torch 0.12 · OCaml Package

Layer Types and Conversions

type t

A layer takes as input a tensor and returns a tensor through the forward function. Layers can hold variables, these are created and registered using a Var_store.t when creating the layer.

type t_with_training

A layer of type t_with_training is similar to a layer of type t except that it is also given a boolean argument when applying it to a tensor that specifies whether the layer is currently used in training or in testing mode. This is typically the case for batch normalization or dropout.

val with_training : t -> t_with_training

with_training t returns a layer using the is_training argument from a standard layer. The is_training argument is discarded. This is useful when sequencing multiple layers via fold.

Basic Layer Creation

val id : t

The identity layer. forward id tensor returns tensor.

val id_ : t_with_training

The identity layer with an is_training argument.

val of_fn : (Tensor.t -> Tensor.t) -> t

of_fn f creates a layer based on a function from tensors to tensors.

val of_fn_ : (Tensor.t -> is_training:Base.bool -> Tensor.t) -> t_with_training

of_fn_ f creates a layer based on a function from tensors to tensors. f also has access to the is_training flag.

val sequential : t Base.list -> t

sequential ts applies sequentially a list of layers ts.

val sequential_ : t_with_training Base.list -> t_with_training

sequential_ ts applies sequentially a list of layers ts.

val forward : t -> Tensor.t -> Tensor.t

forward t tensor applies layer t to tensor.

val forward_ : t_with_training -> Tensor.t -> is_training:Base.bool -> Tensor.t

forward_ t tensor ~is_training applies layer t to tensor with the specified is_training flag.

Linear and Convolution Layers

type activation =

| Relu
| Softmax
| Log_softmax
| Tanh
| Leaky_relu
| Sigmoid

The different kind of activations supported by the various layers below.

val linear : 
  Var_store.t ->
  ?activation:activation ->
  ?use_bias:Base.bool ->
  ?w_init:Var_store.Init.t ->
  input_dim:Base.int ->
  Base.int ->
  t

linear vs ~input_dim output_dim returns a linear layer. When using forward, the input tensor has to use a shape of batch_size * input_dim. The returned tensor has a shape batch_size * output_dim.

val conv2d : 
  Var_store.t ->
  ksize:(Base.int * Base.int) ->
  stride:(Base.int * Base.int) ->
  ?activation:activation ->
  ?use_bias:Base.bool ->
  ?w_init:Var_store.Init.t ->
  ?padding:(Base.int * Base.int) ->
  ?groups:Base.int ->
  input_dim:Base.int ->
  Base.int ->
  t

conv2d vs ~ksize ~stride ~input_dim output_dim returns a 2 dimension convolution layer. ksize specifies the kernel size and stride the stride. When using forward, the input tensor should have a shape batch_size * input_dim * h * w and the returned tensor will have a shape batch_size * output_dim * h' * w'.

val conv2d_ : 
  Var_store.t ->
  ksize:Base.int ->
  stride:Base.int ->
  ?activation:activation ->
  ?use_bias:Base.bool ->
  ?w_init:Var_store.Init.t ->
  ?padding:Base.int ->
  ?groups:Base.int ->
  input_dim:Base.int ->
  Base.int ->
  t

conv2d_ is similar to conv2d but uses the same kernel size, stride, and padding on both the height and width dimensions, so a single integer needs to be specified for these parameters.

val conv_transpose2d : 
  Var_store.t ->
  ksize:(Base.int * Base.int) ->
  stride:(Base.int * Base.int) ->
  ?activation:activation ->
  ?use_bias:Base.bool ->
  ?w_init:Var_store.Init.t ->
  ?padding:(Base.int * Base.int) ->
  ?output_padding:(Base.int * Base.int) ->
  ?groups:Base.int ->
  input_dim:Base.int ->
  Base.int ->
  t

conv_transpose2d creates a 2D transposed convolution layer, this is sometimes also called 'deconvolution'.

val conv_transpose2d_ : 
  Var_store.t ->
  ksize:Base.int ->
  stride:Base.int ->
  ?activation:activation ->
  ?use_bias:Base.bool ->
  ?w_init:Var_store.Init.t ->
  ?padding:Base.int ->
  ?output_padding:Base.int ->
  ?groups:Base.int ->
  input_dim:Base.int ->
  Base.int ->
  t

conv_transpose2d_ is similar to conv_transpose2d but uses a single value for the height and width dimension for the kernel size, stride, padding and output padding.

Normalization

val batch_norm2d : 
  Var_store.t ->
  ?w_init:Var_store.Init.t ->
  ?cudnn_enabled:Base.bool ->
  ?eps:Base.float ->
  ?momentum:Base.float ->
  Base.int ->
  t_with_training

batch_norm2d vs dim creates a batch norm 2D layer. This layer applies Batch Normalization over a 4D input batch_size * dim * h * w. The returned tensor has the same shape.

val layer_norm : 
  Var_store.t ->
  ?cudnn_enable:Base.bool ->
  ?eps:Base.float ->
  Base.int ->
  t

Recurrent Neural Networks

module Lstm : sig ... end

A Long Short Term Memory (LSTM) recurrent neural network.

module Gru : sig ... end

Embeddings

val embeddings : 
  ?sparse:Base.bool ->
  ?scale_grad_by_freq:Base.bool ->
  Var_store.t ->
  num_embeddings:Base.int ->
  embedding_dim:Base.int ->
  t