Merge pull request #8 from Fung-Lab/feature/installation

documentation: environment set-up and installation

Author: vxfung

.gitignore

@@ -109,6 +109,7 @@ venv/
 ENV/
 env.bak/
 venv.bak/
+env.yaml
 
 # PyCharm (JetBrains)
 .idea/**/workspace.xml

README.md

@@ -1,6 +1,40 @@
-## Development
+## Usage
+
+### Installation
+
+1. Pre-install step:
+   ```
+   pip install conda-merge
+   ```
+2. Create conda environment:
+   1. GPU machines:
+      ```
+      conda-merge env.common.yaml env.gpu.yaml > env.yaml
+      conda env create -f env.yaml
+      ```
+   
+   2. CPU-only machines:
+
+      1. M1 Macs (see https://github.com/pyg-team/pytorch_geometric/issues/4549):
+         ```
+         conda env create -f env.common.yaml
+         conda activate matdeeplearn
+         MACOSX_DEPLOYMENT_TARGET=10.9 CC=clang CXX=clang++ pip install torch-scatter torch-sparse torch-geometric -f https://data.pyg.org/whl/torch-1.12.0+cpu.html
+         ```
+         Note: if pip is using cached wheels and the build is failing, add `--no-cache-dir` flag.
 
-Install matdeeplearn with `pip install -e .`
+      2. Everyone else:
+         ```
+         conda-merge env.common.yaml env.cpu.yaml > env.yaml
+         conda env create -f env.yaml
+         ```
+
+3. Install package:
+   ```
+   pip install -e .
+   ```
+
+## Development
 
 #### Code Quality
 This project uses flake8, black, and isort for linting.

env.common.yaml

@@ -0,0 +1,16 @@
+name: matdeeplearn
+channels:
+  - pytorch
+  - conda-forge
+  - defaults
+dependencies:
+  - python=3.9.*
+  - pip
+  - pytorch=1.12.1
+  - pip:
+    - pre-commit
+    - numpy
+    - scipy
+    - ase=3.21.*
+    - black
+    - pandas

env.cpu.yaml

@@ -0,0 +1,7 @@
+dependencies:
+  - cpuonly
+  - pip:
+    - -f https://data.pyg.org/whl/torch-1.12.0+cpu.html
+    - torch-scatter
+    - torch-sparse
+    - torch-geometric

env.gpu.yaml

@@ -0,0 +1,7 @@
+dependencies:
+  - cudatoolkit=11.3
+  - pip:
+    - -f https://data.pyg.org/whl/torch-1.12.0+cu113.html
+    - torch-scatter
+    - torch-sparse
+    - torch-geometric