Merge pull request #1 from sacs-epfl/pets-review

Pets review

Dockerfile

@@ -1,6 +1,13 @@
 # Use the NVIDIA CUDA image for the specified platform
 FROM nvidia/cuda:12.3.2-devel-ubuntu22.04
 
+# Pytorch requires the CUDA SDK architecture version to be specified
+#  for Docker-based Pytorch builds to work
+# Set this argument as:
+#  docker build --build-arg TORCH_CUDA_ARCH_LIST=Turing .
+ARG TORCH_CUDA_ARCH_LIST
+ENV TORCH_CUDA_ARCH_LIST=${TORCH_CUDA_ARCH_LIST:-}
+
 # Update and install required packages
 RUN apt-get update && apt-get install -y \
     openssl \
@@ -21,47 +28,27 @@ RUN apt-get update && apt-get install -y \
     libgl1-mesa-glx \
     git-lfs
 
-# Set the home directory
-ENV HOME=/root
-# ENV TORCH_CUDA_ARCH_LIST=Turing
 # Create a directory called 'shatter' in the home directory
-RUN mkdir -p $HOME/shatter
-
-# Copy the contents of the current folder into the 'shatter' directory
-COPY . $HOME/shatter
-
-# Unzip ImangeNet Validation Data
-WORKDIR $HOME/shatter/artifact_scripts/gradientInversion/rog/
-RUN unzip data.zip
-RUN rm data.zip
+RUN mkdir -p /root/shatter
 
-WORKDIR $HOME
-# Install Miniconda
-ENV CONDA_PREFIX=${HOME}/.conda
-ENV CONDA=${CONDA_PREFIX}/condabin/conda
-
-RUN wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda.sh
-RUN bash miniconda.sh -b -p ${CONDA_PREFIX}
-
-RUN ${CONDA} config --set auto_activate_base false
-RUN ${CONDA} init bash
-RUN rm miniconda.sh
+# Set the home directory
+ENV SHATTER_HOME=/root/shatter
 
-# Set the working directory to the 'shatter' directory
-WORKDIR $HOME/shatter
+# Copy the contents of the current folder into the 'shatter' directory
+COPY . $SHATTER_HOME
 
-# Install the required packages
-RUN ${CONDA} create -n venv python=3.10
+WORKDIR $SHATTER_HOME
+# This is required to install Conda
+ENV CONDA_PREFIX=${SHATTER_HOME}/.conda
 
-ENV VENV_PATH=${CONDA_PREFIX}/envs/venv/bin
+RUN bash prerequisites.sh
 
-RUN ${CONDA} run -n venv pip install --upgrade pip
-RUN ${CONDA} run -n venv pip install --upgrade setuptools
-RUN ${CONDA} run -n venv pip install -r requirements.txt
-# WORKDIR $HOME/shatter/artifact_scripts/gradientInversion/rog/
-# RUN ${CONDA} run -n venv pip install -r requirements.txt
+RUN bash testing-script.sh
 
-WORKDIR $HOME/shatter
+# Run experiment 1
+#RUN cd $SHATTER_HOME/artifact_scripts/gradientInversion/rog && ./run.sh
+## Run experiment 2
+#RUN cd $SHATTER_HOME/artifact_scripts/small_scale && ./run_all
 
 # Default command to open a bash shell
 CMD ["/bin/bash"]

README.rst

@@ -9,7 +9,7 @@ Decentralized Learning (DL) enables collaborative learning without a server and
 Installation
 ============
 
-Before cloning, ensure that ypou have Git LFS installed. If not, you can install it using the following command:
+Before cloning, ensure that you have Git LFS installed. If not, you can install it using the following command:
 
 .. code-block:: bash
 
@@ -19,7 +19,7 @@ To install the necessary dependencies, use the following command:
 
 .. code-block:: bash
 
-    pip install -r requirements.txt && pip install --editable . && pip install -r artifact_scripts/gradientInversion/rog/requirements.txt
+    pip install -r requirements_base.txt && pip install -r requirements_all.txt
 
 Usage
 =====

artifact_scripts/gradientInversion/rog/run.sh

@@ -1,31 +1,26 @@
 #!/bin/bash
 
-# Check if the number of arguments is exactly 1
-if [ "$#" -ne 1 ]; then
-    echo "Usage: $0 <environment python executable folder, e.g. ~/.conda/envs/venv/bin/>"
-    exit 1
-fi
-
-env_python=$1/python
 num_clients=1
 
 # Compute the results
 
+cd $SHATTER_HOME/artifact_scripts/gradientInversion/rog
+
 config_file=config_fedavg_lenet.yaml
-$env_python attack_fedavg.py $config_file $num_clients
+python attack_fedavg.py $config_file $num_clients
 
 echo "Finished EL Reconstructions!"
 sleep 2
 
 config_file=config_topk_lenet.yaml
-$env_python attack_fedavg.py $config_file $num_clients
+python attack_fedavg.py $config_file $num_clients
 echo "Finished TopK Reconstructions!"
 sleep 2
 
 for num_chunks in 2 4 8 16
 do
     config_file=config_vnodes"${num_chunks}"_lenet.yaml
-    $env_python attack_vnodes.py $config_file $num_clients $num_chunks
+    python attack_vnodes.py $config_file $num_clients $num_chunks
     echo "Finished Shatter with k=$num_chunks Reconstructions!"
     sleep 2
 done
@@ -35,6 +30,6 @@ echo "Reconstructed images can be found in experiments/"
 
 echo "Aggregating Metrics"
 
-$env_python aggregateLPIPS.py
+python aggregateLPIPS.py
 
 echo "LPIPS plot and CSV generated!"