nnf_gumbel_softmax

Samples from the Gumbel-Softmax distribution and
optionally discretizes.

Provides functionality to define and train neural networks similar to
'PyTorch' by Paszke et al (2019) <arXiv:1912.01703> but written entirely in R
using the 'libtorch' library. Also supports low-level tensor operations and
'GPU' acceleration.

Daniel Falbel

torch

Tensors and Neural Networks with 'GPU' Acceleration

Javier Luraschi

Dmitriy Selivanov

Athos Damiani

Christophe Regouby

Krzysztof Joachimiak

Hamada S. Badr

 RStudio

nnf_gumbel_softmax function

<dl><dt>logits</dt>
<dd><code>[..., num_features]</code> unnormalized log probabilities</dd>
<dt>tau</dt>
<dd>non-negative scalar temperature</dd>
<dt>hard</dt>
<dd>if <code>True</code>, the returned samples will be discretized as one-hot vectors, but will be differentiated as if it is the soft sample in autograd</dd>
<dt>dim</dt>
<dd>(int) A dimension along which softmax will be computed. Default: -1.</dd></dl>

Arguments

Gumbel_softmax — nnf_gumbel_softmax

<dl>

<dt>logits</dt>
<dd><code>[..., num_features]</code> unnormalized log probabilities</dd>


<dt>tau</dt>
<dd>non-negative scalar temperature</dd>


<dt>hard</dt>
<dd>if <code>True</code>, the returned samples will be discretized as one-hot vectors, but will be differentiated as if it is the soft sample in autograd</dd>


<dt>dim</dt>
<dd>(int) A dimension along which softmax will be computed. Default: -1.</dd>

</dl>

nnf_gumbel_softmax: Gumbel_softmax

Description

Usage

Arguments