Function at::_flash_attention_forward¶

Function Documentation¶

inline ::std::tuple<at::Tensor, at::Tensor, at::Tensor, at::Tensor, at::Tensor> at::_flash_attention_forward(const at::Tensor &query, const at::Tensor &key, const at::Tensor &value, const ::std::optional<at::Tensor> &cum_seq_q, const ::std::optional<at::Tensor> &cum_seq_k, int64_t max_q, int64_t max_k, double dropout_p, bool is_causal, bool return_debug_mask, ::std::optional<double> scale = ::std::nullopt, ::std::optional<int64_t> window_size_left = ::std::nullopt, ::std::optional<int64_t> window_size_right = ::std::nullopt, const ::std::optional<at::Tensor> &seqused_k = {}, const ::std::optional<at::Tensor> &alibi_slopes = {})¶