Add a ResNet example from NVIDIA

samples/nvidia-resnet/LICENSE

@@ -0,0 +1,25 @@
+Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions
+are met:
+ * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+   notice, this list of conditions and the following disclaimer in the
+   documentation and/or other materials provided with the distribution.
+ * Neither the name of NVIDIA CORPORATION nor the names of its
+   contributors may be used to endorse or promote products derived
+   from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

samples/nvidia-resnet/README.md

@@ -0,0 +1,38 @@
+# A simple NVIDIA-accelerated ResNet Kubeflow pipeline
+### This example demonstrates a simple end-to-end training & deployment of a Keras Resnet model on the CIFAR10 dataset utilizing the following NVIDIA technologies:
+* [NVIDIA-Docker2](https://github.com/NVIDIA/nvidia-docker) to make the Docker containers GPU aware.
+* [NVIDIA device plugin](https://github.com/NVIDIA/k8s-device-plugin) to allow Kubernetes to access GPU nodes.
+* [TensorFlow-19.02](https://ngc.nvidia.com/catalog/containers/nvidia:tensorflow) containers from NVIDIA GPU Cloud container registry.
+* [TensorRT](https://docs.nvidia.com/deeplearning/dgx/integrate-tf-trt/index.html) for optimizing the Inference Graph in FP16 for leveraging the dedicated use of Tensor Cores for Inference.
+* [TensorRT Inference Server](https://github.com/NVIDIA/tensorrt-inference-server) for serving the trained model.
+
+## System Requirements
+* Ubuntu 16.04 and above
+* NVIDIA GPU
+
+## Quickstart
+* Install NVIDIA Docker, Kubernetes and Kubeflow on your local machine:
+    * `sudo ./install_kubeflow_and_dependencies.sh`
+* Mount persistent volume to Kubeflow:
+    * `sudo ./mount_persistent_volume.sh`
+* Build the Preprocessing, Training, Serving, and Pipeline containers using the following script:
+    * First, modify `build.sh` in `preprocess`, `train`, and `serve` directories to point to a container registry that is accessible to you
+    * Then, `sudo ./build_pipeline.sh`
+    * Note the `pipeline.py.tar.gz` file that appears on your working directory
+* Determine the ambassador port using this command:
+    * `sudo kubectl get svc -n kubeflow ambassador`
+* Open the Kubeflow Dashboard on:
+    * https://local-machine-ip-address:port-determined-from-previous-step
+    * E.g. https://10.110.210.99:31342
+* Click on the tab Pipeline Dashboard, upload the `pipeline.py.tar.gz` file under you working directory and create a run
+* Once the training has completed (should take about 20 minutes for 50 epochs) and the model is being served, port forward the port of the serving pod (8000) to the local system:
+    * Determine the name of the serving pod by selecting it on the Kubeflow Dashboard
+    * Modify accordingly the variable `SERVING_POD` within `portforward_serving_port.sh`
+    * Then, `sudo ./portforward_serving_port.sh`
+* Build the client container and start a local server for the demo web UI on the host machine (about 15 mins):
+    * `sudo ./test_trtis_client.sh`
+* Now you have successfully set up the client through which you can ping the server with an image URL and obtain predictions:
+    * Open the demo client UI on a web browser with the following IP address:
+https://local-machine-ip-address:8050 
+    * The port is specified in `demo_client_ui.py` and can be changed as needed
+    * Copy an image URL (for one of the 10 CIFAR classes) and paste it in the UI

samples/nvidia-resnet/build_pipeline.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+WORK_DIR=$(pwd)
+
+# Build and push images of kubeflow pipeline components
+cd $WORK_DIR/preprocess && ./build.sh && \
+cd $WORK_DIR/train && ./build.sh && \
+cd $WORK_DIR/serve && ./build.sh && \
+
+# Compile kubeflow pipeline tar file 
+cd $WORK_DIR/pipeline && ./build.sh
+
+

samples/nvidia-resnet/install_kubeflow_and_dependencies.sh

@@ -0,0 +1,84 @@
+#!/bin/bash
+# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+# Install nvidia-docker 2
+sudo tee /etc/docker/daemon.json <<EOF
+{
+    "default-runtime": "nvidia",
+    "runtimes": {
+        "nvidia": {
+            "path": "/usr/bin/nvidia-container-runtime",
+            "runtimeArgs": []
+        }
+    }
+}
+EOF
+sudo pkill -SIGHUP dockerd
+
+# Install MiniKube
+curl -Lo minikube https://storage.googleapis.com/minikube/releases/latest/minikube-linux-amd64 \
+  && chmod +x minikube
+sudo cp minikube /usr/local/bin && rm minikube
+
+# Install kubectl
+sudo apt-get update && sudo apt-get install -y apt-transport-https
+curl -s https://packages.cloud.google.com/apt/doc/apt-key.gpg | sudo apt-key add -
+sudo touch /etc/apt/sources.list.d/kubernetes.list
+echo "deb http://apt.kubernetes.io/ kubernetes-xenial main" | sudo tee -a /etc/apt/sources.list.d/kubernetes.list
+sudo apt-get update
+sudo apt-get install -y kubectl
+
+# Install ksonnet
+export KS_VER=0.13.1
+wget https://github.com/ksonnet/ksonnet/releases/download/v${KS_VER}/ks_${KS_VER}_linux_amd64.tar.gz -O ksonnet.tar.gz
+mkdir -p ksonnet
+tar -xvf ksonnet.tar.gz -C ksonnet --strip-components=1
+sudo cp ksonnet/ks /usr/local/bin
+rm -fr ksonnet
+rm ksonnet.tar.gz
+
+# Start Minikube
+sudo minikube start --vm-driver=none --extra-config=kubelet.resolv-conf=/run/systemd/resolve/resolv.conf
+
+# Add the NVIDIA device plugin
+kubectl create -f https://raw.githubusercontent.com/NVIDIA/k8s-device-plugin/v1.11/nvidia-device-plugin.yml
+
+# Install Kubeflow
+export KUBEFLOW_SRC=/tmp/kubeflow
+export KFAPP=kfapp
+mkdir ${KUBEFLOW_SRC}
+cd ${KUBEFLOW_SRC}
+export KUBEFLOW_TAG=v0.4.1
+curl https://raw.githubusercontent.com/kubeflow/kubeflow/${KUBEFLOW_TAG}/scripts/download.sh | bash
+KUBEFLOW_REPO=${KUBEFLOW_SRC} ${KUBEFLOW_SRC}/scripts/kfctl.sh init ${KFAPP} --platform minikube
+cd ${KFAPP}
+${KUBEFLOW_SRC}/scripts/kfctl.sh generate all
+${KUBEFLOW_SRC}/scripts/kfctl.sh apply all
+
+# Open up ambassador’s port
+kubectl patch svc ambassador -n kubeflow --type='json' -p '[{"op":"replace","path":"/spec/type","value":"NodePort"}]'

samples/nvidia-resnet/mount_persistent_volume.sh

@@ -0,0 +1,36 @@
+#!/bin/bash
+# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+# Modify according to your persistent-volume.yaml
+PERSISTENT_VOL_PATH=/mnt/workspace
+
+sudo mkdir $PERSISTENT_VOL_PATH
+
+kubectl create -f pipeline/persistent-volume.yaml
+kubectl create -f pipeline/persistent-volume-claim.yaml
+

samples/nvidia-resnet/pipeline/Dockerfile

@@ -0,0 +1,31 @@
+# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+FROM python:3.6
+
+# Install Kubeflow Pipelines
+RUN pip install https://storage.googleapis.com/ml-pipeline/release/0.1.12/kfp.tar.gz --upgrade

samples/nvidia-resnet/pipeline/build.sh

@@ -0,0 +1,35 @@
+#!/bin/bash
+# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+IMAGE="kubeflow-pipeline"
+WORK_DIR=$(pwd)
+CMD="python $WORK_DIR/pipeline.py"
+
+docker build -t $IMAGE . && \
+docker run --rm -v $WORK_DIR:$WORK_DIR $IMAGE $CMD && \
+mv -f *.tar.gz ../

samples/nvidia-resnet/pipeline/persistent-volume-claim.yaml

@@ -0,0 +1,11 @@
+kind: PersistentVolumeClaim
+apiVersion: v1
+metadata:
+  name: nvidia-workspace-read-claim
+spec:
+  storageClassName: manual
+  accessModes:
+          - ReadWriteOnce
+  resources:
+    requests:
+      storage: 20Gi

samples/nvidia-resnet/pipeline/persistent-volume.yaml

@@ -0,0 +1,14 @@
+kind: PersistentVolume
+apiVersion: v1
+metadata:
+  name: nvidia-workspace
+  labels:
+    type: local
+spec:
+  storageClassName: manual
+  capacity:
+    storage: 100Gi
+  accessModes:
+    - ReadWriteOnce
+  hostPath:
+    path: "/mnt/workspace"