Minor documentation updates.

Author: vbharadwaj-bk

docs/conf.py

@@ -26,7 +26,7 @@
 # https://www.sphinx-doc.org/en/master/usage/configuration.html#options-for-html-output
 
 html_theme = "furo"
-html_static_path = ["_static"]
+#html_static_path = ["_static"]
 
 extensions = [
     "sphinx.ext.autodoc",

docs/installation.rst

@@ -26,7 +26,7 @@ much faster since this extension is cached.
 
 
 Compiling the Integrated PyTorch Extension
------------------------------
+------------------------------------------
 To support ``torch.compile``, ``torch.export``, and
 JITScript, OpenEquivariance needs to compile a C++ extension
 tightly integrated with PyTorch. If you see a warning that
@@ -50,13 +50,13 @@ These configuration steps are required only ONCE after
 installation (or upgrade) with pip. 
 
 Configurations on Major Platforms 
------------------------------
+---------------------------------
 OpenEquivariance has been tested on both supercomputers and lab clusters.
 Here are some tested environment configuration files. If use OpenEquivariance
 on a widely-used platform, send us a pull request to add your configuration! 
 
 NERSC Perlmutter (NVIDIA A100)
-""""""""""""""""""""""
+""""""""""""""""""""""""""""""
 
 .. code-block:: bash
     :caption: env.sh (last updated June 2024)
@@ -73,7 +73,7 @@ NERSC Perlmutter (NVIDIA A100)
 
 
 OLCF Frontier (AMD MI250x)
-""""""""""""""""""""""
+""""""""""""""""""""""""""
 You need to install a HIP-enabled verison of PyTorch to use our package. 
 To do this, follow the steps `here <https://docs.olcf.ornl.gov/software/analytics/pytorch_frontier.html>`_.
 

docs/supported_ops.rst

@@ -45,7 +45,7 @@ Our code includes correctness checks, but the configuration space is large. If y
 a bug, let us know in a GitHub issue. We'll try our best to correct it or document the problem here.
 
 Unsupported Tensor Product Configurations 
---------------------
+-----------------------------------------
 
 We do not (yet) support:
 
@@ -74,7 +74,7 @@ TorchBind in earlier versions.
 
 
 Multiple Devices and Streams 
---------------------
+----------------------------
 OpenEquivariance compiles kernels based on the compute capability of the
 first visible GPU. On heterogeneous systems, our kernels
 will only execute correctly on devices that share the compute capability 

openequivariance/implementations/convolution/TensorProductConv.py

@@ -17,7 +17,9 @@ class TensorProductConv(torch.nn.Module, LoopUnrollConv):
     :math:`y_1...y_{|E|}`, and weights :math:`W_1...W_{|E|}`, computes
 
     .. math::
+
         z_i = \sum_{(i, j, e) \in \mathcal{N}(i)} W_e (x_j \otimes_{\\textrm{CG}} y_e)
+
     where :math:`(i, j, e) \in \mathcal{N}(i)` indicates that node :math:`i` is connected to node :math:`j`
     via the edge indexed :math:`e`.
 
@@ -29,7 +31,7 @@ class TensorProductConv(torch.nn.Module, LoopUnrollConv):
     :param deterministic: if ``False``, uses atomics for the convolution. If ``True``, uses a deterministic
            fixup-based algorithm. `Default`: ``False``.
     :param kahan: if ``True``, uses Kahan summation to improve accuracy during aggregation. To use this option,
-              the input tensors must be in float32 precision AND you must set ``deterministic=True``. *Default*: ``False``.
+           the input tensors must be in float32 precision AND you must set ``deterministic=True``. *Default*: ``False``.
 
     """