#pragma once

// @generated by torchgen/gen.py from Function.h

#include <ATen/Context.h>
#include <ATen/DeviceGuard.h>
#include <ATen/TensorUtils.h>
#include <ATen/TracerMode.h>
#include <ATen/core/Generator.h>
#include <ATen/core/Reduction.h>
#include <ATen/core/Tensor.h>
#include <c10/core/Scalar.h>
#include <c10/core/Storage.h>
#include <c10/core/TensorOptions.h>
#include <c10/util/Deprecated.h>
#include <c10/util/Optional.h>



#include <ATen/ops/_efficient_attention_forward_ops.h>

namespace at {


// aten::_efficient_attention_forward(Tensor query, Tensor key, Tensor value, Tensor? bias, Tensor? cu_seqlens_q, Tensor? cu_seqlens_k, int? max_seqlen_q, float dropout_p, int custom_mask_type, bool compute_log_sumexp=False, *, float? scale=None, Tensor? causal_diagonal=None, Tensor? seqlen_k=None) -> (Tensor output, Tensor logsumexp, Tensor philox_seed, Tensor philox_offset)
inline ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> _efficient_attention_forward(const at::Tensor & query, const at::Tensor & key, const at::Tensor & value, const c10::optional<at::Tensor> & bias, const c10::optional<at::Tensor> & cu_seqlens_q, const c10::optional<at::Tensor> & cu_seqlens_k, c10::optional<int64_t> max_seqlen_q, double dropout_p, int64_t custom_mask_type, bool compute_log_sumexp=false, c10::optional<double> scale=c10::nullopt, const c10::optional<at::Tensor> & causal_diagonal={}, const c10::optional<at::Tensor> & seqlen_k={}) {
    return at::_ops::_efficient_attention_forward::call(query, key, value, bias, cu_seqlens_q, cu_seqlens_k, max_seqlen_q, dropout_p, custom_mask_type, compute_log_sumexp, scale, causal_diagonal, seqlen_k);
}

}