Skip to content

feat: enable SM103 (B300 / Blackwell Ultra) across the attention SM allowlists #207

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
filipemartinsubrobotics wants to merge 1 commit into
base: main
Choose a base branch
from
Open

feat: enable SM103 (B300 / Blackwell Ultra) across the attention SM allowlists #207

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 1 addition & 1 deletion cpp/kernels/contextAttentionKernels/cuteDslFMHARunner.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ namespace

bool isSupportedBlackwellFmha(int32_t smVersion)
{
return smVersion == 100 || smVersion == 101 || smVersion == 110;
return smVersion == 100 || smVersion == 101 || smVersion == 103 || smVersion == 110;
}

template <auto moduleLoader, auto moduleUnloader, typename Module>
Expand Down
3 changes: 2 additions & 1 deletion cpp/kernels/contextAttentionKernels/cuteDslFMHAV2Runner.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,7 +64,8 @@ namespace
bool isFMHAV2SM(int32_t smVersion)
{
return smVersion == 80 || smVersion == 86 || smVersion == 87 || smVersion == 89 || smVersion == 90
|| smVersion == 100 || smVersion == 101 || smVersion == 110 || smVersion == 120 || smVersion == 121;
|| smVersion == 100 || smVersion == 101 || smVersion == 103 || smVersion == 110 || smVersion == 120
|| smVersion == 121;
}

bool isFMHAV2PagedLlmHeadSize(int32_t headSize)
Expand Down
2 changes: 1 addition & 1 deletion cpp/kernels/decodeAttentionKernels/decoderXQAJitCompiler.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -251,7 +251,7 @@ bool canCompileXQAKernel(int32_t numQHeads, int32_t numKVHeads, int32_t headSize
nvinfer1::DataType dataType, nvinfer1::DataType kvDataType) noexcept
{
// SM versions with an XQA decode kernel at all.
constexpr std::array<int32_t, 9> kALLOWED_SM_VERSIONS{80, 86, 87, 89, 90, 100, 101, 120, 121};
constexpr std::array<int32_t, 10> kALLOWED_SM_VERSIONS{80, 86, 87, 89, 90, 100, 101, 103, 120, 121};
// Of those, the ones with native FP8 converts. Listed explicitly rather
// than tested as `smVersion >= 89`: a numeric threshold silently admits
// every future SM added to kALLOWED_SM_VERSIONS, and whether that SM has
Expand Down
Loading