[GitHub] ykim362 closed pull request #9254: Add MKL documentation

[GitBox <gi...@apache.org>]

ykim362 closed pull request #9254: Add MKL documentation
URL: https://github.com/apache/incubator-mxnet/pull/9254
 
 
   

This is a PR merged from a forked repository.
As GitHub hides the original diff on merge, it is displayed below for
the sake of provenance:

As this is a foreign pull request (from a fork), the diff is supplied
below (as it won't show otherwise due to GitHub magic):

diff --git a/docs/how_to/build_mxnet_with_intelmkl.md b/docs/how_to/build_mxnet_with_intelmkl.md
new file mode 100644
index 0000000000..f303201c2d
--- /dev/null
+++ b/docs/how_to/build_mxnet_with_intelmkl.md
@@ -0,0 +1,114 @@
+# Building MXNet with MKL-based optimizations:
+
+MXNet can be installed and used with several combinations of development 
+tools and libraries on a variety of platforms. This tutorial provides one 
+such recipe describing steps to build and install MXNet with Intel MKL on 
+an Ubuntu-based system, version 16.04.
+
+## Installing
+
+1. Update your operating system and install dependencies:
+
+   ```bash
+    $ sudo apt-get update && apt-get install -y build-essential git libopencv-dev curl gcc libatlas-base-dev python python-pip python-dev python-opencv graphviz python-scipy python-sklearn
+   ```
+
+2. Delete any old or outdated versions of Intel MKL from `/tmp` before 
+   downloading and copying the newest version to `/usr/local`:
+
+   ```bash
+   $ cd /tmp && rm -rf mklml_lnx_*   
+   $ curl -L https://github.com/01org/mkl-dnn/releases/download/v0.10/mklml_lnx_2018.0.20170908.tgz | tar xz
+   $ cp -a mklml_lnx_*/* /usr/local/.
+   ```
+
+3. Clone the latest version of MXNet from the Apache Incubator GitHub repository 
+   and update submodules recursively:
+
+   ```bash
+   $ git clone --recursive git@github.com:apache/incubator-mxnet.git
+   $ cd incubator-mxnet
+   $ git submodule update --recursive
+   ```
+
+4. Build MXNet for CPU with support for the latest version of MKL:
+
+   ```bash
+   $ make -j $(nproc) USE_OPENCV=1 USE_MKL2017=1 USE_BLAS=atlas
+   ```
+
+5. Set your `PYTHONPATH` variable to location with the `incubator-mxnet/python` 
+   submodule. For example, if you cloned it to the `/opt` directory, you'd update 
+   or add `PYTHONPATH` as follows. Note that your regular `PATH` environment 
+   variable should also be set to read where you cloned incubator-mxnet: 
+
+   ```bash
+   $ export PYTHONPATH=/opt/incubator-mxnet/python
+   ```
+
+6. Lastly, we can install some common Python tools for deep learning, and 
+   verify that everything installed correctly by testing printing of some sample 
+   output:     
+
+   ```bash
+   $ pip install --upgrade pip --user
+   $ pip install --upgrade jupyter graphviz cython pandas bokeh matplotlib opencv-python requests --user
+   $ python -c "import mxnet as mx;print((mx.nd.ones((2,3))*2).asnumpy());" 
+   [[2.2.2.]
+    [2.2.2.]]
+    ```
+  
+## Benchmarks
+  
+A range of standard image classification benchmarks can be found under the 
+  
+```bash
+example/image-classification
+``` 
+  
+directory. We?ll focus on running a benchmark meant to test inference across 
+a range of topologies. 
+  
+### Running Inference Benchmark
+  
+The provided `benchmark_score.py` will run a variety of standard topologies 
+(AlexNet, Inception, ResNet, etc) at a range of batch sizes and report the 
+image-per-second results. Prior to running, set the following environment 
+variables for optimal performance:
+  
+#### For Intel? Xeon? Processors
+  
+```
+export OMP_NUM_THREADS=$(($(grep 'core id' /proc/cpuinfo | sort -u | wc -l)*2))
+```
+   
+   
+#### For Intel? Xeon Phi? Processors
+   
+```
+export OMP_NUM_THREADS=$(($(grep 'core id' /proc/cpuinfo | sort -u | wc -l)))export KMP_AFFINITY=granularity=fine,compact,1,0
+```
+    
+Then run the benchmark with:
+    
+```
+$ python example/image-classification/benchmark_score.py
+ ```
+     
+If everything installed correctly, we should see the image-per-second numbered 
+output for the different topologies and batch sizes:
+     
+```console
+INFO:root:network: alexnet
+INFO:root:device: cpu(0)
+INFO:root:batch size  1, image/sec: XXX
+INFO:root:batch size  2, image/sec: XXX
+...
+INFO:root:batch size 32, image/sec: XXX
+INFO:root:network: vgg
+INFO:root:device: cpu(0)
+INFO:root:batch size  1, image/sec: XXX
+```
+     
+ 
+[Compile with MKL-DNN performance library ]:compile_ml_libs 
\ No newline at end of file


 

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services