mirror of
https://github.com/NVIDIA/TensorRT-LLM.git
synced 2026-01-14 06:27:45 +08:00
db4edea1e1
* Update TensorRT-LLM --------- Co-authored-by: Kota Tsuyuzaki <bloodeagle40234@gmail.com> Co-authored-by: Pzzzzz <hello-cd.plus@hotmail.com> Co-authored-by: Patrick Reiter Horn <patrick.horn@gmail.com>
73 lines
3.0 KiB
C++
73 lines
3.0 KiB
C++
/*
|
|
* Copyright (c) 2022-2024, NVIDIA CORPORATION. All rights reserved.
|
|
*
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
*/
|
|
|
|
#include "tensorrt_llm/kernels/buildRelativeAttentionBiasKernel.h"
|
|
#include "tensorrt_llm/thop/thUtils.h"
|
|
|
|
namespace th = torch;
|
|
namespace tl = tensorrt_llm;
|
|
namespace tk = tensorrt_llm::kernels;
|
|
|
|
namespace torch_ext
|
|
{
|
|
|
|
template <typename T>
|
|
void handleInvokeRelativeAttentionBias(th::Tensor& relative_attention_bias, th::Tensor& relative_attention_bias_table,
|
|
int64_t const num_head, int64_t const max_seq_len, int64_t const num_bucket, bool const is_bidirectional,
|
|
int64_t const max_distance, cudaStream_t stream)
|
|
{
|
|
|
|
T* relative_attention_bias_ptr = get_ptr<T>(relative_attention_bias);
|
|
T const* relative_attention_bias_table_ptr = get_ptr<T>(relative_attention_bias_table);
|
|
|
|
tk::invokeBuildRelativeAttentionBias<T>(relative_attention_bias_ptr, relative_attention_bias_table_ptr, num_head,
|
|
(max_seq_len + 1), num_bucket, is_bidirectional, max_distance, stream);
|
|
}
|
|
|
|
void buildRelativeAttentionBias(
|
|
th::Tensor& relative_attention_bias, // sizeof(T) * num_head * (max_seq_len + 1) * (max_seq_len + 1)
|
|
th::Tensor& relative_attention_bias_table, // sizeof(T) * num_head * num_bucket
|
|
int64_t const num_head, int64_t const max_seq_len, int64_t const num_bucket, bool const is_bidirectional,
|
|
int64_t const max_distance)
|
|
{
|
|
auto stream = at::cuda::getCurrentCUDAStream().stream();
|
|
switch (relative_attention_bias_table.scalar_type())
|
|
{
|
|
case at::ScalarType::Float:
|
|
handleInvokeRelativeAttentionBias<float>(relative_attention_bias, relative_attention_bias_table, num_head,
|
|
max_seq_len, num_bucket, is_bidirectional, max_distance, stream);
|
|
break;
|
|
case at::ScalarType::Half:
|
|
handleInvokeRelativeAttentionBias<half>(relative_attention_bias, relative_attention_bias_table, num_head,
|
|
max_seq_len, num_bucket, is_bidirectional, max_distance, stream);
|
|
break;
|
|
#ifdef ENABLE_BF16
|
|
case at::ScalarType::BFloat16:
|
|
handleInvokeRelativeAttentionBias<__nv_bfloat16>(relative_attention_bias, relative_attention_bias_table,
|
|
num_head, max_seq_len, num_bucket, is_bidirectional, max_distance, stream);
|
|
break;
|
|
#endif
|
|
default: throw std::runtime_error("Unimplemented scalar type");
|
|
}
|
|
|
|
sync_check_cuda_error();
|
|
}
|
|
|
|
} // namespace torch_ext
|
|
|
|
static auto relative_attention_bias
|
|
= torch::RegisterOperators("tensorrt_llm::relative_attention_bias", &torch_ext::buildRelativeAttentionBias);
|