diff options
| author |  Vijay Vasudevan <vrv@google.com> | 2015-11-12 16:47:36 -0800 |
|---|---|---|
| committer | Vijay Vasudevan <vrv@google.com> | 2015-11-12 16:47:36 -0800 |
| commit | d50565b35e886e7c3a201ea2f088790ed4b28de4 (patch) | |
| tree | fa6bfce7311467e6c03ec314bb7947a49df7dd8c /tensorflow/g3doc/tutorials/mnist/pros/index.md | |
| parent | 4dffee7f62d81ec9173aba1b0ef6b96e47f8037c (diff) | |
TensorFlow: Upstream changes from afternoon.
Changes:
- Ptrdiff -> DenseIndex change by @jiayq
- Fix to scoping the logging in logging.py by @dga
- Improvement to Conv2DBackpropFilter on CPU by Andy
- Remove lookup table wrappers for the time being (wasn't in our
public API yet) by Yukata
- Add a check similar to numpy to make sure the user isn't in the
tensorflow src directory by @vrv
- More changes for python 3 compat by @girving
- Make dropout preserve shape info from input (@mrry)
- Significant speed improvements by @zheng-xq to BFC allocator to bring
on par (CPU overhead-wise) to the region allocator. Make BFC
allocator the default now that it's working well for a variety
of models.
- Fix a bunch of typos reported by users (@vrv)
- Enable concat for bfloat16 on GPU by Ashish.
Base CL: 107733123
Diffstat (limited to 'tensorflow/g3doc/tutorials/mnist/pros/index.md')
| -rw-r--r-- | tensorflow/g3doc/tutorials/mnist/pros/index.md | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/tensorflow/g3doc/tutorials/mnist/pros/index.md b/tensorflow/g3doc/tutorials/mnist/pros/index.md index a83d3eabd5..0a7d1aeae0 100644 --- a/tensorflow/g3doc/tutorials/mnist/pros/index.md +++ b/tensorflow/g3doc/tutorials/mnist/pros/index.md @@ -39,7 +39,7 @@ Tensorflow relies on a highly efficient C++ backend to do its computation. The connection to this backend is called a session. The common usage for TensorFlow programs is to first create a graph and then launch it in a session. -Here we instead use the convenience `InteractiveSession` class, which +Here we instead use the convenient `InteractiveSession` class, which makes TensorFlow more flexible about how you structure your code. It allows you to interleave operations which build a @@ -232,7 +232,7 @@ print accuracy.eval(feed_dict={x: mnist.test.images, y_: mnist.test.labels}) ## Build a Multilayer Convolutional Network <a class="md-anchor" id="AUTOGENERATED-build-a-multilayer-convolutional-network"></a> Getting 91% accuracy on MNIST is bad. It's almost embarrassingly bad. In this -section, we'll fix that, jumping from a very simple model to something moderatly +section, we'll fix that, jumping from a very simple model to something moderately sophisticated: a small convolutional neural network. This will get us to around 99.2% accuracy -- not state of the art, but respectable. |