Note 7.1.1 Numbered Emoji Sectioning.
-
Helps to make long scripts scannable.
-
Meant to be used as semantic anchor points, not decoration.
Subsection 7.1.3 Fundamental Features
Note 7.1.2 if name == "__main__".
-
Placed at the bottom of the script.
-
✔️ Only use for:
-
Running smoke tests.
-
Minimal demonstrations.
-
Sanity checks.
-
-
❌ Never use for:
-
Running large-scale experiments.
-
Running training loops.
-
Running inference pipelines.
-
Note 7.1.3 def main() -> None:.
-
Always exists.
-
Orchestrates calls to public API for smoke tests, demos, sanity checks.
-
Never contains math or logic explicitly.
def main() -> None:
_smoke_test()
Note 7.1.4 "Big functions" to bundle logic..
Use large functions to bundle related logic together.
Note 7.1.5 Type Annotations Everywhere (within reason).
🔑 Key Habits to Remember:
-
Use
Tensor, nottorch.Tensor, in type annotations. -
Annotate scalars (
float | Tensor) when physically meaningul. -
Use Final for constants:
omega: float | Final[Tensor] = 1.0
Note 7.1.6 NumPy-Style Docstrings.
These should always be present on:
-
public functions.
-
public classes.
-
nontrivial private helpers.
-
Minimal viable structure:
def public_function(x: Tensor) -> Tensor: """ Short description. Parameters ---------- x : Tensor Description of x. Returns ------- Tensor Description of return value. """ pass
