Shortcuts

Struct EmbeddingBagOptions

Page Contents

Struct Documentation

struct EmbeddingBagOptions

Options for the EmbeddingBag module.

Example:

EmbeddingBag model(EmbeddingBagOptions(10,
2).max_norm(2).norm_type(2.5).scale_grad_by_freq(true).sparse(true).mode(torch::kSum));

Public Functions

EmbeddingBagOptions(int64_t num_embeddings, int64_t embedding_dim)
inline auto num_embeddings(const int64_t &new_num_embeddings) -> decltype(*this)

The size of the dictionary of embeddings.

inline auto num_embeddings(int64_t &&new_num_embeddings) -> decltype(*this)
inline const int64_t &num_embeddings() const noexcept
inline int64_t &num_embeddings() noexcept
inline auto embedding_dim(const int64_t &new_embedding_dim) -> decltype(*this)

The size of each embedding vector.

inline auto embedding_dim(int64_t &&new_embedding_dim) -> decltype(*this)
inline const int64_t &embedding_dim() const noexcept
inline int64_t &embedding_dim() noexcept
inline auto max_norm(const c10::optional<double> &new_max_norm) -> decltype(*this)

If given, each embedding vector with norm larger than max_norm is renormalized to have norm max_norm.

inline auto max_norm(c10::optional<double> &&new_max_norm) -> decltype(*this)
inline const c10::optional<double> &max_norm() const noexcept
inline c10::optional<double> &max_norm() noexcept
inline auto norm_type(const double &new_norm_type) -> decltype(*this)

The p of the p-norm to compute for the max_norm option. Default 2.

inline auto norm_type(double &&new_norm_type) -> decltype(*this)
inline const double &norm_type() const noexcept
inline double &norm_type() noexcept
inline auto scale_grad_by_freq(const bool &new_scale_grad_by_freq) -> decltype(*this)

If given, this will scale gradients by the inverse of frequency of the words in the mini-batch.

Default false. Note: this option is not supported when mode="kMax".

inline auto scale_grad_by_freq(bool &&new_scale_grad_by_freq) -> decltype(*this)
inline const bool &scale_grad_by_freq() const noexcept
inline bool &scale_grad_by_freq() noexcept
inline auto mode(const EmbeddingBagMode &new_mode) -> decltype(*this)

"kSum", "kMean" or "kMax".

Specifies the way to reduce the bag. "kSum" computes the weighted sum, taking per_sample_weights into consideration. "kMean" computes the average of the values in the bag, "kMax" computes the max value over each bag.

inline auto mode(EmbeddingBagMode &&new_mode) -> decltype(*this)
inline const EmbeddingBagMode &mode() const noexcept
inline EmbeddingBagMode &mode() noexcept
inline auto sparse(const bool &new_sparse) -> decltype(*this)

If true, gradient w.r.t.

weight matrix will be a sparse tensor. Note: this option is not supported when mode="kMax".

inline auto sparse(bool &&new_sparse) -> decltype(*this)
inline const bool &sparse() const noexcept
inline bool &sparse() noexcept
inline auto _weight(const torch::Tensor &new__weight) -> decltype(*this)

The learnable weights of the module of shape (num_embeddings, embedding_dim)

inline auto _weight(torch::Tensor &&new__weight) -> decltype(*this)
inline const torch::Tensor &_weight() const noexcept
inline torch::Tensor &_weight() noexcept
inline auto include_last_offset(const bool &new_include_last_offset) -> decltype(*this)

If true, offsets has one additional element, where the last element is equivalent to the size of indices.

This matches the CSR format.

inline auto include_last_offset(bool &&new_include_last_offset) -> decltype(*this)
inline const bool &include_last_offset() const noexcept
inline bool &include_last_offset() noexcept
inline auto padding_idx(const c10::optional<int64_t> &new_padding_idx) -> decltype(*this)

If specified, the entries at padding_idx do not contribute to the gradient; therefore, the embedding vector at padding_idx is not updated during training, i.e.

it remains as a fixed “pad”. For a newly constructed EmbeddingBag, the embedding vector at padding_idx will default to all zeros, but can be updated to another value to be used as the padding vector. Note that the embedding vector at padding_idx is excluded from the reduction.

inline auto padding_idx(c10::optional<int64_t> &&new_padding_idx) -> decltype(*this)
inline const c10::optional<int64_t> &padding_idx() const noexcept
inline c10::optional<int64_t> &padding_idx() noexcept

Docs

Access comprehensive developer documentation for PyTorch

View Docs

Tutorials

Get in-depth tutorials for beginners and advanced developers

View Tutorials

Resources

Find development resources and get your questions answered

View Resources