Skip to content

Conversation

@randgun
Copy link
Contributor

@randgun randgun commented Nov 21, 2025

  1. Add lightning indexer op and test file, this op can run at both eager mode and graph mode, max capture graph num is 1024.
  2. Add an GE helper tool, which provides a set of utilities to adapt engineering of GE framework

Accuracy test passed on GSM8K dataset
image

@gemini-code-assist
Copy link
Contributor

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

bool curActSeqLenIsZero = false;
bool needDealActS1LessThanS1 = false; // S1的实际长度小于shape的S1长度时,是否需要清理输出
uint32_t actMBaseSize = 0U; // m轴(gS1)方向实际大小
uint32_t mBasicSizeTail = 0U; // gS1方向循环的尾基本块大小
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Need to translate Chinese code comment to English

@randgun randgun force-pushed the br_indexer branch 3 times, most recently from 24c5522 to 65ec442 Compare November 26, 2025 06:35
@randgun
Copy link
Contributor Author

randgun commented Nov 27, 2025

/gemini check

@gemini-code-assist
Copy link
Contributor

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

@@ -1,5 +1,6 @@
# set the library output dir to the python dir for wheel package build
set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${PROJECT_SOURCE_DIR}/python/sgl_kernel_npu/sgl_kernel_npu/lib)
set(ASCEND_INCLUDE_DIR ${ASCEND_HOME_PATH}/aarch64-linux/include)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

pass env or option to replace ${ASCEND_HOME_PATH}/aarch64-linux/include

at::TensorOptions().dtype(at::kByte).device(query.options().device()));

if (actualCaptureNum >= MAX_CAPTURE_NUM) {
static auto preillTilingBuffer =
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

preillTilingBuffer => prefill
It is uncached or uncaptured tilingBuffer. not only prefill

return;
}
}
// throw std::runtime_error("SetAttrStr failed, attrName not exists");
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

dont comment this check

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants