v0.4.0

v0.4.0 (2025-01-10)

Feature

• feat: add convenience functions for working with one layer (4c24c8a)

• feat: add more nonlinearities (f997ab2)

v0.3.0

v0.3.0 (2024-11-08)

Feature

• feat: provide an immutable sequential module (2c8bcad)

• feat: monkeypatch Tensor repr to only show shape (63aa908)

Fix

• fix: interpret an empty layer iterable as an identity module (3cb36a5)

v0.2.0

v0.2.0 (2024-09-17)

Chore

• chore: remove TODO file (722f7eb)

Feature

• feat: don't unnecessarily wrap modules

Instead of providing a Layers class, provide a module_from_layers().
If all the layers put together produce only a single module, this makes
it possible to simply return that module, without any wrapping. (7e6fdc1)

• feat: allow verbose modules to take multiple arguments (0579427)

v0.1.0

v0.1.0 (2024-06-05)

Chore

• chore: recommit release workflow (4cdb711)

• chore: apply cookiecutter (abe4f57)

Documentation

• docs: explain why this library is useful (f208c68)

Feature

• feat: support pooling in dynamic factories (d639d0b)

• feat: add dynamically-generated layer factories (0052046)

• feat: rename project to 'torchyield' (e7a1898)

• feat: skip ReLU after last layer in MLP (bb70ce2)

• feat: initial commit (38a7a77)

Fix

• fix: restore the option (f3a2581)