Add pytorch demo

.github/workflows/demo_test.yml

@@ -308,6 +308,30 @@ jobs:
     - name: Run Tensorflow-lite benchmark
       run: docker exec tflite_test bash -c "cd /root/occlum/demos/tensorflow_lite && SGX_MODE=SIM ./run_tflite_in_occlum.sh benchmark"
 
+  Pytorch_test:
+    runs-on: ubuntu-18.04
+    steps:
+    - uses: actions/checkout@v1
+      with:
+        submodules: true
+
+    - name: Get occlum version
+      run: echo "OCCLUM_VERSION=$(grep "Version =" src/pal/include/occlum_version.h |  awk '{print $4}')" >> $GITHUB_ENV
+
+    - name: Create container
+      run: docker run -itd --name=pytorch_test -v $GITHUB_WORKSPACE:/root/occlum occlum/occlum:${{ env.OCCLUM_VERSION }}-ubuntu18.04
+
+    - name: Build dependencies
+      run: docker exec pytorch_test bash -c "cd /root/occlum; make submodule"
+
+    - name: Make install
+      run: docker exec pytorch_test bash -c "source /opt/intel/sgxsdk/environment; cd /root/occlum; OCCLUM_RELEASE_BUILD=1 make install"
+
+    - name: Build python and pytorch
+      run: docker exec pytorch_test bash -c "cd /root/occlum/demos/pytorch; ./install_python_with_conda.sh"
+
+    - name: Run pytorch test
+      run: docker exec pytorch_test bash -c "cd /root/occlum/demos/pytorch; SGX_MODE=SIM ./run_pytorch_on_occlum.sh"
 
 # Below tests needs test image to run faster
   Grpc_test:

demos/README.md

@@ -20,6 +20,7 @@ This set of demos shows how real-world apps can be easily run inside SGX enclave
 * [https_server](https_server/): A HTTPS file server based on [Mongoose Embedded Web Server Library](https://github.com/cesanta/mongoose).
 * [grpc](grpc/): A client and server communicating through [gRPC](https://grpc.io).
 * [openvino](openvino/) A benchmark of [OpenVINO Inference Engine](https://docs.openvinotoolkit.org/2019_R3/_docs_IE_DG_inference_engine_intro.html).
+* [pytorch](pytorch/): A demo of [PyTorch](https://pytorch.org/).
 * [redis](redis/): A demo of [redis](https://redis.io).
 * [sqlite](sqlite/) A demo of [SQLite](https://www.sqlite.org) SQL database engine.
 * [tensorflow_lite](tensorflow_lite/): A demo and benchmark of [Tensorflow Lite](https://www.tensorflow.org/lite) inference engine.

demos/pytorch/README.md

@@ -0,0 +1,32 @@
+# Use PyTorch with Python and Occlum
+
+This project demonstrates how Occlum enables _unmodified_ [PyTorch](https://pytorch.org/) programs running in SGX enclaves, on the basis of _unmodified_ [Python](https://www.python.org).
+
+## Sample Code: Linear model
+
+Use the nn package to define our model as a sequence of layers. nn.Sequential is a Module which contains other Modules, and applies them in sequence to produce its output. Each Linear Module computes output from input using a linear function, and holds internal Tensors for its weight and bias.
+
+## How to Run
+
+This tutorial is written under the assumption that you have Docker installed and use Occlum in a Docker container.
+
+Occlum is compatible with glibc-supported Python, we employ miniconda as python installation tool. You can import PyTorch packages using conda. Here, miniconda is automatically installed by install_python_with_conda.sh script, the required python and PyTorch packages for this project are also loaded by this script. Here, we take occlum/occlum:0.22.0-ubuntu18.04 as example.
+
+Step 1 (on the host): Start an Occlum container
+```
+docker pull occlum/occlum:0.22.0-ubuntu18.04
+docker run -it --name=pythonDemo --device /dev/sgx/enclave occlum/occlum:0.22.0-ubuntu18.04 bash
+```
+
+Step 2 (on the host): Download miniconda and install python to prefix position.
+```
+cd /root/occlum/demos/pytorch
+bash ./install_python_with_conda.sh
+```
+
+Step 3 (on the host): Run the sample code on Occlum
+```
+cd /root/occlum/demos/pytorch
+bash ./run_pytorch_on_occlum.sh
+```
+

demos/pytorch/demo.py

@@ -0,0 +1,56 @@
+# Tutorial code from https://pytorch.org/tutorials/beginner/pytorch_with_examples.html#pytorch-nn.
+
+import torch
+
+# N is batch size; D_in is input dimension;
+# H is hidden dimension; D_out is output dimension.
+N, D_in, H, D_out = 64, 1000, 100, 10
+
+# Create random Tensors to hold inputs and outputs
+x = torch.randn(N, D_in)
+y = torch.randn(N, D_out)
+
+# Use the nn package to define our model as a sequence of layers. nn.Sequential
+# is a Module which contains other Modules, and applies them in sequence to
+# produce its output. Each Linear Module computes output from input using a
+# linear function, and holds internal Tensors for its weight and bias.
+model = torch.nn.Sequential(
+    torch.nn.Linear(D_in, H), torch.nn.ReLU(), torch.nn.Linear(H, D_out),
+)
+
+# The nn package also contains definitions of popular loss functions; in this
+# case we will use Mean Squared Error (MSE) as our loss function.
+loss_fn = torch.nn.MSELoss(reduction="sum")
+
+learning_rate = 1e-4
+print("Training...")
+for t in range(500):
+    # Forward pass: compute predicted y by passing x to the model. Module objects
+    # override the __call__ operator so you can call them like functions. When
+    # doing so you pass a Tensor of input data to the Module and it produces
+    # a Tensor of output data.
+    y_pred = model(x)
+
+    # Compute and print loss. We pass Tensors containing the predicted and true
+    # values of y, and the loss function returns a Tensor containing the
+    # loss.
+    loss = loss_fn(y_pred, y)
+    if t % 100 == 99:
+        print(t, loss.item())
+
+    # Zero the gradients before running the backward pass.
+    model.zero_grad()
+
+    # Backward pass: compute gradient of the loss with respect to all the learnable
+    # parameters of the model. Internally, the parameters of each Module are stored
+    # in Tensors with requires_grad=True, so this call will compute gradients for
+    # all learnable parameters in the model.
+    loss.backward()
+
+    # Update the weights using gradient descent. Each parameter is a Tensor, so
+    # we can access its gradients like we did before.
+    with torch.no_grad():
+        for param in model.parameters():
+            param -= learning_rate * param.grad
+
+print("Done")

demos/pytorch/install_python_with_conda.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+set -e
+script_dir="$( cd "$( dirname "${BASH_SOURCE[0]}"  )" >/dev/null 2>&1 && pwd )"
+
+# 1. Init occlum workspace
+[ -d occlum_instance ] || occlum new occlum_instance
+
+# 2. Install python and dependencies to specified position
+[ -d Miniconda3-latest-Linux-x86_64.sh ] || wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
+[ -d miniconda ] || bash ./Miniconda3-latest-Linux-x86_64.sh -b -p $script_dir/miniconda
+$script_dir/miniconda/bin/conda create --prefix $script_dir/occlum_instance/image/opt/python-occlum -y  python=3.7 pytorch torchvision -c pytorch

demos/pytorch/run_pytorch_on_occlum.sh

@@ -0,0 +1,33 @@
+#!/bin/bash
+set -e
+
+BLUE='\033[1;34m'
+NC='\033[0m'
+
+script_dir="$( cd "$( dirname "${BASH_SOURCE[0]}"  )" >/dev/null 2>&1 && pwd )"
+python_dir="$script_dir/occlum_instance/image/opt/python-occlum"
+
+if [ ! -d $python_dir ];then
+    echo "Error: cannot stat '$python_dir' directory"
+    exit 1
+fi
+
+cd occlum_instance
+# Copy files into Occlum Workspace and build
+if [ ! -d "image/lib/python3" ];then
+    ln -s /opt/python-occlum/bin/python3 image/bin/python3
+    cp -f /opt/occlum/glibc/lib/libdl.so.2 image/opt/occlum/glibc/lib/
+    cp -f /opt/occlum/glibc/lib/libutil.so.1 image/opt/occlum/glibc/lib/
+    cp -f /opt/occlum/glibc/lib/librt.so.1 image/opt/occlum/glibc/lib/
+    cp -f ../demo.py image
+    new_json="$(jq '.resource_limits.user_space_size = "6000MB" |
+                    .resource_limits.kernel_space_heap_size = "256MB" |
+                    .process.default_mmap_size = "4000MB" |
+                    .env.default += ["PYTHONHOME=/opt/python-occlum"]' Occlum.json)" && \
+    echo "${new_json}" > Occlum.json
+    occlum build
+fi
+
+# Run the python demo
+echo -e "${BLUE}occlum run /bin/python3 demo.py${NC}"
+occlum run /bin/python3 demo.py