TensorRT-LLMs/cpp/tensorrt_llm/thop/fp4xFp8GemmTrtllmGen.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

131 lines
4.5 KiB
C++

/*
* Copyright (c) 2020-2025, 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/multiHeadAttentionCommon.h"
#include "tensorrt_llm/kernels/trtllmGenKernels/gemm/KernelRunner.h"
#include "tensorrt_llm/thop/thUtils.h"
#include <ATen/cuda/EmptyTensor.h>
#include <ATen/native/cuda/Resize.h>
#include <cuda_fp16.h>
#include <cstdint>
TRTLLM_NAMESPACE_BEGIN
namespace torch_ext
{
namespace
{
namespace tg = gemm::trtllm::gen;
template <tg::Dtype outDtype>
void runGemm(at::Tensor& out, at::Tensor const& mat1, at::Tensor const& mat2, at::Tensor const& mat2Scale,
at::Tensor const& globalScale, int64_t m, int64_t n, int64_t k)
{
tensorrt_llm::kernels::TrtllmGenGemmRunnerOptions options = {.eltTypeA = tg::Dtype::E2m1,
.eltTypeB = tg::Dtype::E4m3,
.outputType = outDtype,
.deepSeekFp8 = false,
.transposeMmaOutput = true};
tensorrt_llm::kernels::TrtllmGenGemmRunner runner(options);
int64_t const numBytesWorkspace = runner.getWorkspaceSizeInBytes(m, n, k);
at::Tensor workspace
= at::detail::empty_cuda({numBytesWorkspace}, at::ScalarType::Char, torch::kCUDA, std::nullopt);
auto stream = at::cuda::getCurrentCUDAStream(mat1.get_device());
float* mat2ScalePtr = static_cast<float*>(mat2Scale.data_ptr());
float* outScalePtr = globalScale.data_ptr<float>();
runner.run(m, n, k, mat1.const_data_ptr(), nullptr, mat2.const_data_ptr(), /* bScale */ mat2ScalePtr,
out.data_ptr(), outScalePtr, /* cScalePtr */ nullptr, workspace.data_ptr(), stream.stream(), mat1.get_device());
}
at::Tensor fp4_fp8_gemm_impl(at::Tensor const& mat1, at::Tensor const& mat2, at::Tensor const& mat2Scale,
at::Tensor const& globalScale, std::optional<c10::ScalarType> out_dtype)
{
using tensorrt_llm::kernels::Data_type;
CHECK_INPUT(mat1, c10::ScalarType::Float8_e4m3fn);
CHECK_INPUT(mat2, FLOAT4_E2M1X2);
CHECK_INPUT(mat2Scale, c10::ScalarType::Float8_e4m3fn);
CHECK_INPUT(globalScale, c10::ScalarType::Float);
TORCH_CHECK(mat1.dim() == 2, "mat1 must be a matrix");
TORCH_CHECK(mat2.dim() == 2, "mat2 must be a matrix");
TORCH_CHECK(mat1.sizes()[1] == mat2.sizes()[1] * 2, "mat1 and mat2 shapes cannot be multiplied (", mat1.sizes()[0],
"x", mat1.sizes()[1], " and ", mat2.sizes()[0], "x", mat2.sizes()[1], ")");
auto const m = mat1.sizes()[0];
auto const n = mat2.sizes()[0];
auto const k = mat1.sizes()[1];
if (!out_dtype)
{
out_dtype = torch::kHalf;
}
TORCH_CHECK(out_dtype == torch::kFloat8_e4m3fn || out_dtype == torch::kHalf || out_dtype == torch::kBFloat16,
"out_dtype must be one of fp8/fp16/bf16. It defaults to fp16.");
at::Tensor out = at::detail::empty_cuda({m, n}, out_dtype.value(), mat1.device(), std::nullopt);
switch (out_dtype.value())
{
case at::ScalarType::Float8_e4m3fn:
runGemm<tg::Dtype::E4m3>(out, mat1, mat2, mat2Scale, globalScale, m, n, k);
break;
case at::ScalarType::Half: runGemm<tg::Dtype::Fp16>(out, mat1, mat2, mat2Scale, globalScale, m, n, k); break;
case at::ScalarType::BFloat16:
runGemm<tg::Dtype::Bfloat16>(out, mat1, mat2, mat2Scale, globalScale, m, n, k);
break;
default: C10_THROW_ERROR(NotImplementedError, "out_dtype must be one of fp8/fp16/bf16.");
}
return out;
}
} // namespace
at::Tensor fp4_fp8_gemm_trtllmgen(at::Tensor const& mat1, at::Tensor const& mat2, at::Tensor const& mat2Scale,
at::Tensor const& globalScale, std::optional<c10::ScalarType> out_dtype)
{
return fp4_fp8_gemm_impl(mat1, mat2, mat2Scale, globalScale, out_dtype);
}
} // namespace torch_ext
TRTLLM_NAMESPACE_END
TORCH_LIBRARY_FRAGMENT(trtllm, m)
{
m.def(
"fp4_fp8_gemm_trtllmgen(Tensor mat1, Tensor mat2, Tensor mat2Scale, Tensor globalScale, "
"ScalarType? out_dtype=None) -> Tensor");
}
TORCH_LIBRARY_IMPL(trtllm, CUDA, m)
{
m.impl("fp4_fp8_gemm_trtllmgen", &tensorrt_llm::torch_ext::fp4_fp8_gemm_trtllmgen);
}