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>
112 lines
3.4 KiB
C++
112 lines
3.4 KiB
C++
/*
|
|
* Copyright (c) 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.
|
|
*/
|
|
|
|
#pragma once
|
|
|
|
#include "tensorrt_llm/common/cudaUtils.h"
|
|
#include "tensorrt_llm/runtime/common.h"
|
|
|
|
namespace tensorrt_llm::runtime
|
|
{
|
|
|
|
class SpeculativeDecodingModule
|
|
{
|
|
public:
|
|
explicit SpeculativeDecodingModule(
|
|
SizeType32 maxDraftPathLen, SizeType32 maxDecodingDraftTokens, SizeType32 maxNumPaths) noexcept
|
|
: mMaxDraftPathLen(maxDraftPathLen)
|
|
, mMaxDecodingDraftTokens(maxDecodingDraftTokens)
|
|
, mMaxNumPaths(maxNumPaths)
|
|
{
|
|
computeNumPackedMasks();
|
|
}
|
|
|
|
explicit SpeculativeDecodingModule() noexcept
|
|
: SpeculativeDecodingModule(0, 0, 0)
|
|
{
|
|
}
|
|
|
|
virtual ~SpeculativeDecodingModule() = default;
|
|
|
|
SpeculativeDecodingModule(SpeculativeDecodingModule const& o) = default;
|
|
SpeculativeDecodingModule& operator=(SpeculativeDecodingModule const& o) = default;
|
|
|
|
/// @return max number of draft tokens that can be accepted by one step of the decoder
|
|
[[nodiscard]] SizeType32 getMaxDraftPathLen() const noexcept
|
|
{
|
|
return mMaxDraftPathLen;
|
|
}
|
|
|
|
/// @return max number of tokens that a request can grow in one step of the decoder
|
|
/// @details one more than draft path len for prediction from primary head
|
|
[[nodiscard]] SizeType32 getMaxPathLen() const noexcept
|
|
{
|
|
return getMaxDraftPathLen() + 1;
|
|
}
|
|
|
|
/// @return max number of draft tokens processed by one step of the decoder
|
|
[[nodiscard]] SizeType32 getMaxDecodingDraftTokens() const noexcept
|
|
{
|
|
return mMaxDecodingDraftTokens;
|
|
}
|
|
|
|
/// @return max number of tokens processed by one step of the decoder
|
|
/// @details one more than decoding draft tokens for prediction from primary head
|
|
[[nodiscard]] SizeType32 getMaxDecodingTokens() const noexcept
|
|
{
|
|
return getMaxDecodingDraftTokens() + 1;
|
|
}
|
|
|
|
[[nodiscard]] SizeType32 getNumPackedMasks() const noexcept
|
|
{
|
|
return mMaxNumPackedMasks;
|
|
}
|
|
|
|
[[nodiscard]] SizeType32 getMaxNumPaths() const noexcept
|
|
{
|
|
return mMaxNumPaths;
|
|
}
|
|
|
|
void setMaxDraftTokens(SizeType32 maxDraftTokens) noexcept
|
|
{
|
|
mMaxDecodingDraftTokens = maxDraftTokens;
|
|
computeNumPackedMasks();
|
|
}
|
|
|
|
void setMaxDraftPathLen(SizeType32 maxDraftPathLen) noexcept
|
|
{
|
|
mMaxDraftPathLen = maxDraftPathLen;
|
|
}
|
|
|
|
void setMaxNumPaths(SizeType32 maxNumPaths) noexcept
|
|
{
|
|
mMaxNumPaths = maxNumPaths;
|
|
}
|
|
|
|
private:
|
|
void computeNumPackedMasks() noexcept
|
|
{
|
|
mMaxNumPackedMasks = tensorrt_llm::common::divUp(mMaxDecodingDraftTokens, 32);
|
|
}
|
|
|
|
private:
|
|
SizeType32 mMaxDraftPathLen; // max length per path (or ray/branch)
|
|
SizeType32 mMaxDecodingDraftTokens; // max combined length of all paths (or rays/branches)
|
|
SizeType32 mMaxNumPaths; // max number of paths (or rays/branches)
|
|
SizeType32 mMaxNumPackedMasks;
|
|
};
|
|
} // namespace tensorrt_llm::runtime
|