mirror of
https://github.com/NVIDIA/TensorRT-LLM.git
synced 2026-01-14 06:27:45 +08:00
94dd456bd0
Simplify StatefulGptDecoderBatched constructor: - Remove speculative decoding mode parameter - Initialize with default mode=None - Update GptSession class accordingly Simplify setup method signatures in StatefulGptDecoder and StatefulGptDecoderBatched: - Remove maxTokensPerStep parameter - Initialize decoders with default maxTokensPerStep=1 - Update GptSession class accordingly Signed-off-by: Robin Kobus <19427718+Funatiq@users.noreply.github.com>
120 lines
4.3 KiB
C++
120 lines
4.3 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.
|
|
*/
|
|
|
|
#pragma once
|
|
|
|
#include "tensorrt_llm/executor/types.h"
|
|
#include "tensorrt_llm/runtime/bufferManager.h"
|
|
#include "tensorrt_llm/runtime/cudaEvent.h"
|
|
#include "tensorrt_llm/runtime/gptDecoder.h"
|
|
#include "tensorrt_llm/runtime/iStatefulGptDecoder.h"
|
|
|
|
#include <memory>
|
|
|
|
namespace tensorrt_llm::runtime
|
|
{
|
|
|
|
//! GPT decoder class with support for in-flight batching
|
|
class StatefulGptDecoder : public IStatefulGptDecoder
|
|
{
|
|
public:
|
|
StatefulGptDecoder(std::size_t vocabSize, std::size_t vocabSizePadded, CudaStreamPtr stream);
|
|
|
|
//! Setup the decoder before calling `forward()`
|
|
void setup(executor::DecodingMode const& mode, SizeType32 maxBatchSize, SizeType32 maxBeamWidth,
|
|
SizeType32 maxAttentionWindow, SizeType32 sinkTokenLength, SizeType32 maxSequenceLength,
|
|
nvinfer1::DataType dtype, ModelConfig const& modelConfig, WorldConfig const& worldConfig) override;
|
|
|
|
//! @brief Initialize the decoder with new batch of inputs.
|
|
void newBatch(GenerationInput const& input, GenerationOutput const& output, SamplingConfig const& samplingConfig,
|
|
ModelConfig const& modelConfig) override;
|
|
|
|
void forwardAsync(decoder::Output& output, decoder::Input const& input) override;
|
|
|
|
void forwardSync() override;
|
|
|
|
//! @brief Gather final results for all requests.
|
|
void finalize(SamplingConfig const& samplingConfig) const override;
|
|
|
|
//! @param step index within tokens generated in one step
|
|
//! @returns [batchSize, maxBeamWidth, maxInputLength + maxNewTokens], contains input token ids and generated token
|
|
//! ids without padding, on gpu
|
|
[[nodiscard]] TensorPtr getIds() const override
|
|
{
|
|
return mDecodingOutput->ids;
|
|
}
|
|
|
|
// This implementation is here to satisfy the interface requirement. Returns ids instead
|
|
[[nodiscard]] TensorPtr getGatheredIds() const override
|
|
{
|
|
return mDecodingOutput->ids;
|
|
}
|
|
|
|
//! @returns [batchSize, maxBeamWidth], cumulative log probabilities (per beam), on gpu
|
|
[[nodiscard]] TensorPtr getCumLogProbs() const override
|
|
{
|
|
return mDecodingOutput->cumLogProbs;
|
|
}
|
|
|
|
//! @returns [batchSize, maxBeamWidth], cumulative log probabilities (per beam), on gpu
|
|
[[nodiscard]] TensorPtr getLogProbs() const override
|
|
{
|
|
return mDecodingOutput->logProbs;
|
|
}
|
|
|
|
//! @brief Get tokens generated in one step of last forward pass
|
|
//! @param iter The iteration within [0; maxTokensPerStep) for which to get the tokens
|
|
//! @returns [batchSize, beamWidth], tokens generated in `iter` (per beam), on gpu
|
|
[[nodiscard]] TensorPtr getNewTokens(SizeType32 iter = 0) const override
|
|
{
|
|
TLLM_CHECK(iter == 0);
|
|
return mDecodingOutput->newTokens;
|
|
}
|
|
|
|
//! @returns [1], number of finished sequences, in pinned host memory
|
|
[[nodiscard]] TensorPtr getNbFinished() const override
|
|
{
|
|
return mFinishedSum;
|
|
}
|
|
|
|
private:
|
|
void reshapeBuffers(SizeType32 batchSize, SizeType32 beamWidth, SizeType32 mMaxAttentionWindow,
|
|
SizeType32 mSinkTokenLength, SizeType32 maxSequenceLength);
|
|
|
|
private:
|
|
std::size_t const mVocabSize;
|
|
std::size_t const mVocabSizePadded;
|
|
CudaStreamPtr mStream;
|
|
BufferManager mBufferManager;
|
|
|
|
using GptDecoderPtr = std::unique_ptr<IGptDecoder>;
|
|
GptDecoderPtr mDecoder;
|
|
using DecodingInputPtr = std::unique_ptr<DecodingInput>;
|
|
DecodingInputPtr mDecodingInput;
|
|
using DecodingOutputPtr = std::unique_ptr<DecodingOutput>;
|
|
DecodingOutputPtr mDecodingOutput;
|
|
CudaEvent mDecodedEvent{};
|
|
|
|
TensorPtr mFinishedSum;
|
|
TensorPtr mSetupBatchSlots;
|
|
|
|
SizeType32 mNbSteps;
|
|
SizeType32 mMaxSequenceLength{};
|
|
SizeType32 mMaxAttentionWindow{};
|
|
SizeType32 mSinkTokenLength{};
|
|
};
|
|
} // namespace tensorrt_llm::runtime
|