine, hello, rotate, zero, seven, aberdeenshire, fao,
Nearest to these: circ, restricted, corporation, afar, the, onyx, often, latinus,
Nearest to states: pedals, intersecting, drinks, gb, circ, two, northwestern, kid,
Nearest to united: antwerp, ray, involving, climatic, anything, matched, assistive, coke,
Nearest to will: against, neologisms, macedonians, onions, xvii, gland, amateurs, ecuador,
Nearest to from: in, and, by, with, antwerp, of, on, for,
Average loss at step  22000 :  7.031585393786431
Average loss at step  24000 :  6.858001116275787
Average loss at step  26000 :  6.695932538747788
Average loss at step  28000 :  6.394147111177444
Average loss at step  30000 :  5.914885578513146
Nearest to an: the, dasyprocta, abiathar, concern, rights, sponsors, initiator, gb,
Nearest to more: deciduous, third, thessalonica, sufferers, hydride, officeholders, cfa, bones,
Nearest to be: have, by, are, was, conventions, accessories, is, conductivity,
Nearest to over: five, invented, discussed, levy, determination, machined, agouti, terms,
Nearest to at: in, from, for, wasn, and, on, with, anthropic,
Nearest to first: agouti, dynasties, pelagius, coke, peptide, equestrian, forerunner, mosque,
Nearest to to: for, with, nine, not, and, in, will, can,
Nearest to he: it, they, she, boulevard, there, starts, no, encryption,
Nearest to some: many, the, vaginal, all, this, circ, agave, tombs,
Nearest to been: by, was, he, long, once, rawlinson, levitt, be,
Nearest to years: six, operatorname, hashes, rotate, leq, versions, exists, io,
Nearest to these: primigenius, circ, restricted, the, afar, corporation, aberdeenshire, often,
Nearest to states: pedals, intersecting, drinks, gb, northwestern, rebounded, circ, ventricular,
Nearest to united: antwerp, ray, earthly, adhd, involving, assaults, matched, assistive,
Nearest to will: to, would, can, against, neologisms, could, ecuador, should,
Nearest to from: in, by, abet, with, antwerp, for, on, at,
Average loss at step  32000 :  5.958182794988155
Average loss at step  34000 :  5.725539566159249
Average loss at step  36000 :  5.798080120563507
Average loss at step  38000 :  5.5125210976600645
Average loss at step  40000 :  5.238765237927437
Nearest to an: the, concern, sponsors, rights, abiathar, tau, pars, initiator,
Nearest to more: deciduous, most, officeholders, biosphere, third, sufferers, thessalonica, cfa,
Nearest to be: have, by, was, were, are, been, accessories, conventions,
Nearest to over: determination, invented, discussed, levy, five, furry, akimbo, terms,
Nearest to at: in, on, from, wasn, six, for, peptide, bckgr,
Nearest to first: agouti, pelagius, forerunner, dynasties, peptide, coke, massif, third,
Nearest to to: nine, will, not, for, can, with, viewers, zero,
Nearest to he: it, she, they, there, who, boulevard, starts, zero,
Nearest to some: many, the, all, vaginal, these, this, circ, agave,
Nearest to been: by, be, was, once, he, long, tal, epistles,
Nearest to years: soar, four, versions, accumulator, operatorname, six, bouvet, rotate,
Nearest to these: restricted, some, afar, circ, primigenius, many, the, corporation,
Nearest to states: pedals, intersecting, drinks, hemorrhagic, gb, ventricular, northwestern, rebounded,
Nearest to united: antwerp, involving, ray, of, earthly, assaults, matched, adhd,
Nearest to will: can, would, to, could, should, against, neologisms, ecuador,
Nearest to from: in, abet, on, with, agouti, antwerp, and, for,
Average loss at step  42000 :  5.3715492960214615
Average loss at step  44000 :  5.232873802542686
Average loss at step  46000 :  5.233447861671448
Average loss at step  48000 :  5.260792678833008

pi@raspberrypi:~ $ pip3 install keras
# Installing collected packages: scipy, keras-preprocessing, pyyaml, h5py, keras-applications, keras

# python3なので Python 3系用
sudo apt-get install python3-pandas
# sudo apt-get install python3-numpy (1:1.12.1-3) ※ Keras 2.2.2は numpy-1.15.0
# sudo apt-get install python3-yaml (3.12-1) ※ Keras 2.2.2は PyYAML-3.13
# sudo apt-get install python3-scipy (0.18.1-2) ※ Keras 2.2.2は scipy-1.1.0
# sudo apt-get install python3-h5py (2.7.0-1) ※ Keras 2.2.2は h5py-2.8.0

pi@raspberrypi:~ $ python3
Python 3.5.3 (default, Jan 19 2017, 14:11:04)
[GCC 6.3.0 20170124] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import keras
Using TensorFlow backend.
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: compiletime version 3.4 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.5
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: builtins.type size changed, may indicate binary incompatibility. Expected 432, got 412
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 56, got 52
  return f(*args, **kwds)

>>> print(keras.__version__)
2.2.2

>>> quit()

wget https://raw.githubusercontent.com/keras-team/keras/2.2.2/examples/deep_dream.py

mkdir src
mkdir dest

wget https://github.com/google/deepdream/raw/master/flowers.jpg
mv flowers.jpg ./src

# Python 3で Kerasする
python3 deep_dream.py src/flowers.jpg dest/flowers

pi@raspberrypi:~ $ python deep_dream.py src/flowers.jpg dest/flowers
Traceback (most recent call last):
  File "deep_dream.py", line 14, in <module>
    from keras.preprocessing.image import load_img, save_img, img_to_array
ImportError: No module named keras.preprocessing.image

pi@raspberrypi:~ $ python3 deep_dream.py src/flowers.jpg dest/flowers
Using TensorFlow backend.
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: compiletime version 3.4 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.5
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: builtins.type size changed, may indicate binary incompatibility. Expected 432, got 412
  return f(*args, **kwds)
/usr/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 56, got 52
  return f(*args, **kwds)

Downloading data from https://github.com/fchollet/deep-learning-models/releases/download/v0.5/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5
87916544/87910968 [==============================] - 102s 1us/step
Model loaded.
WARNING:tensorflow:Variable += will be deprecated. Use variable.assign_add if you want assignment to the variable value or 'x = x + y' if you want a new python Tensor object.
Processing image shape (122, 163)
Bus error

# pythonなので Python 2系用
sudo apt-get -y install python-pandas
sudo apt-get -y install python-numpy
sudo apt-get -y install python-yaml
sudo apt-get -y install python-scipy
sudo apt-get -y install python-h5py

pip install keras

# install the toolchain
sudo apt-get install build-essential

cd
git clone https://github.com/tensorflow/tensorflow

cd tensorflow
./tensorflow/contrib/lite/download_dependencies.sh

./tensorflow/contrib/lite/build_rpi_lib.sh

# compile a static library
# tensorflow/contrib/lite/gen/lib/rpi_armv7/libtensorflow-lite.a

pi@raspberrypi:~/tensorflow $ ./tensorflow/contrib/lite/build_rpi_lib.sh
+ set -e
+++ dirname ./build_rpi_lib.sh
++ cd .
++ pwd
+ SCRIPT_DIR=/home/pi/tensorflow/tensorflow/contrib/lite
+ cd /home/pi/tensorflow/tensorflow/contrib/lite/../../..
+ CC_PREFIX=arm-linux-gnueabihf-
+ make -j 3 -f tensorflow/contrib/lite/Makefile TARGET=RPI TARGET_ARCH=armv7
/bin/sh: 1: [[: not found
arm-linux-gnueabihf-gcc --std=c++11 -O3 -DNDEBUG -march=armv7-a -mfpu=neon-vfpv4 -funsafe-math-optimization
...
./tensorflow/contrib/lite/schema/schema_generated.h:1214:33: error: no matching function for call to ‘flatbuffers::Verifier::Verify(const flatbuffers::Vector<float>*)’
            verifier.Verify(min()) &&

pi@raspberrypi:~/tensorflow/tensorflow/contrib/lite $ [[
> ^C

pi@raspberrypi:~/tensorflow/tensorflow/contrib/lite $ /bin/bash -c "type [["
[[ is a shell keyword

pi@raspberrypi:~/tensorflow/tensorflow/contrib/lite $ head ./build_rpi_lib.sh
#!/bin/bash -x
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#

test.sh
#!/bin/bash
if [[ "string" == "string" ]]; then
    echo This is Bash
fi

$ readlink -f $(which sh)
/bin/dash

$ sh test.sh
test.sh: 2: test.sh: [[: not found

$ bash test.sh
This is Bash

$ chmod 755 test.sh
$ ./test.sh
This is Bash

pi@raspberrypi:~/tensorflow $ ./tensorflow/contrib/lite/build_rpi_lib.sh
...
./tensorflow/contrib/lite/schema/schema_generated.h:4471:39: error: no matching function for call to ‘flatbuffers::Verifier::Verify(const flatbuffers::Vector<flatbuffers::Offset<tflite::SubGraph> >*)’
            verifier.Verify(subgraphs()) &&
...
./tensorflow/contrib/lite/schema/schema_generated.h:4474:41: error: no matching function for call to ‘flatbuffers::Verifier::Verify(const flatbuffers::String*)’
            verifier.Verify(description()) &&
...
./tensorflow/contrib/lite/schema/schema_generated.h:4476:37: error: no matching function for call to ‘flatbuffers::Verifier::Verify(const flatbuffers::Vector<flatbuffers::Offset<tflite::Buffer> >*)’
            verifier.Verify(buffers()) &&
...

+ pwd
/home/pi/tensorflow


CC_PREFIX=arm-linux-gnueabihf- make -j 1 -f tensorflow/contrib/lite/Makefile TARGET=RPI TARGET_ARCH=armv7

mv /home/pi/tensorflow/tensorflow/contrib/lite/downloads/flatbuffers/ /home/pi/tensorflow/tensorflow/contrib/lite/downloads/flatbuffers_/

#
mkdir /home/pi/tensorflow/tensorflow/contrib/lite/downloads/flatbuffers
mkdir /home/pi/tensorflow/tensorflow/contrib/lite/downloads/flatbuffers/include
cp -r /usr/local/include/flatbuffers /home/pi/tensorflow/tensorflow/contrib/lite/downloads/flatbuffers/include/

cd
git clone https://github.com/google/flatbuffers.git
cd flatbuffers

cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=Release

make

sudo make install

pi@raspberrypi:~/flatbuffers $ cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=Release
-- The C compiler identification is GNU 6.3.0
-- The CXX compiler identification is GNU 6.3.0
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Configuring done
-- Generating done
-- Build files have been written to: /home/pi/flatbuffers

pi@raspberrypi:~/flatbuffers $ make
Scanning dependencies of target flatc
[  1%] Building CXX object CMakeFiles/flatc.dir/src/code_generators.cpp.o
[  3%] Building CXX object CMakeFiles/flatc.dir/src/idl_parser.cpp.o
[  5%] Building CXX object CMakeFiles/flatc.dir/src/idl_gen_text.cpp.o
[  7%] Building CXX object CMakeFiles/flatc.dir/src/reflection.cpp.o
[  9%] Building CXX object CMakeFiles/flatc.dir/src/util.cpp.o
...
Scanning dependencies of target flathash
[ 98%] Building CXX object CMakeFiles/flathash.dir/src/flathash.cpp.o
[100%] Linking CXX executable flathash
[100%] Built target flathash

pi@raspberrypi:~/flatbuffers $ sudo make install
[ 45%] Built target flatc
[ 62%] Built target flattests
[ 68%] Built target flatsamplebinary
[ 84%] Built target flatsampletext
[ 96%] Built target flatbuffers
[100%] Built target flathash
Install the project...
-- Install configuration: "Release"
-- Installing: /usr/local/include/flatbuffers
-- Installing: /usr/local/include/flatbuffers/reflection.h
-- Installing: /usr/local/include/flatbuffers/util.h
-- Installing: /usr/local/include/flatbuffers/hash.h
-- Installing: /usr/local/include/flatbuffers/flatbuffers.h
-- Installing: /usr/local/include/flatbuffers/code_generators.h
-- Installing: /usr/local/include/flatbuffers/stl_emulation.h
-- Installing: /usr/local/include/flatbuffers/grpc.h
-- Installing: /usr/local/include/flatbuffers/reflection_generated.h
-- Installing: /usr/local/include/flatbuffers/registry.h
-- Installing: /usr/local/include/flatbuffers/idl.h
-- Installing: /usr/local/include/flatbuffers/base.h
-- Installing: /usr/local/include/flatbuffers/minireflect.h
-- Installing: /usr/local/include/flatbuffers/flatc.h
-- Installing: /usr/local/include/flatbuffers/flexbuffers.h
-- Installing: /usr/local/lib/cmake/flatbuffers/FlatbuffersConfig.cmake
-- Installing: /usr/local/lib/cmake/flatbuffers/FlatbuffersConfigVersion.cmake
-- Installing: /usr/local/lib/libflatbuffers.a
-- Installing: /usr/local/lib/cmake/flatbuffers/FlatbuffersTargets.cmake
-- Installing: /usr/local/lib/cmake/flatbuffers/FlatbuffersTargets-release.cmake
-- Installing: /usr/local/bin/flatc
-- Installing: /usr/local/lib/cmake/flatbuffers/FlatcTargets.cmake
-- Installing: /usr/local/lib/cmake/flatbuffers/FlatcTargets-release.cmake

Raspberry Piで Google謹製の Bazelビルドツールをビルドして TensorFlowを自己ビルドする方法

Raspberry Piで TensorFlow Deep Learning Frameworkを自己ビルドする方法

cd
git clone https://github.com/tensorflow/tensorflow

cd tensorflow
# where Branch is the desired branch
git checkout r1.7

# Python 2.7
sudo apt-get install python-numpy python-dev python-pip python-wheel

# Python 3.n
sudo apt-get install python3-numpy python3-dev python3-pip python3-wheel

./configure

pi@raspberrypi:~/tensorflow $ ./configure
Cannot find bazel. Please install bazel.
Configuration finished

pi@raspberrypi:~/tensorflow $ sudo apt-get install bazel
Reading package lists... Done
Building dependency tree
Reading state information... Done
E: Unable to locate package bazel

pi@raspberrypi:~/tensorflow $ apt-cache search bazel
（bazelパッケージが無い）

cd
git clone https://github.com/bazelbuild/bazel.git --depth 1

cd bazel

# scripts/bootstrap/compile.sh
export BAZEL_JAVAC_OPTS="-J-Xmx500M"

  # Use BAZEL_JAVAC_OPTS to pass additional arguments to javac, e.g.,
  # export BAZEL_JAVAC_OPTS="-J-Xmx2g -J-Xms200m"
  # Useful if your system chooses too small of a max heap for javac.
  # We intentionally rely on shell word splitting to allow multiple
  # additional arguments to be passed to javac.
  run "${JAVAC}" -classpath "${classpath}" -sourcepath "${sourcepath}" \
      -d "${output}/classes" -source "$JAVA_VERSION" -target "$JAVA_VERSION" \
      -encoding UTF-8 ${BAZEL_JAVAC_OPTS} "@${paramfile}"

pi@raspberrypi:~/bazel $ ./compile.sh
🍃  Building Bazel from scratch
ERROR: Must specify PROTOC if not bootstrapping from the distribution artifact

pi@raspberrypi:~/bazel $ ls -l /usr/bin/protoc
-rwxr-xr-x 1 root root 9856 Nov 25  2016 /usr/bin/protoc

export PROTOC=/usr/bin/protoc

pi@raspberrypi:~/bazel $ ./compile.sh
 Building Bazel from scratch
ERROR: Must specify GRPC_JAVA_PLUGIN if not bootstrapping from the distribution artifact

# boilerplate code
from __future__ import print_function
import os
from io import BytesIO
import numpy as np
from functools import partial
import PIL.Image
from IPython.display import clear_output, Image, display, HTML

import tensorflow as tf


!wget -nc https://storage.googleapis.com/download.tensorflow.org/models/inception5h.zip && unzip -n inception5h.zip


model_fn = 'tensorflow_inception_graph.pb'

# creating TensorFlow session and loading the model
graph = tf.Graph()
sess = tf.InteractiveSession(graph=graph)
with tf.gfile.FastGFile(model_fn, 'rb') as f:
    graph_def = tf.GraphDef()
    graph_def.ParseFromString(f.read())
t_input = tf.placeholder(np.float32, name='input') # define the input tensor
imagenet_mean = 117.0
t_preprocessed = tf.expand_dims(t_input-imagenet_mean, 0)
tf.import_graph_def(graph_def, {'input':t_preprocessed})


layers = [op.name for op in graph.get_operations() if op.type=='Conv2D' and 'import/' in op.name]
feature_nums = [int(graph.get_tensor_by_name(name+':0').get_shape()[-1]) for name in layers]

print('Number of layers', len(layers))
print('Total number of feature channels:', sum(feature_nums))


# Helper functions for TF Graph visualization

def strip_consts(graph_def, max_const_size=32):
    """Strip large constant values from graph_def."""
    strip_def = tf.GraphDef()
    for n0 in graph_def.node:
        n = strip_def.node.add()
        n.MergeFrom(n0)
        if n.op == 'Const':
            tensor = n.attr['value'].tensor
            size = len(tensor.tensor_content)
            if size > max_const_size:
                tensor.tensor_content = tf.compat.as_bytes("<stripped %d bytes>"%size)
    return strip_def

def rename_nodes(graph_def, rename_func):
    res_def = tf.GraphDef()
    for n0 in graph_def.node:
        n = res_def.node.add()
        n.MergeFrom(n0)
        n.name = rename_func(n.name)
        for i, s in enumerate(n.input):
            n.input[i] = rename_func(s) if s[0]!='^' else '^'+rename_func(s[1:])
    return res_def

def show_graph(graph_def, max_const_size=32):
    """Visualize TensorFlow graph."""
    if hasattr(graph_def, 'as_graph_def'):
        graph_def = graph_def.as_graph_def()
    strip_def = strip_consts(graph_def, max_const_size=max_const_size)
    code = """
        <script>
          function load() {{
            document.getElementById("{id}").pbtxt = {data};
          }}
        </script>
        <link rel="import" href="https://tensorboard.appspot.com/tf-graph-basic.build.html" onload=load()>
        <div style="height:600px">
          <tf-graph-basic
              id="{id}"
              ></tf-graph-basic>
        </div>
    """.format(data=repr(str(strip_def)), id='graph'+str(np.random.rand()))

    iframe = """
        <iframe seamless style="width:800px;height:620px;border:0"
            srcdoc="{}"
            ></iframe>
    """.format(code.replace('"', '&quot;'))
    display(HTML(iframe))

# Visualizing the network graph. Be sure expand the "mixed" nodes to see their
# internal structure. We are going to visualize "Conv2D" nodes.
tmp_def = rename_nodes(graph_def, lambda s:"/".join(s.split('_',1)))
show_graph(tmp_def)


# Picking some internal layer. Note that we use outputs before applying the ReLU nonlinearity
# to have non-zero gradients for features with negative initial activations.
layer = 'mixed4d_3x3_bottleneck_pre_relu'
channel = 139 # picking some feature channel to visualize

# start with a gray image with a little noise
img_noise = np.random.uniform(size=(224,224,3)) + 100.0

def showarray(a, fmt='jpeg'):
    a = np.uint8(np.clip(a, 0, 1)*255)
    f = BytesIO()
    PIL.Image.fromarray(a).save(f, fmt)
    display(Image(data=f.getvalue()))

def visstd(a, s=0.1):
    '''Normalize the image range for visualization'''
    return (a-a.mean())/max(a.std(), 1e-4)*s + 0.5

def T(layer):
    '''Helper for getting layer output tensor'''
    return graph.get_tensor_by_name("import/%s:0"%layer)

def render_naive(t_obj, img0=img_noise, iter_n=20, step=1.0):
    t_score = tf.reduce_mean(t_obj) # defining the optimization objective
    t_grad = tf.gradients(t_score, t_input)[0] # behold the power of automatic differentiation!

    img = img0.copy()
    for i in range(iter_n):
        g, score = sess.run([t_grad, t_score], {t_input:img})
        # normalizing the gradient, so the same step size should work
        g /= g.std()+1e-8         # for different layers and networks
        img += g*step
        print(score, end = ' ')
    clear_output()
    showarray(visstd(img))

render_naive(T(layer)[:,:,:,channel])


def tffunc(*argtypes):
    '''Helper that transforms TF-graph generating function into a regular one.
    See "resize" function below.
    '''
    placeholders = list(map(tf.placeholder, argtypes))
    def wrap(f):
        out = f(*placeholders)
        def wrapper(*args, **kw):
            return out.eval(dict(zip(placeholders, args)), session=kw.get('session'))
        return wrapper
    return wrap

# Helper function that uses TF to resize an image
def resize(img, size):
    img = tf.expand_dims(img, 0)
    return tf.image.resize_bilinear(img, size)[0,:,:,:]
resize = tffunc(np.float32, np.int32)(resize)


def calc_grad_tiled(img, t_grad, tile_size=512):
    '''Compute the value of tensor t_grad over the image in a tiled way.
    Random shifts are applied to the image to blur tile boundaries over
    multiple iterations.'''
    sz = tile_size
    h, w = img.shape[:2]
    sx, sy = np.random.randint(sz, size=2)
    img_shift = np.roll(np.roll(img, sx, 1), sy, 0)
    grad = np.zeros_like(img)
    for y in range(0, max(h-sz//2, sz),sz):
        for x in range(0, max(w-sz//2, sz),sz):
            sub = img_shift[y:y+sz,x:x+sz]
            g = sess.run(t_grad, {t_input:sub})
            grad[y:y+sz,x:x+sz] = g
    return np.roll(np.roll(grad, -sx, 1), -sy, 0)


def render_multiscale(t_obj, img0=img_noise, iter_n=10, step=1.0, octave_n=3, octave_scale=1.4):
    t_score = tf.reduce_mean(t_obj) # defining the optimization objective
    t_grad = tf.gradients(t_score, t_input)[0] # behold the power of automatic differentiation!

    img = img0.copy()
    for octave in range(octave_n):
        if octave>0:
            hw = np.float32(img.shape[:2])*octave_scale
            img = resize(img, np.int32(hw))
        for i in range(iter_n):
            g = calc_grad_tiled(img, t_grad)
            # normalizing the gradient, so the same step size should work
            g /= g.std()+1e-8         # for different layers and networks
            img += g*step
            print('.', end = ' ')
        clear_output()
        showarray(visstd(img))

render_multiscale(T(layer)[:,:,:,channel])


k = np.float32([1,4,6,4,1])
k = np.outer(k, k)
k5x5 = k[:,:,None,None]/k.sum()*np.eye(3, dtype=np.float32)

def lap_split(img):
    '''Split the image into lo and hi frequency components'''
    with tf.name_scope('split'):
        lo = tf.nn.conv2d(img, k5x5, [1,2,2,1], 'SAME')
        lo2 = tf.nn.conv2d_transpose(lo, k5x5*4, tf.shape(img), [1,2,2,1])
        hi = img-lo2
    return lo, hi

def lap_split_n(img, n):
    '''Build Laplacian pyramid with n splits'''
    levels = []
    for i in range(n):
        img, hi = lap_split(img)
        levels.append(hi)
    levels.append(img)
    return levels[::-1]

def lap_merge(levels):
    '''Merge Laplacian pyramid'''
    img = levels[0]
    for hi in levels[1:]:
        with tf.name_scope('merge'):
            img = tf.nn.conv2d_transpose(img, k5x5*4, tf.shape(hi), [1,2,2,1]) + hi
    return img

def normalize_std(img, eps=1e-10):
    '''Normalize image by making its standard deviation = 1.0'''
    with tf.name_scope('normalize'):
        std = tf.sqrt(tf.reduce_mean(tf.square(img)))
        return img/tf.maximum(std, eps)

def lap_normalize(img, scale_n=4):
    '''Perform the Laplacian pyramid normalization.'''
    img = tf.expand_dims(img,0)
    tlevels = lap_split_n(img, scale_n)
    tlevels = list(map(normalize_std, tlevels))
    out = lap_merge(tlevels)
    return out[0,:,:,:]

# Showing the lap_normalize graph with TensorBoard
lap_graph = tf.Graph()
with lap_graph.as_default():
    lap_in = tf.placeholder(np.float32, name='lap_in')
    lap_out = lap_normalize(lap_in)
show_graph(lap_graph)


def render_lapnorm(t_obj, img0=img_noise, visfunc=visstd,
                   iter_n=10, step=1.0, octave_n=3, octave_scale=1.4, lap_n=4):
    t_score = tf.reduce_mean(t_obj) # defining the optimization objective
    t_grad = tf.gradients(t_score, t_input)[0] # behold the power of automatic differentiation!
    # build the laplacian normalization graph
    lap_norm_func = tffunc(np.float32)(partial(lap_normalize, scale_n=lap_n))

    img = img0.copy()
    for octave in range(octave_n):
        if octave>0:
            hw = np.float32(img.shape[:2])*octave_scale
            img = resize(img, np.int32(hw))
        for i in range(iter_n):
            g = calc_grad_tiled(img, t_grad)
            g = lap_norm_func(g)
            img += g*step
            print('.', end = ' ')
        clear_output()
        showarray(visfunc(img))

render_lapnorm(T(layer)[:,:,:,channel])


render_lapnorm(T(layer)[:,:,:,65])


render_lapnorm(T('mixed3b_1x1_pre_relu')[:,:,:,101])


render_lapnorm(T(layer)[:,:,:,65]+T(layer)[:,:,:,139], octave_n=4)


def render_deepdream(t_obj, img0=img_noise,
                     iter_n=10, step=1.5, octave_n=4, octave_scale=1.4):
    t_score = tf.reduce_mean(t_obj) # defining the optimization objective
    t_grad = tf.gradients(t_score, t_input)[0] # behold the power of automatic differentiation!

    # split the image into a number of octaves
    img = img0
    octaves = []
    for i in range(octave_n-1):
        hw = img.shape[:2]
        lo = resize(img, np.int32(np.float32(hw)/octave_scale))
        hi = img-resize(lo, hw)
        img = lo
        octaves.append(hi)

    # generate details octave by octave
    for octave in range(octave_n):
        if octave>0:
            hi = octaves[-octave]
            img = resize(img, hi.shape[:2])+hi
        for i in range(iter_n):
            g = calc_grad_tiled(img, t_grad)
            img += g*(step / (np.abs(g).mean()+1e-7))
            print('.',end = ' ')
        clear_output()
        showarray(img/255.0)


img0 = PIL.Image.open('pilatus800.jpg')
img0 = np.float32(img0)
showarray(img0/255.0)


render_deepdream(tf.square(T('mixed4c')), img0)


render_deepdream(T(layer)[:,:,:,139], img0)

NVIDIA Jetson Nano 開発者キットを買ってみた。メモリ容量 4GB LPDDR4 RAM

Raspberry Piでメモリを馬鹿食いするアプリ用に不要なサービスを停止してフリーメモリを増やす方法

【成功版】最新版の Darknetに digitalbrain79版の Darknet with NNPACKの NNPACK処理を適用する

【成功版】Raspberry Piで NNPACK対応版の Darknet Neural Network Frameworkをビルドする方法

【成功版】Raspberry Piで Darknet Neural Network Frameworkをビルドする方法

【成功版】Raspberry Piに TensorFlow Deep Learning Frameworkをインストールする方法

Raspberry Piで TensorFlow Deep Learning Frameworkを自己ビルドする方法

Raspberry Piで Caffe Deep Learning Frameworkで物体認識を行なってみるテスト

【ビルド版】Raspberry Piで DeepDreamを動かしてキモイ絵をモリモリ量産 Caffe Deep Learning Framework

【インストール版】Raspberry Piで DeepDreamを動かしてキモイ絵をモリモリ量産 Caffe Deep Learning

Raspberry Piで Caffe2 Deep Learning Frameworkをソースコードからビルドする方法

Orange Pi PC 2の 64bitのチカラで DeepDreamしてキモイ絵を高速でモリモリ量産してみるテスト

Raspberry Piに Jupyter Notebookをインストールして拡張子 ipynb形式の IPythonを動かす

Raspberry Piで Deep Learningフレームワーク Chainerをインストールしてみる

Raspberry Piで DeepBeliefSDKをビルドして画像認識フレームワークを動かす方法

Raspberry Piで Microsoftの ELLをビルドする方法

Raspberry Piで MXNet port of SSD Single Shot MultiBoxを動かして画像の物体検出をする方法

Raspberry Piで Apache MXNet Incubatingをビルドする方法

Raspberry Piで OpenCVの Haar Cascade Object Detectionでリアルタイムにカメラ映像の顔検出を行なってみる

Raspberry Piで NNPACKをビルドする方法

Raspberry Pi 3の Linuxコンソール上で使用する各種コマンドまとめ