change version number

.github/workflows/release.yml

@@ -0,0 +1,29 @@
+name: Upload Python Package
+
+on:
+  release:
+    types: [created]
+
+jobs:
+  deploy:
+
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v2
+    - name: Set up Python
+      uses: actions/setup-python@v2
+      with:
+        python-version: 3.8
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install setuptools wheel twine
+    - name: Build and publish
+      env:
+        TWINE_USERNAME: ${{ secrets.PYPI_USERNAME }}
+        TWINE_PASSWORD: ${{ secrets.PYPI_PASSWORD }}
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      run: |
+        python setup.py sdist bdist_wheel
+        twine upload dist/*

ChangeLog.md

@@ -0,0 +1,12 @@
+# Changelog
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [Unreleased]
+
+## [0.4.0] - 2022-03-19
+
+### Added
+- Changelog

README.md

@@ -7,7 +7,7 @@ Disclamier: This is a work in progress. No part of the library can be considered
 ## What is different from the pytorch version?
 
 - no more `shared_weights` and `internal_weights` in `TensorProduct`. Extensive use of `jax.vmap` instead (see example below)
-- support of python list of `jnp.array` for the data. This allows to avoid unnecessary `jnp.concatenante` followed by indexing to reverse the concatenation
+- support of python structure `IrrepsData` that contains a contiguous version of the data and a list of `jnp.array` for the data. This allows to avoid unnecessary `jnp.concatenante` followed by indexing to reverse the concatenation
 - support of `None` in the list of `jnp.array` to avoid unnecessary computation with zeros
 
 ## Example
@@ -32,6 +32,32 @@ irreps.D_from_angles(alpha=0.0, beta=0.0, gamma=0.0, k=1)  # the matrix that app
 #              [ 0., -1.]], dtype=float32)
 ```
 
+`IrrepsData` contains both the irreps and the data.
+Here is the example of the tensor product of the two vectors.
+```python
+from e3nn_jax import full_tensor_product, IrrepsData
+
+out = full_tensor_product(
+    IrrepsData.from_contiguous("1o", jnp.array([2.0, 0.0, 0.0])),
+    IrrepsData.from_contiguous("1o", jnp.array([0.0, 2.0, 0.0]))
+)
+# out is of type `IrrepsData` and contains the following fields:
+
+out.irreps
+# 1x0e+1x1e+1x2e
+
+out.contiguous
+# DeviceArray([0.  , 0.  , 0.  , 2.83, 0.  , 2.83, 0.  , 0.  , 0.  ], dtype=float32)
+
+out.list
+# [DeviceArray([[0.]], dtype=float32),
+#  DeviceArray([[0.  , 0.  , 2.83]], dtype=float32),
+#  DeviceArray([[0.  , 2.83, 0.  , 0.  , 0.  ]], dtype=float32)]
+```
+
+The two fields `contiguous` and `list` contain the same information under different forms.
+This is not a performence issue, we rely on `jax.jit` to optimize the code and get rid of the unused operations.
+
 ## Shared weights
 
 `torch` version ([e3nn](github.com/e3nn/e3nn) repo):

e3nn_jax/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.3.1"
+__version__ = "0.4.0"
 
 from ._graph_util import index_add, radius_graph
 from ._rotation import (