Added documentation with pydoc.

Makefile

@@ -0,0 +1,2 @@
+docs: 
+	python -m pdoc --docformat numpy --logo "https://avatars.githubusercontent.com/u/88346553?s=200&v=4" dect -o ./docs

README.md

@@ -33,17 +33,22 @@ with or without CUDA support. The DECT implementation also relies on the
 of a vector given an index vector. 
 
 - First install `torch` and `torch-scatter`.
-   ```{python}
+
+   ```python
    pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117
    pip install torch-scatter -f https://data.pyg.org/whl/torch-2.1.0+${CUDA}.html
    ```
+
 - Then install our package in either one of two ways:
    - As a git submodule 
-   ```{pyton}
+
+   ```sh
    git submodule add https://github.com/aidos-lab/dect.git
    ```
+
    - Or as a pip installable package. 
-   ```{python}
+
+   ```sh
    pip install git+https://github.com/aidos-lab/dect.git
    ```
 
@@ -54,7 +59,7 @@ ECT of the gif of this readme.
 The code is provided on an as is basis. You are cordially invited to both contribute and 
 provide feedback. Do not hesitate to contact us.
 
-```{python}
+```python
 import torch
 from torch_geometric.data import Data, Batch
 from dect.ect import EctConfig, EctLayer

dect/__init__.py

@@ -0,0 +1,3 @@
+"""
+.. include:: ../README.md
+"""

dect/directions.py

@@ -7,7 +7,9 @@
 import torch
 
 
-def generate_uniform_directions(num_thetas: int = 64, d: int = 3, device: str = "cpu"):
+def generate_uniform_directions(
+    num_thetas: int = 64, d: int = 3, device: str = "cpu"
+):
     """
     Generate randomly sampled directions from a sphere in d dimensions.
 
@@ -49,8 +51,12 @@ def generate_uniform_2d_directions(num_thetas: int = 64, device: str = "cpu"):
     """
     v = torch.vstack(
         [
-            torch.sin(torch.linspace(0, 2 * torch.pi, num_thetas, device=device)),
-            torch.cos(torch.linspace(0, 2 * torch.pi, num_thetas, device=device)),
+            torch.sin(
+                torch.linspace(0, 2 * torch.pi, num_thetas, device=device)
+            ),
+            torch.cos(
+                torch.linspace(0, 2 * torch.pi, num_thetas, device=device)
+            ),
         ]
     )
 
@@ -88,3 +94,19 @@ def generate_multiview_directions(num_thetas: int, bump_steps: int, d: int):
         v[idx, idx_pair[1], :] = w[1]
 
     return v
+
+
+def test_fn():
+    """Yields
+    ------
+    err_code : int
+        Non-zero value indicates error code, or zero on success.
+    err_msg : str or None
+        Human readable error message, or None on success.
+
+    Returns
+    -------
+    int
+        Description of anonymous integer return value.
+
+    """

dect/ect.py

@@ -81,15 +81,17 @@ def compute_ecc(
     """
     ecc = torch.nn.functional.sigmoid(scale * torch.sub(lin, nh))
 
-    # Due to (I believe) a bug in segment_add_coo, we have to first transpose and
-    # then apply segment add. In the original code movedim was applied after
+    # Due to (I believe) a bug in segment_add_coo, we have to first transpose
+    # and then apply segment add. In the original code movedim was applied after
     # and that yields an bug in the backwards pass. Will have to be reported to
     # pytorch eventually.
     ecc = ecc.movedim(0, 2).movedim(0, 1)
     return segment_add_coo(ecc, index)
 
 
-def compute_ect_points(batch: Batch, v: torch.FloatTensor, lin: torch.FloatTensor):
+def compute_ect_points(
+    batch: Batch, v: torch.FloatTensor, lin: torch.FloatTensor
+):
     """Computes the Euler Characteristic Transform of a batch of point clouds.
 
     Parameters
@@ -109,7 +111,9 @@ def compute_ect_points(batch: Batch, v: torch.FloatTensor, lin: torch.FloatTenso
     return compute_ecc(nh, batch.batch, lin)
 
 
-def compute_ect_edges(data: Batch, v: torch.FloatTensor, lin: torch.FloatTensor):
+def compute_ect_edges(
+    data: Batch, v: torch.FloatTensor, lin: torch.FloatTensor
+):
     """Computes the Euler Characteristic Transform of a batch of graphs.
 
     Parameters
@@ -145,7 +149,9 @@ def compute_ect_edges(data: Batch, v: torch.FloatTensor, lin: torch.FloatTensor)
     )
 
 
-def compute_ect_faces(data: Batch, v: torch.FloatTensor, lin: torch.FloatTensor):
+def compute_ect_faces(
+    data: Batch, v: torch.FloatTensor, lin: torch.FloatTensor
+):
     """Computes the Euler Characteristic Transform of a batch of meshes.
 
     Parameters

docs/index.html

@@ -0,0 +1,7 @@
+<!doctype html>
+<html>
+<head>
+    <meta charset="utf-8">
+    <meta http-equiv="refresh" content="0; url=./dect.html"/>
+</head>
+</html>

tests/test_ecc.py

@@ -10,7 +10,8 @@
 class TestEct:
     """
     1. When normalized, the ect needs to be normalized.
-    2. The dimensions need to correspond. e.g. the batches need not to be mixed up.
+    2. The dimensions need to correspond. e.g. the batches need not to
+        be mixed up.
     3. Test that when one of the inputs has a gradient the out has one too.
     """