Skip to content

Latest commit

 

History

History
65 lines (50 loc) · 1.88 KB

CHANEGLOG.md

File metadata and controls

65 lines (50 loc) · 1.88 KB
Raw

Changelog

V0.x

Changes

  • Moved image related layers to serket.image

  • ScanRNN changes:

    • cell.init_state is deprecated use sk.tree_state(cell, ...) instead.

  • Naming changes:

    • ***_init_func -> ***_init shorter and more concise
    • gamma_init_func -> weight_init
    • beta_init_func -> bias_init
    • act_func -> act

  • MLP produces smaller jaxprs and are faster to compile. for my use case -higher order differentiation through PINN- the new MLP is faster to compile.

  • kernel_dilation -> dilation

  • input_dilation -> Removed.

  • p -> drop_rate in all dropout layers

  • FlipLeftRight2D -> HorizontalFlip2D

  • FlipUpDown2D -> VerticalFlip2D

  • sk.nn.{Sequential,RandomChoice} to sk.{Sequential,RandomChoice}. as they are applicable to other modules and not specfic to nn

Additions

  • tree_eval: a dispatcher to define layers evaluation rule. for example Dropout is changed to Identity when tree_eval is applied.

    @sk.tree_eval.def_eval(sk.nn.Dropout)
def dropout_evaluation(_) -> sk.nn.Identity:
    return sk.nn.Identity()

  • tree_state: a dispatcher to define state intialization for BatchNorm, RNN cells.

    @sk.tree_state.def_state(sk.nn.SimpleRNNCell)
def simple_rnn_init_state(cell: SimpleRNNCell, array: jax.Array  | None) -> SimpleRNNState:
    del kwargs # unused
    return SimpleRNNState(jnp.zeros([cell.hidden_features]))

  • MultiHeadAttention

  • BatchNorm

  • RandomHorizontalShear2D

  • RandomPerspective2D

  • RandomRotate2D

  • RandomVerticalShear2D

  • Rotate2D

  • VerticalShear2D

  • Pixelate2D

  • Solarize2D

  • Posterize2D

  • RandomJigSaw2D

  • FFTAvgBlur2D

  • FFTGaussianBlur2D

Deprecations

  • Bilinear is deprecated, use Multilinear((in1_features, in2_features), out_features)
  • HistogramEqualization2D
  • Remove .blocks, and will move it to examples