package kaun

You can search for identifiers within the package.

in-package search v0.2.0

Neural networks for OCaml

Install

dune-project

Dependency

github.com Readme Changelog Edit opam file Versions (4)

Authors

Thibaut Mattio

Maintainers

Thibaut Mattio

Sources

raven-1.0.0.alpha3.tbz

sha256=96d35ce03dfbebd2313657273e24c2e2d20f9e6c7825b8518b69bd1d6ed5870f

sha512=90c5053731d4108f37c19430e45456063e872b04b8a1bbad064c356e1b18e69222de8bfcf4ec14757e71f18164ec6e4630ba770dbcb1291665de5418827d1465

doc/kaun.datasets/Kaun_datasets/index.html

Module `Kaun_datasets`Source

Dataset loaders for kaun.

Datasets are downloaded on demand and cached locally under $RAVEN_CACHE_ROOT/datasets/ (or $XDG_CACHE_HOME/raven/datasets/).

Source

val mnist : 
  ?fashion:bool ->
  ?normalize:bool ->
  ?data_format:[ `NCHW | `NHWC ] ->
  unit ->
  (Nx.float32_t * Nx.int32_t) * (Nx.float32_t * Nx.int32_t)

mnist () is ((x_train, y_train), (x_test, y_test)).

Images are float32 in [0, 1] (when normalize is true, the default). Labels are int32 class indices.

fashion selects Fashion-MNIST when true. Defaults to false. data_format defaults to `NCHW.

Tensor shapes:

`NCHW: images [N; 1; 28; 28], labels [N]
`NHWC: images [N; 28; 28; 1], labels [N]

Raises Failure on download or parsing errors.

Source

val cifar10 : 
  ?normalize:bool ->
  ?data_format:[ `NCHW | `NHWC ] ->
  unit ->
  (Nx.float32_t * Nx.int32_t) * (Nx.float32_t * Nx.int32_t)

cifar10 () is ((x_train, y_train), (x_test, y_test)).

Images are float32 in [0, 1] (when normalize is true, the default). Labels are int32 class indices (0--9: airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck).

data_format defaults to `NCHW.

Tensor shapes:

`NCHW: images [N; 3; 32; 32], labels [N]
`NHWC: images [N; 32; 32; 3], labels [N]

Raises Failure on download, extraction, or parsing errors.