Simplified first run process + docs [needs test]

docker/Dockerfile

@@ -1,8 +1,4 @@
-ARG CUDA="10.2"
-ARG CUDNN="7"
-
-# TODO: conda installs its own version of cuda
-FROM nvidia/cuda:${CUDA}-cudnn${CUDNN}-devel-ubuntu18.04
+FROM nvidia/cuda:10.1-cudnn7-devel-ubuntu18.04
 
 ENV CUDA_PATH /usr/local/cuda
 ENV CUDA_ROOT /usr/local/cuda/bin
@@ -38,26 +34,28 @@ ENV PATH "/home/runner/miniconda3/bin:$PATH:$CUDA_ROOT"
 RUN conda update -n base conda -yq
 RUN conda install python=3.8
 RUN conda install numpy pyyaml mkl mkl-include setuptools cmake cffi typing boost
-RUN conda install pytorch torchvision cudatoolkit=10.2 -c pytorch
+RUN conda install pytorch torchvision cudatoolkit=10.1 -c pytorch
 RUN conda install scipy pandas scikit-learn scikit-image -yq
 RUN conda install cython tqdm jupyter sqlalchemy -yq
 
+RUN git clone https://github.com/NVIDIA/apex \
+ && cd apex \
+ && pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./ \
+ && cd ..
+
 RUN python -m pip install opencv-python simpleitk h5py -q
 RUN python -m pip install tb-nightly wandb -q
 RUN python -m pip install monai memcnn -q
 RUN python -m pip install omegaconf loguru -q
 
+
 USER root
 
-# Create runner for input and output
-RUN mkdir /runner && chmod 777 /runner
-RUN mkdir /runner/ganslate && chmod 777 /runner/ganslate
-RUN mkdir /runner/datasets && chmod 777 /runner/datasets
+# Create folders for input and output
+RUN mkdir /ganslate && chmod 777 /ganslate
+RUN mkdir /data && chmod 777 /data
 USER runner
 
-# Copy files into the docker
-COPY [".", "/runner"]
-
-WORKDIR /runner
+WORKDIR /ganslate
 
 ENTRYPOINT $0 $@

docs/installation.md

@@ -1,35 +1,47 @@
 # Installation
 
+You can begin the install by cloning the repo,
+```
+git clone https://github.com/Maastro-CDS-Imaging-Group/ganslate.git
+```
+
+Once the repo is fetched, you can either use a docker-based install or a conda-based install on your local system.
+
+!!! note 
+	You can get to the repository root level after cloning,
+	```
+	cd ganslate
+	```
+
+
 ## Docker
 *Supported operating systems: Linux, [Windows with WSL](https://docs.nvidia.com/cuda/wsl-user-guide/index.html)*
 
 Dockerized setup is the easiest way to get started with the framework. If you do not have docker installed, you can follow instructions [here](https://docs.docker.com/engine/install/ubuntu/#install-using-the-repository)
 
 
-Here is a step-by-step walkthrough of how to install the framework with docker,
-
-1. The docker image can be built using,
+First, you can build the docker image by running the following commands at the repository root level
 
 ```console
 cd docker/
 docker build -t ganslate:latest .
 ```
 
-2.  Once the docker image is built, it can be run using,
+Once the docker image is built, it can be run using,
 
 ```console
 docker run --gpus all -it \
-	--shm-size=24gb --volume=<data_dir>:/data --volume=<code_dir>:/code \
+	--shm-size=24gb --volume=<data_dir>:/data --volume=<repo_dir>:/ganslate \
 	--name=ganslate ganslate:latest /bin/bash
 ```
 
-
-
 The docker container [mounts volumes from the host system](https://docs.docker.com/storage/volumes/) to allow easier persistence of data. 
 
-`<data_dir>` must be replaced with the full path of a directory where your data is located. It can also point to an empty directory during setup (Data can be moved into this directory later as docker mounts the directory to the container). 
+`<data_dir>` must be replaced with the full path of a directory where your data is located.
+`<repo_dir>` must be replaced with the path to the `ganslate` repository was cloned. 
 
-`<code_dir>` must be replaced with the path to the `ganslate` repository. 
+!!! note
+	`<data_dir>` can initially point to an empty directory to simplify setup. Data can be moved into this directory later as docker mounts the directory to the container. 
 
 ## Conda
 

docs/tutorials/horse2zebra.md

@@ -1,29 +1,25 @@
 # Your First Run (horse2zebra)
 
-
-## For Docker Install
-If you haven't already, please follow the installation using [Docker](../installation.md#docker)
-
-
+For both types of install, running the basic horse2zebra example is similar.
 #### Data download
-The horse2zebra dataset can be downloaded to the `<data_dir>` in the host system using instructions below,
+The horse2zebra dataset can be downloaded using instructions below,
 1. Open a terminal inside the cloned repository and run, 
 
 ```console
 cd projects/horse2zebra
-bash download_cyclegan_dataset.sh horse2zebra <data_dir>
+bash download_cyclegan_dataset.sh horse2zebra .
 ```
 
-### Training
-Next, the following commands are to be run in the docker container shell 
-
-![docker_commands](../imgs/your_first_run_docker.png)
+!!! note
+    The horse2zebra dataset will be downloaded at the root-level of the ganslate directory. This can be changed by providing a <data_dir> to the command `bash download_cyclegan_dataset.sh horse2zebra <data_dir>. However, the yaml files need to be manually changed. For non-advanced users, it is best to stick to the default location
 
-Run the training using,
+### Training
+Next, you can run the training using the command below,
 
 ```console
 cd /code
-python tools/train.py config=projects/horse2zebra/experiments/default_docker.yaml
+python tools/train.py config=projects/horse2zebra/experiments/default.yaml
 ```
 
-NOTE: If you have more than one GPU, then you can either run the training in distributed mode or [set CUDA_VISIBLE_DEVICES environment variable](https://developer.nvidia.com/blog/cuda-pro-tip-control-gpu-visibility-cuda_visible_devices/) to use only single GPUs.
+!!! note
+    If you have more than one GPU, then you can either run the training in distributed mode or [set CUDA_VISIBLE_DEVICES environment variable](https://developer.nvidia.com/blog/cuda-pro-tip-control-gpu-visibility-cuda_visible_devices/) to use only single GPUs.