Skip to content

[None][fix] Split mContextChunkSize into per-target/draft fields #12058

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Hrithvik-Alex wants to merge 3 commits into
base: main
Choose a base branch
from
+17 −10
Open

[None][fix] Split mContextChunkSize into per-target/draft fields #12058

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
5e7e8a8
[None][fix] Split mContextChunkSize into per-target/draft fields
Hrithvik-Alex Mar 7, 2026
1dcee26
Merge branch 'main' into fix-separate-kv-cache-chunk-size
Hrithvik-Alex Mar 10, 2026
3ee1d88
Merge branch 'main' into fix-separate-kv-cache-chunk-size
Hrithvik-Alex Mar 10, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 17 additions & 10 deletions cpp/include/tensorrt_llm/batch_manager/llmRequest.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -178,7 +178,8 @@ class GenericLlmRequest
, mLoraConfig(std::move(loraConfig))
, mLookaheadConfig(std::move(lookaheadConfig))
, mKvCacheRetentionConfig(std::move(kvCacheRetentionConfig))
, mContextChunkSize{mPromptLen}
, mContextChunkSizeTarget{mPromptLen}
, mContextChunkSizeDraft{mPromptLen}
, mLogProbs(samplingConfig.beamWidth)
, mCumLogProbs(samplingConfig.beamWidth)
, mDraftTokens(draftTokens.value_or(std::make_shared<VecTokens>()))
Expand Down Expand Up @@ -256,7 +257,8 @@ class GenericLlmRequest
, mLoraWeights(std::move(loraWeights))
, mLoraConfig(std::move(loraConfig))
, mLookaheadConfig(lookaheadConfig)
, mContextChunkSize(mPromptLen)
, mContextChunkSizeTarget(mPromptLen)
, mContextChunkSizeDraft(mPromptLen)
, mLogProbs(samplingConfig.beamWidth)
, mCumLogProbs(samplingConfig.beamWidth)
, mDraftTokens(std::make_shared<VecTokens>(draftTokens.value_or(VecTokens())))
Expand Down Expand Up @@ -293,7 +295,8 @@ class GenericLlmRequest
, mOrigPromptLen(mPromptLen)
, mNumPreDecodedTokens(mSamplingConfig.beamWidth, 0)
, mMaxSentTokenLen(mPromptLen)
, mContextChunkSize{mPromptLen}
, mContextChunkSizeTarget{mPromptLen}
, mContextChunkSizeDraft{mPromptLen}
, mLogProbs(mSamplingConfig.beamWidth)
, mCumLogProbs(mSamplingConfig.beamWidth)
, mDraftTokens(std::make_shared<VecTokens>())
Expand Down Expand Up @@ -861,7 +864,8 @@ class GenericLlmRequest
mContextCurrentPositionDraft = 0;
mPrepopulatedPromptLenTarget = 0;
mPrepopulatedPromptLenDraft = 0;
mContextChunkSize = mPromptLen;
mContextChunkSizeTarget = mPromptLen;
mContextChunkSizeDraft = mPromptLen;
mSeqSlot.reset();
}

Expand Down Expand Up @@ -1590,7 +1594,7 @@ class GenericLlmRequest
TLLM_CHECK_WITH_INFO(
isContextInitState() || isDisaggGenerationInitState() || isDisaggGenerationTransmissionComplete(),
"getContextChunkSize is only possible during the context phase or generation init phase.");
return mContextChunkSize;
return mUseDraftModel ? mContextChunkSizeDraft : mContextChunkSizeTarget;
}

/// To set the context chunk size, throw an exception when the chunk size is negative. If the chunk
Expand All @@ -1602,7 +1606,8 @@ class GenericLlmRequest
isContextInitState() || isDisaggGenerationInitState() || isDisaggGenerationTransmissionComplete(),
"setContextChunkSize is only possible during the context phase or generation init phase.");
TLLM_CHECK_WITH_INFO(size >= 0, "The chunk size of context (%d) can't be negative.", size);
mContextChunkSize = std::min(size, getContextRemainingLength());
auto& contextChunkSize = mUseDraftModel ? mContextChunkSizeDraft : mContextChunkSizeTarget;
contextChunkSize = std::min(size, getContextRemainingLength());
}

/// Determines whether the current position is only one chunk away from the end of the context.
Expand All @@ -1625,9 +1630,10 @@ class GenericLlmRequest
{
TLLM_CHECK_WITH_INFO(isContextInitState(), "Chunking is only possible during the context phase.");

mContextCurrentPositionDraft += getContextChunkSize();
mContextCurrentPositionTarget += getContextChunkSize();
setContextChunkSize(0);
mContextCurrentPositionDraft += mContextChunkSizeDraft;
mContextCurrentPositionTarget += mContextChunkSizeTarget;
mContextChunkSizeDraft = 0;
mContextChunkSizeTarget = 0;
}

[[nodiscard]] executor::PriorityType priority() const noexcept
Expand Down Expand Up @@ -1987,7 +1993,8 @@ class GenericLlmRequest
// Paged-KV-Cache must be enabled while enabling Chunked-Context.
// The size of the context chunk must be multiple of the KV-Cache block size except the last one.
// Value `0` means Chunked-Context is disabled.
SizeType32 mContextChunkSize{0};
SizeType32 mContextChunkSizeTarget{0};
SizeType32 mContextChunkSizeDraft{0};
SizeType32 mContextCurrentPositionTarget{0};
SizeType32 mContextCurrentPositionDraft{0};

Expand Down