File size: 339 Bytes
c1af2fa |
1 2 3 4 5 6 7 8 9 10 11 12 13 |
#pragma once
#include <ATen/core/TensorBase.h>
#include <optional>
namespace at::cuda::detail {
TORCH_API void bf16bf16_grouped_mm(
at::Tensor mat_a, // bf16
at::Tensor mat_b, // bf16
std::optional<at::Tensor> offs,
std::optional<at::Tensor> bias, // BF16
at::Tensor& out);
} // namespace at::cuda::detail
|