print([name for name in dir(torch.distributions)
if not name.startswith('_')][:20])Dive into Deep Learning · §1.7
APIs change; the skill of looking things up doesn’t
discover · inspect · read · verify.
Motivation
No book covers a whole framework, and the libraries keep changing under us. The durable skill is a four-move loop, run without leaving the notebook, repeated until the call does what you want.
The official docs remain the source of truth: bookmark your framework’s reference and tutorial pages.
Discover
Know roughly where a tool should live, but not its name? dir lists a module’s contents; the names alone sketch what is on offer:
['AbsTransform', 'AffineTransform', 'Bernoulli', 'Beta', 'Binomial', 'CatTransform', 'Categorical', 'Cauchy', 'Chi2', 'ComposeTransform', ...
Skip the _-prefixed internals. In a notebook, module. + Tab gives the same list, filtered as you type, usually the fastest way to turn up a name.
Inspect · read
help(...) prints the docstring: arguments, defaults, return value, often an example.
In Jupyter, ones? opens the docstring in a side pane, and ones?? shows the source code: the final word when a docstring is terse or ambiguous.
Verify
Docstrings drift out of date; a running call does not lie:
tensor([1., 1., 1., 1.])
Exactly the promised shape and values. This discover → inspect → read → verify loop outlives every function in this book.
Assistants
An assistant usually produces a plausible function and a working call in seconds. Treat the suggestion like a knowledgeable colleague’s tip: a good starting point that still gets a two-line check before you build on it.
Glance at the signature (help / ?), then run a small example. A suggestion that survives both is one you can rely on.
Wrap-up
dir (or Tab-completion).help / ?; read the source with ??.