For anyone landing here, the following answer was provided (by a googler) on: https://stackoverflow.com/questions/42922948/why-use-tensorflow-gfile-for-file-i-o/


&gt; The main roles of the tf.gfile module are:
&gt; 
&gt; 1. To provide an API that is close to Python&#39;s file objects, and
&gt; 
&gt; 2. To provide an implementation based on TensorFlow&#39;s C++ FileSystem API.
&gt; 
&gt; The C++ FileSystem API supports multiple file system implementations,
&gt; including local files, Google Cloud Storage (using a `gs://` prefix),
&gt; and HDFS (using an `hdfs://` prefix). TensorFlow exports these as
&gt; `tf.gfile`, so that you can use these implementations for saving and
&gt; loading checkpoints, writing TensorBoard logs, and accessing training
&gt; data (among other uses). However, if all of your files are local, you
&gt; can use the regular Python file API without any problem.



As you correctly point out `tf.gfile` is an abstraction for accessing the filesystem and is documented [here](https://www.tensorflow.org/api_docs/python/tf/gfile). It is recommended over using plain python API since it provides some level of portability.

I&#39;m using [react-router v4](https://reacttraining.com/react-router/) and [material-ui](http://www.material-ui.com/) in my React app. I was wondering how to change the URL once there is a click on a `GridTile` within [GridList](http://www.material-ui.com/#/components/grid-list).

My initial idea was to use a handler for `onTouchTap`. However, the only way I can see to redirect is by using the components `Redirect` or `Link`. How could I change the URL without rendering those two components?

I&#39;ve tried `this.context.router.push(&#39;/foo&#39;)` but it doesn&#39;t seem to work.

Changing the URL in react-router v4 without using Redirect or Link

Is there any difference if i use one space, two or four spaces per indent level in **YAML**?

Are there any specific rules for space numbers per Structure type??

For example **4** spaces for nesting **maps** , **1** space per **list item** etc??


I am writing a yaml configuration file for elastic beanstalk .ebextensions and i am having really hard time constructing this correctly. Although i have valid yaml in [YAML Validator][1] elastic beanstalk seems to understand a different structure.


  [1]: http://www.yamllint.com/

YAML How many spaces per indent?

I&#39;ve seen people using several functions from `tf.gfile` such as `tf.gfile.GFile` or `tf.gfile.Exists`. I have the idea that `tf.gfile` deals with files. However, I haven&#39;t been able to find the official documentation to see what else it offers. 

It&#39;d be great if you could help me with it.

What does tf.gfile do in TensorFlow?

<p>I've seen people using several functions from <code>tf.gfile</code> such as <code>tf.gfile.GFile</code> or <code>tf.gfile.Exists</code>. I have the idea that <code>tf.gfile</code> deals with files. However, I haven't been able to find the official documentation to see what else it offers.</p>
<p>It'd be great if you could help me with it.</p>


I want to insert a key-value pair into dict if key not in dict.keys().
Basically I could do it with:

    if key not in d.keys():
        d[key] = value

But is there a better way? Or what&#39;s the pythonic solution to this problem?

Python update a key in dict if it doesn&#39;t exist

I want to install Tensorflow 1.o for python on windows.

**This is information for my system.**

    D:\&gt;python --version
    Python 3.5.2 :: Anaconda 4.2.0 (32-bit)
    
    D:\&gt;pip3 --version
    pip 9.0.1 from d:\web\anaconda\lib\site-packages (python 3.5)&#39;

**But, when I execute below command,**

    D:\&gt;pip3 install tensorflow
    Collecting tensorflow
      Could not find a version that satisfies the requirement tensorflow (from versions: )
    No matching distribution found for tensorflow

I don&#39;t understand what the problem is...

And I tried another way...

**This is case when I use Conda**

    (tensorflow) D:\&gt;pip install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/windows/cpu/tensorflow-1.0.0-cp35-cp35m-win_x86_64.whl
    tensorflow-1.0.0-cp35-cp35m-win_x86_64.whl is not a supported wheel on this platform.

What is the problem?

Tensorflow r1.0 : could not a find a version that satisfies the requirement tensorflow

https://github.com/affinelayer/pix2pix-tensorflow/tree/master/tools

An error occurred when compiling &quot;process.py&quot; on the above site.

     python tools/process.py --input_dir data --            operation resize --outp
    ut_dir data2/resize
    data/0.jpg -&gt; data2/resize/0.png

  Traceback (most recent call last):

    File &quot;tools/process.py&quot;, line 235, in &lt;module&gt;
      main()
    File &quot;tools/process.py&quot;, line 167, in main
      src = load(src_path)
    File &quot;tools/process.py&quot;, line 113, in load
      contents = open(path).read()
          File&quot;/home/user/anaconda3/envs/tensorflow_2/lib/python3.5/codecs.py&quot;, line 321, in decode
      (result, consumed) = self._buffer_decode(data, self.errors, final)
    UnicodeDecodeError: &#39;utf-8&#39; codec can&#39;t decode     byte 0xff in position 0: invalid start byte



What is the cause of the error?
Python&#39;s version is 3.5.2.

error UnicodeDecodeError: &#39;utf-8&#39; codec can&#39;t decode byte 0xff in position 0: invalid start byte

I have a Python dataframe with about 1,500 rows and 15 columns. With one specific column I would like to remove the first 3 characters of each row. As a simple example here is a dataframe:

    import pandas as pd

    d = {
        &#39;Report Number&#39;:[&#39;8761234567&#39;, &#39;8679876543&#39;,&#39;8994434555&#39;],
        &#39;Name&#39;         :[&#39;George&#39;, &#39;Bill&#39;, &#39;Sally&#39;]
         }

    d = pd.DataFrame(d)

I would like to remove the first three characters from each field in the `Report Number` column of dataframe `d`.

Remove first x number of characters from each row in a column of a Python dataframe

How can a pre-existing conda environment be updated with another .yml file. This is extremely helpful when working on projects that have multiple requirement files, i.e. `base.yml, local.yml, production.yml`, etc.

For example, below is a `base.yml` file has conda-forge, conda, and pip packages:

base.yml

    name: myenv
    channels:
      - conda-forge
    dependencies:
      - django=1.10.5
      - pip:
        - django-crispy-forms==1.6.1

The actual environment is created with:
`conda env create -f base.yml`.

Later on, additional packages need to be added to `base.yml`. Another file, say `local.yml`, needs to import those updates.

Previous attempts to accomplish this include:

creating a `local.yml` file with an import definition:

    channels:

    dependencies:
      - pip:
        - boto3==1.4.4
    imports:
      - requirements/base. 

And then run the command:
`conda install -f local.yml`. 

This does not work. Any thoughts?

How to update an existing Conda environment with a .yml file

It is really straightforward to see and understand the scalar values in TensorBoard. However, it&#39;s not clear how to understand histogram graphs. 

For example, they are the histograms of my network weights.

[![enter image description here][1]][2]

(After fixing a bug thanks to sunside)
[![enter image description here][2]][2]
What is the best way to interpret these? Layer 1 weights look mostly flat, what does this mean?

I added the network construction code here.

    X = tf.placeholder(tf.float32, [None, input_size], name=&quot;input_x&quot;)
    x_image = tf.reshape(X, [-1, 6, 10, 1])
    tf.summary.image(&#39;input&#39;, x_image, 4)
    
    # First layer of weights
    with tf.name_scope(&quot;layer1&quot;):
        W1 = tf.get_variable(&quot;W1&quot;, shape=[input_size, hidden_layer_neurons],
                             initializer=tf.contrib.layers.xavier_initializer())
        layer1 = tf.matmul(X, W1)
        layer1_act = tf.nn.tanh(layer1)
        tf.summary.histogram(&quot;weights&quot;, W1)
        tf.summary.histogram(&quot;layer&quot;, layer1)
        tf.summary.histogram(&quot;activations&quot;, layer1_act)
    
    # Second layer of weights
    with tf.name_scope(&quot;layer2&quot;):
        W2 = tf.get_variable(&quot;W2&quot;, shape=[hidden_layer_neurons, hidden_layer_neurons],
                             initializer=tf.contrib.layers.xavier_initializer())
        layer2 = tf.matmul(layer1_act, W2)
        layer2_act = tf.nn.tanh(layer2)
        tf.summary.histogram(&quot;weights&quot;, W2)
        tf.summary.histogram(&quot;layer&quot;, layer2)
        tf.summary.histogram(&quot;activations&quot;, layer2_act)
    
    # Third layer of weights
    with tf.name_scope(&quot;layer3&quot;):
        W3 = tf.get_variable(&quot;W3&quot;, shape=[hidden_layer_neurons, hidden_layer_neurons],
                             initializer=tf.contrib.layers.xavier_initializer())
        layer3 = tf.matmul(layer2_act, W3)
        layer3_act = tf.nn.tanh(layer3)
    
        tf.summary.histogram(&quot;weights&quot;, W3)
        tf.summary.histogram(&quot;layer&quot;, layer3)
        tf.summary.histogram(&quot;activations&quot;, layer3_act)
    
    # Fourth layer of weights
    with tf.name_scope(&quot;layer4&quot;):
        W4 = tf.get_variable(&quot;W4&quot;, shape=[hidden_layer_neurons, output_size],
                             initializer=tf.contrib.layers.xavier_initializer())
        Qpred = tf.nn.softmax(tf.matmul(layer3_act, W4)) # Bug fixed: Qpred = tf.nn.softmax(tf.matmul(layer3, W4))
        tf.summary.histogram(&quot;weights&quot;, W4)
        tf.summary.histogram(&quot;Qpred&quot;, Qpred)
    
    # We need to define the parts of the network needed for learning a policy
    Y = tf.placeholder(tf.float32, [None, output_size], name=&quot;input_y&quot;)
    advantages = tf.placeholder(tf.float32, name=&quot;reward_signal&quot;)
    
    # Loss function
    # Sum (Ai*logp(yi|xi))
    log_lik = -Y * tf.log(Qpred)
    loss = tf.reduce_mean(tf.reduce_sum(log_lik * advantages, axis=1))
    tf.summary.scalar(&quot;Q&quot;, tf.reduce_mean(Qpred))
    tf.summary.scalar(&quot;Y&quot;, tf.reduce_mean(Y))
    tf.summary.scalar(&quot;log_likelihood&quot;, tf.reduce_mean(log_lik))
    tf.summary.scalar(&quot;loss&quot;, loss)
    
    # Learning
    train = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(loss)


  [1]: https://i.stack.imgur.com/RVOA5.jpg
  [2]: https://i.stack.imgur.com/IttNH.jpg

Understanding TensorBoard (weight) histograms

&gt; Note : this question was initially [asked on github](https://github.com/tensorflow/tensorflow/issues/7648#issuecomment-280866214), but it was asked to be here instead

I&#39;m having trouble running tensorflow on gpu, and it does not seems to be the usual cuda&#39;s configuration problem, because everything seems to indicate cuda is properly setup.

The main symptom: when running tensorflow, my gpu is not detected ([the code being run](https://gist.github.com/oelmekki/cafda411bf5c2ea695d984fa98e0995b), and [its output](https://gist.github.com/oelmekki/77235c6b0dde99b3438f190eb557f40f)).

What differs from usual issues is that cuda seems properly installed and running `./deviceQuery` from cuda samples is successful ([output](https://gist.github.com/oelmekki/fe65a15daec45aa90ec33b10b51d3aae)).

I have two graphical cards:

* an old GTX 650 used for my monitors (I don&#39;t want to use that one with tensorflow)
* a GTX 1060 that I want to dedicate to tensorflow

I use:

* [tensorflow-1.0.0](https://pypi.python.org/pypi/tensorflow)
* cuda-8.0 ([ls -l /usr/local/cuda/lib64/libcud*](https://gist.github.com/oelmekki/6e5e9d7d1ea871e1d73efae307efe9ce))
* cudnn-5.1.10
* python-2.7.12
* nvidia-drivers-375.26 (this was installed by cuda and replaced my distro driver package)


I&#39;ve tried:

* adding `/usr/local/cuda/bin/` to `$PATH`
* forcing gpu placement in tensorflow script using `with tf.device(&#39;/gpu:1&#39;):` (and `with tf.device(&#39;/gpu:0&#39;):` when it failed, for good measure)
* whitelisting the gpu I wanted to use with `CUDA_VISIBLE_DEVICES`, in case the presence of my old unsupported card did cause problems
* running the script with sudo (because why not)

Here are the outputs of [nvidia-smi](https://gist.github.com/oelmekki/7bdcb5cc2f791cea561a60f8b21e87b5) and [nvidia-debugdump -l](https://gist.github.com/oelmekki/b83a5a0a72e8924aeb44b70b3598f9b4), in case it&#39;s useful.

At this point, I feel like I have followed all the breadcrumbs and have no idea what I could try else. I&#39;m not even sure if I&#39;m contemplating a bug or a configuration problem. Any advice about how to debug this would be greatly appreciated. Thanks!

**Update**: with the help of Yaroslav on github, I gathered more debugging info by raising log level, but it doesn&#39;t seem to say much about the device selection : https://gist.github.com/oelmekki/760a37ca50bf58d4f03f46d104b798bb

**Update 2**: Using theano detects gpu correctly, but interestingly it complains about cuDNN being too recent, then fallback to cpu ([code ran](https://gist.github.com/oelmekki/34b6e41a0ff2b17ff9f39bcf56d0635a), [output](https://gist.github.com/oelmekki/11626d6b34058337dae64f1915e5a9fe)). Maybe that could be the problem with tensorflow as well?

tensorflow on GPU: no known devices, despite cuda&#39;s deviceQuery returning a &quot;PASS&quot; result

I have a `.tfrecord` but I don&#39;t know how it is structured. How can I inspect the schema to understand what the `.tfrecord` file contains?

All Stackoverflow answers or documentation seem to assume I know the structure of the file.

    reader = tf.TFRecordReader()
    file = tf.train.string_input_producer(&quot;record.tfrecord&quot;)
    _, serialized_record = reader.read(file)
    
    ...HOW TO INSPECT serialized_record...

How to inspect a Tensorflow .tfrecord file?

To save a model in Keras, what are the differences between the output files of:

 1. `model.save()` 
 2. `model.save_weights()`
 3. `ModelCheckpoint()` in the callback

The saved file from `model.save()` is larger than the model from `model.save_weights()`, but significantly larger than a JSON or Yaml model architecture file.  Why is this?  

Restating this: Why is size(model.save()) + size(something) = size(model.save_weights()) + size(model.to_json()), what is that &quot;something&quot;?

Would it be more efficient to just `model.save_weights()` and `model.to_json()`, and load from these than to just do `model.save()` and `load_model()`?  

What are the differences?

Content Type	Original Author	Original Content on Stackoverflow
Question	Chip Huyen	View Question on Stackoverflow
Solution 1 - Python	Yuval Atzmon	View Answer on Stackoverflow
Solution 2 - Python	drpng	View Answer on Stackoverflow

What does tf.gfile do in TensorFlow?

Python Problem Overview

Python Solutions

Solution 1 - Python

Solution 2 - Python

YAML How many spaces per indent?

Changing the URL in react-router v4 without using Redirect or Link

Attributions