TensorRT-LLMs/cpp/tensorrt_llm/thop/thUtils.cpp
Yihan Wang 9df4dad3b6
[None][fix] Introduce inline namespace to avoid symbol collision (#9541)
Signed-off-by: Yihan Wang <yihwang@nvidia.com>
2025-12-12 23:32:15 +08:00

118 lines
3.6 KiB
C++

/*
* Copyright (c) 2020-2023, 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/thop/thUtils.h"
#include <NvInferRuntime.h>
#include <array>
TRTLLM_NAMESPACE_BEGIN
namespace torch_ext
{
tensorrt_llm::runtime::ITensor::Shape convert_shape(torch::Tensor tensor)
{
constexpr auto trtMaxDims = nvinfer1::Dims::MAX_DIMS;
auto const torchTensorNumDims = tensor.dim();
TLLM_CHECK_WITH_INFO(torchTensorNumDims <= trtMaxDims,
"TensorRT supports at most %i tensor dimensions. Found a Torch tensor with %li dimensions.", trtMaxDims,
torchTensorNumDims);
auto result = nvinfer1::Dims{};
result.nbDims = static_cast<int32_t>(torchTensorNumDims);
for (int i = 0; i < torchTensorNumDims; i++)
{
result.d[i] = static_cast<int64_t>(tensor.size(i));
}
return result;
}
template <typename T>
tensorrt_llm::runtime::ITensor::UniquePtr convert_tensor(torch::Tensor tensor)
{
return tensorrt_llm::runtime::ITensor::wrap(
get_ptr<T>(tensor), tensorrt_llm::runtime::TRTDataType<T>::value, convert_shape(tensor));
}
// Template instantiations
template tensorrt_llm::runtime::ITensor::UniquePtr convert_tensor<int32_t*>(torch::Tensor tensor);
template tensorrt_llm::runtime::ITensor::UniquePtr convert_tensor<int32_t>(torch::Tensor tensor);
template tensorrt_llm::runtime::ITensor::UniquePtr convert_tensor<uint8_t>(torch::Tensor tensor);
template tensorrt_llm::runtime::ITensor::UniquePtr convert_tensor<int8_t>(torch::Tensor tensor);
template tensorrt_llm::runtime::ITensor::UniquePtr convert_tensor<float>(torch::Tensor tensor);
template tensorrt_llm::runtime::ITensor::UniquePtr convert_tensor<half>(torch::Tensor tensor);
#ifdef ENABLE_BF16
template tensorrt_llm::runtime::ITensor::UniquePtr convert_tensor<__nv_bfloat16>(torch::Tensor tensor);
#endif
template tensorrt_llm::runtime::ITensor::UniquePtr convert_tensor<bool>(torch::Tensor tensor);
std::optional<float> getFloatEnv(char const* name)
{
char const* const env = std::getenv(name);
if (env == nullptr)
{
return std::nullopt;
}
try
{
float value = std::stof(env);
return {value};
}
catch (std::invalid_argument const& e)
{
return std::nullopt;
}
catch (std::out_of_range const& e)
{
return std::nullopt;
}
};
int nextPowerOfTwo(int v)
{
--v;
v |= v >> 1;
v |= v >> 2;
v |= v >> 4;
v |= v >> 8;
v |= v >> 16;
return ++v;
}
cudaDataType_t convert_torch_dtype(torch::ScalarType dtype)
{
switch (dtype)
{
case torch::kInt32: return CUDA_R_32I;
case torch::kFloat: return CUDA_R_32F;
#ifdef ENABLE_BF16
case torch::kBFloat16: return CUDA_R_16BF;
#endif
case torch::kHalf: return CUDA_R_16F;
#ifdef ENABLE_FP8
case torch::kFloat8_e4m3fn: return CUDA_R_8F_E4M3;
case torch::kFloat8_e5m2: return CUDA_R_8F_E5M2;
#endif
case torch::kByte: return CUDA_R_8U;
case torch::kChar: return CUDA_R_8I;
case torch::kBool: return CUDA_R_8U;
default: TLLM_THROW("Unsupported torch dtype %s to convert", c10::toString(dtype));
}
}
} // namespace torch_ext
TRTLLM_NAMESPACE_END