Minor docs fix for math rendering

Author: kavanase

docs/conf.py

@@ -33,6 +33,11 @@
     "sphinx_rtd_theme",
     "myst_parser",
 ]
+myst_enable_extensions = [
+    "html_admonition",
+    "dollarmath",  # "amsmath", # to parse Latex-style math
+]
+
 autodoc_member_order = "bysource"
 autosummary_generate = True
 source_suffix = [".rst", ".md"]

docs/guide/conventions.md

@@ -22,11 +22,13 @@ Training labels for stress in the original dataset must be pre-processed by the
 ```
 
 Stress also includes an arbitrary sign convention, for which we adopt the choice that `virial = -stress x volume  <=>  stress = (-1/volume) * virial`. This is the most common sign convention in the literature (e.g. adopted by `ASE`), but notably differs from that used by VASP (see [here](https://www.vasp.at/wiki/index.php/ISIF)). In the sign convention used by `nequip`, stress is defined as the derivative of the energy $E$ with respect to the strain tensor $\eta_{ji}$:
+
 $$\sigma_{ij} = \frac{\delta E} {\delta \eta_{ji}}$$
+
 such that a positive in the diagonals implies the system is _under tensile strain_ and wants to compress, while a negative value implies the system is _under compressive strain_ and wants to expand. When VASP results are parsed by `ASE`, the sign is flipped to match the `nequip` convention.
 
 ```{warning}
 Training labels for stress in the original dataset must be pre-processed by the user to be in **this sign convention**, which they may or may not already be depending on their origin.
 ```
 
-Users that have data with virials but seek to train on stress can use the data transform `nequip.data.transforms.VirialToStressTransform`.
+Users that have data with virials but seek to train on stress can use the data transform `nequip.data.transforms.VirialToStressTransform`.