Page Arithmetics

WIP

Page Automatic Differentiation (Autodiff)

WIP

This is not currently supported

WIP

Namespace dl::detail

refine to check that T is actually a Model

Member dl::Embedding::forward (const dl::TensorPtr input)

implement

Member dl::InitializerTensor< T >::InitializerTensor (std::initializer_list< InitializerTensor > &&value) noexcept

Check if all values have the same size

Member dl::Model< R(Args...)>::~Model ()=default

For later: these const member functions make sense to indicate that we know at compile time that the instance is not modified (e.g. since it is not part of the computation graph for auto differentiation.

Member dl::Trainer< Model, Dataset, Optimizer >::fit (Model &model, auto adapter)

log the loss

with batching increment by batch size

Struct fmt::formatter< std::vector< T > >

remove when formatting ranges is supported

Page Getting Started

WIP

Example training, validating and evaluating on MNIST Digits Create a new file called main.cpp in your src directory with the following contents:

Member nlp::BERT::forward (const dl::TensorPtr &input) override

implement

Member nlp::BERTEmbeddings::forward (const dl::TensorPtr &inputIds, const dl::TensorPtr &inputTokenTypes)

implement

Page Transformer

WIP