Neural Network Inference Using Intel® OpenVINO™

Deploying trained neural network models for inference on different platforms is a challenging task. The inference environment is usually different than the training environment which is typically a data center or a server farm. The inference platform may be power constrained and limited from a software perspective. The model might be trained using one of the many available deep learning frameworks such as Tensorflow, PyTorch, Keras, Caffe, MXNet, etc. Intel® OpenVINO™ provides tools to convert trained models into a framework agnostic representation, including tools to reduce the memory footprint of the model using quantization and graph optimization. It also provides dedicated inference APIs that are optimized for specific hardware platforms, such as Intel® Programmable Acceleration Cards, and Intel® Movidius™ Vision Processing Units.


The Intel® OpenVINO™ toolkit


  1. Model Optimizer

The Model Optimizer is a cross-platform command-line tool that facilitates the transition between the training and deployment environment, performs static model analysis, and adjusts deep learning models for optimal execution on end-point target devices. It is a Python script which takes as input a trained Tensorflow/Caffe model and produces an Intermediate Representation (IR) which consists of a .xml file containing the model definition and a .bin file containing the model weights.

2. Inference Engine

The Inference Engine is a C++ library with a set of C++ classes to infer input data (images) and get a result. The C++ library provides an API to read the Intermediate Representation, set the input and output formats, and execute the model on devices. Each supported target device has a plugin which is a DLL/shared library. It also has support for heterogenous execution to distribute workload across devices. It supports implementing custom layers on a CPU while executing the rest of the model on a accelerator device.


  1. Using the Model Optimizer, convert a trained model to produce an optimized Intermediate Representation (IR) of the model based on the trained network topology, weights, and bias values.
  2. Test the model in the Intermediate Representation format using the Inference Engine in the target environment with the validation application or the sample applications.
  3. Integrate the Inference Engine into your application to deploy the model in the target environment.

Using the Model Optimizer to convert a Keras model to IR

The model optimizer doesn’t natively support Keras model files. However, because Keras uses Tensorflow as its backend, a Keras model can be saved as a Tensorflow checkpoint which can be loaded into the model optimizer. A Keras model can be converted to an IR using the following steps

  1. Save the Keras model as a Tensorflow checkpoint. Make sure the learning phase is set to 0. Get the name of the output node.

import tensorflow as tf

from keras.applications import Resnet50

from keras import backend as K

from keras.models import Sequential, Model

K.set_learning_phase(0) # Set the learning phase to 0

model = ResNet50(weights=‘imagenet’, input_shape=(256, 256, 3))

config = model.get_config()

weights = model.get_weights()

model = Sequential.from_config(config)

output_node =‘:’)[0] # We need this in the next step

graph_file =

ckpt_file =

saver = tf.train.Saver(sharded=True)

, graph_file), ckpt_file)

2. Run the Tensorflow freeze_graph program to generate a frozen graph from the saved checkpoint.

tensorflow/bazel-bin/tensorflow/python/tools/freeze_graph –input_graph=./resnet50_graph.pb –input_checkpoint=./resnet50.ckpt –output_node_names=Softmax –output_graph=resnet_frozen.pb

3. Use the script and the frozen graph to generate the IR. The model weights can be quantized to FP16.

python –input_model=resnet50_frozen.pb –output_dir=./ –input_shape=[1,224,224,3] — data_type=FP16


The C++ library provides utilities to read an IR, select a plugin depending on the target device, and run the model.

  1. Read the Intermediate Representation – Using the InferenceEngine::CNNNetReader class, read an Intermediate Representation file into a CNNNetwork class. This class represents the network in host memory.
  2. Prepare inputs and outputs format – After loading the network, specify input and output precision, and the layout on the network. For these specification, use the CNNNetwork::getInputInfo() and CNNNetwork::getOutputInfo()
  3. Select Plugin – Select the plugin on which to load your network. Create the plugin with the InferenceEngine::PluginDispatcher load helper class. Pass per device loading configurations specific to this device and register extensions to this device.
  4. Compile and Load – Use the plugin interface wrapper class InferenceEngine::InferencePlugin to call the LoadNetwork() API to compile and load the network on the device. Pass in the per-target load configuration for this compilation and load operation.
  5. Set input data – With the network loaded, you have an ExecutableNetwork object. Use this object to create an InferRequest in which you signal the input buffers to use for input and output. Specify a device-allocated memory and copy it into the device memory directly, or tell the device to use your application memory to save a copy.
  6. Execute – With the input and output memory now defined, choose your execution mode:
    • Synchronously – Infer() method. Blocks until inference finishes.
    • Asynchronously – StartAsync() method. Check status with the wait() method (0 timeout), wait, or specify a completion callback.
  7. Get the output – After inference is completed, get the output memory or read the memory you provided earlier. Do this with the InferRequest GetBlob API.

The classification_sample and classification_sample_async programs perform inference using the steps mentioned above. We use these samples in the next section to perform inference on an Intel® FPGA.

Using the Intel® Programmable Acceleration Card with Intel® Arria® 10GX FPGA for inference

The OpenVINO toolkit supports using the PAC as a target device for running low power inference. The steps for setting up the card are detailed here. The pre-processing and post-processing is performed on the host while the execution of the model is performed on the card. The toolkit contains bitstreams for different topologies.

  1. Programming the bitstream

aocl program <device_id> <open_vino_install_directory>/a10_dcp_bitstreams/2-0-1_RC_FP16_ResNet50-101.aocx

2. The Hetero plugin can be used with CPU as the fallback device for layers that are not supported by the FPGA. The -pc flag prints performance details for each layer

./classification_sample_async -d HETERO:FPGA,CPU -i <path/to/input/image.png> –m <path/to/ir>/resnet50_frozen.xml


Intel® OpenVINO™ toolkit is a great way to quickly integrate trained models into applications and deploy them in different production environments. The complete documentation for the toolkit can be found at


Training an AI Radiologist with Distributed Deep Learning

The potential of neural networks to transform healthcare is evident. From image classification to dictation and translation, neural networks are achieving or exceeding human capabilities. And they are only getting better at these tasks as the quantity of data increases.

But there’s another way in which neural networks can potentially transform the healthcare industry: Knowledge can be replicated at virtually no cost. Take radiology as an example: To train 100 radiologists, you must teach each individual person the skills necessary to identify diseases in x-ray images of patients’ bodies. To make 100 AI-enabled radiologist assistants, you take the neural network model you trained to read x-ray images and load it into 100 different devices.

The hurdle is training the model. It takes a large amount of cleaned, curated, labeled data to train an image classification model. Once you’ve prepared the training data, it can take days, weeks, or even months to train a neural network. Even once you’ve trained a neural network model, it might not be smart enough to perform the desired task. So, you try again. And again. Eventually, you will train a model that passes the test and can be used out in the world.

neural-network-workflow.pngWorkflow for Developing Neural Network Models

In this post, I’m going to talk about how to reduce the time spent in the Train/Test/Tune cycle by speeding up the training portion with distributed deep learning, using a test case we developed in Dell EMC’s HPC and AI Innovation Lab to classify pathologies in chest x-ray images. Through a combination of distributed deep learning, optimizer selection, and neural network topology selection, we were able to not only speed the process of training models from days to minutes, we were also able to improve the classification accuracy significantly.

We began by surveying the landscape of AI projects in healthcare, and Andrew Ng’s group at Stanford University provided our starting point. CheXNet was a project to demonstrate a neural network’s ability to accurately classify cases of pneumonia in chest x-ray images.

The dataset that Stanford used was ChestXray14, which was developed and made available by the United States’ National Institutes of Health (NIH). The dataset contains over 120,000 images of frontal chest x-rays, each potentially labeled with one or more of fourteen different thoracic pathologies. The data set is very unbalanced, with more than half of the data set images having no listed pathologies.

Stanford decided to use DenseNet, a neural network topology which had just been announced as the Best Paper at the 2017 Conference on Computer Vision and Pattern Recognition (CVPR), to solve the problem. The DenseNet topology is a deep network of repeating blocks over convolutions linked with residual connections. Blocks end with a batch normalization, followed by some additional convolution and pooling to link the blocks. At the end of the network, a fully connected layer is used to perform the classification.


An Illustration of the DenseNet Topology (source: Kaggle)

Stanford’s team used a DenseNet topology with the layer weights pretrained on ImageNet and replaced the original ImageNet classification layer with a new fully connected layer of 14 neurons, one for each pathology in the ChestXray14 dataset.

Building CheXNet in Keras

It’s sounds like it would be difficult to setup. Thankfully, Keras (provided with TensorFlow) provides a simple, straightforward way of taking standard neural network topologies and bolting-on new classification layers.

from tensorflow import kerasfrom keras.applications import DenseNet121orig_net = DenseNet121(include_top=False, weights=’imagenet’, input_shape=(256,256,3))

Importing the base DenseNet Topology using Keras

In this code snippet, we are importing the original DenseNet neural network (DenseNet121) and removing the classification layer with the include_top=False argument. We also automatically import the pretrained ImageNet weights and set the image size to 256×256, with 3 channels (red, green, blue).

With the original network imported, we can begin to construct the classification layer. If you look at the illustration of DenseNet above, you will notice that the classification layer is preceded by a pooling layer. We can add this pooling layer back to the new network with a single Keras function call, and we can call the resulting topology the neural network’s filters, or the part of the neural network which extracts all the key features used for classification.

from keras.layers import GlobalAveragePooling2Dfilters = GlobalAveragePooling2D()(orig_net.output)

Finalizing the Network Feature Filters with a Pooling Layer

The next task is to define the classification layer. The ChestXray14 dataset has 14 labeled pathologies, so we have one neuron for each label. We also activate each neuron with the sigmoid activation function, and use the output of the feature filter portion of our network as the input to the classifiers.

from keras.layers import Denseclassifiers = Dense(14, activation=’sigmoid’, bias_initializer=’ones’)(filters)

Defining the Classification Layer

The choice of sigmoid as an activation function is due to the multi-label nature of the data set. For problems where only one label ever applies to a given image (e.g., dog, cat, sandwich), a softmax activation would be preferable. In the case of ChestXray14, images can show signs of multiple pathologies, and the model should rightfully identify high probabilities for multiple classifications when appropriate.

Finally, we can put the feature filters and the classifiers together to create a single, trainable model.

from keras.models import Modelchexnet = Model(inputs=orig_net.inputs, outputs=classifiers)

The Final CheXNet Model Configuration

With the final model configuration in place, the model can then be compiled and trained.

To produce better models sooner, we need to accelerate the Train/Test/Tune cycle. Because testing and tuning are mostly sequential, training is the best place to look for potential optimization.

How exactly do we speed up the training process? In Accelerating Insights with Distributed Deep Learning, Michael Bennett and I discuss the three ways in which deep learning can be accelerated by distributing work and parallelizing the process:

  • Parameter server models such as in Caffe or distributed TensorFlow,
  • Ring-AllReduce approaches such as Uber’s Horovod, and
  • Hybrid approaches for Hadoop/Spark environments such as Intel BigDL.

Which approach you pick depends on your deep learning framework of choice and the compute environment that you will be using. For the tests described here we performed the training in house on the Zenith supercomputer in the Dell EMC HPC & AI Innovation Lab. The ring-allreduce approach enabled by Uber’s Horovod framework made the most sense for taking advantage of a system tuned for HPC workloads, and which takes advantage of Intel Omni-Path (OPA) networking for fast inter-node communication. The ring-allreduce approach would also be appropriate for solutions such as the Dell EMC Ready Solutions for AI, Deep Learning with NVIDIA.


The MPI-RingAllreduce Approach to Distributed Deep Learning

Horovod is an MPI-based framework for performing reduction operations between identical copies of the otherwise sequential training script. Because it is MPI-based, you will need to be sure that an MPI compiler (mpicc) is available in the working environment before installing horovod.

Adding Horovod to a Keras-defined Model

Adding Horovod to any Keras-defined neural network model only requires a few code modifications:

  1. Initializing the MPI environment,
  2. Broadcasting initial random weights or checkpoint weights to all workers,
  3. Wrapping the optimizer function to enable multi-node gradient summation,
  4. Average metrics among workers, and
  5. Limiting checkpoint writing to a single worker.

Horovod also provides helper functions and callbacks for optional capabilities that are useful when performing distributed deep learning, such as learning-rate warmup/decay and metric averaging.

Initializing the MPI Environment

Initializing the MPI environment in Horovod only requires calling the init method:

import horovod.keras as hvdhvd.init()

This will ensure that the MPI_Init function is called, setting up the communications structure and assigning ranks to all workers.

Broadcasting Weights

Broadcasting the neuron weights is done using a callback to the Keras method. In fact, many of Horovod’s features are implemented as callbacks to, so it’s worthwhile to define a callback list object for holding all the callbacks.

callbacks = [ hvd.callbacks.BroadcastGlobalVariablesCallback(0) ]

You’ll notice that the BroadcastGlobalVariablesCallback takes a single argument that’s been set to 0. This is the root worker, which will be responsible for reading checkpoint files or generating new initial weights, broadcasting weights at the beginning of the training run, and writing checkpoint files periodically so that work is not lost if a training job fails or terminates.

Wrapping the Optimizer Function

The optimizer function must be wrapped so that it can aggregate error information from all workers before executing. Horovod’s DistributedOptimizer function can wrap any optimizer which inherits Keras’ base Optimizer class, including SGD, Adam, Adadelta, Adagrad, and others.

import keras.optimizersopt = hvd.DistributedOptimizer(keras.optimizers.Adadelta(lr=1.0))

The distributed optimizer will now use the MPI_Allgather collective to aggregate error information from training batches onto all workers, rather than collecting them only to the root worker. This allows the workers to independently update their models rather than waiting for the root to re-broadcast updated weights before beginning the next training batch.

Averaging Metrics

Between steps error metrics need to be averaged to calculate global loss. Horovod provides another callback function to do this called MetricAverageCallback.

callbacks = [ hvd.callbacks.BroadcastGlobalVariablesCallback(0), hvd.callbacks.MetricAverageCallback() ]

This will ensure that optimizations are performed on the global metrics, not the metrics local to each worker.

Writing Checkpoints from a Single Worker

When using distributed deep learning, it’s important that only one worker write checkpoint files to ensure that multiple workers writing to the same file does not produce a race condition, which could lead to checkpoint corruption.

Checkpoint writing in Keras is enabled by another callback to However, we only want to call this callback from one worker instead of all workers. By convention, we use worker 0 for this task, but technically we could use any worker for this task. The one good thing about worker 0 is that even if you decide to run your distributed deep learning job with only 1 worker, that worker will be worker 0.

callbacks = [ ... ]if hvd.rank() == 0: callbacks.append(keras.callbacks.ModelCheckpoint(‘./checkpoint-{epoch].h5’))

Once a neural network can be trained in a distributed fashion across multiple workers, the Train/Test/Tune cycle can be sped up dramatically.

The figure below shows exactly how dramatically. The three tests shown are the training speed of the Keras DenseNet model on a single Zenith node without distributed deep learning (far left), the Keras DenseNet model with distributed deep learning on 32 Zenith nodes (64 MPI processes, 2 MPI processes per node, center), and a Keras VGG16 version using distributed deep learning on 64 Zenith nodes (128 MPI processes, 2 MPI processes per node, far right). By using 32 nodes instead of a single node, distributed deep learning was able to provide a 47x improvement in training speed, taking the training time for 10 epochs on the ChestXray14 data set from 2 days (50 hours) to less than 2 hours!


Performance comparisons of Keras models with distributed deep learning using Horovod

The VGG variant, trained on 128 Zenith nodes, was able to complete the same number of epochs as was required for the single-node DenseNet version to train in less than an hour, although it required more epochs to train. It also, however, was able to converge to a higher-quality solution. This VGG-based model outperformed the baseline, single-node model in 4 of 14 conditions, and was able to achieve nearly 90% accuracy in classifying emphysema.


Accuracy comparison of baseline single-node DenseNet model vs VGG variant with distributed deep learning


In this post we’ve shown you how to accelerate the Train/Test/Tune cycle when developing neural network-based models by speeding up the training phase with distributed deep learning. We walked through the process of transforming a Keras-based model to take advantage of multiple nodes using the Horovod framework, and how these few simple code changes, coupled with some additional compute infrastructure, can reduce the time needed to train a model from days to minutes, allowing more time for the testing and tuning pieces of the cycle. More time for tuning means higher-quality models, which means better outcomes for patients, customers, or whomever will benefit from the deployment of your model.

Lucas A. Wilson, Ph.D. is the Lead Data Scientist in Dell EMC’s HPC & AI Engineering group. (Twitter: @lucasawilson)


  • No Related Posts

Model Compatibility using Intel BigDL

Deep learning has exploded over the landscape of both the popular and business media landscapes. Current and upcoming technology capable of powering the calculations required by deep learning algorithms has enabled a rapid transition from new theories to new applications. One of current supporting technologies that is expanding at an increasing rate is in the area of faster and more use case specific hardware accelerators for deep learning such as GPUs with tensor cores and FPGAs hosted inside of servers. Another foundational deep learning technology that has advanced very rapidly is the software that enables implementations of complex deep learning networks. New frameworks, tools and applications are entering the landscape quickly to accomplish this, some compatible with existing infrastructure and others that require workflow overhauls.

As organizations begin to develop more complex strategies for incorporating deep learning they are likely to start to leverage multiple frameworks and application stacks for specific use cases and to compare performance and accuracy. But training models is time consuming and ties up expensive compute resources. In addition, adjustments and tuning can vary between frameworks, creating a large number of framework knobs and levers to remember how to operate. What if there was a framework that could just consume these models right out the box?

BigDL is a distributed deep learning framework with native Spark integration, allowing it to leverage Spark during model training, prediction, and tuning. One of the things that I really like about Intel BigDL is how easy it is to work with models built and/or trained in Tensorflow, Caffe and Torch. This rich interop support for deep learning models allows BigDL applications to leverage the plethora of models that currently exist with little or no additional effort. Here are just a few ways this might be used in your applications:

  • Efficient Scale Out – Using BigDL you can scale out a model that was trained on a single node or workstation and leverage it at scale with Apache Spark. This can be useful for training on a large distributed dataset that already exists in your HDFS environment or for performing inferencing such as prediction and classification on a very large and often changing dataset.

  • Transfer Learning – Load a pretrained model with weights and then freeze some layers, append new layers and train / retrain layers. Transfer learning can improve accuracy or reduce training time by allowing you to start with a model that is used to do one thing, such as classify a different objects, and use it to accelerate development of a model to classify something else, such as specific car models.

  • High Performance on CPU – GPUs get all of the hype when it comes to deep learning. By leveraging Intel MKL and multi threading Spark tasks you can achieve better CPU driven performance leveraging BigDL than you would see with Tensorflow, Caffe or Torch when using Xeon processors.

  • Dataset Access – Designed to run in Hadoop, BigDL can compute where your data already exists. This can save time and effort since data does not need to be transferred to a seperate GPU environment to be used with the deep learning model. This means that your entire pipeline from ingest to model training and inference can all happen in one environment, Hadoop.

Real Data + Real Problem

Recently I had a chance to take advantage of the model portability feature of BigDL. After learning of an internal project here at Dell EMC, leveraging deep learning and telemetry data to predict component failures, my team decided we wanted to take our Ready Solution for AI – Machine Learning with Hadoop and see how it did with the problem.

The team conducting the project for our support organization was using Tensorflow with GPU accelerators to train an LSTM model. The dataset was sensor readings from internal components collected at 15 minute intervals showing all kinds of metrics like temperature, fan speeds, runtimes, faults etc.

Initially my team wanted to focus on testing out two use cases for BigDL:

  • Using BigDL model portability to perform inference using the existing tensorflow model
  • Implement an LSTM model in BigDL and train it with this dataset

As always, there were some preprocessing and data cleaning steps that had happened before we could get to modeling and inference. Luckily for us though we received the clean output of those steps from our support team to get started quickly. We received the data in the form of multiple csv files, already balanced with records of devices that did fail and those that did not. We got over 200,000 rows of data that looked something like this:



Converting the data to a tfrecord format used by Tensorflow was being done with Python and pandas dataframes. Moving this process to be done in Spark is another area we knew we wanted to dig in to, but to start we wanted to focus on our above mentioned goals. When we started the pipeline looked like this:

From Tensorflow to BigDL

For BigDL, instead of creating tfrecords we needed to end up with an RDD of Sample(s). Each Sample is one record of your dataset in the form of feature, label. Feature and label are in the form of one or more tensors and we create the sample from ndarray. Looking at the current pipeline we were able to simple take the objects created before writing to tfrecord and instead wrote a function that took these arrays and formed our RDD of Sample for BigDL.

def convert_to(x, y): sequences = x labels = y record = zip(x,y) record_rdd = sc.parallelize(record) sample_rdd = x:Sample.from_ndarray(x[0], x[1])) return sample_rddtrain = convert_to(x_train,y_train)val = convert_to(x_val,y_val)test = convert_to(x_test,y_test)

After that we took the pb and bin files representing the pretrained models definition and weights and loaded it using the BigDL Model.load_tensorflow function. It requires knowing the input and output names for the model, but the tensorflow graph summary tool can help out with that. It also requires a pb and bin file specifically, but if what you have is a ckpt file from tensorflow that can be converted with tools provided by BigDL.

model_def = "tf_modell/model.pb"model_variable = "tf_model/model.bin"inputs = ["Placeholder"]outputs = ["prediction/Softmax"]trained_tf_model = Model.load_tensorflow(model_def, inputs, outputs, byte_order = "little_endian", bigdl_type="float", bin_file=model_variable)

Now with our data already in the correct format we can go ahead and inference against our test dataset. BigDL provides Model.evaluate and we can pass it our RDD as well as the validation method to use, in this case Top1Accuracy.

results = trained_tf_model.evaluate(test,128,[Top1Accuracy()])

Defining a Model with BigDL

After testing out loading the pretrained tensorflow model the next experiment we wanted to conduct was to train an LSTM model defined with BigDL. BigDL provides a Sequential API and a Functional API for defining models. The Sequential API is for simpler models, with the Functional API being better for complex models. The Functional API describes the model as a graph. Since our model is LSTM we will use the Sequential API.

Defining an LSTM model is as simple as:

def build_model(input_size, hidden_size, output_size): model = Sequential() recurrent = Recurrent() recurrent.add(LSTM(input_size, hidden_size)) model.add(InferReshape([-1, input_size], True)) model.add(recurrent) model.add(Select(2, -1)) model.add(Linear(hidden_size, output_size)) return modellstm_model = build_model(n_input, n_hidden, n_classes)

After creating our model the next step is the optimizer and validation logic that our model will use to train and learn.

Create the optimizer:

optimizer = Optimizer( model=lstm_model, training_rdd=train, criterion=CrossEntropyCriterion(), optim_method=Adam(), end_trigger=MaxEpoch(50), batch_size=batch_size)

Set the validation logic:

optimizer.set_validation( batch_size=batch_size, val_rdd=val, trigger=EveryEpoch(), val_method=[Top1Accuracy()])

Now we can do trained_model = optimizer.optimize() to train our model, in this case for 50 epochs. We also set our TrainSummary folder so that the data was logged. This allowed us to also get visualizations in Tensorboard, something that BigDL supports.

At this point we had completed the two initial tasks we had set out to do, load a pretrained Tensorflow model using BigDL and train a new model with BigDL. Hopefully you found some of this process interesting, and also got an idea for how easy BigDL is for this use case. The ability to leverage deep learning models inside Hadoop with no specialized hardware like Infiniband, GPU accelerators etc provides a great tool that is sure to change up the way you currently view your existing analytics.


Accelerating Insights with Distributed Deep Learning

By: Lucas A. Wilson, Ph.D. and Michael Bennett

Artificial intelligence (AI) is transforming the way businesses compete in today’s marketplace. Whether it’s improving business intelligence, streamlining supply chain or operational efficiencies, or creating new products, services, or capabilities for customers, AI should be a strategic component of any company’s digital transformation.

Deep neural networks have demonstrated astonishing abilities to identify objects, detect fraudulent behaviors, predict trends, recommend products, enable enhanced customer support through chatbots, convert voice to text and translate one language to another, and produce a whole host of other benefits for companies and researchers. They can categorize and summarize images, text, and audio recordings with human-level capability, but to do so they first need to be trained.

Deep learning, the process of training a neural network, can sometimes take days, weeks, or months, and effort and expertise is required to produce a neural network of sufficient quality to trust your business or research decisions on its recommendations. Most successful production systems go through many iterations of training, tuning and testing during development. Distributed deep learning can speed up this process, reducing the total time to tune and test so that your data science team can develop the right model faster, but requires a method to allow aggregation of knowledge between systems.

There are several evolving methods for efficiently implementing distributed deep learning, and the way in which you distribute the training of neural networks depends on your technology environment. Whether your compute environment is container native, high performance computing (HPC), or Hadoop/Spark clusters for Big Data analytics, your time to insight can be accelerated by using distributed deep learning. In this article we are going to explain and compare systems that use a centralized or replicated parameter server approach, a peer-to-peer approach, and finally a hybrid of these two developed specifically for Hadoop distributed big data environments.

Container native (e.g., Kubernetes, Docker Swarm, OpenShift, etc.) have become the standard for many DevOps environments, where rapid, in-production software updates are the norm and bursts of computation may be shifted to public clouds. Most deep learning frameworks support distributed deep learning for these types of environments using a parameter server-based model that allows multiple processes to look at training data simultaneously, while aggregating knowledge into a single, central model.

The process of performing parameter server-based training starts with specifying the number of workers (processes that will look at training data) and parameter servers (processes that will handle the aggregation of error reduction information, backpropagate those adjustments, and update the workers). Additional parameters servers can act as replicas for improved load balancing.


Parameter server model for distributed deep learning

Worker processes are given a mini-batch of training data to test and evaluate, and upon completion of that mini-batch, report the differences (gradients) between produced and expected output back to the parameter server(s). The parameter server(s) will then handle the training of the network and transmitting copies of the updated model back to the workers to use in the next round.

This model is ideal for container native environments, where parameter server processes and worker processes can be naturally separated. Orchestration systems, such as Kubernetes, allow neural network models to be trained in container native environments using multiple hardware resources to improve training time. Additionally, many deep learning frameworks support parameter server-based distributed training, such as TensorFlow, PyTorch, Caffe2, and Cognitive Toolkit.

High performance computing (HPC) environments are generally built to support the execution of multi-node applications that are developed and executed using the single process, multiple data (SPMD) methodology, where data exchange is performed over high-bandwidth, low-latency networks, such as Mellanox InfiniBand and Intel OPA. These multi-node codes take advantage of these networks through the Message Passing Interface (MPI), which abstracts communications into send/receive and collective constructs.

Deep learning can be distributed with MPI using a communication pattern called Ring-AllReduce. In Ring-AllReduce each process is identical, unlike in the parameter-server model where processes are either workers or servers. The Horovod package by Uber (available for TensorFlow, Keras, and PyTorch) and the mpi_collectives contributions from Baidu (available in TensorFlow) use MPI Ring-AllReduce to exchange loss and gradient information between replicas of the neural network being trained. This peer-based approach means that all nodes in the solution are working to train the network, rather than some nodes acting solely as aggregators/distributors (as in the parameter server model). This can potentially lead to faster model convergence.


Ring-AllReduce model for distributed deep learning

The Dell EMC Ready Solutions for AI, Deep Learning with NVIDIA allows users to take advantage of high-bandwidth Mellanox InfiniBand EDR networking, fast Dell EMC Isilon storage, accelerated compute with NVIDIA V100 GPUs, and optimized TensorFlow, Keras, or Pytorch with Horovod (or TensorFlow with tensorflow.contrib.mpi_collectives) frameworks to help produce insights faster.

Hadoop and other Big Data platforms achieve extremely high performance for distributed processing but are not designed to support long running, stateful applications. Several approaches exist for executing distributed training under Apache Spark. Yahoo developed TensorFlowOnSpark, accomplishing the goal with an architecture that leveraged Spark for scheduling Tensorflow operations and RDMA for direct tensor communication between servers.

BigDL is a distributed deep learning library for Apache Spark. Unlike Yahoo’s TensorflowOnSpark, BigDL not only enables distributed training – it is designed from the ground up to work on Big Data systems. To enable efficient distributed training BigDL takes a data-parallel approach to training with synchronous mini-batch SGD (Stochastic Gradient Descent). Training data is partitioned into RDD samples and distributed to each worker. Model training is done in an iterative process that first computes gradients locally on each worker by taking advantage of locally stored partitions of the training data and model to perform in memory transformations. Then an AllReduce function schedules workers with tasks to calculate and update weights. Finally, a broadcast syncs the distributed copies of model with updated weights.


BigDL implementation of AllReduce functionality

The Dell EMC Ready Solutions for AI, Machine Learning with Hadoop is configured to allow users to take advantage of the power of distributed deep learning with Intel BigDL and Apache Spark. It supports loading models and weights from other frameworks such as Tensorflow, Caffe and Torch to then be leveraged for training or inferencing. BigDL is a great way for users to quickly begin training neural networks using Apache Spark, widely recognized for how simple it makes data processing.

One more note on Hadoop and Spark environments: The Intel team working on BigDL has built and compiled high-level pipeline APIs, built-in deep learning models, and reference use cases into the Intel Analytics Zoo library. Analytics Zoo is based on BigDL but helps make it even easier to use through these high-level pipeline APIs designed to work with Spark Dataframes and built in models for things like object detection and image classification.

Regardless of whether you preferred server infrastructure is container native, HPC clusters, or Hadoop/Spark-enabled data lakes, distributed deep learning can help your data science team develop neural network models faster. Our Dell EMC Ready Solutions for Artificial Intelligence can work in any of these environments to help jumpstart your business’s AI journey. For more information on the Dell EMC Ready Solutions for Artificial Intelligence, go to

Lucas A. Wilson, Ph.D. is the Lead Data Scientist in Dell EMC’s HPC & AI Engineering group. (Twitter: @lucasawilson)

Michael Bennett is a Senior Principal Engineer at Dell EMC working on Ready Solutions.


Machine Learning 300 – why a run-time compiler for data science makes sense.

My last article was on frameworks and includes information about framework optimizations for specific platforms and accelerators. After reading that material you might look at this title and wonder, don’t we have enough complexity to deal with already?. The answer is clearly yes, but, I have been following this topic for a while and wanted to introduce you to a recent development in just-in-time or run time compilers for machine learning that I hope you’ll also find interesting. I want to be clear that this information is most useful to software developers but I think it will help data scientists understand more about how to do large scale model training and inferencing and I promise to keep it short.

To understand the types of problems that a run-time compiler for data science could solve let’s examine some of the details of how TensorFlow is constructed. TensorFlow supports both CPU and GPU device types and is distributed in two separate versions. One of the most common operations used by deep learning frameworks like TensorFlow is matrix multiplication (matmul). TensorFlow has both CPU and GPU implementations of matmul that are optimized for their respective target devices. On a system with one or more GPUs, the GPU with the lowest ID will be selected to run matmul by default.

Every framework developer must consider what devices to support and what operations to optimize for those devices. Without the ability to do optimization at run-time, each framework developer must specifically code optimizations for each back end device or accelerator they choose to support. With the rapidly expanding number of both frameworks (m) plus hardware devices (n) this results in a large number of optimization coding efforts ( m * n) for compete coverage.


In order to address these challenges, Intel recently announced the open source release of nGraph, a C++ library, compiler and run-time suite for running Deep Neural Networks on a variety of frameworks and devices. The picture above shows where the nGraph core sits in the deep learning hardware and software stack. Since all frameworks could be optimized for all hardware platforms using a run-time compiler like nGraph, the engineering could be reduced to an m+n scale effort. The current version of nGraph is available from the following GitHub repository – NervanaSystems/ngraph.

How nGraph works. Every deep learning framework implements a framework-specific symbolic representation for their computations, called a computational graph. The nGraph community has developed a framework bridge for each supported framework. Developers can install the nGraph library and compile a framework with the library in order to specify nGraph as the framework back end you want to use. The appropriate framework bridge will convert the framework specific computation definitions into a framework-independent intermediate representation (IR) that can then be compiled into a form that can be executed on any supported back end device. The IR layer handles all the device abstraction details and lets developers focus on their data science, algorithms and models, rather than on machine code.

Community contributions to nGraph are always welcome. If you have an idea how to improve the library share your proposal via GitHub issues.

Resources for this article:

Thanks for reading,

Phil Hummel



What Is a Deep Learning Framework?

Picture1.jpgData science requires a mix of computer programming, mathematics/statistics and domain knowledge. This article focuses on the intersection of the first two requirements. Comprehensive software packages for classical machine learning, primarily supported by statistical algorithms, have been widely available for decades. There are many mature offerings available from both the open source software community and commercial software makers. Modern deep learning is less mature, still experiencing rapid innovation, and so the software landscape is more dynamic. Data scientists engaged in deep learning must get more involved in programming than typically required for classical machine learning. The remainder of this article explains the why and how of that effort.

First, let me start with a high-level summary of what deep learning is. Computer scientists have been studying ways to perform speech recognition, image recognition, natural language processing, including translation, relationship identification, recommendation systems and other forms of data relationship discovery since computers were first invented. After decades of parallel research in mathematics and software development, researchers discovered a methodology called artificial neural networks (ANNs) that could be used to solve these types of problems and many more using a common set of tools. The building blocks of ANNs are layers. Each layer typically accepts structured data (tensors) as inputs, then perform a type of transformation on that data, and finally, sends the transformed data to the next layer until the output layer is processed. The layers typically used in ANNs can be grouped into categories, for example

  • Input Layers
  • Learnable Layers
  • Activation Layers
  • Pooling Layers
  • Combination Layers
  • Output Layers

The number and definition of the layers, how they are connected, and the data structures used between layers are called the model structure. In addition to defining the structure, a data scientist must specify how the model is to be executed including the function to be optimized and optimization method. Given the complexity of the mathematics and the need to efficiently process large data sets, the effort to create a deep learning software program is a significant development effort, even for professional computer scientists.

Deep learning frameworks were developed to make software for deep learning available to the wider community of programmers and data scientists. Most of today’s popular frameworks are developed through open source software initiatives in each of which attract dozens of active developers. The rate of innovation in the deep learning framework space is both impressive and somewhat overwhelming.

To further complicate the world of deep learning (yes, that is possible) despite the many similar capabilities of the most popular deep learning frameworks, there are also significant differences that lead to a need for careful evaluation for compatibility once a project is defined. Based on a sample of the many comparisons of deep learning frameworks that can be found in just the last couple of years, I estimate that there are between 15-20 viable alternatives today.

The Intel® AI Academy recently published a comparison summary focused on frameworks that have versions optimized by Intel and that can effectively run on CPUs optimized for matrix multiplication. The table below is a sample of the type of analysis and data that was collected.


# of GitHub Stars

# of GitHub Forks










Microsoft Cognitive Toolkit*















neon™ framework



Source Intel AI Academy 2017

The NVIDIA Deep Learning AI website has a summary of deep learning frameworks such as Caffe2, Cognitive toolkit, MXNet, PyTorch, TensorFlow and others that support GPU-accelerated libraries such as cuDNN and NCCL to deliver high-performance multi-GPU accelerated training. The page also includes links to learning and getting started resources.


  1. Don’t be surprised if the data science team proposes projects using different frameworks.
  2. Get curious if every project requires a different framework.
  3. Plan to age out some frameworks over time and bring in new ones.
  4. Allocate time for new framework investigation.
  5. Look for platforms that support multiple frameworks to reduce silos.
  6. Check online reviews from reputable sources to see how others rate a framework before adopting it for a project.

Thanks for reading,

Phil Hummel



Developing cognitive IoT solutions for anomaly detection by using deep learning, Part 5: Using Keras and TensorFlow for anomaly detection

This article is the fifth in a five-part series, “Developing cognitive
IoT solutions for anomaly detection by using deep learning.” This article
demonstrates a deep learning solution using Keras and TensorFlow and how it is
used to analyze the large amount of data that IoT sensors gather.


7 tips for getting started with R language

This article gives you 7 tips to make it easier and faster to start
using the powerful R language. Learn about R packages, how to structure data in R, and the powerful
RStudio IDE. Discover how the apply function can save you loads of time when running complicated scripts
and how ggplot2 can really make your visuals pop. Finally, see how the Rcpp package allows you to import
C++ functions into an R script.